Datasets:

Hendrik-a

/

arxiv

like 0

we have used the caltech submillimeter observatory ( cso ) to follow - up a sample of wise - selected , hyperluminous galaxies , so called w1w2-dropout galaxies . this is a rare ( all - sky ) population of galaxies at high redshift ( peaks at=2 - 3 ) , that are faint or undetected by wise at 3.4 and 4.6 m , yet are clearly detected at 12 and 22 m . the optical spectra of most of these galaxies show significant agn activity . we observed 14 high - redshift ( ) w1w2-dropout galaxies with sharc - ii at 350 to 850 m , with 9 detections ; and observed 18 with bolocam at 1.1 mm , with five detections . warm _ spitzer _ follow - up of 25 targets at 3.6 and 4.5 , as well as optical spectra of 12 targets are also presented in the paper . combining wise data with observations from warm _ spitzer _ and cso , we constructed their mid - ir to millimeter spectral energy distributions ( seds ) . these seds have a consistent shape , showing significantly higher mid - ir to submm ratios than other galaxy templates , suggesting a hotter dust temperature . we estimate their dust temperatures to be k using a single - temperature model . their infrared luminosities are well over 10l . these seds are not well fitted with existing galaxy templates , suggesting they are a new population with very high luminosity and hot dust . they are likely among the most luminous galaxies in the universe . we argue that they are extreme cases of luminous , hot dust - obscured galaxies ( dogs ) , possibly representing a short evolutionary phase during galaxy merging and evolution . a better understanding of their long - wavelength properties needs alma as well as _ herschel _ data .

1208.5518

the redshift @xmath82 - 3 epoch stands out as a unique era for studying galaxy formation and evolution . at this epoch , the cosmic star formation rate reaches its peak ( heavens et al . 2004 ; hopkins & beacom 2006 ; reddy et al . 2008 ) , and ultra - luminous infrared galaxies ( ulirgs , @xmath9 , sanders & mirabel 1996 ) contribute a significant fraction to the infrared luminosity density ( elbaz et al . 2002 , chapman et al . 2005 , caputi et al . 2007 , reddy et al . 2008 , magnelli et al . 2009 ) . the cosmic quasar density also peaks around @xmath82 ( hopkins et al . 2007 , assef et al . 2011 ) . a framework of galaxy evolution through major mergers has been gradually built up by theorists ( barnes & hernquist 1992 , schweizer 1998 , jogee 2006 , hopkins et al . 2006 , 2008 ) . in one of the most popular scenarios ( e.g. hopkins et al . 2008 ) , the tidal torques generated by major mergers funnel gas into the center of galaxies , leading to a central starburst and rapid growth of a supermassive black hole ( smbh ) . black hole and supernova feedback terminate further star formation , evacuating the residual gas and dust , leaving a visible quasar and remnant spheroid . this picture establishes the evolutionary connections between ulirgs , quasars , and massive elliptical galaxies . submillimeter galaxies ( smgs ) are thought to be the analogues of local ulirgs at high redshift ( blain et al . 2002 , tacconi et al . smgs are selected by their strong cold dust emission at 850 @xmath2 m ( @xmath105 mjy ) . they are characterized by very high star formation rates ( 100 - 1000 m@xmath11yr@xmath12 ) and infrared luminosity ( @xmath13 , chapman et al . 2005 , magnelli et al 2012 ) . although most smgs host growing black holes ( e.g. , alexander et al . 2005 , 2008 ) , their luminosities are normally dominated by star formation ( swinbank et al . 2004 ; menndez - delmestre et al . 2007 , younger et al . 2008 , hainline et al . the redshift distribution of smgs strongly peaks at @xmath1=2 - 3 ( chapman et al . 2005 ) , and the surface density of smgs is several hundred per square degree . an 850@xmath2 m selected sample ( smgs ) may be biased toward ulirgs with large amounts of dust , but miss a substantial population of ulirgs with a smaller amount of ( but warmer ) dust , which can be found by surveys at shorter wavelengths . a series of surveys using bright _ spitzer _ 24@xmath2 m emission combined with optically faint photometry have been carried out to probe the ulirg population with emission from smaller and warmer dust grains ( e.g. , rigby et al . 2004 , donley et al . 2007 , yan et al . 2007 , farrah et al . 2008 , soifer et al . 2008 , lonsdale et al . 2009 , huang et al . one of the simplest search criteria is given as @xmath14 mjy , and @xmath15 > 14 $ ] ( where @xmath16 and [ 24 ] are vega magnitudes for @xmath16 band and _ spitzer _ 24 @xmath2 m ) , or roughly @xmath17 ( dey et al . 2008 , fiore et al . 2008 ) , leading to a well defined @xmath182 population which is referred to as dust obscured galaxies ( dey et al . 2008 , hereafter d08 ) . the most luminous dogs have star formation rates ( 500 - 1000 m@xmath11yr@xmath12 or more ) and infrared luminosities ( @xmath19 , bussmann et al 2009 , tyler et al . 2009 , melbourne et al . 2012 ) that are comparable to smgs . it has been proposed that both smgs and dogs are an early phase of galaxy merging , with smgs representing an earlier , starburst - dominant phase , while luminous dogs are in a transitional phase from starburst - dominated to agn - dominated ( e.g. narayanan 2010 ) . the bolometric luminosities also reach their maximum during these phases , making the most luminous galaxies in these phases also among the most luminous objects in the universe . looking for the most luminous galaxies in the universe is one of the major goals of nasa s wide - field infrared survey explorer ( wise , wright et al . wise surveyed the entire sky at 3.4 , 4.6 , 12 and 22 @xmath2 m ( hereafter w1 , w2 , w3 , w4 ) in 2010 . the wise dataset is well suited to investigate the starburst - agn phase of distant , infrared luminous galaxies . at @xmath18 2 - 3 , starburst- or agn - heated hot dust can be traced by 12 and 22 @xmath2 m emission , while the rest near infrared ( nir ) obscuration is sampled by 3.4 and 4.6 @xmath2 m continuum . studies of luminous infrared galaxies with the wise w1 , w2 and w4 bands can take advantage of existing knowledge and techniques developed by earlier studies with _ spitzer _ at similar wavelengths ( irac at 3.6 and 4.5 @xmath2 m , and mips at 24 @xmath2 m ) . observing w4 selected galaxies with wise is similar to observing 24 @xmath2 m bright galaxies with _ spitzer _ , but with the surveyed area enlarged from a few tens of square degrees covered by existing dog surveys to the entire sky . in order to search for hyperluminous infrared galaxies ( hylirgs , @xmath20 ) from the wise dataset , the wise team has explored multiple methods to select candidates . the most productive method so far has been to search for more heavily obscured galaxies , whose w1 ( 3.4 @xmath2 m ) and w2 ( 4.6 @xmath2 m ) flux densities become faint or undetected by wise , while remaining easily detectable at 12 and/or 22 @xmath2 m , with typical w4 ( 22 @xmath2 m ) flux densities @xmath21 . we call this population `` w1w2-dropouts '' ( eisenhardt et al 2012 ) or for brevity `` w12drops '' . follow - up spectroscopy of more than 100 w12drop galaxies at large telescopes ( this paper , eisenhardt et al . ( 2012 ) , see also bridge et al . 2012 ) reveals that a large fraction ( @xmath22 ) of these galaxies are at high redshift ( @xmath23 ) , with the highest at @xmath1=4.6 . most of the redshifts are between 2 and 3 , suggesting they also trace the peak epoch of cosmic star formation and qso activity . at these redshifts , such high flux densities at 22 @xmath2 m imply extremely high luminosities . they are potentially hyperluminous galaxies . in order to understand the dust properties and calculate the total luminosities of these unusual galaxies , continuum measurements at longer wavelengths are crucial . as the first high redshift examples were identified , we began follow - up 0.35 - 1.1 mm continuum observations using the caltech submillimeter observatory ( cso ) , in order to construct their seds and to explore the nature of w12drop galaxies . in this paper , we report the initial results of this follow - up study . the wise data are described in section 2.1 , and the w12drop population followed up with the cso and reported here is listed in table 1 . sections 2.2 and 2.3 describe the cso data , while section 2.4 describes the follow - up optical spectroscopy , which is summarized in table 2 . section 2.5 describes spitzer follow - up observations at 3.6 @xmath2 m and 4.5 @xmath2 m of the w12drops , which were selected to be difficult to detect by wise at w1 ( 3.4 @xmath2 m ) and w2 ( 4.6 @xmath2 m ) , and the photometry for the sources is presented in table 3 . section 3 presents luminosity and dust temperatures constraints from the photometry , while section 4 compares w12drop properties to those of dogs and smgs , and section 5 summarizes the findings . throughout this paper we assume a cdm cosmology with @xmath24 km s@xmath12 mpc@xmath12 , @xmath25 , and @xmath26 . wise began surveying the sky a month after it launched on 2009 december 14 , completing its first coverage of the sky six months later , and continued surveying until february 1 , 2011 . the wise all - sky data release was made public on 2012 march 14 , and its content and characteristics are documented in the explanatory supplement . wise has achieved much better sensitivity than previous all - sky survey missions ( 5@xmath27 point source sensitivities are better than 0.07 , 0.1 , 0.9 and 5.4 mjy in w1 , w2 , w3 and w4 bands ( explanatory supplement ) , and identified hundreds of millions of sources . the selection criteria for w12drops use wise catalog photometry , which provides psf - ftting ( i.e. total ) magnitudes and uncertainties in the vega system . the criteria are w1 @xmath28 17.4 , and either : a ) w4 @xmath297.7 and w2@xmath30w4 @xmath28 8.2 ; or b ) w3 @xmath31 10.6 and w2@xmath30w3 @xmath28 5.3 ) . additional details are given in eisenhardt et al . this selection yields only about 1000 targets over the full sky . w3 and w4 flux densities and uncertainties for the w12drops discussed in this paper are provided in table 3 , converted from catalog magnitudes using zero - points of 29.04 and 8.284 jy for w3 and w4 respectively ( wright et al . no color corrections have been made to these zero - points , because the results presented here are not sensitive to such corrections . the submillimeter high angular resolution camera ii ( sharc - ii ) installed at the 10.4 m cso telescope ( dowell et al . 2002 ) is a background - limited 350 and 450 @xmath2 m facility camera , and it is also equipped with a filter that allows 850 @xmath2 m continuum observations . it adopts a `` ccd - style '' bolometer array with 12 @xmath32 32 pixels , resulting in a 0.97@xmath33 field of view . the fwhm beam sizes of sharc - ii at 350 @xmath2 m , 450 @xmath2 m , and 850 @xmath2 m are @xmath34 , @xmath35 , and @xmath36 , respectively . we used sharc - ii to follow - up 14 high redshift w12drop galaxies during runs in july and september of 2010 , and in february and september of 2011 ( see table 1 ) . most of the targets were only observed at 350 @xmath2 m , with a few also observed at 450 @xmath2 m and 850 @xmath2 m . examples of sharc - ii images of detected w12drop galaxies are presented in figure 1 . since the 350 and 450 @xmath2 m atmospheric transmission is very sensitive to the weather , we only observed targets under good weather conditions ( i.e. when the opacity at 225 ghz @xmath37 , which occurs @xmath38 20% of the time ) . the targets and observing information are listed in table 1 . the dish surface optimization system ( dsos , leong 2006 ) was used to correct the dish surface figure for imperfections and gravitational deformations as the dish moved in elevation during observations . we used the comprehensive reduction utility for sharc - ii ( crush , kovcs 2006 ) , version crush2.01 - 4 , to reduce the sharc - ii data . option `` -deep '' in crush was applied to optimize the signal - to - noise ratio ( s / n ) for faint ( @xmath31 100 mjy ) point sources . planets uranus and neptune , when available , were used for flux calibration , focus correction and pointing correction , and secondary calibrators such as k3 - 50 and crl618 were used when these planets were not available . in order to flux calibrate , we used the starlink s `` astro '' package to calculate the flux density of the calibrator within the proper beam size for a sharc - ii band , and the `` show '' package in crush to obtain the readout flux density of the observed calibrator . the calibration factors ( peaks ) were derived by dividing the real peak flux density of the calibrator by the readout peak flux density from crush within one beam . the flux density of the target and the rms noise were then derived by applying this calibration factor to the readout of the peak position of the target and off - peak positions 1 - 2 beams from the peak , using the crush `` show '' package in the same way as for the calibrator , and convolving with the same beam size . the statistics of the calibration factor over all our runs indicates a calibration uncertainty of 20% . we used the sweep mode for the sharc - ii observations , in which the telescope moves in a lissajous pattern that keeps the central regions of the maps fully sampled . the edges are much noisier than the central regions , and to compensate for this , we used `` imagetool '' in crush to eliminate the regions of each map that had a total exposure time less than 25@xmath39 of the maximum . pointing was checked every hour with planets and secondary calibrators . the pointing drift was normally less than 3@xmath40 in both the azimuth and zenith directions between pointing checks , and a pointing correction has been applied during the data stacking . the uncertainty on the centroid position of an object will be the quadrature sum of the pointing uncertainty and the measurement uncertainty on the centroid , which approximately equals to the beam fwhm divided by s / n . given our sharc - ii detections normally have s / n of 3 - 4 , and beam size at 350@xmath4 is @xmath34 , the measurement uncertainty on the centroid for our targets is about 2@xmath40 - 3 @xmath40 . therefore , the uncertainty of offset between the sharc - ii centroids and wise positions should be less than 5@xmath40 - 6 @xmath40 , which is consistent with our sharc - ii detections ( figure 1 ) . bolocam is a large format camera at the cso with 144 detectors , a circular eight - arcmin - diameter field of view ( fov ) , an observing band centered at 1.1 mm , and a point - spread function with a 30 arcsec full - width at half - maximum ( fwhm ) ( haig et al . eighteen w12drop galaxies were observed with bolocam during runs in june 2010 , december 2010 , febuary 2011 , and september 2011 , with observing information listed in table 1 . the observations were made by scanning the cso in a lissajous pattern , with scanning parameters chosen to keep the source within the fov 100% of the time ( sayers et al . 2011 ) while scanning at an average speed of 2@xmath41/sec in june 2010 , december 2010 and february 2011 and an average speed of 4 @xmath41/sec in september 2011 . a pointing model , accurate to 5@xmath40 , was created from frequent observations of bright objects within @xmath42 degrees of the target galaxies following the methods described in sayers et al . ( 2009 ) . the flux calibration , in nv / jy , was determined according to the procedure described in laurent et al . ( 2005 ) , based on observations of neptune and k3 - 50a in june 2010 , irc 10216 in december 2010 , g34.3 and ngc 2071ir in february 2011 , and uranus in september 2011 ( sandell 1994 ; griffin and orton 1993 ) . the uranus calibration model of griffin and orton 1993 was updated based on the 143 ghz bolocam results described in sayers et al . 2012 . we estimate the flux calibration to be accurate to @xmath43 . our atmospheric noise subtraction procedure was based on the algorithms described in sayers et al . ( 2011 ) , with the following modifications : 1 ) an adaptive principal component analysis ( pca ) algorithm was used to subtract the correlated atmospheric signal over the fov ( laurent et al . 2005 , downes et al . 2012 ) and 2 ) the data were then high - pass filtered at a characteristic frequency of 400 mhz , which corresponds to an angular scale of 5@xmath41 and 10@xmath41 for the data collected at 2@xmath41/sec and 4@xmath41/sec . as described in detail in sayers et al . ( 2011 ) , the atmospheric noise subtraction also attenuated the astronomical signal . to account for this signal attenuation , observations of the flux calibration sources were processed in an identical way prior to determining the flux calibration . although the adaptive pca algorithm is non - linear , we verified via simulation that the combination of adaptive pca and a 400 mhz high - pass filter results in a constant fractional amount of signal attenuation for point - like objects with flux densities @xmath44 jy , which is well above the flux density of any of our flux calibration sources . we estimated the noise in our images via jackknife realizations of our data following the methods described in sayers et al . these jackknifes involve multiplying a randomly selected subset of half of our data by -1 . by adding these jackknifes together we preserve the noise properties of the data while removing the astronomical signal . in addition to instrumental and atmospheric noise , some of our images also contain a non - negligible amount of noise due to fluctuations in the astronomical ( back)foregrounds . using the galaxy number counts model in bethermin & dole ( 2011 ) , we estimate the confusion noise ( quantified by the rms noise fluctuations on beam size scales ) to be 0.6 mjy , which is approximately the confusion noise measured at the same wavelength / resolution with aztec ( scott et al . 2010 ) . the total uncertainty on the flux density of a galaxy is then given by the quadrature sum of instrument / measurement noise and confusion noise . the noise fluctuations in the map are gaussian within our ability to measure them . for our non - detections we quote 95% confidence level upper limits based on the formalism given in feldman and cousins ( 1998 ) , who provided a rigorous method for quoting upper limits from a measurement with gaussian noise and a physical constraint that the true underlying value is non - negative ( which is the case for our measurements , since negative flux densities are unphysical ) . specifically , our upper limits are computed from the values given in table 10 of feldman and cousins ( 1998 ) . we obtained optical spectroscopy of the wise - selected sources over the course of four observing runs between july 2010 and february 2011 . optical spectroscopic results for 14 sources in table 1 are reported in eisenhardt et al . ( 2012 ) and bridge et al . ( in preparation ) , as noted in table 1 . table 2 lists the primary observing parameters , including telescope , instrument , observing date , and integration time , for the remaining 12 w12drop galaxies . all of the targets were observed with multiple exposures , which were generally dithered along the slit to improve the reduction quality . table 2 also lists the resultant redshifts , most of which are based on multiple features and are therefore considered secure . most targets were observed with the dual - beam low resolution imaging spectrometer ( lris ; oke et al . 1995 ) on the keck i telescope . all of the lris observations used the 15 wide longslit , the 5600 dichroic , and the 400 @xmath45 grating on the red arm of the spectrograph ( blazed at 8500 ; spectral resolving power @xmath46 for objects filling the slit ) . the july 2010 observations used the 600 @xmath47 grism on the blue arm of the spectrograph ( @xmath48 ; @xmath49 ) , while the 2011 observations used the slightly lower resolution 400 @xmath45 blue grism ( @xmath50 ; @xmath51 ) . observations were generally obtained at a position angle that placed a brighter offset star on the slit . since lris has an atmospheric dispersion corrector , there are no issues with lost light due to observing at non - parallactic angles . the final 2 sources , w0220 + 0137 and w0116@xmath300505 , were observed with the blue channel spectrograph ( bcs ) on the 6.5-m mmt telescope in non - photometric conditions on ut 2010 december 4 . these observations used the 15 wide longslit and the 500 @xmath45 grating ( @xmath52 ; @xmath53 ) , and were obtained at a position angle of 47.3 @xmath54 . we processed the data using standard procedures , including bias subtraction , gain correction , cosmic ray removal , sky subtraction and stacking the 2-dimensional spectra ( e.g. , stern et al . 2010 ) . the spectra were extracted using a 15 aperture and wavelength calibrated using internal arc lamps . as a final step in the wavelength calibration , we shifted the wavelength solution based on telluric emission and absorption lines , conservatively providing wavelength solutions that are robust to better than 1 . we flux calibrated the spectra using observations of standard stars from massey & gronwall ( 1990 ) , generally observed on the same night as the science observations . for photometric nights , we estimate that the flux calibration is accurate to 10% . for non - photometric data , which includes the two sources observed with mmt , the spectrophotometry is less accurate . the final , reduced spectra are presented in figure 2 . the sources range in redshift from @xmath55 to @xmath56 and all but three of the sources ( w0338 + 1941 , w0926 + 4232 and w1830 + 6504 ) are clearly agns as evidenced by strong , high - ionization emission lines such as , , and/or ] . one of the three outliers , w0338 + 1941 , has an unusually broad ly@xmath57 profile , indicating that it is also likely an agn . note the diversity of spectroscopic features . most of the source spectra are dominated by strong , narrow ly@xmath57 emission . some of the sources are clearly type-2 agns , with only narrow emission features visible ( e.g. , w1422 + 5613 ) . we highlight two sources with unusual spectra . w0542@xmath302705 shows a large number of moderate - width ( @xmath58 ) emission features . of particular note is the strong @xmath59 emission , which is stronger than the ] @xmath60 emission ( / ] @xmath61 ) . in contrast , the corresponding line ratio for the vanden berk et al . ( 2001 ) sdss quasar composite is / @xmath62 , approximately two orders of magnitude weaker . the other unusual spectrum is w0926 + 4232 which does not show any emission features , though multiple absorption features as well as a lyman forest break clearly indicate a redshift of @xmath63 , analogous to the the spectrum of w1814 + 3412 reported in eisenhardt et al . ( 2012 ) . warm _ _ spitzer _ _ observations at 3.6 and 4.5 @xmath2 m of all the galaxies except one ( w0211@xmath302422 ) in table 1 were obtained under program 70162 between november 2010 and july 2011 . the sources were observed using five exposures of 30s in each irac band ( irac has a 5@xmath64 field of view with @xmath65 pixels ) , with the medium scale reuleaux dither pattern . the _ _ spitzer _ _ pipeline post - bcd processed images , which are resampled onto @xmath66 pixels , were used for photometry . all targets in this paper with warm _ _ spitzer _ _ follow - up were detected in both irac bands . flux densities at 3.6 and 4.5 @xmath2 m were measured in @xmath67 diameter apertures and are listed in table 3 . the aperture corrections are 1.205 and 1.221 for 3.6 and 4.5 @xmath2 m , respectively . we did not apply a color correction , since the results presented here are not sensitive to such corrections . for w0211@xmath302422 , which was not observed by _ _ spitzer _ _ , we list its w1 and w2 flux densities in table 3 . of the 14 high redshift w12drop galaxies observed with sharc - ii at 350 @xmath2 m , nine were detected at @xmath38 3 @xmath27 or above . the relatively high detection rate of these w12drop galaxies at 350 @xmath2 m implies they are a submillimeter bright population with high infrared luminosity . we also observed three of the 14 targets using the sharc - ii 450 @xmath2 m band , with one detection and one marginal detection ( 2@xmath27 - 3@xmath27 ) ; and observed one ( w0410@xmath300913 , which has the brightest 350@xmath2 m flux density ) in the 850 @xmath2 m band , with a detection . the flux densities of the detections are presented in table 3 . we used bolocam to follow - up 18 w12drop galaxies at [email protected] in the 1.1 mm band , including six of the galaxies that were observed with sharc - ii . we obtained five detections , and 13 useful upper limits . flux densities of the detected targets and 95@xmath39 confidence ( 2 @xmath27 ) upper limits for undetected targets are presented in table 3 . we also tabulate in table 3 _ _ spitzer _ _ 3.6 and 4.5 @xmath2 m , and wise 12 and 22 @xmath2 m measurements . w0149 + 2350 was observed with the submillimeter array ( sma ) at 1.3 mm ( wu , bussmann , et al . in prep ) . we list this sma measurement in the 1.1 mm column in table 3 . taking the 3.6 @xmath2 m to 1.1 mm measurements from table 3 , we construct the mid - ir to millimeter sed for w12drop galaxies . figure 3 shows seds for the 9 sharc - ii detected w12drop galaxies compared with galaxy models . in the first panel we overlay a wide range of templates at the corresponding redshift for one w12drop galaxy , normalized to the same w4 flux densities . these templates include the starburst - dominated galaxy arp220 , the agn - starburst blend mrk231 , type i ( unobscured ) , type ii ( obscured ) agn models ( qso 1 and qso 2 ) from polletta et al . ( 2007 ) , a torus model ( polletta et al . 2006 ) , and simulation models of dogs ( narayanan et al . 2010 ) . in the remaining panels we overlay only the arp220 and mark231 templates . the most notable feature is the apparently flat seds extending from the mid - ir to the submillimeter in all of these w12drop galaxies . at 3.6 and 4.5 @xmath2 m , large visual extinction must be added to the comparison templates to match the data , suggesting they are highly obscured . if we normalize all seds at their 22 @xmath2 m flux densities , then the submm emission of w12drop galaxies is much fainter than expected for any other population , indicating their mid - ir to submm flux ratio is unusually high . starburst models miss these seds by a large margin . qso models are better , and the closest match is from the agn dust torus model , but the fit is still poor . to quantitatively show the high mid - ir to submm ratio , we compare @xmath68l@xmath68(24@xmath4)/@xmath68l@xmath68(350@xmath4 ) in w12drops to this ratio in smgs and dogs in figure 4a . figure 5 shows the seds of all submm detected w12drops plotted at their rest frame wavelength in @xmath68l@xmath69 units , normalized by their total luminosity ( see the next section ) . this figure shows a fairly consistent sed for all the w12drop galaxies reported in this paper . this sed has a power - law in the mid - ir ( 1 - 5 @xmath2 m ) , a mid - ir bump that dominates the total luminosity contribution , and becomes flat in the mid - ir to submm . the typical sed of w12drops is quite different from any existing galaxy templates , indicating they may be a new type of galaxy . their seds apparently peak at significantly shorter wavelengths than other galaxy templates , indicating hotter average dust temperatures . in order to understand the nature of the w12drop galaxies , we need to estimate their luminosities and dust temperatures . the standard method to do this is to fit several black - body models with wavelength - dependent opacities to fit sed points along a large range of wavelengths , to constrain both temperatures and luminosities . we already know that w12drop galaxies have unusually high mid - ir to submm ratios , and that the major luminosity comes from 24@xmath4 to 350@xmath4 emission ( see figure 5 ) , so this is clearly the key wavelength range to characterize . at redshift 2 - 3 , the irac [ 3.6 ] and [ 4.5 ] bands ( rest wavelength 0.8 - 1.5 @xmath2 m ) may be significantly affected by stellar light , and at longer wavelengths , we see indications that these w12drop galaxies may have components in addition to a hot - dust component ( see section 4 ) . in many cases infrared - luminous galaxies are dominated by one major dust component , and can be approximated by a single - temperature modified blackbody model . for example , a single dust temperature model provides a good description of the far - ir and submm / mm seds of smgs ( e.g. magnelli et al . 2012 ) , with typical dust temperatures of 30 - 40 k ( chapman et al . 2005 , kovcs et al . 2006 , coppin et al . 2008 , wu et al . 2009 , magnelli et al . 2012 ) . most dogs can also be described by a single - temperature model with dust temperatures of 20 - 40 k ( melbourne et al . 2012 ) . because the peak of the w12drop seds is not well sampled in the data presented here , it is unclear whether their 24@xmath4 to 350@xmath4 emission can be well described by a single - temperature model , and we are obtaining _ herschel _ data to better determine this . in this paper we use a single - temperature model to describe the bulk of the emission from w12drops . we apply a single temperature , modified black - body model combined with a power - law model , to fit the mid - ir to mm seds . at lower frequencies we use @xmath70 , where @xmath71 is the plank function and @xmath72 is the dust emissivity index with @xmath73 , attached smoothly to which at higher frequencies is a power - law with @xmath74 . the two portions are joined at the frequency when the modified black - body slope equals that of the power - law ( @xmath57 ) . the @xmath57 parameter and dust temperature @xmath75 are constrained by w3 , w4 and 350@xmath2 m data , as well as 450@xmath2 m , 850@xmath2 m and 1.1 mm data when available . these data do not absolutely determine the shape of the sed , but the combination of the @xmath57 and @xmath75 parameters provide a reasonable measure of the peak frequency of the sed . in table 4 , we give the derived dust temperatures and mid - ir power - law indexes from this model . the derived dust temperatures of w12drops range from 60 k to 123 k , with a median value of 94 k. we applied this same model to calculate dust temperatures for dogs in bo@xmath76tes field , using similar data from melbourne et al . we used their reported irac 8@xmath4 and mips 24@xmath4 data in place of w3 and w4 to calculate @xmath57 , together with their spire 350 and 500 @xmath4 data to estimate the dust temperature . we obtained similar dust temperatures to the 20 - 40 k reported by melbourne et al . this confirms that w12drop galaxies are much hotter than typical dogs . our single - temperature model also provides a luminosity when @xmath57 , @xmath72 and t are fixed . however this luminosity is sensitive to the data points close to the peak , which is not well sampled here . therefore instead , we use a simple , but conservative method to estimate the total luminosity , which is to simply connect the data points of all the available sed points with power - laws , and integrate the total flux densities . this method may miss the luminosity close to the peak of the sed , so it provides a lower bound to the true luminosity . we list the conservative total luminosity calculated between 2 - 1000 @xmath2 m in table 4 . the derived total luminosities range from 1.7 @xmath77 ( w0211@xmath302242 ) , to 1.8 @xmath78 ( w0410@xmath300913 ) , confirming that these galaxies are very luminous , well above the @xmath79 threshold for hylirgs . our cso follow - up observations of 26 w12drop galaxies show that their luminosities are very high , with some over @xmath80 , and a median and mean of 5.7 and 6.1 @xmath77 , all using the conservative power - law method . this is roughly an order of magnitude brighter than the typical smg ( with median luminosity @xmath81 , chapman et al . 2005 , kovcs et al . 2006 ) , or dog ( with median and mean luminosity @xmath82 and @xmath83 , melbourne et al . 2012 ) , and is comparable to the brightest known optically selected quasars ( schneider et al . 2005 ) . extremely luminous infrared galaxies are often found to be magnified by galaxy - galaxy lensing ( blain 1996 , eisenhardt et al . 1996 , solomon & vanden bout 2005 , vieira et al . 2010 , negrello et al . an immediate concern about the hyperluminous w12drop galaxies is whether they are lensed , too . however , high - resolution imaging follow - up of selected w12drops does not indicate gravitational lensing ( eisenhardt et al . 2012 , bridge et al . in prep . ) , so that the derived @xmath84 luminosities are consistent with being intrinsic based on the current data . additional high - resolution follow - up observations with hubble space telescope are currently underway , and should reveal if these w12drop galaxies are not lensed . if the lack of lensing is confirmed , these galaxies are one of the most luminous populations in the universe . their unusually high dust temperatures and extremely high luminosities make w12drop galaxies of great interest for studying galaxy formation and evolution . how do they become so luminous ? are they experiencing special evolutionary events ? what is their relationship to other well - established galaxy populations , such as smgs and dogs ? classical smgs are defined with strong 850@xmath2 m emission ( @xmath28 5mjy ) which normally indicates significant cold - dust content . table 3 gives examples of some w12drop galaxies that meet this criterion . hence some w12drop galaxies would be selected as smgs . however , the relatively low detection rate with bolocam at 1.1 mm implies that many w12drop galaxies are not as bright as smgs at longer wavelengths . this is understandable given that w12drop galaxies are dominated by emission from hotter dust . dogs normally have both agn and starburst contributions , with warmer dust grains than smgs . in dey et al ( 2008 ) , dogs are defined as galaxies with @xmath14 mjy , and @xmath15 > 14 $ ] ( in vega magnitudes ) , where the @xmath16 photometry is centered at 6393 . since the w4 band at 22 @xmath2 m is similar to the _ spitzer _ 24 @xmath2 m band , our w4 @xmath28 7 mjy selection corresponds to much higher flux densities at 24 @xmath2 m than normal dogs . to make a comparison between w12drops and typical dogs , we obtained @xmath85-band ( centered at 6231 @xmath86 ) photometry from sdss ( dr8 ) , as listed in table 3 , and used the @xmath85-w4 color to approximate the @xmath16-[24 ] color . taking the average power - law index @xmath57 of 2.09 from table 4 and extrapolating to @xmath85-band and 24 @xmath2 m , the difference between the @xmath85-w4 and @xmath16-[24 ] color ranges from 0.2 to 0.24 mag as @xmath16-[24 ] changes from 14 to 17 . all 18 targets in table 3 that are covered by sdss dr8 meet the @xmath85-w4 @xmath28 14 dog criterion , with @xmath85-w4 ranging from 14.4 to 16.1 for @xmath85-band detected sources , and @xmath85-w4 @xmath28 15.3 for @xmath85-band undetected sources ( using @xmath85=22.9 vega mag as the sdss detection limit ) . in figure 6 we compare the distribution of @xmath16 vs. @xmath16-[24 ] for these high - redshift w12drops to dogs in d08 . clearly all w12drop galaxies in table 3 qualify as dogs , with similar colors , but are much brighter at 24 @xmath2 m . although many and maybe most w12drop galaxies can be classified as dogs , their properties are quite different from normal dogs . comparing to the dogs reported in d08 , w12drop galaxies have an order of magnitude higher luminosity , although their redshift distributions are similar ( eisenhardt et al . 2012 ) . bussmann et al . ( 2009 ) used sharc - ii at 350 @xmath2 m to follow - up a subset of dogs with the brightest 24 @xmath2 m flux densities from d08 ( fig 6 ) , obtaining infrared luminosities ( 8 - 1000@xmath2 m , @xmath87 ) and dust temperatures ( @xmath2830 - 60 k ) , still significantly lower than for the w12drop galaxies reported here . since the d08 survey covered only @xmath88 9 deg@xmath89 ( bo@xmath76tes field ) , the dog surface density is @xmath38 320 dogs per square degree with 24 @xmath2 m fluxes density greater than 0.3 mjy . the w12drop selection requires w4 @xmath28 7 mjy , which is at the high 24 @xmath2 m flux density end of the d08 sample , and only selects @xmath38 1000 targets over the whole sky . the typical dogs are 20 times fainter than the w12drops , but the latter are about 10000 times rarer . the high - z w12drop galaxies are apparently extreme cases of dogs with very high dust obscuration and hotter dust temperatures , and appear to be hyperluminous , hot dogs . the very low surface density of w12drop galaxies suggests either they are intrinsically extremely rare , or are only seen during a very short phase of galaxy evolution . dogs are thought to be the transitional phase of mergers between starburst - dominated mode and agn - dominated modes . based on the mid - ir sed ( 3.6 - 24 @xmath2 m ) , d08 classified dogs into two categories , those which have a distinct `` bump '' in their sed between 3 - 10 @xmath2 m attributed to the redshifted starlight from rest - frame 1.6 @xmath2 m , and those whose mid - ir sed is a power - law . bump dogs are thought to be dominated by starbursts ( yan et al . 2005 , sajina et al . 2007 , farrah et al . 2008 , desai et al . 2009 ) , and tend to have fainter 24 @xmath2 m flux densities ( dey et al . 2008 ) , while power - law dogs are thought to be dominated by agn in the mid - ir ( weedman et al . 2006 , donley et al . 2007 , yan et al . 2007 , murphy et al . 2009 ) , and make up most of the bright end of the 24@xmath2 m flux density distribution . the fraction of power - law dogs increases from 10% at @xmath90 mjy to 60% at @xmath91 mjy in the _ spitzer _ deep , wide - field survey ( ashby et al . the mid - ir ( mips 24 @xmath4 ) to submm ( spire 250 @xmath4 ) flux density ratio for power - law dogs is found to be similar to the agn dominated ulirg mrk231 ( melbourne et al . the irac1 to w4 seds of w12drops ( fig . 5 ) are more like the mid - ir seds of dogs rather than smgs ( hainline et al . 2009 ) . they show typical power - law shapes with no obvious bumps , and are very bright at 24 @xmath2 m . consequently , it is plausible that w12drop galaxies are also dominated by very powerful agns . these powerful , highly obscured agns can heat the surrounding dust cocoon to a very high temperature . although the seds of w12drops are dominated by emission from very hot dust components ( figure 5 ) that are likely contributed by powerful agns , a hot agn component alone ca nt explain all of the observed sed from the mid - ir to millimeter bands . it is likely that the seds are composed of multiple components with different temperatures . a more detailed model to decompose seds with multiple - temperature components needs a more complete set of sed data , which will become available from our ongoing _ herschel _ program . but the 350@xmath2 m and 1.1 mm data reported in this paper can give a useful constraint on the coldest component , if we assume that the 350 @xmath2 m to 1.1 mm sed is tracing the coldest dust in these galaxies . in figure 7 , we plot the modeled flux density ratios of 350 @xmath2 m to 1.1 mm continuum versus the redshifts , for models with a single - temperature black - body times a wavelength - dependent opacity , with various dust temperatures and emissivities . w12drop galaxies with available 350 @xmath2 m and 1.1 mm measurements are plotted in the figure . for comparison , models based on galaxies with significant starburst components ( arp220 and mrk231 ) are also plotted . for @xmath72 between 1.5 and 2.0 , the 350 @xmath2 m to 1.1 mm ratios of w12drop galaxies in figure 7 apparently favor a model with t@xmath92 less than 50 k , in addition to the @xmath38 100 k hot dust component that dominates the mid - ir . this temperature of colder dust is comparable to the typical dust temperature of @xmath3835 k associated with starburst galaxies ( such as arp220 ) . considering that the very hot agn component will contribute more to the continuum at 350@xmath2 m than at 1.1 mm , the actual 350 @xmath2 m to 1.1 mm ratios that trace the coldest dust could be lower , and therefore closer to the track of arp220 or mrk231 in figure 7 . this may imply that the cold dust component in these galaxies is not very different from those in starburst galaxies , and is possibly related to star formation . for example , in the detailed study of the first discovered w12drop galaxy ( w1814 + 3412 , eisenhardt et al . 2012 ) , a significant starburst is found , although only contributing a small fraction to the overall luminosity . the cold dust properties of w12drop galaxies may also be different from known obscured qsos ( e.g. martnez - sansigre et al . the ratio of 350 @xmath2 m to 1.1 mm in ams16 , a high - z obscured quasar , is lower than for w12drops , as plotted in figure 7 . a detailed study of the long - wavelength properties for these w12drop galaxies , for instance , to distinguish the contribution and distribution of cold dust ( star formation ) and hot dust ( agn ) , will need observations from alma as well as _ the similarity between the optical to 22 @xmath2 m seds of dogs and w12drop galaxies , with the latter being much brighter , suggests w12drops may be the high luminosity tail of the dog distribution . but the high mid - ir to submm luminosity ratio of w12drops implies they are much hotter than typical dogs . are these w12drops merely luminous dogs , or a distinct population ? or do they have any evolutionary connection ? some theoretical models for dogs ( e.g. narayanan et al . 2010 ) propose that smgs , bump dogs , and power - law dogs may form an evolutionary sequence , representing the transition of merging galaxies from a starburst - dominated phase to an agn - dominated phase , although direct observational support for this is still rare . in figure 4a , we see a strong correlation between the mid - ir flux density and the mid - ir to submm luminosity ratio that supports such a sequence , with w12drops at the highest luminosities . the correlation is roughly linear , suggesting the cold dust component ( traced by 350@xmath4 emission which may be from a starburst ) does nt change significantly during this process , as clearly shown in figure 4b , while the hot dust component ( traced by 24@xmath4 ) becomes stronger , possibly tracing the growth of an embedded smbh . in this scenario , w12drops represent a late phase of this evolution , with more massive smbhs and similar cold - dust components to smgs and dogs . if so , the low surface density of w12drops suggests either such a phase is very short , or not every galaxy goes through this stage . a better understanding of whether the populations have an evolutionary connection will need a fuller study of the w12drop population , and of the luminosity function of all these populations . wise has discovered a possibly new type of object , the w12drop galaxies . the results of our cso submm / mm follow - up observations for a subsample of w12drop galaxies are as follows : \2 . the seds constructed from wise , warm _ spitzer _ and cso data reveal consistent features for w12drop galaxies . these seds show a power - law shape in the mid - ir , and are apparently flat from the mid - ir to submm / mm . their seds have unusually high mid - ir to submm luminosity ratios , indicating a hotter dust temperature than other populations . their seds can not be well fit with existing galaxy templates , indicating they are likely a new population . using power - laws to connect the sed data points , we estimate their total luminosities to be at least 1.7 @xmath77 to 1.8 @xmath78 . using a single - temperature modified blackbody model with @xmath72=1.5 , we estimate their dust temperatures to be 60 k to 120 k , much hotter than other infrared luminous galaxies . besides the hot dust component , they may also have colder dust components that are similar to starburst galaxies . w12drop galaxies in this paper would also be selected as dogs , but are at least 10 times more luminous and 10000 times rarer . they may be the extreme cases of very luminous , hot dogs , and may represent a short evolutionary phase during galaxy merging , following the phase of smg , bump dog and power - law dog . j. wu and r. assef were supported by an appointment to the nasa postdoctoral program at the jet propulsion laboratory , administered by oak ridge associated universities through a contract with nasa . n. evans acknowledges support from nsf grant ast-1109116 . this publication makes use of data products from the wide - field infrared survey explorer , which is a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration . this work is based on observations made with the caltech submillimeter observatory , which is operated by the california institute of technology under funding from the national science foundation , contract ast 90 - 15755 . this work uses data obtained from the _ spitzer _ space telescope , which is operated by the jet propulsion laboratory , california institute of technology under contract with nasa . some of the data presented herein were obtained at the w.m . keck observatory , which is operated as a scientific partnership among caltech , the university of california and nasa . the keck observatory was made possible by the generous financial support of the w.m . keck foundation . some data reported here were obtained at the mmt observatory , a joint facility of the university of arizona and the smithsonian institution . this paper uses data from sdss ( dr 8) . funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s . department of energy office of science . the sdss - iii web site is http://www.sdss3.org/. dey , a. , soifer , b. t. , desai , v. , brand , k. , le floch , e. , brown , m. j. i. , jannuzi , b. t. , armus , l. , bussmann , s. , brodwin , m. , bian , c. , eisenhardt , p. , higdon , s. j. , weedman , d. , willner , s. p. 2008 , , 677 , 943 sayers , j. , golwala , s. r. , rossinot , p .. , ade , p. a. r. , aguirre , j. e. , bock , j. j. , edgington , s. f. , glenn , j. , goldin , a. , haig , d. , lange , a. e. , laurent , g. t. , mauskopf , p. d. , nguyen , h. t. 2009 , , 690 , 1597 w0026 + 2015 & 00:26:09.24 & + 20:15:56.2 & 1.990 & 350 @xmath2 m & 2011 sep 9 & 1.0 & 0.045 + w0116@xmath300505 & 01:16:01.41 & -05:05:04.1 & 3.173 & 350 @xmath2 m & 2011 sep 10 & 1.0 & 0.045 + & & & & 1100 @xmath2 m & 2011 sep 18 & 2.3 & 0.15 w0149 + 2350 & 01:49:46.16 & + 23:50:14.6 & 3.228@xmath93 & 350 @xmath2 m & 2010 jul 28,31 & 2.0 & 0.05 + & & & & 450 @xmath2 m & 2010 sep 12 & 0.7 & 0.04 w0211@xmath302242 & 02:11:34.63 & -22:42:23.4 & 1.746@xmath93 & 350 @xmath2 m & 2011 sep 10 & 1.0 & 0.05 + w0220 + 0137 & 02:20:52.12 & + 01:37:11.6 & 3.122 & 350 @xmath2 m & 2010 sep 12 , 2011 sep 10 & 1.8 & 0.045 + & & & & 1100 @xmath2 m & 2011 sep 19,20 & 3.7 & 0.16,0.07 w0243 + 4158 & 02:43:44.18 & + 41:58:09.1 & 2.010 @xmath93 & 350 @xmath2 m & 2011 sep 10 & 0.8 & 0.05 + w0248 + 2705 & 02:48:58.81 & + 27:05:29.8 & 2.210 & 350 @xmath2 m & 2010 sep 13 , 2011 sep 9 & 2.8 & 0.045 + & & & & 1100 @xmath2 m & 2011 feb 19,20,22 , sep 18 - 20 & 7.4 & 0.15,0.07 w0338 + 1941 & 03:38:51.33 & + 19:41:28.6 & 2.123 & 350 @xmath2 m & 2011 sep 10 & 0.8 & 0.05 + w0410@xmath300913 & 04:10:10.60 & -09:13:05.2 & 3.592 & 350 @xmath2 m & 2011 feb 21 , 2011 sep 9 & 2.3 & 0.95 , 0.045 + & & & & 850 @xmath2 m & 2011 feb 16 - 18 & 4.3 & 0.12 & & & & 1100 @xmath2 m & 2010 dec 12 - 14 & 7.3 & 0.14,0.1 w0422@xmath301028 & 04:22:48.82 & -10:28:32.0 & 2.227@xmath93 & 1100 @xmath2 m & 2011 sep 17 & 2.7 & 0.21 + w0542@xmath302705 & 05:42:30.90 & -27:05:40.5 & 2.532 & 350 @xmath2 m & 2010 sep 13 & 0.4 & 0.05 + w0757 + 5113 & 07:57:25.07 & + 51:13:19.7 & 2.277 @xmath93 & 1100 @xmath2 m & 2011 feb 18,22 & 4.0 & 0.21 + w0851 + 3148 & 08:51:24.78 & + 31:48:56.1 & 2.640 & 1100 @xmath2 m & 2010 dec 13,14 & 5.3 & 0.13,0.1 + w0856 + 0005 & 08:56:28.08 & + 00:05:48.7 & 2.519 @xmath93 & 1100 @xmath2 m & 2011 feb 20 & 2.7 & 0.20 + w0859 + 4823 & 08:59:29.94 & + 48:23:02.3 & 3.245@xmath93 & 1100 @xmath2 m & 2010 dec 13 - 14,2011 feb 16 - 18 & 6.7 & 0.11,0.13 + w0926 + 4232 & 09:26:25.44 & + 42:32:51.9 & 2.498 & 1100 @xmath2 m & 2011 feb 19 & 4.0 & 0.28 + w1146 + 4129 & 11:46:12.87 & + 41:29:14.3 & 1.772 @xmath93 & 1100 @xmath2 m & 2011 feb 16 & 3.0 & 0.15 + w1316 + 3512 & 13:16:28.53 & + 35:12:35.1 & 1.956 @xmath93 & 1100 @xmath2 m & 2011 feb 19 & 2.7 & 0.21 + w1409 + 1335 & 14:09:25.56 & + 13:35:02.1 & 3.048 @xmath93 & 1100 @xmath2 m & 2011 feb 17,22 & 2.3 & 0.1 , 0.2 + w1422 + 5613 & 14:22:28.86 & + 56:13:55.6 & 2.524 & 1100 @xmath2 m & 2011 feb 18,20,21 & 9.0 & 0.2 + w1603 + 2745 & 16:03:57.39 & + 27:45:53.3 & 2.633 @xmath93 & 350 @xmath2 m & 2010 sep 13 & 0.5 & 0.04 + w1814 + 3412 & 18:14:17.30 & + 34:12:25.0 & 2.452 @xmath94 & 350 @xmath2 m & 2010 jul 13,23 & 2.7 & 0.06 , 0.04 + & & & & 450 @xmath2 m & 2010 sep 12,13 & 1.8 & 0.045 & & & & 1100 @xmath2 m & 2010 jun 17,18 & 6.0 & 0.1 w1830 + 6504 & 18:30:13.53 & + 65:04:20.5 & 2.653 & 350 @xmath2 m & 2011 sep 9,10 & 2.7 & 0.05 + w1835 + 4355 & 18:35:33.71 & + 43:55:49.1 & 2.298 @xmath93 & 350 @xmath2 m & 2010 sep 12 & 1.0 & 0.04 + & & & & 450 @xmath2 m & 2010 sep 13 & 0.6 & 0.04 w2207 + 1939 & 22:07:43.84 & + 19:39:40.3 & 2.022 @xmath93 & 1100 @xmath2 m & 2010 dec 14 & 1.8 & 0.1 + w2238 + 2653 & 22:38:10.20 & + 26:53:19.8 & 2.405 & 1100 @xmath2 m & 2011 sep 20 & 1.7 & 0.075 + w0026@xmath952015 & keck / lris & 2010 nov 08 & 600 + 300 & 1.990 & ly@xmath57 , nv , civ , heii , [ neiv ] , mgii + w0116@xmath300505 & mmt / bcs & 2010 dec 04 & @xmath96 & 3.173 & ly@xmath72,ly@xmath57,nv , siiv / oiv + w0220@xmath950137 & mmt / bcs & 2010 dec 04 & @xmath96 & 3.122 & ly@xmath72,ly@xmath57,nv , siiv / oiv + w0248@xmath952705 & keck / lris & 2010 nov 09 & @xmath97 & 2.210 & ly@xmath57,cii],mgii + w0338@xmath951941 & keck / lris & 2011 feb 02 & @xmath98 & 2.123 & ly@xmath57 + w0410@xmath300910 & keck / lris & 2010 nov 08 & @xmath99 & 3.592 & civ , heii ( ly@xmath57 at dichroic ) + w0542@xmath302705 & keck / lris & 2010 nov 08 & @xmath97 & 2.532 & ciii , ovi , ly@xmath57,nv , ... + w0851@xmath953148 & keck / lris & 2011 nov 08 & @xmath97 & 2.640 & ovi , ly@xmath57,civ , heii , ciiii],cii ] + w0926@xmath954232 & keck / lris & 2011 feb 02 & @xmath98 & 2.498 & absorption lines only + w1422@xmath955613 & keck / lris & 2010 jul 15 & @xmath97 & 2.524 & ly@xmath57,nv , civ + w1830@xmath956504 & keck / lris & 2010 jul 13 & @xmath99 & 2.653 & likely ly@xmath57 + w2238@xmath952653 & keck / lris & 2010 nov 08 & @xmath98 & 2.405 & ly@xmath57 , nv , siiv / oiv , civ + llcccccccc w0026 + 2015 & 22.10 & 40.1(2.0 ) & 80.5(2.0 ) & 2.77(0.13 ) & 15.05(0.97 ) & @xmath100 & ... & ... & ... + w0116@xmath300505 & 21.38 & 50.7(2.2 ) & 89.4(2.3 ) & 2.39(0.13 ) & 12.96(1.00 ) & 36(12 ) & ... & ... & @xmath101 + w0149 + 2350 & @xmath3122.9 & 19.7(1.7 ) & 34.7(1.5 ) & 1.77(0.10 ) & 9.18(0.76 ) & 29(8 ) & 35(9 ) & ... & 2(0.4)@xmath94 + w0211@xmath302242 & & 30.5(11)@xmath102 & 78.0(10)@xmath102 & 3.31(0.11 ) & 11.31(0.73 ) & 56(15 ) & ... & ... & ... + w0220 + 0137 & 21.84 & 25.2(1.8 ) & 38.4(1.4 ) & 1.78(0.10 ) & 11.98(0.81 ) & 43(9 ) & ... & ... & 6.2(2.0 ) + w0243 + 4158 & ... & 23.1(1.7 ) & 70.7(2.0 ) & 2.56(0.13 ) & 9.02(0.93 ) & 38(13 ) & ... & ... & ... + w0248 + 2705 & ... & 31.5(1.9 ) & 52.0(1.6 ) & 2.04(0.14 ) & 11.11(1.05 ) & 32(8 ) & ... & ... & @xmath103 + w0338 + 1941 & & 12.2(1.6 ) & 37.5(1.6 ) & 1.97(0.14 ) & 10.27(0.99 ) & @xmath104 & ... & ... & ... + w0410@xmath300913 & ... & 26.7(1.8 ) & 46.1(1.5 ) & 2.45(0.14 ) & 12.35(0.99 ) & 118(17 ) & ... & 40(14 ) & 13.6(2.6 ) + w0422@xmath301028 & ... & 18.0(1.6 ) & 55.0(1.6 ) & 2.74(0.13 ) & 10.70(1.02 ) & ... & ... & ... & @xmath105 + w0542@xmath302705 & & 24.6(1.7 ) & 29.2(1.2 ) & 2.55(0.11 ) & 14.08(0.90 ) & @xmath106 & ... & ... & ... + w0757 + 5113 & 22.27 & 20.0(1.7 ) & 35.0(1.4 ) & 1.46(0.11 ) & 9.31(0.84 ) & ... & ... & ... & @xmath107 + w0851 + 3148 & 21.64 & 41.4(2.0 ) & 88.5(2.2 ) & 3.51(0.16 ) & 14.73(1.02 ) & ... & ... & ... & @xmath108 + w0856 + 0005 & @xmath3122.9 & 52.7(2.2 ) & 74.2(2.0 ) & 2.94(0.13 ) & 15.06(0.96 ) & ... & ... & ... & @xmath109 + w0859 + 4823 & @xmath3122.9 & 16.4(1.6 ) & 44.8(1.4 ) & 2.22(0.10 ) & 11.83(0.91 ) & ... & ... & ... & 6.2(1.5 ) + w0926 + 4232 & @xmath3122.9 & 18.6(1.6 ) & 28.6(1.2 ) & 1.45(0.11 ) & 7.78(0.93 ) & ... & ... & ... & @xmath110 + w1146 + 4129 & @xmath3122.9 & 23.6(1.7 ) & 45.7(1.5 ) & 3.90(0.13 ) & 20.35(1.05 ) & ... & ... & ... & @xmath111 + w1316 + 3512 & 22.56 & 22.0(1.7 ) & 49.4(1.5 ) & 3.00(0.12 ) & 12.62(0.94 ) & ... & ... & ... & @xmath112 + w1409 + 1335 & @xmath3122.9 & 6.9(0.4 ) & 16.4(1.0 ) & 1.63(0.09 ) & 9.44(0.73 ) & ... & ... & ... & 5.7(2.1 ) + w1422 + 5613 & @xmath3122.9 & 28.6(1.8 ) & 74.3(1.9 ) & 3.07(0.09 ) & 11.95(0.68 ) & ... & ... & ... & @xmath113 + w1603 + 2745 & @xmath3122.9 & 29.9(1.8 ) & 47.6(1.5 ) & 3.15(0.12 ) & 9.53(0.87 ) & @xmath114 & ... & ... & @xmath115 + w1814 + 3412 & 23.00 & 20.8(2.1 ) & 26.5(1.9 ) & 1.86(0.10 ) & 14.38(0.86 ) & 33(9 ) & @xmath116 & ... & @xmath117 + w1830 + 6504 & @xmath3122.9 & 14.2(1.6 ) & 41.9(1.4 ) & 2.25(0.05 ) & 7.56(0.34 ) & @xmath104 & ... & ... & ... + w1835 + 4355 & ... & 51.5(2.2 ) & 142.8(3.0 ) & 6.13(0.13 ) & 27.05(0.87 ) & 46(16 ) & 31(14 ) & ... & ... + w2207 + 1939 & @xmath3122.9 & 41.7(2.0 ) & 57.8(1.7 ) & 1.49(0.11 ) & 10.27(0.90 ) & ... & ... & ... & @xmath101 + w2238 + 2653 & 22.77 & 42.2(2.0 ) & 58.8(1.6 ) & 2.35(0.11 ) & 17.15(0.98 ) & ... & ... & ... & 6.0(2.2 ) + w0116@xmath300505 & 123@xmath1188 & [email protected] & 7.4 @xmath119 + w0149 + 2350 & 100@xmath1185 & [email protected] & 5.7 @xmath119 + w0211@xmath302242 & 60@xmath1185 & [email protected] & 1.7 @xmath119 + w0220 + 0137 & 118@xmath1186 & [email protected] & 7.1 @xmath119 + w0243 + 4158 & 68@xmath1185 & [email protected] & 2.0 @xmath119 + w0248 + 2705 & 87@xmath1188 & [email protected] & 2.8 @xmath119 + w0410@xmath300913 & 82@xmath1185 & [email protected] & 1.8 @xmath120 + w1814 + 3412 & 113@xmath1187 & [email protected] & 4.0 @xmath119 + w1835 + 4355 & 94@xmath11810 & [email protected] & 6.5 @xmath119 +

we discuss the asymptotic behavior of regulated field commutators for linearly polarized , cylindrically symmetric gravitational waves and the mathematical techniques needed for this analysis . we concentrate our attention on the effects brought about by the introduction of a physical cut - off in the study of the microcausality of the model and describe how the different physically relevant regimes are affected by its presence . specifically we discuss how genuine quantum gravity effects can be disentangled from those originating in the introduction of a regulator .

gr-qc0412028

linearly polarized cylindrical waves , also known as einstein - rosen waves @xcite , provide a symmetry reduction of general relativity that can be used as a test bed for the quantization of the theory . this system displays several interesting features that contribute to its relevance . on one hand it has an infinite number of local degrees of freedom and , hence , it is a genuine quantum field theory ( in contradistinction to other symmetry reductions , such as bianchi models , that have a finite number of global degrees of freedom ) . on the other , the system is tractable both classically and quantum mechanically , thus allowing to derive exact consequences independent of any approximation scheme @xcite . the main reason behind this success and tractability is the fact that the gravitational degrees of freedom of the model are encoded in an free , massless , axially symmetric , scalar field that evolves in an auxiliary minkowskian background . in previous papers we have analyzed the issue of microcausality in this system ; in particular , we have studied in detail the smearing of light cones owing to the quantization of the gravitational field @xcite . the main tool for this type of analysis is the study in vacuo of the field commutator evaluated at different spacetime points . as is well known the commutator of quantum fields reflects the causal structure of spacetime ( minkowskian spacetime in ordinary perturbative quantum field theory ) in the sense that the quantum fields in spatially separated spacetime points commute this is true for all standard types of quantum fields , i.e. scalar , fermion , or vector fields though issues related to gauge invariance must be carefully considered in this last case. in the specific model that we are interested in , gauge invariance has been discussed in ref . the authors of that paper conclude that it is correct to use the ashtekar - pierri gauge fixed action , written in terms of the axially symmetric scalar field , to derive gauge invariant information about the model . in a recent work @xcite we discussed the situation when no cut - off is introduced in the system , studying the unregulated commutator . the main results reached were the following . first , one can clearly see that light cones are smeared by quantum gravity effects ; in fact it is possible to obtain a quantitative measure of this smearing and show how sharp light cones are recovered in the limit of large distances as compared to the natural length scale of the model the planck length. it is also interesting to point out that the asymptotic behavior of the commutator in the different physically relevant regimes strongly depends on the causal relationship between the different spacetime points involved . second , one finds a singularity structure in the commutator that differs from that of the free theory ; in particular the field commutator for equal values of the radial coordinate @xmath0 is singular . finally one observes that , in the case when one of the spacetime points that appear in the commutator corresponds to the symmetry axis , there are quantum effects that persist for large values of the difference of the time coordinates . this effect is reminiscent of the large quantum gravity effects first discussed by ashtekar @xcite . the purpose of this paper is to study how the conclusions of ref . @xcite are changed by the introduction of a cut - off . as is well known regulators are generally necessary in order to have well defined quantum field theories . one can justify its use , for example , by noticing that the action of the field operator on the vacuum in a fock space is not a vector in the hilbert space because it has infinite norm . in order to have a well defined action of the field operator one regulates it by introducing smearing functions that render the norms of these states finite . the problem then consists in removing these regulators ( or rather showing that the physical results are independent of them ) . in principle , it is possible to argue that the results derived in the absence of regulators somehow approximate those derived after their introduction ; this is straightforward to see in the case of cut - offs . in the presence of a cut - off @xmath1 , the improper integrals that define the field commutator become proper because the integration region is a closed interval @xmath2 $ ] . for a given value of the parameters that appear in the integral ( involving the values of the spacetime coordinates of the quantum fields and the gravitational constant ) it is always possible to choose a value for @xmath1 such that the integral with the cut - off is well approximated by the integral extended to @xmath3 . of course it is conceivable that the cut - off is not just a mathematical device but rather a physical scale defining a fundamental limit for the resolution of our measurements . if spacetime becomes discrete at short distances ( such as the planck length ) the continuum spacetime picture breaks down and , certainly , it would be difficult to justify the extension of the integrals involved in the definition of field commutators ( or @xmath4 matrix elements , for that matter ) to infinite intervals in momenta ( inverse length ) . our point of view here is that the introduction of a cut - off can mimic some of the effects appearing after a successful quantization of gravity ( for example , in the loop quantum gravity approach ) and hence we plan to study its effect within the consistent framework provided by the einstein - rosen waves . it is also interesting to point out here that the cut - off by itself can produce some of the effects expected from quantum gravity . in particular , it is possible to show that light cones are also smeared by cut - offs . in our opinion this makes it necessary to study in detail how the effects of the cut - off and quantum gravity can be disentangled . the paper is organized as follows . after this introduction , we briefly review the main results about microcausality in quantum cylindrical gravitational waves and introduce the commutators that we will discuss in the rest of the paper . we will then study the field commutators in the presence of a cut - off with the help of the asymptotic techniques already employed in ref . @xcite . here the situation is simpler because we will only have to consider integrals over closed intervals . we will discuss one by one the asymptotic behaviors in the different parameters involved . in sec . vi we will derive a power series expansion in the gravitational constant for the commutator in the presence of a cut - off in the spirit of ordinary perturbative quantum field theory , and discuss the uniform convergence of this series under appropriate conditions on the cut - off in sec . we end the paper with a discussion of our results and our conclusions . einstein - rosen waves describe topologically trivial spacetimes with two linearly independent , commuting , spacelike , and hypersurface orthogonal killing vector fields @xcite endowed with a metric that can be written as @xmath5 here we use the coordinates @xmath6 , and @xmath7 and @xmath8 are functions only of @xmath0 and @xmath9 . the einstein field equations are very simple . the scalar field @xmath7 satisfies the wave equation for a massless , axially symmetric scalar field in three dimensions @xmath10 and the function @xmath8 can be expressed in terms of this field @xcite as @xmath11.\ ] ] we will use in the following a system of units such that @xmath12 and define @xmath13 , where @xmath14 denotes the gravitational constant per unit length in the direction of the symmetry axis has dimensions of length in this system of units . ] . the function @xmath15 ( apart from a factor of @xmath16 ) has a simple physical interpretation : it is the energy of the scalar field in a ball of radius @xmath0 whereas @xmath17 denotes @xmath18 ( the energy of the whole two - dimensional flat space ) . it is also possible to show @xcite that @xmath19 coincides with the hamiltonian @xmath20 of the system obtained by a linearization of the metric ( [ metric ] ) . in order to have a unit asymptotic timelike killing vector and a physical notion of energy ( per unit length ) we introduce the coordinates @xmath21 defined by @xmath22 . in these coordinates the metric takes the form @xcite @xmath23 by taking a sufficiently fast fall - off for @xmath7 as @xmath24 , this metric describes asymptotically flat cylindrical spacetimes such that @xmath25 is a unit timelike killing vector in the asymptotic region . in the 2 + 1 dimensional framework these spacetimes are asymptotically flat at spacelike and null infinities @xcite ( the appropriate introduction of null infinity will be needed in order to study the @xmath4 matrix of the model ) . it is also worthwhile noting that these spacetimes have a non - zero deficit angle . the einstein field equations can be obtained from a hamiltonian action principle @xcite . a remarkable ( and useful ) feature of the physical hamiltonian @xmath26 ( associated with the physical time @xmath27 ) is the fact that it is a function of the hamiltonian corresponding to the free scalar field , @xmath20 : @xmath28 in terms of the @xmath9-time and imposing regularity at the axis @xmath29 @xcite , the classical solutions for the field @xmath7 can be written as @xmath30\end{aligned}\ ] ] where @xmath31 and its complex conjugate @xmath32 are fixed by the initial conditions . the free hamiltonian @xmath20 can be written now as @xmath33 using this expression , we obtain the @xmath27-evolution of the field @xmath34.\;\end{aligned}\ ] ] the quantization can be carried out in the usual way by introducing a fock space where @xmath35 , the quantum counterpart of @xmath36 , is the operator - valued distribution @xcite given by tag . ] @xmath37\!. \label{psi0}\end{aligned}\ ] ] its action on fock space is determined by those of @xmath38 and @xmath39 the annihilation and creation operators with non - vanishing commutators given by @xmath40=\delta(k_1,k_2)$ ] . we can regulate the field by introducing suitable functions @xmath41 that render finite the norms of the states obtained by acting with the quantum field on fock space vectors . in the following we will make the simplest choice @xmath42}(k)$ ] ( here @xmath43}$ ] denotes the characteristic function of the interval @xmath44 $ ] ) . by doing this the integration region in ( [ psi0 ] ) becomes compact and we have @xmath45\!. \label{psi0lambda}\end{aligned}\ ] ] evolution in @xmath9 is given by the unitary operator @xmath46 where @xmath47 is the quantum hamiltonian operator of a three - dimensional , axially symmetric scalar field . the cut - off regulated quantum scalar field in the heisenberg picture is hence given by @xmath48 \!.\hspace*{.3cm}\nonumber\end{aligned}\ ] ] if we describe the evolution in our model in terms of the physical time @xmath27 , the quantum hamiltonian is @xmath49 and unitary evolution is given by @xmath50 . with this time evolution the annihilation and creation operators in the heisenberg picture are @xmath51\!\hat{a}(k),\nonumber\\ \hat{a}_e^\dagger(k , t)\!\!&=&\!\hat{a}^\dagger(k)\,\exp\!\left[ite(k ) e^{-4g\hat{h}_0}\right],\nonumber\end{aligned}\ ] ] where @xmath52 , and the regulated field operator evolved with the physical hamiltonian [ that we denote as @xmath53 is given by @xmath54.\end{aligned}\ ] ] the field commutator @xmath55 $ ] can be computed from these expressions @xcite . since we are dealing with an effectively interacting theory this operator is not proportional to the identity in the fock space basis that we are using and , hence , we have to consider its matrix elements . as in previous work we will concentrate on the vacuum expectation value . if a cut - off @xmath56 is introduced , this is given by @xmath57\,|0\rangle= \int_0^{\lambda_k}\!\!\!\!\!\!dk\,j_0(r_1 k)j_0(r_2 k)\sin\left[\frac{t_2-t_1}{4g}(1\!-\!e^{-4gk})\right]\ ! { , } \hspace*{.4cm}&&\label{integral}\end{aligned}\ ] ] that can be seen to depend on the time coordinates only through their difference @xmath58 which we will assume in the following to be positive. notice that it depends symmetrically on @xmath59 and @xmath60 . the functional dependence in @xmath61 is less trivial , a fact that requires especial attention when studying the limit in which the relevant lengths and time differences are much larger than the planck length @xcite . it is convenient to refer the dimensional parameters of these integrals to another length scale , that we choose as @xmath59 . we hence introduce @xmath62 , @xmath63 , and @xmath64 and rewrite ( [ integral ] ) as @xmath65\,|0\rangle= \frac{\lambda}{r_1}\im\bigg\{\int_0^{\lambda_q}dqj_0(\lambda q)j_0(\rho \lambda q)e^{i\tau\lambda(1-e^{-q})}\bigg\}\label{newintegral}\ ] ] after introducing the new variable @xmath66 . here @xmath67 denotes the imaginary part and @xmath68 . physically , the cut - off @xmath56 ( which in principle has the dimensionality of an inverse length ) could be interpreted as having its origin in the existence of a minimum length . this comes out naturally in loop quantum gravity where space is discrete and the area and volume operators have minimum eigenvalues of the order of the planck area and volume , respectively . in fact , the existence of a minimum length ( of the size of the planck scale ) can be considered a generic feature of essentially every quantum theory @xcite . the interpretation of the adimensional cut - off @xmath69 would follow from that of @xmath56 , so it may be reasonable to expect that it be a number of order unity ; nevertheless we will treat it as a free parameter in the following . let us start by considering the behavior of ( [ newintegral ] ) when the parameter @xmath70 grows to infinite or approaches @xmath71 . this integral can be written as a standard @xmath72-transform @xcite by the change of variables @xmath73 . the most convenient way to get its asymptotic behavior in @xmath74 is by rewriting it in the form @xmath75{.}\label{split}\ ] ] one can then use the asymptotic behavior obtained for the first integral in ref . @xcite , and find the asymptotics of the second integral by standard integration by parts [ employing the fact that terms forces us to split the original integral in two as in ( [ split ] ) to avoid singularities at @xmath76 . ] @xmath77 . by doing this one gets the following two contributions @xmath78}+o(\frac{1}{\rho^7}){,}\ ] ] @xmath79+o(\frac{1}{\rho^{5/2}}).\ ] ] the first one is cut - off independent but subdominant with respect to the second , hence we see that the presence of a cut - off changes the asymptotic behavior in @xmath70 . this is the kind of behavior that one would expect even in a lorentz covariant theory after the introduction of a cut - off because of the breaking of the lorentz symmetry . the novel feature here is the presence of cut - off independent terms . although the cut - off dependent one dominates in the asymptotic limit , there may be a transient regime whose onset will be controlled by the value of @xmath56 in which the asymptotic behavior is given by the first term . this will be most evident when @xmath80 . in the @xmath81 limit we get @xmath82 as a result of the continuity at @xmath71 of the integral defining the commutator ( [ newintegral ] ) . the integral in ( [ newintegral ] ) has the convenient form of an @xmath72-transform and , hence , it can be studied by standard mellin transform methods @xcite if the asymptotic parameter is chosen to be @xmath70 ; however this is no longer true if the asymptotic parameter is taken to be @xmath83 ( which corresponds to considering large separations in the time coordinates ) . this fact introduces some mathematical difficulties in the asymptotic analysis . in this case one has to consider the cases @xmath71 and @xmath84 separately . if @xmath71 one finds that the asymptotic behavior when @xmath85 without the cut - off is given by @xcite @xmath86 } e^{\frac{\pi}{2}\lambda}\gamma(i\lambda)+ e^{-i[\frac{\pi}{4}-\tau\lambda-\lambda\log(\tau\lambda ) ] } e^{-\frac{\pi}{2}\lambda}\gamma(-i\lambda)\big\}+ o(\frac{1}{\log\tau})\label{tauasimpr0}\ ] ] whereas for @xmath84 we get @xmath87 } e^{\frac{\pi}{2}\lambda(1+\rho ) } \gamma[i\lambda(1+\rho)]+e^{-i[\frac{\pi}{2}- \tau\lambda-\lambda(1+\rho)\log(\tau\lambda ) ] } e^{-\frac{\pi}{2}\lambda(1+\rho ) } \gamma[-i\lambda(1+\rho)]\hspace{6mm}&&\label{tauasimprneq0}\\ { + } e^{i[\tau\lambda-\lambda(1-\rho)\log(\tau\lambda ) ] } e^{\frac{\pi}{2}\lambda(1-\rho ) } \gamma[i\lambda(1-\rho)]+e^{i[\tau\lambda- \lambda(\rho-1)\log(\tau\lambda ) ] } e^{\frac{\pi}{2}\lambda(\rho-1 ) } \gamma[i\lambda(\rho-1)]\big\}+o ( \frac{1}{(\log\tau)^2}){.}\nonumber\end{aligned}\ ] ] the most interesting feature of these expressions is their unusual dependence on the asymptotic parameter @xmath83 ; in fact the dependence on inverse powers of logarithms ( especially on the inverse square root of @xmath88 ) can not be obtained by direct application of the usual asymptotic expressions derived by mellin transform techniques @xcite . it is also remarkable how slowly the commutator decays in @xmath83 in particular in the axis @xmath71 , a fact that is suggestive of the large quantum gravity effects discussed by ashtekar @xcite . outside the axis the decay is faster but still quite slow . a consequence of the different asymptotic behaviors in @xmath83 for @xmath71 and @xmath84 is the impossibility to recover ( [ tauasimpr0 ] ) as the limit when @xmath89 of ( [ tauasimprneq0 ] ) . as we can see , the frequency of the oscillations of the commutator in @xmath83 is controlled by @xmath90 ( proportional to the inverse of @xmath61 ) in such a way that although the amplitude of the oscillations decays very slowly they will average to zero on scales larger than the planck length . when we introduce a cut - off @xmath56 , the above asymptotic expressions change to @xmath91\big\}+o(\frac{1}{\tau^2}),\label{tauasimptlamb}\ ] ] valid both for @xmath71 and @xmath84 . this can be obtained by straightforward integration by parts . an interesting situation develops at this point because the asymptotic behavior of the integral in @xmath83 behaves in a discontinuous way in the cut - off . in the analysis carried out to study the asymptotic behavior in @xmath70 we found out that the cut - off dependent term , in spite of being dominant , goes to zero in the limit @xmath80 . here the situation is different : taking now @xmath80 in ( [ tauasimptlamb ] ) does not lead to the asymptotic expressions corresponding to @xmath92 . that is , the asymptotic behavior of the improper integral in ( [ integral ] ) is not the limit when @xmath80 of ( [ newintegral ] ) . as in the case of the asymptotics in @xmath70 , one expects that there must be a transient regime in which the behavior in @xmath83 of ( [ newintegral ] ) is given by ( [ tauasimpr0 ] ) and ( [ tauasimprneq0 ] ) . we will not consider here a precise characterization of this transient behavior for arbitrary values of the relevant parameters because its main properties can be conveniently discussed , at least for large @xmath90 , by looking at the @xmath93 asymptotics of the commutator in the ( @xmath94 ) plane . the @xmath95 limit is easy to analyze . in fact what we find , both with and without the cut - off , is that the series obtained by expanding @xmath96 in powers of @xmath97 , exchanging integration and infinite sum , and computing the resulting integrals gives a series that converges to the value of the commutator . the asymptotic behavior in @xmath90 is studied by following the procedure described in ref . it is worth remarking that the limit @xmath93 of the regulated field commutator can not be identified with that in which the gravitational constant @xmath61 vanishes if one admits that the dimensionful cut - off @xmath98 is kept constant in principle . on the contrary , the two limits could be considered equivalent only under the assumption that @xmath56 increases as the inverse of @xmath61 for small gravitational constant , so that its dimensionless counterpart @xmath68 may remain fixed . the analysis of the asymptotics in @xmath90 when the cut - off is present is simultaneously simpler in some respects and more complicated in others compared with the case when no cut - off is introduced . it is simpler because the lengthy analysis needed to discuss the asymptotics of the improper integral is not necessary now . it is more complicated in the sense that the final asymptotic expressions contain additional terms and also because the number of regions with different @xmath90-asymptotic regimes in the @xmath99 plane increases . we have to consider now the cases @xmath71 and @xmath84 separately . let us consider first @xmath71 and write the r.h.s . of ( [ newintegral ] ) as @xcite @xmath100}{\bigg\}}\label{bintegral}\ ] ] after using the usual integral representation for the bessel functions @xmath101 ( @xmath102 ) @xmath103 where @xmath8 is a closed , positively oriented , simple path in the complex plane surrounding the origin . notice that we are integrating an integrable function in a compact region , so we can write the integrals in any order we want . the asymptotic analysis of ( [ bintegral ] ) can be carried out by following the same steps as in ref . @xcite . as we did there , it is useful to introduce neutralizers to split the integral in three pieces @xmath104 , @xmath105 , and choose appropriate contours for each of them . these integrals are @xmath106}\big\}\end{aligned}\ ] ] where we have introduced the neutralizer functions @xmath107 , @xmath105 , satisfying @xmath108 in @xmath109 $ ] and @xmath110,\\ & & \nu_1(q)=0\quad \textrm{if}\quad q\in[\alpha_2,\lambda_q],\\ & & \nu_2(q)=0\quad \textrm{if}\quad q\in[0,\alpha_1]\cup[\beta_2,\lambda_q],\\ & & \nu_2(q)=1\quad \textrm{if}\quad q\in[\alpha_2,\beta_1],\\ & & \nu_3(q)=0\quad \textrm{if}\quad q\in[0,\beta_1],\\ & & \nu_3(q)=1\quad \textrm{if}\quad q\in[\beta_2,\lambda_q],\end{aligned}\ ] ] with @xmath111 ( these parameters are chosen as in ref . @xcite ) . by doing this the effective integration regions in @xmath112 are @xmath113 $ ] , @xmath114 $ ] , and @xmath115 $ ] and the boundary @xmath116 appears only in the first . the asymptotics in @xmath90 of the integral @xmath117 is best obtained by choosing an integration contour satisfying @xmath118 ( that passes necessarily through @xmath119 and @xmath120 ) . by using the same method of ref . @xcite we see that the first two relevant terms are given by the contour integrals @xmath121 @xmath122 whose sum gives @xmath123 and @xmath124 although the second term will be subdominant with respect to some of the contributions coming from @xmath125 and @xmath126 , it improves the approximation of the full commutator obtained from the asymptotics in @xmath90 in the region @xmath127 . the contribution of @xmath126 to the asymptotics in @xmath90 is obtained from the contour integral ( corresponding to the boundary at @xmath128 ) @xmath129 the asymptotics in @xmath90 of this integral can be easily studied by using the method of steepest descents . this gives @xmath130\big\}.\ ] ] finally the integral @xmath125 ( for which we choose for @xmath8 the curve @xmath131 ) only contributes when the stationary points of the exponent are in the integration region . this happens only when @xmath132 . the contribution to the first relevant order in @xmath90 is @xcite @xmath133 adding up the different terms we get @xmath134\big\}{,}\label{asinteje}\end{aligned}\ ] ] where @xmath135 denotes the step function . we see that the final result consists of several contributions : the free commutator for an infinite cut - off , a @xmath136 correction for @xmath127 , the term with the @xmath137 dependence for @xmath132 , and a cut - off dependent contribution for all values of @xmath83 that falls off to zero when @xmath138 . if the cut - off goes to infinity the commutator can be approximated by the one obtained in ref . @xcite ; however , if it is of order one ( as would be the case if it is defined by the planck length ) , then that approximation is no longer valid . notice that the values of @xmath139 are those for which the asymptotics provided by the unregulated commutator is a correct approximation . this is roughly the transient region in the @xmath83 parameter mentioned in the previous subsection . for the field commutator over @xmath140 as a function of @xmath83 for @xmath71 , @xmath141 , and @xmath142 . we compare it both with a numerical computation of the integral that defines it and with the unregulated free commutator . as we can see , the approximation is good except at the points where the asymptotic behavior changes ( @xmath143 and @xmath144 ) . notice the difference in the amplitude of the oscillations for @xmath145 and @xmath146.,width=623 ] for the field commutator over @xmath140 as a function of @xmath83 for @xmath71 , @xmath141 , and @xmath147 ( compared both with a numerical computation of the integral that defines it and with the unregulated free commutator ) . as we can see , the asymptotic approximation is good except at @xmath143 . notice that the cut - off introduces a modulation of the amplitude in the region @xmath132 . the singularity of the asymptotic approximation at @xmath144 lies outside the plotted region.,width=623 ] for the field commutator over @xmath140 as a function of @xmath83 for @xmath71 , @xmath141 , and @xmath148 ( compared both with a numerical computation of the integral that defines it and with the unregulated free commutator ) . as we can see , the asymptotic approximation is good except at @xmath143 . the asymptotic approximation obtained in ref . @xcite is good in a large region in the @xmath83 axis . the singularity of the asymptotic approximation at @xmath144 lies outside the plotted region.,width=623 ] figures [ fig1:propeje]-[fig1:propejec ] show the behavior of the field commutator ( over @xmath140 ) when @xmath71 as a function of @xmath83 for several values of @xmath69 . as we can see , the asymptotic approximation becomes singular between regions with different asymptotic regimes , but approximates well the exact value of the commutator ( obtained by numerical methods ) for the remaining values of @xmath83 . notice that the singularity at @xmath144 of the asymptotic expansion lies outside the plotted region in figs . [ fig1:propejeb ] and [ fig1:propejec ] . in order to study the @xmath84 case we start by writing the r.h.s . of ( [ newintegral ] ) as @xmath149}{\big\}}\label{trintegral}\ ] ] after employing the usual integral representation for the bessel functions @xmath101 ( @xmath102 ) . again it is helpful to introduce the same neutralizers as above to split the integral in three pieces @xmath104 , @xmath105 . the integral @xmath117 gives the following two contributions : @xmath150 and @xmath151 ^ 3}\bigg\}.\label{doble}\ ] ] both integrals can be computed exactly @xcite in terms of complete elliptic integrals of first and second kind . the first one gives the contribution of the unregulated free commutator ( i.e. , with infinite cut - off ) . plane used in the discussion of the @xmath90 asymptotics and the free commutator . region i is defined by @xmath152 , region ii by @xmath153 , and region iii by @xmath154.,width=321 ] in order to describe it we define the regions i , ii , and iii by @xmath152 , @xmath153 , and @xmath154 respectively . they are shown in fig . [ fig4:regions ] . in region i the free commutator is zero , whereas in regions ii and iii it is given by : region ii @xmath155 region iii @xmath156 the second contribution ( [ doble ] ) can be computed by the method outlined in appendix iv of ref . @xcite , obtaining : region i @xmath157\sqrt{(1+\rho)^2-\tau^2 } } { ( 1+\rho-\tau)^2(1-\rho+\tau)^2(-1+\rho+\tau)^2(1+\rho+\tau)^2 } e\left(\sqrt{\frac{4\rho}{(1+\rho)^2-\tau^2}}\right)\right.\nonumber\\ & & \hspace{2.0cm}\left.-\frac{2\tau^2 } { [ \rho^4+(\tau^2 - 1)^2 - 2(1+\tau^2)\rho^2]\sqrt{(1+\rho)^2-\tau^2 } } k\left(\sqrt{\frac{4\rho}{(1+\rho)^2-\tau^2}}\right)\right\}\end{aligned}\ ] ] region ii @xmath158[(1+\rho)^2-\tau^2]^2 } k\left(\sqrt{\frac{(1+\rho)^2-\tau^2}{4\rho}}\right)\right.\nonumber\\ & & \hspace{1.1cm}\left.+\frac{4\sqrt{\rho } [ 1 - 2\rho^2+\rho^4 + 2\tau^2 - 3\tau^4 + 2\rho^2\tau^2 ] } { [ \rho^4+(\tau^2 - 1)^2 - 2(1+\tau^2)\rho^2]^2 } e\left(\sqrt{\frac{(1+\rho)^2-\tau^2}{4\rho}}\right)\right\}{.}\end{aligned}\ ] ] the value of ( [ doble ] ) in region iii is zero . the integral @xmath126 , on the other hand , can also be studied by the methods described in ref . the first relevant term to its asymptotic expansion in inverse powers of @xmath90 is derived from the double contour integral @xmath159e^{\lambda[\frac{{\lambda_q}}{2}(t_1-\frac{1}{t_1})+ \rho\frac{{\lambda_q}}{2}(t_2-\frac{1}{t_2})- i\tau e^{-\lambda_q}]}}{\lambda_q^2[(t_1+\frac{1}{\displaystyle t_1})^2 + \rho^2(t_2+\frac{1}{\displaystyle t_2})^2]-[t_1-\frac{1}{\displaystyle t_1}+ \rho(t_2-\frac{1}{\displaystyle t_2})+2i\tau e^{-\lambda_q}]^2}\bigg\}{,}\label{borderlambda}\ ] ] corresponding to @xmath128 and whose asymptotic behavior can be determined by employing standard techniques for multiple integrals @xcite . in this way we obtain the following contribution : @xmath160}{1+\rho-\tau e^{-\lambda_q}}-\frac{\sin[\lambda\lambda_q(1+\rho)+ \tau\lambda(1-e^{-\lambda_q})]}{1+\rho+\tau e^{-\lambda_q}}\hspace{5cm}&&\label{asimptbord}\\ \hspace{4cm}+\frac{\cos[\lambda\lambda_q(1-\rho)- \tau\lambda(1-e^{-\lambda_q})]}{1-\rho-\tau e^{-\lambda_q}}-\frac{\cos[\lambda\lambda_q(1-\rho)+ \tau\lambda(1-e^{-\lambda_q})]}{1-\rho+\tau e^{-\lambda_q}}\bigg\}.&&\nonumber\end{aligned}\ ] ] finally , the integral @xmath125 is written in terms of a neutralizer that vanishes at @xmath116 and @xmath128 . this integral is best studied by choosing the unit circumference centered in the origin of the complex plane as the integration contour @xmath8 . the contributions of this integral to the asymptotics of ( [ trintegral ] ) come from the stationary points of the exponent in the integrand whenever they are within the integration region . this fact is controlled by the value of the cut - off @xmath69 . the result is @xmath161\theta(\rho-1)+ \theta(\tau+\rho-1)\theta[(1-\rho)e^{\lambda_q}-\tau]\theta(1-\rho)\ } \im\bigg\{\frac{e^{-i\frac{\pi}{4 } } e^{i\lambda[\tau+|\rho-1|(1+\log\frac{\tau}{|\rho-1| } ) ] } } { r_1\sqrt{2\pi\lambda \rho|1-\rho|}\log\frac{\tau}{|1-\rho|}}\bigg\}\ ] ] @xmath162\im\bigg\ { \frac{e^{i\frac{\pi}{4}}e^{i\lambda[\tau+(\rho+1)(\log\frac{1+\rho}{\tau}-1 ) ] } } { r_1\sqrt{2\pi\lambda \rho(1+\rho)}\log\frac{\tau}{1+\rho}}\bigg\},\hspace{8cm}\ ] ] where the step functions define the regions where the different stationary points contribute . as we can see and it is explained in fig . [ fig3:regcutoff ] , there are two contributions in some parts of the @xmath99 plane , only one in some other parts , and no contribution in the remaining ones . plane used in the discussion of the @xmath90 asymptotics in the presence of a cut - off @xmath163 . the label of each region indicates how many critical points contribute to the asymptotic expansion in @xmath90.,width=321 ] notice that , whenever they differ from zero , these contributions are dominant with respect to those coming from the other integrals . several points are now in order . first it is interesting to realize that the singularity at @xmath164 that exists when the cut - off is taken to be infinite ( and is obviously absent now ) shows up as the region defined by the lines @xmath165 , @xmath166 , and @xmath167 shrinks with growing @xmath69 . another interesting feature of the commutator when the cut - off is present is the appearance of some regions where the leading asymptotic behavior is not given by the expressions obtained in ref . @xcite for infinite cut - off , namely the regions with @xmath168 labeled 0 in fig . [ fig3:regcutoff ] and the region labeled 1 that connects them . on the contrary , there are two regions where two stationary points contribute to the asymptotics just as in the @xmath138 case , showing the characteristic slow decay in the @xmath83 direction . of these two regions , the one closer to the axis is bounded , whereas the second one [ defined by the lines @xmath166 and @xmath169 is not . the effect of the symmetry axis is evident in the sense that it is precisely there where one of the lines that limits the boundary of this region starts , namely @xmath170 . as we can see , the influence of the cut - off is important in some parts of the ( @xmath94 ) plane , but there are others where the asymptotic behavior is described at leading order(s ) by the unregulated @xmath93 limit . the consideration of these different regions helps in describing the intermediate regimes where the infinite cut - off approximation is expected to work , at least for large values of @xmath90 . finally we want to point out that the most dramatic quantum effect observed when the cut - off is infinite the very slow fall - off of the commutator at the axis in the @xmath83 direction is no longer present after introducing a regulator . this casts some doubts about the observability " of large quantum gravitational fluctuations at the axis . these behaviors can be visually appreciated in figs . [ fig5:regcutoffdens]-[fig6:propneq0 - 3 ] . and different values of @xmath69 . comparing the results with those of ref . @xcite , we can see that the commutator in the regions labeled 2 in fig . [ fig4:regions ] is essentially equal to the one corresponding to an infinite cut - off . notice also the process by which the singularity at @xmath59 appears.,width=623 ] for the field commutator over @xmath140 as a function of @xmath83 for @xmath141 , @xmath171 , and @xmath163 compared with a numerical approximation . the regions with different asymptotic regimes are shown . the asymptotic approximation is good except at the boundaries between these regions . the different types of behavior are also evident.,width=623 ] for the field commutator over @xmath140 as a function of @xmath83 for @xmath141 , @xmath172 , and @xmath147 compared with a numerical approximation . , width=623 ] for the field commutator over @xmath140 as a function of @xmath83 for @xmath141 , @xmath172 , and @xmath148 compared with a numerical approximation.,width=623 ] in the above sections we have discussed the asymptotics of the regulated field commutator as a function of @xmath70 , @xmath90 , and @xmath83 . these are dimensionless parameters obtained by using @xmath59 as a length scale . we want to discuss now the possibility of expanding this commutator as a power series in @xmath61 . the main motivation to consider this issue is that one would expect to arrive at an expansion of this kind when adopting a standard perturbative approach for the treatment of the problem . as we will see , this can be done in a rather straightforward way if a cut - off is introduced in the system . however , our description breaks down when the cut - off is removed . let us analyze then the expansion of the vacuum expectation value of the commutator in powers of the quantum gravitational constant$ ] as a power series in @xmath173 . this yielded a convergent series representation for the commutator . nonetheless , the complicated dependence of this expansion in @xmath61 did not allow us to obtain a power series in the gravitational constant . ] @xmath174 . to this end we rewrite ( [ integral ] ) as @xmath65\,|0\rangle= \int_0^{\lambda_k}\!\!\!\!\!\!dk\,j_0(r_1 k)j_0(r_2 k)\sin\left[k{r_1\tau } \frac{(1\!-\!e^{-q})}{q}\right]\!. \hspace*{.4cm}\label{vacom}\ ] ] with our conventions , both @xmath175 and @xmath176 are dimensionless , whereas @xmath177 can be regarded to have dimensions of an inverse length . note that all the dependence on @xmath61 is contained in @xmath112 , accepting that the cut - off @xmath56 is fixed . thus , in order to arrive at the desired series , we will expand the integrand in powers of the variable @xmath112 . at this point , it is worth remarking that , had we described the regulated commutator by means of the dimensionless cut - off @xmath68 as in previous sections ( as well as @xmath177 and @xmath178 ) by multiplication with an arbitrary , fixed length scale independent of the gravitational constant , so that no spurious dependence on @xmath61 is introduced in the process . in this sense , one may always view @xmath56 either as a @xmath61-independent dimensionful cut - off or equivalently as a fixed dimensionless parameter . ] , it would not have been possible to single out the dependence on the gravitational constant via that on @xmath112 . we will use the following formulas for the taylor expansion of the functions involved in our expression ( [ vacom ] ) and the composition of the resulting series , assuming for the moment their convergence : @xmath179}^m&=&\left[\sum_{n=1}^{\infty } \frac{(-q)^n}{(n+1)!}\right]^m=\sum_{p = m}^{\infty}a_p[m]\,(-q)^p,\\ \label{apms } a_p[m]&:=&\sum_{\sigma(p|m)}\prod_{i=1}^{m}\frac{1}{(n_i+1)\,!}\,,\end{aligned}\ ] ] where the range of the last sum extends to the sets of @xmath180-integers @xmath181 given by @xmath182 interchanging the sum and integration orders , one then obtains the formal series @xmath183|0 \rangle=\ ! \sum_{p=0}^{\infty}\int_0^{{\lambda_k}}\!\!dkj_0(r_1k)j_0(r_2k)(-4gk)^p \,f_p(kr_1\tau),\end{aligned}\ ] ] with @xmath184}\frac{(-1)^m}{(2m)!}a_p[2m]\,(kr_1\tau)^{2m}\nonumber\\ & & \hspace{2cm}{+}\cos{(kr_1\tau)}\sum_{m=0}^{{\rm int}[(p-1)/2]}\frac{(-1)^m}{(2m+1)!}a_p[2m+1]\,(kr_1\tau)^{2m+1},\;\;\ ; \;p\geq 1,\\ \!\!\!\!&&f_0(kr_1\tau):=\sin{(kr_1\tau)}\nonumber.\end{aligned}\ ] ] here , the function @xmath185 $ ] is the integer part of @xmath186 , and the sum over @xmath180 that multiplies the function @xmath187 is understood to vanish when @xmath188 . note that , in the case of infinite cut - off , the first ( @xmath189 ) term reproduces the commutator of the free - field theory . moreover , then all the @xmath190 additions to the free field contribution are integrals over @xmath3 of oscillating , non - bounded functions and , hence , at best conditionally convergent . so , in the unregulated theory ( @xmath92 ) , the above expansion should be taken only as a formal expression , and therefore we expect that the corresponding vacuum expectation value of the field commutator is not analytic in @xmath61 . of course these problems disappear when we admit the existence of a finite cut - off @xmath191 . taking into account that all the functions @xmath192 are analytic in @xmath177 around the positive real axis , because @xmath193 is a finite combination of products of analytic functions , and that so are the zeroth - order bessel functions that appear in the integrals of ( [ vaser ] ) , it is easy to conclude that all those integrals are well - defined when they are restricted to a compact interval @xmath194 $ ] . thus , each term in the power series ( [ vaser ] ) is finite for any finite positive value of @xmath56 . in the rest of this section , we will discuss the formal manipulations that we have carried out with infinite sums in order to deduce the above expansion . first , notice that the taylor series in ( [ expa ] ) , which is obtained from that of the exponential function , has an infinite convergence radius . when this series is substituted in ( [ vacom ] ) , one obtains a trigonometric function similar to that on the l.h.s . of ( [ sina ] ) , but with @xmath195 expanded in powers of @xmath112 . on the other hand , relation ( [ sina ] ) is just the formula for the sine of the sum of two angles , with the resulting functions @xmath196 and @xmath197 replaced with their taylor expansion . the series compositions @xmath198}$ ] and @xmath199}$ ] can then be rearranged without problems employing for @xmath200}^m$ ] the value given in ( [ compo ] ) because @xmath201 ( that we recall that converges for all @xmath202 ) is always smaller than the convergence radii of the sine and cosine series , which are in fact infinite . in this way , one arrives at a expectation value of the regulated commutator that is equal to an integral over the interval @xmath203 $ ] of the series of functions @xmath204 , with @xmath205 since the functions @xmath206 are clearly continuous in @xmath207 $ ] ( for all allowed values of @xmath59 , @xmath60 and @xmath83 ) and this interval is compact , they are all integrable in that region . as a consequence , it is sufficient that the considered series of functions converges uniformly in @xmath207 $ ] to guarantee that the integration can be interchanged with the infinite sum . we will postpone to the next section the proof of this uniform convergence , at least for a convenient choice of the cut - off . in conclusion , we have seen that the field commutator in vacuo , regulated with a ( dimensionful ) fixed cut - off , can be expanded as a power series in the gravitational constant @xmath61 , each term in the series being finite . besides , all the manipulations performed to deduce this series are rigorously justified provided that the cut - off is chosen so that the series @xmath204 converges uniformly in @xmath207 $ ] . furthermore , in fact this requirement of uniform convergence automatically ensures that the corresponding integrated power series ( [ vaser ] ) converges , and that it does so to the actual value of the expectation value of the regulated commutator . we want to demonstrate that there exists a non - zero value of the cut - off for which the series @xmath208 converges uniformly in @xmath207 $ ] for any fixed non - negative value of @xmath59 , @xmath60 , and @xmath83 . let us start by finding a convenient upper bound for the coefficients @xmath209 $ ] , with @xmath210 , defined in ( [ apms ] ) . first , note that @xmath209=0 $ ] unless @xmath211 , because no set of the form @xmath212 exists with @xmath213 if @xmath214 . in addition , since @xmath215 for @xmath213 , we have that @xmath216\leq \frac{1}{2^m}\,\sum_{\sigma(p|m)}1.\ ] ] from our definition ( [ sigma ] ) , the last sum equals the different ways to arrange @xmath217 non - distinguishable elements [ namely , the excess about its minimum of the sum of @xmath180 elements @xmath213 , which equals @xmath217 for @xmath212 ] between @xmath180 different sets [ which correspond to the @xmath180 integers @xmath181 ] . the result is given by the permutations of @xmath220 elements [ the latter @xmath221 elements representing movable delimiters between the @xmath180 sets ] with possible repetition in @xmath217 ( the genuine , non - distinguishable elements ) and in @xmath221 ( the imaginary delimiters ) . thus , @xmath222\leq \frac{1}{2^m}\,\frac{(p-1)\,!}{(p - m)\,!(m-1)\,!}\leq \frac{1}{2^m}\,\frac{p\,!}{(p - m)!}.\ ] ] in the last inequality we have employed that @xmath223 for all @xmath224 . using that the absolute value of the sine and the cosine is never greater than the unity , it is not difficult then to deduce from ( [ fms ] ) the following bound for @xmath225 , with @xmath226 , @xmath227\leq \sum_{m=1}^p \left(\frac{k{r_1\tau}}{2}\right)^m \left(\begin{array}{c}p\\m\end{array}\right)=\left(1+\frac{k{r_1 \tau}}{2}\right)^p -1.\ ] ] in the last step , we have employed the formula of the binomial expansion . likewise , since the zeroth - order bessel function is bounded by the unity in the positive real axis , we get that , for all non - negative values of @xmath59 and @xmath60 , @xmath228\leq \left[4gk\left(1+\frac{k{r_1\tau}}{2}\right ) \right]^p.\ ] ] the last inequality is trivial , given that @xmath229 . note also that the bound on the r.h.s . is valid even in the case @xmath189 , taking into account ( [ fmrs ] ) . finally , since @xmath230 is a strictly increasing function of @xmath177 in @xmath194 $ ] , we obtain a bound independent of the variable @xmath177 in the interval considered , @xmath231^p.\ ] ] to obtain the desired convergence properties , it will suffice to require that @xmath232 we then get that both @xmath233 and its integral over @xmath194 $ ] are small corrections for large @xmath234 , which tend to zero in the limit @xmath235 . in order to prove the uniform convergence of the series @xmath236 , we have to check that , for each @xmath237 , there exists an integer @xmath238 such that , for every @xmath207 $ ] , @xmath239 taking into account inequality ( [ pert ] ) , it is clear that , given @xmath237 , we can always find a sufficiently large integer @xmath238 for which @xmath240^p<\epsilon\left[1 - 4g\lambda_k \left(1+\frac{\lambda_k{r_1\tau}}{2}\right ) \right].\ ] ] note that the choice of this @xmath238 depends only on the values of @xmath241 , @xmath56 , @xmath61 and @xmath242 . using ( [ fmrsb2 ] ) , we then have @xmath243^p = \frac{\left[4g\lambda_k\left(1+\frac{\lambda_k{r_1\tau}}{2 } \right ) \right]^p}{1 - 4g\lambda_k\left(1+\frac{\lambda_k{r_1\tau}}{2 } \right)}\leq \epsilon.\nonumber\end{aligned}\ ] ] so , inequality ( [ unico ] ) is valid for all @xmath177 in the considered interval @xmath194 $ ] , as we wanted to prove . we have thus shown that , for a given @xmath83 , every choice of the cut - off @xmath244 that satisfies condition ( [ pert ] ) leads to a convergent power series in the gravitational constant @xmath61 for the expectation vacuum of the regulated commutator , regardless of the radial coordinates @xmath59 and @xmath60 . moreover , the power expansion converges indeed to the true value of this regulated commutator in vacuo . we have studied in this paper the issue of microcausality for quantum einstein - rosen waves after a suitable cut - off is introduced to regulate the quantum fields . in more detail , we have considered the introduction of a momentum cut - off @xmath56 ( or its dimensionless counterpart @xmath69 ) . we have discussed first the asymptotic expansions in terms of the dimensionless parameters @xmath70 , @xmath83 , and @xmath90 along the lines of ref . @xcite . owing to the fact that these parameters are defined with the help of @xmath59 , in principle one does not need to make explicit the dependence of the cut - off @xmath56 on @xmath61 in this case . on physical grounds , one could view this cut - off , for example , as the inverse of the planck length . we have seen that the introduction of a finite cut - off modifies some of the conclusions obtained in ref . in particular we have seen that some of the most dramatic effects present when the cut - off is infinite ( in particular the behavior of the field commutators in the symmetry axis ) are now somewhat mitigated . nevertheless we have been able to show that the approximation provided by the unregulated field commutator is a good one in some regions of the @xmath99 plane , and , in fact , there is an unbounded region where that approximation prevails . this indicates that , even though the influence of the cut - off is felt in some regions of the parameter space , it is irrelevant in others . in secs . vi and vii , on the other hand , we have considered the expansion of the field commutator in terms of the gravitational constant @xmath61 . we notice , nonetheless , that condition ( [ pert ] ) on the cut - off @xmath56 , that guarantees the convergence of the series , depends on @xmath61 . at this stage , one possibility would be to admit that the cut - off depends on the gravitational constant ; however , the expansion obtained would then fail to provide a genuine power series in @xmath61 , because this parameter would also enter the different terms in the series via the implicit dependence of @xmath56 on it . another possibility that indeed respects the interpretation of our expansion as a power series in @xmath61 is the following . employing that condition ( [ pert ] ) is an inequality equation for @xmath56 given in terms of a function of @xmath61 that is strictly increasing , it is easy to see that the inequality is satisfied for all values of @xmath61 in a certain interval @xmath245 $ ] if and only if it is satisfied for @xmath246 . something similar happens with respect to the dependence on the value of @xmath247 , so that if we want to consider a whole time interval of the form @xmath248}$ ] , we only have to evaluate our condition at the maximum time lapse . in other words , to ensure the convergence of the series for @xmath249 $ ] and any time difference in @xmath250 $ ] , we only have to demand the requirement ( [ pert ] ) at @xmath251 and @xmath252 , because then @xmath253 in this way we arrive at a cut - off that is independent of the particular values considered for @xmath254 and the gravitational constant ( in the commented intervals ) , and our expansion becomes a true power series in @xmath61 . the above inequality leads to the following positive upper bound for @xmath56 : @xmath255 therefore , with a cut - off that satisfies this condition , the power series ( [ vaser ] ) converges in the interval @xmath245 $ ] for all radial positions @xmath59 and @xmath60 and @xmath256}$ ] . when @xmath257 is small , the bound on @xmath56 is approximately @xmath258 , whereas for large @xmath257 it is nearly equal to @xmath259 . in particular , with this bound the cut - off would have to be vanishingly small if we want a good convergent behavior in an infinitely large time interval ( @xmath260 ) . an open question is whether it is possible or not to find a different , non - zero time - independent cut - off such that the expansion of the regulated commutator converges for any value of the time elapsed , i.e. for all @xmath261 . we expect to encounter convergence problems when the time interval is unbounded ; for instance , one can prove that the series ( [ vaser ] ) does not converge uniformly in @xmath262 with any choice of the cut - off @xmath56 ( for generic @xmath59 and @xmath60 ) . nonetheless , one can in fact consider a kind of semi - classical limit in which @xmath246 tends to zero ( and hence so does the value of the gravitational constant , that had been restricted to @xmath245 $ ] ) , while the time interval where the convergence is granted reaches infinity . in order to do this , one only needs to allow a dependence of @xmath257 on @xmath246 , so that the assumed maximum value of the time difference varies with that of the gravitational constant . suppose , let us say , that @xmath263 with @xmath264 . then , the bound ( [ cutgt ] ) on the cut - off becomes @xmath265 thus , when @xmath246 tends to zero , we get the asymptotic behavior @xmath266 . since @xmath267 and @xmath268 diverge for vanishing @xmath246 , because @xmath264 , we therefore conclude that the cut - off can be removed in the limit @xmath269 while ensuring that the time interval @xmath270 $ ] , where the expansion is well - defined , covers the positive real axis . we finally discuss the physical interpretation of this type of cut - off . it turns out to be intimately related to the maximum resolution that can be reached for the physical time when a certain perturbative approach is adopted to describe the quantum dynamics @xcite . in such an approach , one expands the evolution generator in powers of @xmath61 and regards the free - field hamiltonian as the dominant contribution , with the higher powers seen as corrections . the auxiliary time @xmath9 , associated with the free - field hamiltonian , plays then the role of evolution parameter in the quantum theory , whereas the physical time becomes an operator . it was shown in ref . @xcite that , under these circumstances , a resolution limit @xmath271 emerges for the physical time , @xmath272 ^ 2\geq 4g^2 + 4 g t.\ ] ] employing the inequality @xmath273 for @xmath274 , evaluated at @xmath275 , one can easily check from condition ( [ cutgt ] ) that the inverse of the cut - off satisfies @xmath276 therefore , the bound on @xmath277 equals that on the time resolution @xmath271 for a value @xmath251 of the gravitational constant and a time elapsed @xmath278 ( and thus of the same order as @xmath279 ) . in this sense , one can assign to @xmath277 the interpretation of a genuine resolution limit in the physical time . the future prospects for this line of work will focus on the issue of deriving and obtaining meaningful physical information from the @xmath4 matrix of the model . we feel that the mathematical techniques employed here to study the asymptotics of field commutators , with and without a cut - off , will be also helpful in analyzing this issue . we plan to concentrate on this problem in the future . this work was supported by the spanish mec under the research projects bfm2001 - 0213 and bfm2002 - 04031-c02 - 02 . the authors want to thank a. ashtekar , j. a. torresano , and m. varadarajan for helpful comments , as well as j. klauder for an interesting question .

we present the results of simulations on the detectability of in the atmosphere of earth twins around nearby low mass stars using high resolution transmission spectroscopy . we explore such detectability with each of the three upcoming extremely large telescopes ( elts ) , i.e. gmt , tmt and e - elt , and high resolution spectrographs , assuming such instruments will be available in all elts . with these simulations we extend previous studies by taking into account atmospheric refraction in the transmission spectrum of the exo - earth and observational white and red noise contributions . our studies reveal that the number of transits necessary to detect the in the atmosphere of an earth twin around m - dwarfs is by far higher than the number of transits estimated by snellen et al . ( 2013 ) . in addition , our simulations show that , when accounting for typical noise levels associated to observations in the optical and near - infrared , the a - band at 760 nm is more favorable to detect the exoplanetary signal than the band at 1268 nm for all the spectral types , except m9v . we conclude that , unless unpredicted instrumental limitations arise , the implementation of pre - slit optics such as image slicers appear to be key to significantly improve the yield of this particular science case . however , even in the most optimistic cases , we conclude that the detection of in the atmosphere of an earth twin will be only feasible with the elts if the planet is orbiting a bright close - by ( d 8 pc ) m - dwarf with a spectral type later than m3 .

1312.1585

the path to finding exoplanets either inhabited or amenable for life as we know it on earth includes the detection in their atmospheres of chemical compounds such as @xmath2 , @xmath3 , @xmath4 , and @xmath5 , known as biomarkers ( schindler et al . 2000 , pavlov et al . 2000 , kaltenegger & traub 2007 , traub et al . another important biomarker is gaseous oxygen , @xmath0 , detected by sagan et al . ( 1993 ) while analyzing the spectrum of earth observed by the galileo probe . in that spectrum @xmath0 appeared to be , together with @xmath4 in strong thermodynamical disequilibrium , the strongest indicator of life on earth . @xmath0 has since drawn significant attention because it provides a strong indication of the presence of oxygen - producing forms of life and because it has several spectral absorption bands at visible and near - infrared wavelengths , which can be detected using ground - based telescopes . the most prominent of those bands are the so - called _ @xmath0 a - band _ around 760 nm , and the @xmath0 band at 1268 nm , with the a - band being the strongest with 55 strong lines . several studies have now investigated the detectability of @xmath0 for an earth - like exoplanet in the habitable zone of its host star . the first of those studies was published by schneider ( 1994 ) who analytically calculated the detectability of the @xmath0 a - band by estimating the photometric depth of the absorption produced by the full band , which has a width of @xmath6 3 nm . computing the ratio of the projected surface areas of the planet s atmosphere and the disk of the star during a transit , this study concluded that it was possible to detect the @xmath0 a - band with a 2.4 meter space - bound telescope , assuming the planet orbits a solar type star at a distance of 2 parsecs or a 0.3 @xmath7 star at 10 parsecs . webb & wormleaton ( 2001 ) refined the schneider ( 1994 ) analysis by using empirical observations of the earth s @xmath0 a - band to simulate the transmission spectrum of the exoplanet and by superimposing the resultant spectrum with the spectrum of the host star . their analysis concluded that it is possible to detect the planetary @xmath0 a - band for host stars with r @xmath8 0.3 @xmath7 , although detectability with current 8-meter class telescopes would be limited to m dwarfs with @xmath9 10 . this study also proposed that a way to confirm that the @xmath0 lines originated in the atmosphere of the exoplanet is that they should be offset from the earth s own @xmath0 lines by an amount equal to the peculiar velocity of the host star . however , they did not include the telluric @xmath0 a - band in their analysis . a few years later , kaltenegger & traub ( 2009 ) generated simulations using the earth s atmosphere as proxy to estimate the detectability of biomarkers during the transit of an earth twin around a sun - like star and also around m - dwarfs , for a 6.5-meter telescope in space . their main conclusion was that the signal - to - noise ratio ( s / n ) of a single transit would not be enough to detect any spectral features in the atmosphere of an exo - earth orbiting in the habitable zone of those stars . instead , multiple transits would be needed , with an estimate of 6 to 192 transits per year for m0v to m9v stars . they also concluded that the spectral features of different molecules in the atmosphere of an exo - earth around an m star will be easier to detect in the infrared than in the optical , because the stellar flux peaks in the infrared . just recently , snellen et al . ( 2013 ) extended previous studies by including the effect of the telluric @xmath0 a - band and applying the latest techniques to detect the planet s doppler shift from the ground ( e.g. brogi et al . 2012 , rodler et al . their simulations show that it will be possible to detect the @xmath0 a - band in the atmosphere of an earth twin if a spectrograph with a spectral resolving power of @xmath10 is installed on the next generation of extremely large telescopes ( elts ) . in particular , they concluded that the 39-meter e - elt , will be able to detect the @xmath0 a - band transmission signal of a planet around an m5v star by combining the observations of a few dozen transits . in addition , they remarked that in the case of stars smaller than m7v the @xmath0 might be easier to detect in the @xmath0band at 1268 nm since those stars will be brighter at near - infrared wavelengths than in the visible . in this paper we further refine the snellen et al . ( 2013 ) simulations by 1 ) accounting for atmospheric refraction effects in the exoplanet s transmission spectrum and 2 ) including the effect of random and correlated ( atmospheric and instrumental ) observational noise . in addition , we extend previous studies by exploring the expected performance of instruments for the three upcoming elts , exploring a range of spectral resolutions , and performing simulations for a range of stellar peculiar velocities and spectral types . finally , we explore in detail the feasibility of the observations both in @xmath0 a - band and the @xmath0 band at 1268 nm . in sections 2 and 3 we describe the implementation and results of our simulations . a discussion of those results and the optimal observational set - up is given in section 4 . the first step to perform these @xmath0 detection feasibility studies is to generate a set of hypothetical exo - earth transiting planet configurations and observational setups . the parameters necessary for the simulations are 1 ) a model transmission spectrum of the earth - like planet ( which we assume to be an earth twin , including atmospheric conditions and chemistry ) , 2 ) a model transmission spectrum of earth , 3 ) a model of the spectral emission of the star , including assumptions about the relative velocity of the star with respect to earth , 4 ) a telescope+instrument setup configuration , and 5 ) observational noise models . the model atmospheric transmission spectrum of the exo - earth was calculated adopting a line - by - line radiative transfer model ( lblrtm ) code based of the fascode algorithm ( clough et al . 2005 ) . as molecular database for oxygen models we adopted hitran ( rothman et al . 2009 ) . in those calculations we accounted for the effect of refraction in the transmission spectrum of the exo - earth . as described in garcia - muoz et al . ( 2012 ) , the transmission spectrum of an earth twin observed transiting its host star will differ from the earth s transmission spectrum because of refraction . refraction is the effect by which light crossing the planet s atmosphere near its surface gets deflected by an angle that depends on the properties of the atmosphere itself , the size of the star , and the orbital distance of the planet . this effect causes the atmospheric layers close to the planet s surface to be invisible for a remote observer . as a consequence , any chemical species at heights lower than that limit can not be observed . for species distributed over a large range of atmospheric heights , as is the case of @xmath0 , the depth of their absorption lines will be altered . for e. g. , in the particular case of the earth - sun system atmospheric layers at heights lower than 12 - 14 km will be invisible to a remote observer seeing the system transit . in the case of smaller stars , i.e. m - dwarfs , where both the radius of the star and the orbital distance of an earth - like planet in the habitable zone of the system are smaller , only atmospheric altitudes lower than about 5 km appear invisible . in this work we focus on earth twins orbiting around m - dwarf stars ( see section 2.1.3 ) . therefore , to simulate this refraction effect we only integrate the model transmission spectrum of the exo - earth at atmospheric layers between 5 km to 85 km from the surface . 85 km marks the end of the earth s mesosphere , with over 99@xmath11 of the atmospheric mass contained below that atmospheric height ( lutgens & tarbuck 1995 ) . therefore , the contribution of the atmospheric layers above that height to the transmission spectrum of the planet can be considered minuscule . the resulting model transmission spectrum of the exo - earth around the @xmath0 a - band as well as the @xmath0 band at 1268 nm are illustrated in the top panels figures 1 and 2 . [ f1 ] [ f2 ] the transmission spectrum of earth ( telluric spectrum ) was calculated following the procedure outlined by seifahrt et al . ( 2010 ) . like in the previous section , this procedure uses the lblrtm code developed by clough ( 2005 ) and the transmission molecular absorption database hitran containing the 42 most prominent molecules and isotopes present in the atmosphere of earth . as an input to lblrtm , we adopted a model of the earth s atmosphere that contains meteorological information for temperature and pressure as a function of height for an average night at a given observatory . in our case we retrieved the weather information from the global data assimilation system ( gdas ) . gdas models are available in 3 hour intervals for any location around the globe . by employing lblrtm , we then calculated absorption and emission spectra for a given path through the atmosphere , i.e. for a given airmass . in our model we have assumed that the observations take place at a ground - based observatory at an altitude of about 2500 meters above sea level , and we have accounted for a changing airmass values during the course of the night within the range @xmath12 to @xmath13 . the transmission spectrum around the @xmath0 a - band and the @xmath0 band at 1268 nm for an airmass of 1.3 is illustrated as example in the middle panels of figures 1 and 2 . we note that a few h@xmath14o absorption lines appear in the telluric spectrum around 1290 nm . these lines partly overlap with the o@xmath14 lines and constitute an additional source of noise . in our simulations , we have considered the depth of these lines to be constant for simplicity reasons . in fact , however , the depths of the water absorption lines may strongly vary during the course of a night ( e.g. rodler et al . 2012 ) , an effect that should be implemented in future simulations . to model the emission from the star , we focus on m - type stars with spectral types m1v m9v . the reason for this , as already explained by kaltenegger & traub ( 2009 ) and snellen et al . ( 2013 ) , is that the planetary transit depths are inversely proportional to the squared radius of the star and therefore , the transmission spectrum signal of an exo - earth will be easier to detect around a star with smaller radius . for a star like the sun ( g2v ) , the ratio of the areas between the stellar disk and the atmosphere of an exo - earth s ring is of the order of 5@xmath15 @xmath16 , while for an m1v star that ratio is @xmath6 1.25@xmath15@xmath16 , and for an m9v star the ratio drops to @xmath6 4@xmath15@xmath17 . we adopted stellar model spectra for effective temperatures and surface pressures corresponding to m1v m9v stars using the new library of high - resolution spectra based on the phoenix code and presented in husser et al . this new library has been generated using a new equation of state , which improves the treatment of chemical equilibrium in stellar atmospheres , and updated atomic and molecular line lists . these new models provide a better match to observed spectra of m dwarfs than the latest phoenix ( brott et al . 2005 ) and atlas9 ( castelli et al . 2003 ) models . for simplicity , we assumed solar abundances in all the models . as an example , we show the model spectrum generated for an m4v star with effective temperature @xmath18 k , gravity @xmath19 dex and solar abundances in the bottom panels of figures 1 and 2 . there are three extremely large telescopes ( elts ) with diameters 24 meters or larger expected to start operations in the next decade . the smallest of those elts is the giant magellan telescope ( gmt ; www.gmto.org ) with an effective diameter of 24.5 meters . one of the first light instruments on the gmt will be g - clef ( szentgyorgyi et al . 2012 ) , which will be a high - resolution spectrograph with resolution modes between r = 25,000120,000 and spectral coverage between 350950 nm , suitable to study the @xmath0 a - band . the second largest of the elts is the thirty meter telescope ( tmt ; www.tmt.org ) . among the tmt s expected suite of instruments the best suited to detect @xmath0 in the atmosphere of an earth - like planet is the high resolution optical spectrometer ( hros : crampton et al . 2008 ) , listed among the telescope s _ first decade _ instruments . hros will provide a resolution of @xmath20 for a 1 arcsec slit , or @xmath21 with an image slicer , and spectral coverage between @xmath6 3101000 nm . an image slicer is a device that allows to feed a large fraction of the stellar flux into a narrow slit , thereby avoiding slit losses and at the same time allowing to employ high spectral resolving powers ( see e.g. dekker et al . 2003 ) . the largest of the elts will be the european extremely large telescope ( e - elt ) , with a diameter of 39-meters . in the case of the e - elt , there are two spectrographs with resolution @xmath22 being studied on phase a , codex ( pasquini et al . 2010 ) and simple ( origlia et al . 2010 ) . codex only covers wavelengths up to @xmath6 710 nm , while the wavelength coverage of simple starts at @xmath6 800 nm . therefore none of these instruments include the @xmath0 a - band at 760 nm . simple , on the other hand , will cover the @xmath0 band at 1268 nm . a third high - resolution spectrograph called hires , is now being proposed as one of the first light instruments for the e - elt , although the specifications for this instrument are not defined yet . the last of the simulation inputs is noise . the noise in an observed spectrum can be divided into random ( white ) and correlated ( red ) noise . while the distribution of white noise is poissonian and can be reduced by increasing the s / n of the spectrum , the red noise is not random and can not be reduced by increasing the s / n . however , the red noise can be related to instrumental and atmospheric effects , such as e.g. variations of airmass or changes of the position of the spectrum on the detector . analysis techniques have improved in recent years to levels where it is now possible to significantly reduce the amount of red noise in the data . in our simulations we have added both sources of noise . white noise is simulated using a random distribution function with a maximum amplitude set by the s / n of the particular model ( see section 2.2 ) . in the case of red noise , its pattern can contain a series of frequencies that depend on the specific cause of the noise for a given dataset . therefore , we have simulated red noise using the approach described in ( kasdin 1995 ) , by assuming a functional form of 1/f , where @xmath23 is the frequency . the amplitude of the red noise is set to a fraction of the white noise in an specific model . for our simulations we consider red noise contributions of 0 to 100@xmath11 of the white noise . typical red noise levels of ground based data remain of the order of 20@xmath11 of the white noise in the optical and 50 - 100@xmath11 of the white noise in the near - infrared ( see e.g. pont et al . 2006 , rogers et al . 2009 ) . using the inputs described above we generate model spectra , @xmath24 , using a expression of the form @xmath25 where @xmath26 in these equations , @xmath27 denotes the speed of light , @xmath28 is the spectrum of the host star , @xmath29 is the transmission spectrum of the planetary atmosphere , and @xmath30 is the telluric spectrum of earth for a given airmass . @xmath31 is the instrumental profile ( ip ) of the spectrograph . in our simulations , we assume the ip to be a gaussian function that degrades the spectral resolution of the combined spectrum to a chosen value . the remaining parameters in the equations are @xmath32 , the relative stellar velocity with respect to earth ( i.e. systemic velocity and barycentric velocity ) , @xmath33 , the instantaneous radial velocity of the planet with respect to its host star ( which is close to 0 km s@xmath34 at a primary transit ) , and @xmath35 , the area ratio between the stellar disk and the atmospheric ring of the planet . once we have computed the spectrum @xmath24 , we interpolate it onto a pixel grid , which simulates a given instrumental layout . since the exact parameters of most instruments described in section 2.1.4 are not yet fully defined , we adopt two different pixel grid scales ; one with a velocity resolution of 0.75 km s@xmath34 pixel@xmath34 and the other 1.2 km s@xmath34 pixel@xmath34 . the 0.75 km s@xmath34 pixel@xmath34 grid scale corresponds to the planned scale of g - clef for resolutions @xmath36 ( szentgyorgyi et al . 2012 ) , while the 1.2 km s@xmath34 pixel@xmath34 grid scale corresponds to uves ( dekker et al . 2000 ) , which is currently mounted at the 8.2 very large telescopes ( vlt ) of the european southern observatory ( eso ) . the next step is to adjust the flux of the model spectra based on the expected flux of the exo - earth s host star that is received on earth . given that the potential transiting exo - earth host stars will be at a range of possible distances from earth , and therefore will have different apparent magnitudes , we consider a range of brightnesses for stars of a given spectral type . ( snellen et al . 2013 ) and ( kaltenegger et al . 2009 ) computed the most likely apparent magnitude of a nearby transiting exo - earth host star based on the solar neighbourhood mass function and transit probability of an exo - earth and they provided detectability estimations that are based on that number ( see e. g. section 3 in snellen et al . 2013 ) . here we adopt a different approach and instead of making any assumption about the apparent magnitude of the transiting exo - earth host star , we assume that such star can be at any distance within 20 pc . based on the star s distance , we adjust its apparent magnitude between 1 and 20 pc . the effective temperature and radius of the stars adopted for each spectral type are listed in table 1 . the table also shows the values of the area ratio between the stellar ring and the atmospheric ring of the planet , @xmath35 , the i - band and j - band magnitudes of each star at 10 pc , the orbital period of a planet in the middle of the habitable zone of the star based on the definition by kasting ( 1993 ) , and the duration of the transits of such planet . the flux that we finally measure from a star of a given magnitude will depend on the telescope and instrument throughput . therefore , we estimated those fluxes using online exposure time calculators ( etc ) for each instrument , whenever available . for all etcs , we assumed a median seeing of 0.8 " and airmass values ranging from 1.1 to 2.0 . g - clef already has an etc , while for simple , we adopted the throughput value estimates listed on the project webpage . the instrumental layout of hros has not been published yet ; we therefore assumed that hros will be similar to g - clef , with a factor of 1.5 shorter integration times due to the larger diameter of the tmt with respect to the gmt . for a spectrograph with a uves - like layout , we used the uves etc and scaled the results accordingly to the diameter ratio between the elt and the vlt with 8.2 m. for this instrument , we calculated the s / ns for different slit widths , resulting in different spectral resolving powers . in addition , we also determined the s / n when employing the uves image slicer # 3 , which has a resolution of @xmath37 ( dekker et al . 2003 ) . with the etcs we estimated the exposure time needed to attain the maximum number of counts for each star s apparent magnitude , but at the same time avoiding reaching typical non - linearity regimes of 36 000 adu and 12 000 adu ( in the brightest pixel on the chip ) in a single exposure , respectively for the ccds and near - infrared detectors . we furthermore accounted for the integration time and the dead time between two exposures to estimate the observations duty cycle for each star . while for ccds we assumed a dead time of 15 seconds and a read - out noise of 5 electrons per exposure , we accounted for a dead time of 10 seconds and a read - out noise of 10 electrons for infrared detectors . the maximum single exposure time was chosen to be 600 seconds , since at longer exposures the radial velocity shift of the remote planet would smear out the absorption lines in its atmospheric spectrum . combining the duty cycle of each star with the flux collected per integration we obtained the total number of photons collected during the duration of one transit ( see last column in table 1 ) . the duty cycles for each star , for each instrument , are summarized in tables 2 and 3 . the remaining columns in these tables are explained in section 3.4 . finally , we added poisson noise and red noise to each model based on their flux levels ( i.e. s / n ) , following the description in section 2.1.5 . before adding the noise , we scaled each spectrum in such a way that the actual s / n per spectral pixel corresponded to the square root of the flux at that pixel . this approach ensures that regions of telluric lines as well as stellar absorption lines have a higher noise and a lower s / n than the continuum . we generated data sets simulating transmission spectra of planetary transits using eqs . 1 and 2 with the following range of parameters , @xmath38 to @xmath39 km s@xmath34 , @xmath40 , depending on the spectral type of the host star , and spectral resolutions of @xmath41 , as well as a set of s / n levels including red noise contributions from 0 to 100@xmath11 of the white noise level . in each data set , we simulated the observations during a night . hence , each data set consisted of a series of spectra , which differed from each other by the airmass and the instantaneous radial velocity shift of the planetary transmission spectrum ( @xmath42 ) with respect to its host star during transit . additionally , we factored in the effects of the change of the barycentric velocity of the star plus its planet during the course of a night , which is of the order of a few times 100 m s@xmath34 . furthermore , to rule out selection effects coming from the adopted pixel grid , we also varied the zero - point of the pixel grid up to the average velocity span of one pixel . the duration of the datasets corresponds to twice the duration of the transits listed in the last column of table 1 . to analyze each transit dataset in search for the extraterrestrial signal of @xmath0 , we followed the approach outlined in rodler et al . ( 2012 , 2013 ) and removed both a model telluric spectrum and a model stellar spectrum from each synthetic spectrum , thereby calculating a residual spectrum which ideally only contained the transmission spectrum of the planet , plus noise . we then carried out cross - correlations between the residual spectrum and the transmission model spectrum of the exo - earth , and determined the cross - correlation function . for all spectra of each data set , we then summed up the cross - correlation functions in the rest frame of the planet and determined the radial velocity value that yielded the highest correlation ( candidate feature ) . to determine the number of false positives ( i.e. the confidence level of the candidate feature ) , we employed a bootstrap randomization method ( e.g. rodler et al . random values of the orbital phases were assigned to the observed spectra , thereby creating @xmath43 different data sets ( @xmath44 in our analyses ) . any signal present in the original data was then scrambled in these artificial data sets . for all these randomized data sets , we re - ran the data analysis in the given parameter search ranges and located the candidate feature with its maximum value . the confidence level of the candidate features was estimated to be @xmath45 , where fap is the false alarm probability and @xmath46 is the number of the best fit models having a peak value larger or equal than the peak value of the cross - correlation function found in the original , unscrambled data sets . a detection is considered significant when the signal is at least @xmath47 , which corresponds to a fap of @xmath48 . lccccccc[t1 ] g2v & 5800 & 1 & 520 000 & 4.1 & 3.6 & 365.2 & 13.1 + m1v & 3600 & 0.49 & 125 000 & 7.7 & 6.4 & 43 & 4.0 + m2v & 3400 & 0.44 & 101 000 & 8.3 & 6.5 & 33 & 3.4 + m3v & 3250 & 0.39 & 80 000 & 8.8 & 7.1 & 27 & 3.0 + m4v & 3100 & 0.26 & 35 000 & 10.0 & 7.9 & 16 & 2.1 + m5v & 2800 & 0.20 & 21 000 & 11.2 & 8.6 & 9.5 & 1.5 + m6v & 2600 & 0.15 & 12 000 & 12.4 & 10.1 & 6.0 & 1.1 + m7v & 2500 & 0.12 & 7 500 & 13.6 & 10.7 & 4.1 & 0.78 + m8v & 2400 & 0.10 & 6 000 & 13.9 & 11 & 3.3 & 0.69 + m9v & 2300 & 0.08 & 4 000 & 14.7 & 11.6 & 1.9 & 0.43 + l|cccc|cccc[t2 ] g2v & 840 & 0.05 & 65 & @xmath49 & 429 & 0.08 & 34 & @xmath50 + m1v & 110 & 0.58 & 28 & @xmath51 & 67 & 0.71 & 17 & @xmath52 + m2v & 98 & 0.69 & 29 & @xmath53 & 62 & 0.80 & 19 & @xmath54 + m3v & 88 & 0.79 & 29 & @xmath55 & 57 & 0.87 & 19 & @xmath56 + m4v & 43 & 0.92 & 21 & @xmath57 & 29 & 0.95 & 14 & @xmath58 + m5v & 46 & 0.97 & 31 & @xmath59 & 31 & 0.98 & 21 & @xmath60 + m6v & 43 & 0.98 & 39 & @xmath61 & 30 & 0.98 & 27 & @xmath62 + m7v & 35 & 0.98 & 45 & @xmath63 & 24 & 0.98 & 31 & @xmath64 + m8v & 34 & 0.98 & 48 & @xmath65 & 25 & 0.98 & 36 & @xmath64 + m9v & 29 & 0.98 & 66 & @xmath66 & 22 & 0.98 & 50 & @xmath67 + l|cccc|cccc [ t3 ] g2v & 470 & 0.18 & 37 & @xmath68 & 448 & 0.13 & 35 & @xmath69 + m1v & 133 & 0.86 & 33 & @xmath70 & 97 & 0.81 & 24 & @xmath71 + m2v & 133 & 0.91 & 40 & @xmath72 & 95 & 0.87 & 28 & @xmath73 + m3v & 130 & 0.94 & 44 & @xmath74 & 91 & 0.92 & 30 & @xmath75 + m4v & 70 & 0.98 & 34 & @xmath76 & 48 & 0.97 & 23 & @xmath77 + 5v & 79 & 0.98 & 53 & @xmath78 & 53 & 0.98 & 36 & @xmath77 + m6v & 75 & 0.98 & 68 & @xmath79 & 51 & 0.98 & 46 & @xmath80 + m7v & 61 & 0.98 & 78 & @xmath81 & 41 & 0.98 & 53 & @xmath58 + m8v & 69 & 0.98 & 100 & @xmath81 & 46 & 0.98 & 67 & @xmath58 + m9v & 67 & 0.98 & 154 & @xmath80 & 45 & 0.98 & 104 & @xmath60 + + for a distance of 5 pc l|cccc [ t4 ] m3v & 713 & 0.12 & 238 & @xmath82 + m4v & 173 & 0.41 & 82 & @xmath83 + m5v & 99 & 0.69 & 66 & @xmath84 + m6v & 38 & 0.75 & 35 & @xmath63 + m7v & 21 & 0.84 & 27 & @xmath66 + m8v & 18 & 0.87 & 26 & @xmath85 + m9v & 10 & 0.91 & 24 & @xmath86 + + for a distance of 5 pc using the simulations described above , we have investigated a series of questions pertaining the successful ground - based detection of @xmath0 in the atmosphere of an earth twin around a nearby star . those questions include the optimal spectral resolution of the observations , the limitations imposed by the relative radial velocity of the star and earth , the limitations imposed by the depth of the telluric @xmath0 lines , and the number of transits we need to observe to achieve at least a 3@xmath87 detection in the presence of white and red noise . the answers to those questions are provided in figures 3 - 8 and tables 2 - 4 , and are also described in detail in the text below . the simulations were carried out for g - clef mounted on the 24 m gmt , for hros on the tmt assuming an instrument configuration similar to g - clef , and for three instruments mounted on the e - elt ; the near - infrared spectrograph simple and two hypothetical high - resolution spectrographs in the visual , a g - clef - like instrument and uves - like instrument . we used the uves etc to investigate the efficiency of observations with an uves - like instrument at different spectral resolutions . we carried out simulations for resolutions between 60,000 and 110,000 , rescaling the s / n of the contimuun of each spectrum to the aperture of the e - elt . figure 3 shows the result of those tests for the particular case of an earth twin orbiting around an m4v star at a distance of 5 pc . in this case , the optimal spectral resolution appears to be @xmath88 , which corresponds to the best trade - off between slit losses and the spectral smearing of the exo - earth s signal . the plot also shows , for comparison , the result of the same test when using an image slicer in this particular case we simulated image slicer # 3 described in dekker et al . when using the image slicer , which is only used at resolution @xmath89 , the efficiency of the observations improves by a factor of two with respect to the efficiency of the instrument at that same resolution when using only a slit . we note that this result represents the ideal case , i.e. we did not account for possible changes of the instrumental profile induced by an image slicer , which might hamper the data analysis . the tests also show that the use of the image slicer improves the overall efficiency of the observations , where only about 21 transits are necessary to achieve a 3@xmath87 detection of the exo - earth s atmosphere , while 25 transits are necessary to achieve the same precision level when just observing at a spectral resolution of @xmath88 without an image slicer . the result of these tests are instrument - specific and similar tests will have to be done for other instruments to determine the optimal resolution for this type of science once their final specifications are defined . however , we can argue that employing pre - slit optics , such as image slicers capable of concentrating a larger fraction of the stellar flux into the slit , can significantly reduce the number of transits needed for a successful detection , for all instruments . the detectability of an earth twin depends , among other things , on how many of its @xmath0 lines can be observed . our atmosphere produces the same absorption features ( telluric lines ) , at the same wavelength positions ( cf . figs . 1 and 2 ) , and in addition , the lines have some intrinsic width . therefore , the earth twin spectrum will be only detectable when its lines are doppler shifted with respect to the telluric lines by certain amounts . to investigate this effect we generated a set of simulations in which the spectrum of the earth twin was shifted with respect to the spectrum of earth by values of the relative velocity between the host star and earth between -150 km s@xmath34 and + 150 km s@xmath34 , and determined the fraction of lines that would appear blended in each case . the results of those simulations are illustrated in figure 4 for both the @xmath0 a - band and the @xmath0 band at 1268 nm . in the case of the @xmath0 a - band , we find that most lines will appear blended as long as the relative velocities are @xmath8 @xmath90 15 km s@xmath34 . line blending becomes significant again ( with about 50@xmath11 of the lines blended ) at relative velocities of about @xmath90 50 km s@xmath34 , coinciding with the average separation between individual lines in the @xmath0 a - band spectrum . the optimal relative velocity regimes ( with 10@xmath11 or less of the lines blended ) , appear to be -15 to -30 km s@xmath34 and 15 to 30 km s@xmath34 . in the case of the @xmath0 band at 1268 nm , which are narrower than the a - band lines , we find that relative radial velocities in the regimes around -85 to -65 km s@xmath34 , -10 to 10 km s@xmath34 , and 65 to 85 km s@xmath34 are also significantly blended . stars with any other relative velocity value can be observed with the guarantee that no more than about 20@xmath11 of the @xmath0 lines will be blended . [ f3 ] [ f4 ] a - band at 760 nm and the bottom panel to the @xmath0 band at 1268 nm . based on this plot , observations are best suited when the relative velocity of the star is between @xmath90(15 and 30 ) km s@xmath34 in the case of the a - band , and for velocities outside the ranges @xmath90(85 to 65 ) km s@xmath34 and -10 to 10 km s@xmath34 in the case of the @xmath0 band at 1268 nm . , title="fig : " ] [ f5 ] another effect affecting the detectability of an earth twin is noise introduced by telluric contamination . in other words , telluric lines block a large fraction of the host star s light , causing the s / n of its spectrum to drop significantly around those wavelengths . in the case of the @xmath0 band at 1268 nm , telluric contamination around 1268 nm causes the s / n ratio to drop to 0.3 with respect to the s / n ratio in the continuum . in the case of the @xmath0 a - band , telluric contamination around 760 nm fully blocks all incoming light at some wavelengths ( see middle panels in figures 1 and 2 ) . the best strategy to avoid this source of noise during the data analysis process is to skip wavelength regions where the s / n of the observed spectrum drops below a certain value . therefore , we define a cut - off parameter with respect to the s / n of the spectral continuum for which lines ( or portions of the spectrum ) to keep during the analysis . the optimum value of this cut - off parameter is a trade - off between keeping a large enough number of absorption features from the planet and avoiding noisy regions due to strong telluric contamination . adopting a non - optimal cut - off parameter will result in strongly suppressing the exo - earth s signal and might lead to a non - detection ( i.e. to a detection with confidence level less then 3@xmath87 ) . to determine the value of this cut - off parameter we analyzed different datasets by masking out different wavelength regions in which the telluric absorption lines dropped below a certain amount , e.g. a cut - off parameter of 0.5 corresponds to masking out all the wavelength regions for which the s / n drops by 50@xmath11 with respect to that of the continuum . the result of these tests is shown in figure 5 , where we represent the fap of the exo - earth s detection as a function of the cut - off parameter in s / n units ( i.e. a value of 1 for the cut - off parameter means that even the continuum is masked out ) . the figure indicates that the best value of the cut - off parameter is approximately 0.4 times the s / n of the continuum . the last question we addressed is how many transit observations are necessary to achieve at least a 3@xmath87 detection of an earth twin s atmosphere , and what will be the time span needed to collect those many transits . to estimate the number of transits needed , we focused first on the 760 nm @xmath0 a - band and assumed a transiting planetary system with an earth twin in its habitable zone , at a distance of 5 parsecs from earth and moving away from us with a relative velocity of 20 km s@xmath34 . we determined the duty cycle of the observations by assuming an instrument dead time of 15 sec after each exposure ( see section 2.2 ) . to calculate the number of transits , we divided the observing time by the transit duration of the planet as a function of stellar spectral type . the values for the orbital periods and transit durations were taken from kaltenegger & traub ( 2009 ) and are provided in table 1 . to estimate the overall time span of the observations , we factored in the orbital period of the planet and assumed that only every 9th transit can be observed , after accounting for object visibility from earth , transit occurrence , and the relative radial velocity between the target and earth ( see also section 4 ) . the results are summarised in tables 2 and 3 . table 2 is split into two halves ; the left - hand side lists the results for a spectrograph with a uves - like layout ( i.e. spectral resolving power of @xmath37 , image slicer # 3 , velocity span of 1.2 km s@xmath34 pixel@xmath34 ) installed on the e - elt , while the right - hand side shows the values determined for a spectrograph with a design similar to g - clef ( i.e. @xmath91 , velocity span of 0.75 km s@xmath34 pixel@xmath34 ) , also on the e - elt . g - clef s design seems to be significantly more efficient . in this table , and in table 3 , we also show the times needed for an earth twin orbiting a g2v star . in this case , because of the very low duty cycle and the very shallow transit depth , it would take over 30 transits , and therefore more than 30 years to detect @xmath0 with 3@xmath87 confidence . once we also factor in for the number of observable transits from the ground , these observations would take many decades . for m1v - m3v stars we estimate the time span of the observations to be @xmath6 13 - 19 years . for later type m - dwarfs , where the orbital periods of planets in their habitable zones are shorter , this number drops significantly to about 6 years for m4v stars and about 2 years for m9v stars . for a planet orbiting a m4v star , it would require a minimum of 14 transits to detect @xmath0 with 3@xmath87 confidence in its atmosphere . in the case of m9v stars , which are intrinsically fainter , it would take at least 50 transits . figure 6 shows the number of transits required for a 3@xmath87 detection with the two instrumental setups in table 2 , as a function of distance and for spectral types m3v through m7v . planets around an m3v star need more transits at very low distances due to a decrease in the duty cycle of the observations , i.e. the stars become too bright and the observing times need to be short to avoid saturation . therefore most of the observing time in those cases is spent on instrument readouts . the same applies for earlier type stars . the figure also shows how for stars located at distances larger than about 8 - 10 parsecs , the number of transits necessary to achieve a 3@xmath87 detection becomes too large . table 3 is similar to table 2 , but showing the results for two other observational configurations ; the left - hand side of the table shows the results of our simulations for g - clef mounted on the gmt . the right - hand side shows the result of the simulations for hros mounted on the tmt . for both spectrographs we assume a spectral resolution of @xmath10 and a velocity span of 0.75 km s@xmath34 pixel@xmath34 . the top panel of figure 7 shows the number of transits required for a 3@xmath87 detection with g - clef on the gmt as a function of distance and , as in figure 6 , for spectral types m3v through m7v . as expected , in this case , the number of transits necessary to achieve a detection is larger , by approximately a factor of two , than the number of transits necessary with an instrument like g - clef in the larger aperture e - elt ( also see bottom panel of figure 6 ) . the number of transits in figure 7 decrease by a factor of about 1.5 for the case of hros mounted on the tmt . finally , table 4 summarizes the results of our simulations for the near - ir spectrograph simple to be mounted on the e - elt . in this case the simulations focus on the 1268 nm @xmath0 band , while we still assume a transiting planetary system with an earth twin in its habitable zone , at a distance of 5 parsecs from earth and moving away from us with a relative velocity of 20 km s@xmath34 . we adopted a spectral resolution of @xmath10 , a velocity span of 1.2 km s@xmath34 pixel@xmath34 and a dead time between two subsequent exposures of 10 seconds . we only carried out simulations for stars with spectral types later than m3v , since the number of transits necessary for an m3v star was already over 200 . the results on that table reveal that using the @xmath0 band at 1268 nm to detect @xmath0 in the atmosphere of an earth twin will require a significantly larger amount of telescope time than observations of the @xmath0 a - band for all spectral types earlier than m7v . the bottom panel of figure 7 shows he number of transits required or a 3@xmath87 detection with simple as a function of distance and for spectral types m5v through m9v . in this case , the number of transits decreases steadily with increasing spectral type . [ f6 ] absorption band with @xmath47 confidence versus distance of the planetary system from earth . as instrument , we assume a high - resolution spectrograph mounted on the 39-m e - elt . the upper panel illustrates the results for a g - clef - like instrumental layout which provides a velocity span of 0.75 km s@xmath34 pixel@xmath34 and a spectral resolution of @xmath92 . the lower panel depicts the results for a uves - like spectrograph with velocity span of 1.2 km s@xmath34 pixel@xmath34 , an image slicer and a spectral resolution of @xmath93 . in these plots , we assume no red noise . to account for different red noise levels , the number of transits needs to be multiplied by the values provided in figure 8.,title="fig : " ] [ f78 ] confidence versus distance of the planetary system from earth . _ bottom panel : _ same as above , but for measurements of the 1268 nm oxygen absorption band with the simple - spectrograph employing a spectral resolving power of @xmath92 . in these plots we assume no red noise . to account for different red noise levels , the number of transits needs to be multiplied by the values provided in figure 8.,title="fig : " ] tables 2 , 3 , and 4 , as well as figures 6 and 7 show the results for the ideal case , i.e. without any contributions from correlated ( red ) noise . however , typical red noise levels are 20% and 50% of the white noise level for data recorded in the visual and near - infrared , respectively ( pont et al . 2006 , rogers et al . 2009 ) . the higher red noise level in the near - infrared wavelength regime is due telluric contamination and to a larger extent , due to detector noise . to make the simulations more realistic , we added different levels of red noise to the simulated data sets and analysed them as described before . figure 8 shows the difference in the amount of observing time required to achieve a 3@xmath87 detection as the red noise levels increase . for a red noise level of 20% of the white noise level , the observing time required to attain a @xmath47 detection rises by a factor of @xmath94 this means that for a red noise level of 20% of the white noise level , the observing times as well as the number of transits shown in tables 2 and 3 and in figure 6 and 7 ( top panel ) needs to be multiplied by this value . this effect becomes more severe for higher red noise levels . e.g. for correlated noise levels of 50% and 100% of the white noise level , the time factor increases by @xmath95 and @xmath77 , respectively . when comparing the results determined by accounting for typical red noise levels ( 20% of the white noise level in the visual , 50% of the white noise level at near - infrared wavelengths ) , we find that the lowest time span of observations is achieved with a g - clef - like spectrograph mounted on the e - elt for all m - dwarfs with spectral types m1v - m8v . observations in the @xmath0 band at 1268 nm are more efficient than in the visual only for m9v stars ( see figure 9 ) . [ f7a ] [ f9 ] we have carried out detailed feasibility studies to probe molecular oxygen in the atmosphere of an earth - analogue orbiting m - dwarfs with future instruments to be mounted on upcoming elts . in our simulations , we factored in the stellar spectra for different spectral types , the telluric spectrum of the earth , different instrumental settings of possible future instrumentation , refraction effects in the earth analog s atmosphere , and a variety of noise models , including the effects of correlated noise ( red noise ) in the data . when we compare our work to previous , similar studies , these are the most detailed simulations to date . [ f10 ] bins . . the bins of stars with velocities less than 15 and 10 km s@xmath34 , respectively , are shown as dashed and square shaded areas in the histogram . assuming the chubak et al . ( 2012 ) sample is representative of all nearby m - dwarfs , about 44.3@xmath11 of the stars will have absolute radial velocities smaller than 10 km s@xmath34 , and 56.8@xmath11 will have radial velocities smaller than 15 km s@xmath34.,title="fig : " ] a result of our investigation is that a g - clef - like instrument mounted on an e - elt is the best suited for the detection of @xmath0 in the atmosphere of an earth - like planet orbiting m1v - m8v stars . only for planets orbiting m9v stars , observations carried out in the near - infrared become more efficient than in the visual . this is a result of lower red noise levels at visible wavelengths , but the situation could change in the future if infrared red noise levels can be improved to match noise levels in the visible . in that case , near - infrared observations will be more efficient for planets around m7v stars and smaller . fig . 10 depicts the frequency of m1v to m9v stars in the solar neighborhood . m3 and m4 dwarfs are the most common sub - types , with more than 25 stars of each spectral type within a distance of 10 pc . when combining this distribution with the results of our simulations , we conclude that very close - by m4v stars will be the most likely candidates for the detection of oxygen in the atmosphere of an earth - like planet . in just 40 hours of observing time ( @xmath96 transits ) a 3@xmath87 detection of an object at 5 pc from earth could be attained with a g - clef - like instrument mounted on the 39 m e - elt . for this estimation , we assumed a red noise level of 20% of the white noise level . a typical orbital period of a planet located in the habitable zone around an m4-dwarf is 16 days . once the object visibility , orbital period and the relative radial velocity are taken into account , we estimate that it would nevertheless take a total time span of observations of @xmath57 years to accumulate sufficient data for a detection . this number soars up to 175 hours ( 74 transits ) and a time span of roughly 30 years for an object at a distance of 10 pc , which is the most probable distance to find an earth - analogue around such a star ( kaltenegger & traub 2009 ) . we compare our results to previous studies by other groups . snellen et al . ( 2013 ) carried out simulations with the goal to detect the @xmath0 a - band in the atmosphere of earth - analogues orbiting around m4v and m6v stars , with apparent magnitudes @xmath97 and @xmath98 , respectively . according to their results , it takes about 4 to 15 transits to attain a @xmath47 detection of @xmath0 . they assumed a line contrast ratio between stellar continuum and the deepest absorption lines of @xmath99 , which corresponds to @xmath100 in the case of a spectral resolution @xmath91 , which causes a line depth of 0.6 for the deepest lines with respect to the stellar continuum ( c.f . line depths in fig . 1 , top panel ) . furthermore , they assumed that the transit duration was 1.4 hours for planets orbiting those types of dwarfs . adopting the values for @xmath35 and the transit duration listed in table 1 , our studies reveal that it takes a total of about 42 and 60 transits for a @xmath47 detection of oxygen in the atmosphere of an earth - like planet orbiting an m4v and m6v stars , respectively . these results are for the ideal case , i.e. without red noise and a g - clef - like spectrograph mounted on the 39 m e - elt . if we include a typical 20% red noise contribution in these studies , the numbers rise to 60 and 84 transits ( i.e. 12 - 25 years ) , respectively for m4 and m6 dwarfs . our results are far less optimistic than those presented by snellen et al . ( 2013 ) . ideally , the incoming flux from the planet host star should find its way to the data analysis . to record high - resolution spectra of a star , it is necessary to carry out the observations through a narrow slit ( or fiber ) . this , however , means that only a fraction of the stellar light is used and in case of bad seeing conditions , it might happen that more than 60% of the incoming light is lost . to avoid such an unfavorable situation , pre - slit optics should be employed which enable to feed more than 90% of the photons of the star into the narrow slit . this can be achieved with image slicers or multiple fibers , although in this last case there is some concerns that transmission differences between fibers might affect the results . such potential problems will have to be evaluated before hand . since the designs of most of the high - resolution spectrographs to be mounted on the elts havent been fixed yet , we strongly recommend to include pre - slit optics to collect most of the stellar light and feed it into the spectrograph . based on our simulations in figure 3 , the usage of pre - slit optics could improve the required number of observations by a factor of two . in this estimation , however , we have not accounted for possible irregularities in the ip induced by an image slicer , which might represent a further noise source . at this point , we want to emphasize that our simulations represent a somewhat ideal case , i.e. stable instrument configurations and perfectly fitting models in the data analysis . our simulations need to be redone once the configuration details of the spectrographs are known . for the planning of the observations , we find that a very important parameter to take into account is the relative radial velocity of the star with respect to the earth . that relative radial velocity is the combination of the systemic radial velocity of the star ( @xmath101 ) and the instantaneous barycentric velocity @xmath102 . our atmosphere produces the same line pattern that we search for in the atmosphere of a remote planet . at some wavelengths , the telluric contamination almost absorbs all the incoming flux , leading to very low s / n ratios in those wavelength regions , which need to be masked out in the data analysis . it is therefore crucial to minimize the numbers of absorption line blends by only taking observation when the telluric spectrum and the spectrum of the planet are doppler shifted by a certain amount . for observations in the visual , we find that the optimum velocity range of the planet host star with respect to the earth is limited to two small velocity windows , ranging from @xmath90(15 to 30 ) km s@xmath34 . in figure 11 we show the distribution of absolute values of the barycentric radial velocities for a sample of 176 m - dwarfs reported by chubak et al . ( 2012 ) . assuming that the distribution of radial velocities in that sample is representative of all nearby m - dwarfs , we can conclude that the majority of them have barycentric radial velocities between -15 and 15 km s@xmath34 . therefore , what is the amount of time of optimum visibility of those objects , i.e. when they show relative radial velocities of @xmath90(15 to 30 ) km s@xmath34 ? to answer this question , we need to account for the barycentric velocity of the earth , which causes a periodical radial velocity shift with a semi - amplitude @xmath103 of approximately 29.8 km s@xmath34 for an object located close to the ecliptic plane . however , the semi - amplitude of the barycentric velocity is a function of the ecliptic latitude @xmath104 ( i.e. the angular distance of the object from the orbital plane ) : @xmath105 for the vast majority of stars , the semi - amplitude of the barycentric velocity @xmath103 is between 20 and 29.8 km s@xmath34 . figure 12 shows the amount of time of optimum visibility of an object in units of the yearly observing time available at an ideal observing site . for this estimation , we assumed that the object can be observed at an airmass @xmath106 between 2 hours and 8 hours ( at opposition with earth ) . furthermore , for simplicity we assumed a circular orbit of earth and factored in the instantaneous value of the barycentric velocity of the object . figure 12 depicts that for semi - amplitudes of the barycentric velocity of @xmath107 km s@xmath34 , the atmospheric lines from the planet would be detectable a fraction of 0.2 to 0.4 of the time for the majority of the m - dwarfs , which have instrinsic radial velocities between -15 and 15 km s@xmath34 . to understand possible error sources in the data analysis , we briefly review the main steps : we remove both the telluric spectrum as well as the stellar spectrum from the data , and finally attempt to retrieve the planetary signature from the residual spectrum via cross - correlation . the removal of both spectra is an iterative process , which involves the determination of the ip and the convolution of the stellar + telluric model spectra with the ip ( the method is outlined in rodler et al . while the telluric spectrum can be well reproduced with theoretical model spectra , the modeling of m - dwarf spectra is still in its infancy . in principle , if the instrument is stable during one night , it would be sufficient to take spectra of the star for some hours before and after the transit , subtract the telluric lines , then co - add these spectra to form one high - s / n spectrum for that night , which is then subtracted from the observed spectrum taken during the transit ( cf . procedure used by rodler et al . however , this procedure would drastically boost the need of observing time . detailed studies on the modeling of m - dwarf spectra are clearly required . additionally , it is crucial to discard strong telluric features in the data analysis . we want to make future observers aware that regions which show a s / n ratio less than about 0.4 times the s / n ratio in the continuum of the spectrum should be masked out . adopting a wrong cut - off parameter causes the the weak planetary signal to be suppressed . another possible error noise constitutes the model spectrum of the planet atmosphere , which is adopted in the cross - correlation . we are attempting to detect a very weak planetary signal of the planetary atmosphere ; the lack of some absorption features in the theoretical model would suppress the planetary signal in the cross - correlation function . to avoid this situation , a large numbers of atmospheric models with different mixing ratios and molecular configurations need to be probed . we conclude that successful measurements of @xmath0 in earth - like atmospheres with future ground - based instrumentation will be most likely limited to host stars of spectral types later than m3v , which are located at distances less than 8 pc . our simulations demonstrate , that a @xmath47 discovery of @xmath0 in the atmosphere of such a planet would require even tenths of transits and several years of observing time at elts . we are grateful to the constructive comments of the anonymous referee , who helped us to improve the manuscript . fr acknowledges financial support from the spanish ministry of economy and competitiveness ( mineco ) and the `` fondo europeo de desarrollo regional '' ( feder ) through grant aya2012 - 39612-c03 - 01 . this research has made use of nasa s astrophysics data system bibliographic services . , d. , simard , l. , & silva , d. 2008 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7014 , society of photo - optical instrumentation engineers ( spie ) conference series , h. , dodorico , s. , kaufer , a. , delabre , b. , & kotzlowski , h. 2000 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4008 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . m. iye & a. f. moorwood , 534545 , h. , nissen , p. e. , kaufer , a. , primas , f. , dodorico , s. , & hanuschik , r. w. 2003 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4842 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . e. atad - ettedgui & s. dodorico , 139150 , l. , oliva , e. , maiolino , r. , gustafsson , b. , piskunov , n. , kochucov , o. , vanzi , l. , minniti , d. , zoccali , m. , hatzes , a. , & guenther , e. 2010 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7735 , society of photo - optical instrumentation engineers ( spie ) conference series , l. , cristiani , s. , garca lpez , r. , haehnelt , m. , mayor , m. , liske , j. , manescau , a. , avila , g. , dekker , h. , iwert , o. , delabre , b. , lo curto , g. , dodorico , v. , molaro , p. , viel , m. , vanzella , e. , bonifacio , p. , di marcantonio , p. , santin , p. , comari , m. , cirami , r. , coretti , i. , zerbi , f. m. , span , p. , riva , m. , rebolo , r. , israelian , g. , herrero , a. , zapatero osorio , m. r. , tenegi , f. , carswell , b. , becker , g. , udry , s. , pepe , f. , lovis , c. , naef , d. , dessauges , m. , & mgevand , d. 2010 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 7735 , society of photo - optical instrumentation engineers ( spie ) conference series , l. s. , gordon , i. e. , barbe , a. , benner , d. c. , bernath , p. f. , birk , m. , boudon , v. , brown , l. r. , campargue , a. , champion , j .- p . , chance , k. , coudert , l. h. , dana , v. , devi , v. m. , fally , s. , flaud , j .- m . , gamache , r. r. , goldman , a. , jacquemart , d. , kleiner , i. , lacome , n. , lafferty , w. j. , mandin , j .- y . , massie , s. t. , mikhailenko , s. n. , miller , c. e. , moazzen - ahmadi , n. , naumenko , o. v. , nikitin , a. v. , orphal , j. , perevalov , v. i. , perrin , a. , predoi - cross , a. , rinsland , c. p. , rotger , m. , v simekov , m. , smith , m. a. h. , sung , k. , tashkun , s. a. , tennyson , j. , toth , r. a. , vandaele , a. c. , & vander auwera , j. 2009 , , 110 , 533 , a. , frebel , a. , furesz , g. , hertz , e. , norton , t. , bean , j. , bergner , h. , crane , j. , evans , j. , evans , i. , gauron , t. , jordn , a. , park , s. , uomoto , a. , barnes , s. , davis , w. , eisenhower , m. , epps , h. , guzman , d. , mccracken , k. , ordway , m. , plummer , d. , podgorski , w. , & weaver , d. 2012 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 8446 , society of photo - optical instrumentation engineers ( spie ) conference series

we study here the impact of priorities on conflict resolution in inconsistent relational databases . we extend the framework of , which is based on the notions of repair and consistent query answer . we propose a set of postulates that an extended framework should satisfy and consider two instantiations of the framework : ( locally preferred ) - repairs and ( globally preferred )-repairs . we study the relationships between them and the impact each notion of repair has on the computational complexity of repair checking and consistent query answers .

cs0506063

the main purpose of integrity constraints is to express semantic properties of the data stored in the database . usually , it is the database management system that is responsible for maintaining the integrity of the database . however , in many recent applications the integrity enforcement becomes a problematic issue . for example in the data integration setting , even when the data contained by a data source satisfies the integrity constrains , a different data source may contribute conflicting information . at the same time data sources may be autonomous and it may be impossible to modify their contents in order to remove the conflicts . integrity constraints may also fail to be enforced because of efficiency considerations . finally , in the case of long running operations , integrity violations may be only temporary and will be eliminated by further operations . typically , the user formulates a query with the assumption that the database is consistent ( i.e. satisfies the integrity constraints ) . a simple evaluation of the query over an inconsistent database may return incorrect answers . to address this problem arenas , bertossi , and chomicki @xcite proposed the framework of _ consistent query answers_. they introduced the notion of a _ repair _ : a consistent database that is minimally different from the original one . a _ consistent answer _ to a query is an answer _ true _ in _ every _ repair . the framework of @xcite is used as a foundation for most of the work in the area of querying inconsistent databases @xcite . [ ex : cqa ] consider a database consisting of two tables @xmath1 and @xmath2 whose instance @xmath3 can be found in table [ tab : tab4 ] . name & dept + alice & a + alice & b + dept & name & t + a & mary & 2 + b & bob & 1 + b & mary & 3 + assume that we have two functional dependencies @xmath4 and @xmath5 . this database contains two conflicts : 1 ) in relation @xmath1 between the tuples @xmath6 and @xmath7 ; 2 ) in relation @xmath2 between the tuples @xmath8 and @xmath9 ( note that one person can be the manager of more than one department ) . each of those conflicts can be resolved in two different ways by assuming that one tuple is correct and removing the other . this leads to four different repairs : @xmath10 for example , the repair @xmath11 is obtained by assuming that @xmath12 works in department @xmath13 and the manager of department @xmath14 is @xmath15 . since in every repair @xmath16 is the manager of the department @xmath13 , we can infer that _ true _ is the consistent answer to the query @xmath17 however it is not certain that @xmath12 works in a department managed by @xmath16 , i.e. _ true _ is not the consistent answer to the following query @xmath18 this is because of the repair @xmath19 , where @xmath20 is false . as it is shown in the previous example , each conflict can be resolved in two different ways . the framework of @xcite does not provide any means to favor one way over another . however , in many cases some additional information which can be used to provide a resolution of some conflicts is available . for example : * in e - commerce applications , data are accompanied with the timestamp of creation / last modification the conflicts can be resolved by removing from consideration old , outdated tuples . * in data integration scenarios , it is often possible to provide a ( partial ) order on the sources , capturing the reliability of contributed information the most reliable data can be used to resolve conflicts . * statistics can be used to resolve conflicts created by misspellings . suppose that the column @xmath21 of the table @xmath2 contains for each tuple its creation timestamp ( lower values correspond to older tuples ) . we can use this information to express the preference that if some tuples of @xmath2 are conflicting , the older should be removed from consideration ( but not removed from the database ) . since the tuple @xmath9 is older than @xmath8 , we consider only the repairs containing the latter one : @xmath22 and @xmath23 . in such a case we can also infer that it is certain that @xmath12 works in the department managed by @xmath16 , i.e. _ true _ is the _ preferred _ consistent answer to the query @xmath20 . in this paper we extend the framework of consistent query answers with an additional input consisting of preference information @xmath24 . we use @xmath24 to define the set of _ preferred _ repairs @xmath25 . when we compute consistent answers , instead of considering the set of all repairs @xmath26 , we use the set of preferred repairs . we assume that there exists a ( possibly partial ) operation of extending @xmath24 with some additional preference information and we write @xmath27 when @xmath28 is an _ extension _ of @xmath24 . we consider @xmath24 to be _ maximal _ when it can not be extended further . the main objective of our research is to develop a framework of preferred repairs that fulfills the following postulates : 1 . * non - emptiness * @xmath29 2 . * non - discrimination * : if no preference information is given , then no repair is removed from consideration @xmath30 3 . * monotonicity * : extending preferences can only narrow the set of preferred repairs @xmath31 4 . * categoricity * : given maximal preference information we obtain exactly one repair @xmath32 we note here that the postulates [ eq : p1 ] and [ eq : p3 ] together imply an important property of * conservativeness * : preferred repairs are a subset of the standard repairs . another important goal of our research is to determine the computational implications of introducing preferences . for this purpose we study here two fundamental decision problems in inconsistent databases @xcite : ( i ) _ repair checking _ finding if a given database is a preferred repair ; ( ii ) _ computing consistent answers _ finding if an answer to a query is present in every preferred repair . the main contributions of this paper are : * a general and intuitive framework for incorporating preferences into inconsistency handling based on the notion of priority . * a study of the semantic and computational properties of two instantiations of the framework : ( locally preferred ) - repairs and ( globally preferred ) @xmath0-repairs . in this paper , we work with databases over a schema consisting of only one relation @xmath33 with attributes from @xmath34 . we use @xmath35 to denote elements of @xmath34 and @xmath36 to denote subsets of @xmath34 . we consider two disjoint domains : uninterpreted names @xmath37 and natural numbers @xmath38 . every attribute in @xmath34 is typed . we assume that constants with different names are different and that symbols @xmath39 , @xmath40 , @xmath41 , @xmath42 have the natural interpretation over @xmath38 . the instances of @xmath33 , denoted by @xmath43 , can be seen as finite , first - order structures , that share the domains @xmath37 and @xmath38 . for any tuple @xmath44 from @xmath45 by @xmath46 we denote the value associated with the attribute @xmath13 . in this paper we consider first - order queries over the alphabet consisting of @xmath33 and binary relation symbols @xmath39 , @xmath40 , @xmath41 , and @xmath42 . the limitation to only one relation is made only for the sake of clarity and along the lines of @xcite the framework can be easily extended to handle databases with multiple relations . the class of integrity constraints we study consists of functional dependencies . we use @xmath47 to denote the following constraint : @xmath48 we use this formula to identify tuples creating conflicts . given a set of functional dependencies @xmath49 , two tuples @xmath50 are _ conflicting _ w.r.t @xmath49 , denoted @xmath51 , if and only if there exists a functional dependency @xmath52 such that @xmath53 for all @xmath54 and @xmath55 for some @xmath56 . a database @xmath45 is _ inconsistent _ with a set of constraints @xmath49 if and only if @xmath45 contains some conflicting tuples . otherwise , the database is _ consistent_. in the general framework when repairing a database we consider two operations : adding or removing a tuple . because in the presence of functional dependencies adding new tuples can not remove conflicts , we only consider repairs obtained by deleting tuples from the original instance . [ def : repair ] given a database @xmath45 and a set of integrity constraints @xmath49 , a database @xmath57 is a _ repair _ of @xmath45 w.r.t . @xmath49 if @xmath57 is a maximal subset of @xmath45 consistent with @xmath49 . we denote by @xmath58 the set of all repairs of @xmath45 w.r.t @xmath49 . a repair can be viewed as the result of a process of cleaning the input relation . note that since every conflict can be resolved in two different ways and conflict are often independent , there may be an exponential number of repairs . also , the set of repairs of a consistent relation @xmath45 contains only @xmath45 . @xcite a _ conflict graph _ @xmath59 is a graph whose set of vertices is equal to @xmath45 and two tuples @xmath50 are adjacent only if they are conflicting ( i.e. @xmath51 ) . recall that a maximal independent set of a graph @xmath60 is a maximal set of vertices that contains no edge from @xmath60 . by @xmath61 we denote the set of all maximal independent sets of @xmath60 . the following observation explains why the conflict graph is considered a _ compact representation of all repairs_. for any database @xmath45 and any set of functional dependencies @xmath49 we have that @xmath62 for the clarity of presentation we assume that from now on we work with a fixed database instance @xmath45 and a fixed set of functional dependencies @xmath49 . to represent the preference information , we use ( possibly partial ) orientations of the conflict graph . it allows us to express preferences at the level of single conflicts . a binary relation @xmath63 is a _ priority _ if : 1 . @xmath64 is asymmetric , i.e. @xmath65,\ ] ] 2 . @xmath64 is defined only on conflicting tuples , i.e. @xmath66 if @xmath67 we say that the pair @xmath68 is _ prioritized _ and that @xmath69 _ dominates _ over @xmath70 . a priority @xmath64 is _ total _ if every pair of conflicting tuples is prioritized by @xmath64 . a priority @xmath64 is _ acyclic _ if there does not exist @xmath71 such that @xmath72 , where @xmath73 is the transitive closure of @xmath64 . the first condition of priority demands the preference information to be unambiguous for a single conflict . the second condition ensures that we are given only the relevant preference information . if the second condition is not fulfilled , then it can be easily enforced by intersecting @xmath64 with @xmath74 . this form of preference information allows us to easily define the the preference extension : we orient some conflicting edges that were not oriented before . [ def : extension ] a priority @xmath75 is an _ extension _ of a priority @xmath64 if @xmath75 agrees with @xmath64 where @xmath64 is defined ( i.e. @xmath76 ) . note that @xmath64 can not be extended further only if @xmath64 is total . also an extension @xmath75 of a priority @xmath64 is also a priority and therefore @xmath75 is antisymmetric and defined only on pairs of conflicting tuples . now we present two methods of using a priority to restrict the set of all repairs of a given relation . the first one , - repairs , uses the priority to restrict the ways of constructing a repair ( cleaning the database ) . the process consists of multiple iterative steps and in each of them only a limited number of conflicts is considered . the use of the priority has a local character because the subset of priority used in one step is not used in any further steps . the second method , @xmath0-repairs , uses the priority in a global fashion by selecting most preferred repairs according to an order induced by the priority . recall a general nondeterministic procedure for constructing a maximal independent set of a graph : as long as the graph is not empty , we _ choose _ a vertex , add it to the constructed set , and remove the vertex and all its neighbors from the graph . depending on the choices of vertices we make , we can construct any maximal independent set of the input graph . now , let s look at this procedure from the point of constructing a repair . each choice of a vertex corresponds to taking a single repair action : keeping the corresponding tuple in the relation and removing all tuples conflicting with it . since the choice of the tuple to keep is unconstrained , every conflict can be resolved in several different ways . we use the priority to restrict the possible ways of choosing the tuple that will be kept and whose conflicts will be resolved . the chosen tuple is among those that are not dominated at the given step of the repairing process . we use the _ winnow operator _ @xcite to formally describe the set of tuples that we choose from : @xmath77 algorithm [ alg : alg1 ] implements the construction of preferred repairs . an _ - repair _ ( or a _ locally preferred _ repair ) is any instance @xmath57 we can obtain with this algorithm . we denote the set of all - repairs of @xmath45 w.r.t . @xmath49 and @xmath64 by @xmath78 . [ step : stop_cond ] [ step : choice ] note that an - repair can be characterized by the sequence of choices made in the step [ step : choice ] in algorithm [ alg : alg1 ] ( however there can be more than one such sequence ) . this observation allows us to state an alternative definition of an - repair . [ prop : proc_computation_path ] given a priority @xmath64 , a set of tuples @xmath79 is an - repair , if and only if there exists an ordering @xmath80 of @xmath79 such that for every @xmath81 the following set is non - empty @xmath82 and @xmath83 . the next construction uses the priority directly to compare two repairs . intuitively , one repair is better than another if all the differences between them are justified by the priority . formally , we define @xmath0-repairs in the following way . [ def : decl_repair ] given a priority @xmath64 and two repairs @xmath84 , we say that @xmath85 is _ preferred _ over @xmath86 , and write @xmath87 , if @xmath88 a repair is a @xmath0-repair ( or a _ globally _ preferred repair ) if it is a @xmath89-maximal repair . by @xmath90 we denote the set of all @xmath0-repairs . this particular `` lifting '' of a preference on objects to a preference on sets of objects can be found in other contexts . for example , a similar definition is used for a preference among different models of a logic program @xcite , or for a preference among different worlds @xcite . in this paper , we use a generalized notion of consistent query answers . instead of taking the set of all repairs , as in @xcite , we consider families of preferred repairs . we only study closed first - order logic queries . we can easily generalize our approach to open queries along the lines of @xcite . for a given query @xmath91 we say that _ true _ is an answer to @xmath91 in @xmath45 , if @xmath92 in the standard model - theoretic sense . given a closed query @xmath91 and a family of repairs @xmath93 , _ true _ is the _ @xmath94-consistent query answer _ to a query @xmath91 if for every repair @xmath95 we have @xmath96 . note that we obtain the original notion of consistent query answer @xcite if we take for @xmath94 the whole set of repairs @xmath58 . in this paper , we study the cases when we take for @xmath94 either the set of - repairs or the set of @xmath0-repairs . this gives us two notions : 1 . _ - preferred consistent query answer _ if @xmath97 , 2 . _ @xmath0-preferred consistent query answer _ if @xmath98 . we write @xmath99 ( @xmath100 ) to denote that _ true _ is the - preferred ( resp . @xmath0-preferred ) consistent answer to @xmath91 ( in @xmath45 w.r.t . @xmath49 and @xmath64 ) . before discussing specific properties of preferred repairs , we present reasons for removing cyclic priorities from consideration . [ ex : ex1 ] assume a database schema @xmath101 and a set of functional dependencies @xmath102 . consider the following database @xmath103 and a total cyclic priority @xmath104 . the set of all repairs is @xmath105 as we can easily find @xmath78 is empty . it is also easy to see that @xmath87 and @xmath106 and thus @xmath107 . this violates the postulates [ eq : p1 ] and [ eq : p5 ] . intuitively , a cycle in the conflict graph represents a mutually dependent group of conflicts ( a solution of one conflict may restrict the ways of solving other conflicts ) . our intention is to break the cycle by choosing a @xmath64-maximal element . if @xmath64 is cyclic , then such element does not exist , which makes the construction of a preferred repair impossible . we find this kind of preference information ( cyclic priority ) to be incoherent and we exclude it form our considerations . when we restrict our considerations only to acyclic priorities , the relation @xmath89 has interesting order properties . [ prop : decl_order ] if @xmath64 is an acyclic priority and the binary relation @xmath89 on @xmath108 is defined in terms of @xmath64 as in definition [ def : decl_repair ] , then 1 . @xmath89 is reflexive , 2 . @xmath89 is anti - symmetric , 3 . @xmath89 is transitive , provided that @xmath64 is transitive . before proving the main thesis we will introduce one definition and show its two properties given two sets @xmath109 and a priority @xmath64 , an _ @xmath110-alternating @xmath64-chain _ is a ( possibly infinite ) sequence @xmath111 such that : * every element with even index belongs to @xmath13 @xmath112 * every element with odd index belongs to @xmath14 @xmath113 * @xmath64 holds between every two consecutive elements , i.e. @xmath114 we say that an @xmath110-alternating @xmath64-chain is _ maximal _ if it s not a proper prefix of some @xmath110-alternating @xmath64-chain is a proper prefix of a sequence @xmath115 if and only if @xmath116 and @xmath117 for every @xmath118 . note that @xmath115 can be infinite ( @xmath119 ) , but an infinite sequence can not have a proper prefix . ] . when @xmath64 will be know from the context instead of saying that @xmath120 is an @xmath110-alternating @xmath64-chain we will simply say that @xmath120 is an @xmath110-chain . [ prop : inf_seq ] for any acyclic priority @xmath64 and any two sets @xmath109 every @xmath110-chain is finite . suppose there exists such an infinite @xmath110-chain @xmath120 . because @xmath45 is finite , @xmath120 contains a recurrent element @xmath70 . thus @xmath121 this gives us a contradiction with @xmath64 being acyclic . [ prop : altll ] for any acyclic priority @xmath64 , and any two sets @xmath122 such that @xmath123 ( where @xmath89 is defined in terms of @xmath64 ) , any maximal @xmath124-chain is of even length ( it ends with an element from @xmath125 ) . by previous proposition we have that any @xmath124-chain is finite . assume now that , there exists a maximal @xmath124-chain of odd length ( i.e. ending with an element from @xmath126 ) : @xmath127 since @xmath123 , there exists @xmath128 such that @xmath129 . thus is a prefix of the following @xmath124-chain : @xmath130 this contradicts the maximality of . we also state a trivial fact [ fact : altstart ] for any acyclic priority @xmath64 , any two sets @xmath131 such that @xmath123 , and any @xmath132 there exists an @xmath124-chain that starts with @xmath70 . now , we show the order properties of @xmath89 : 1 . @xmath89 is reflexive . + because universal quantification over empty set is true , then trivially @xmath133 for any set @xmath134 . @xmath89 is asymmetric . + take two different sets @xmath122 such that @xmath123 and @xmath135 , i.e. : @xmath136 w.l.o.g we can assume that @xmath137 . take any @xmath138 . by we are able to find @xmath139 such that @xmath140 . now , by we are able to find @xmath141 such that @xmath142 . this way we can construct an infinite @xmath143-chain . this contradicts proposition [ prop : inf_seq ] . if @xmath64 is transitive , then @xmath89 is transitive . + assume @xmath64 is transitive and take three different sets @xmath144 such that @xmath123 and @xmath145 ( the case when two sets are equal is trivial ) . note that : @xmath146 now we take any @xmath147 and consider two cases depending if @xmath148 or not . + suppose @xmath149 . let @xmath150 be a maximal @xmath151-chain where @xmath152 . ( the existence of such a chain is by proposition [ prop : altll ] and fact [ fact : altstart ] ) . if there exists an element @xmath153 of this chain that belongs to @xmath154 then by transitivity of @xmath64 we have @xmath155 ( which end this path of the proof ) . suppose that none of the elements of the @xmath151-chain belongs to @xmath154 , then in particular @xmath156 belongs to @xmath157 . by there exists @xmath158 such that @xmath159 . moreover @xmath160 or otherwise we get a contradiction of the maximality of the @xmath151-chain . by transitivity of @xmath64 we get @xmath161 and obviously @xmath162 ; + similarly we deal with the case when @xmath163 . take @xmath164 to be a maximal @xmath165-chain , where @xmath166 . if there exists an element @xmath153 of this sequence that belongs to @xmath154 , then by transitivity of @xmath64 we have @xmath167 ( which end this path of the proof ) . suppose that none of the elements of the @xmath124-chain belongs to @xmath154 , then in particular @xmath69 belongs to @xmath168 . by there exists @xmath169 such that @xmath170 . moreover @xmath171 or otherwise we get a contradiction of the maximality of the @xmath124-chain . finally , by transitivity of @xmath64 we get @xmath172 and obviously @xmath173 . this ends the proof . the following example shows that @xmath89 may not be transitive if the underlying priority is not transitive . consider a database @xmath174 over the schema @xmath101 with one functional dependency @xmath175 and with priority @xmath176 . there are three repairs of @xmath45 : @xmath177 the corresponding conflict graph is presented on figure [ fig : fig6 ] . ( 3.4,1.6 ) ( 0.2,.5 ) ( 1.8,.5 ) ( 3.4,.5 ) we note that @xmath178 and @xmath179 but @xmath180 . before we prove the fulfillment of the postulates [ eq : p1][eq : p5 ] we state an important property of the two instantiations of preferred repairs : constructing a repair from the locally best tuples by the notion of - repairs conforms with the global notion of preference ( @xmath0-repairs ) . [ thm : proc_subsumes_decl ] if @xmath64 is an acyclic priority , then @xmath181 induction over the size of @xmath45 . trivial for @xmath182 . assume the hypothesis holds for any proper subset of @xmath45 and there exists @xmath183 such that @xmath123 for some @xmath184 . by proposition [ prop : proc_computation_path ] @xmath185 is non - empty . take then any @xmath186 . @xmath148 or otherwise we receive a contradiction @xmath123 . note that @xmath187 is a repair of @xmath188 and @xmath189 even a - repair of @xmath188 . moreover @xmath190 in terms of the database @xmath188 . thus @xmath191 is not @xmath0-repair of @xmath192 , which is a contradiction of the inductive hypothesis . in the following example we observe that the reverse containment does not hold for an arbitrary acyclic priority , i.e. the construction of - repairs by choosing only the best elements locally ( as in - repairs ) may miss a @xmath0-repair . [ ex : proc_diff_decl ] consider a database @xmath193 over the schema @xmath194 with a set of functional dependencies @xmath195 and a acyclic priority @xmath196 the set of repairs is @xmath197 . as we can easily find @xmath198 . because each of the @xmath199 and @xmath200 is dominated , the @xmath0-repair @xmath201 is not an - repair , and thus @xmath202 . later on we present sufficient conditions under which both instantiations of preferred repairs are equivalent ( theorem [ thm : proc_equals_decl ] ) . we recall that extending priority consists of prioritizing conflicts not prioritized before and a priority that can not be extended further ( i.e. is maximal ) is a total priority . both classes of referred repairs that we consider satisfy the postulates [ eq : p1 ] [ eq : p5 ] : [ thm : proc_postulates ] for every relation instance @xmath45 , set of functional dependencies @xmath49 , and acyclic priority @xmath64 , @xmath78 satisfies [ eq : p1][eq : p5 ] . we receive [ eq : p1 ] from the fact that if @xmath64 is acyclic then @xmath203 is non - empty if and only if @xmath79 is non - empty . [ eq : p3 ] is implied by the fact that @xmath204 is an identity function what makes @xmath205 a generic procedure for constructing all maximal independent sets of @xmath206 . to prove [ eq : p4 ] assume that @xmath207 are acyclic priorities such that @xmath208 . take then any @xmath209 and let @xmath210 be any ordering of @xmath79 from proposition [ prop : proc_computation_path ] . note that since for any set @xmath13 we have @xmath211 then @xmath210 also fulfills conditions of proposition [ prop : proc_computation_path ] in terms of @xmath75 . [ eq : p5 ] is a consequence of [ eq : p1 ] for @xmath205 , theorem [ thm : proc_subsumes_decl ] , and [ eq : p5 ] for @xmath212 . [ thm : decl_postulates ] for every relation instance @xmath45 , set of functional dependencies @xmath49 , and acyclic priority @xmath64 , @xmath90 satisfies [ eq : p1][eq : p5 ] . we get [ eq : p1 ] from the definition . with an empty priority we can not justify @xmath135 for any two different repairs @xmath79 and @xmath213 , what implies [ eq : p3 ] . to show [ eq : p4 ] assume that @xmath207 are acyclic priorities such that @xmath208 , @xmath214 , and suppose there exists @xmath215 such that @xmath213 is preferred over @xmath79 in terms of @xmath75 . but since @xmath216 this implies that @xmath213 is also preferred over @xmath79 in terms of @xmath64 . this is a contradiction . in order to prove [ eq : p5 ] assume there exist two different repairs @xmath79 and @xmath213 in @xmath90 . @xmath217 implies that there exists an element @xmath218 such that for any conflicting with @xmath70 tuple @xmath69 from @xmath125 we have @xmath219 . since @xmath64 is total for any such @xmath69 we have @xmath220 . take all such tuples @xmath221 and by @xmath222 denote any repair that contains the following elements @xmath223 such a repair exists because this set contains no conflicting tuples . obviously @xmath224 and at the same time @xmath225 . this contradicts that @xmath226 . as we showed in example [ ex : proc_diff_decl ] @xmath205 does nt have to be equal to @xmath212 . it suffices , however , to remove from consideration priorities with cyclic extensions to obtain the equivalence of the two notions of preferred repair : [ thm : proc_equals_decl ] if @xmath64 is a priority having only acyclic extensions , then @xmath227 we need to show @xmath228 . take any @xmath214 and construct @xmath75 a total extension of @xmath64 by prioritizing ( un - prioritized by @xmath64 ) conflicts in favor for @xmath79 , i.e. @xmath75 is any total priority such that for any @xmath229 and any @xmath69 if @xmath230 and @xmath219 then @xmath231 . since @xmath64 has only acyclic extensions @xmath75 is acyclic . it should be clear from the construction that @xmath232 . by [ eq : p1 ] , [ eq : p3 ] , [ eq : p5 ] and theorem [ thm : proc_subsumes_decl ] this implies that @xmath233 . this by [ eq : p4 ] gives us that @xmath234 . the following example shows , however , that the requirement of no cyclic extensions is not necessary for the equality above to hold . consider schema @xmath194 together with a set of functional dependencies @xmath195 . suppose we have a database : @xmath235 with a priority @xmath236 . the conflict graph is presented on figure [ fig : fig1 ] . ( 2,1.3 ) ( .4,1.1 ) ( 1.6,1.1 ) ( .4,.1 ) ( 1.6,.1 ) @xmath64 has a cyclic extension @xmath237 . at the same time we study two fundamental problems of handling inconsistencies with priorities : ( i ) _ repair checking _ determining if a database is a preferred repair of a given database ; ( ii ) _ consistent query answers _ checking if _ true _ is an answer to a given query in every preferred repair . we use the notion of _ data complexity _ @xcite which captures the complexity of a problem as a function of the number of tuples in the database . the database schema , the integrity constraints , and the query are assumed to be fixed . recall algorithm [ alg : alg1 ] and note that because the consecutive choices made in the step [ step : choice ] consist of mutually non - conflicting tuples , the state of the computation is independent of the order of the choices . given a repair @xmath57 , we can `` simulate '' its construction by restricting the choices in the step [ step : choice ] to @xmath239 . the simulation succeeds if and only if @xmath57 is an - repair . [ thm : decl_total_tractable ] given a fixed set of functional dependencies @xmath49 , the set @xmath240 is in ptime . it is shown in @xcite that computing consistent answers to conjunctive queries is co - np - complete , but if we consider only ground quantifier - free queries , the problem is in ptime . on the other hand , computing - preferred consistent answers turns out to be an intractable problem even if we consider very simple , single - atom queries . [ thm : proc_answers_intractable ] there exists a set of four functional dependencies @xmath49 and a quantifier - free ground query @xmath91 ( consisting of one atom only ) such that the set @xmath241 is co - np - complete . it s easy to construct a nondeterministic turing machine for @xmath242 following informal description presented here : the machine uses nondeterministic transitions to compute all - preferred repairs of @xmath45 and for each one checks the answer to @xmath91 . note that @xmath243 this allows us to state that the constructed machine decides the complement of @xmath242 . now , consider the schema @xmath244 with the set of functional dependencies @xmath245 and a ground query @xmath246 , where the value of @xmath247 can be found in table [ tab : tab2 ] . we show here a polynomial reduction of the complement of @xmath248 to @xmath249 , i.e. for any boolean formula @xmath91 in @xmath250 we construct a pair @xmath251 of a polynomial size in the size of @xmath91 and such that @xmath252 take then any formula @xmath91 in @xmath250 and let @xmath253 be the number of variables used in @xmath91 and @xmath254 the number of conjuncts of @xmath91 . for simplicity we assume that : * used variables have consecutive indexes @xmath80 , * @xmath255 * each conjunct consists of exactly three literals @xmath256 for @xmath257 . we define two auxiliary functions @xmath258 and @xmath259 on literals in the following fashion : @xmath260 the constructed database contains the following elements : @xmath261 whose exact values can be found in table [ tab : tab2 ] . @xmath262 the priority relation @xmath263 is the unique minimal binary relation on @xmath264 satisfying the following conditions : @xmath265 note that this priority relation is acyclic . also note that construction of @xmath251 can be implemented in time polynomial in the size of the of the input formula @xmath91 . on figure [ fig : fig3 ] we can find a conflict graph of an instance received from reduction of a formula @xmath266 . ( 15,-6 ) ( 1,-1 ) ( 2,-1 ) ( 3,-1 ) ( 4,-1 ) ( 5,-1 ) ( 6,-1 ) ( 7,-1 ) ( 8,-1 ) ( 9,-1 ) ( 10,-1 ) ( 11,-1 ) ( 12,-1 ) ( 13,-1 ) ( 14,-1 ) ( 4,-3 ) ( 8,-3 ) ( 12,-3 ) ( 8,-5.4 ) now , we show that @xmath267 * fist note that since @xmath268 then none of the tuples @xmath269 belongs to @xmath57 . therefore for every @xmath118 either @xmath270 or @xmath271 belongs to @xmath57 . thus the following is a proper definition of a boolean valuation : @xmath272 next , we show that @xmath91 is true for @xmath273 . suppose otherwise , i.e. there exists a conjunct @xmath274 that is not true for @xmath273 . we can assume that @xmath275 . this implies that @xmath276 and thus @xmath277 . + take @xmath278 to be the ordering of @xmath57 from proposition [ prop : proc_computation_path ] . since no @xmath279 tuples are present in @xmath57 , and the tuple @xmath247 is dominated by every @xmath279 tuple ( which in turn is dominated by some @xmath270 and @xmath271 tuples ) then @xmath280 . let @xmath281 be the last index of this sequence that @xmath282 is equal to either @xmath283 , @xmath284 , or @xmath285 . since @xmath286 is dominated only by @xmath287 , @xmath288 , and @xmath289 we have for any @xmath290 @xmath291 this implies that @xmath292 which gives a contradiction . * take any valuation @xmath273 for which @xmath91 is true and construct the following set @xmath293 first , note that @xmath57 is a repair : it contains no conflicting tuples and for every tuple from @xmath294 there exists a conflicting tuple in @xmath57 . + next , we show that @xmath295 . in order to prove that we note that for any subset @xmath296 we have @xmath297 suppose otherwise , i.e. there exists a set @xmath298 and @xmath299 such that @xmath300 w.l.o.g . we can assume that @xmath301 . from the construction of @xmath264 and @xmath263 this implies that @xmath302 which is equivalent with @xmath303 , @xmath304 , and @xmath305 . this implies that @xmath274 is not true for @xmath273 which yields a contradiction with @xmath91 being satisfied by @xmath273 . + the property allows us to use proposition [ prop : proc_computation_path ] ( take any ordering of @xmath57 with @xmath247 on the last position ) to state that @xmath57 is - preferred repair w.r.t @xmath49 and @xmath263 . it should be noted here that adding just one tuple @xmath306 and extending the priority with @xmath307 constructs a reduction of @xmath248 to the complement of @xmath308 . and therefore computing - preferred consistent answers is intractable also for a query consisting only of one positive literal . unlike - repairs , the notion of @xmath0-repairs , because of its global character , can not be captured without an essential use of nondeterminism . [ thm : decl_repair_check_intractable ] there exists a set of five functional dependencies @xmath49 such that the set @xmath309 is co - np - complete . it s easy to construct a nondeterministic turing machine @xmath310 . the machine first checks if @xmath57 is a repair ; if yes the machine nondeterministically computes every repair and checks if any of them ( different than @xmath57 ) is preferred over @xmath57 w.r.t . this machine decides the complement of @xmath310 . now , we show that the problem co - np - hard by reducing the complement of @xmath248 to @xmath310 . consider the database schema @xmath311 with the following set of integrity constraints @xmath312 . for any boolean formula @xmath91 in @xmath250 we construct a triple @xmath313 of size polynomial in the size of @xmath91 and such that @xmath314 moreover the reduction can be implemented in time polynomial in the size of @xmath91 . take then any formula @xmath91 in @xmath250 and let @xmath253 be the number of variables used in @xmath91 and @xmath254 the number of conjuncts of @xmath91 . for simplicity we assume that : * used variables have consecutive indexes @xmath80 , * @xmath255 * each conjunct consists of exactly three literals @xmath256 for @xmath257 . we define two auxiliary functions @xmath258 and @xmath259 on literals as follows : @xmath260 the constructed database contains the following elements @xmath315 whose exact values can be found in table [ tab : tab1 ] . @xmath316 the set @xmath317 consists of the following elements @xmath318 it s easy to note that @xmath317 is a repair of @xmath264 w.r.t . clearly @xmath319 , no two elements of @xmath317 are conflicting , and for every element from the set @xmath320 there exists a conflicting element from @xmath317 ( @xmath281 for @xmath44 and @xmath321 for @xmath270 or @xmath271 ) . the priority relation @xmath263 is the unique minimal binary relation on @xmath264 satisfying the following conditions : @xmath322 note that this priority relation is acyclic . also note that the triple @xmath313 can be constructed in the time polynomial in the size of the formula @xmath91 . on figure [ fig : fig2 ] we can find a conflict graph of the instance received from reduction of the formula @xmath323 . ( 17,-6 ) ( 5,-1 ) ( 7,-1 ) ( 8.5,-1 ) ( 9.5,-1 ) ( 11,-1 ) ( 12,-1 ) ( 13.5,-1 ) ( 14.5,-1 ) ( 16,-1 ) ( 1,-5 ) ( 3,-5 ) ( 5,-5 ) ( 7.75,-5 ) ( 10.25,-5 ) ( 12.75,-5 ) ( 15.25,-5 ) ( .5,-5.5)(16,-4.5 ) ( 9,-5.5)@xmath317 now , we show that for any @xmath91 using variables @xmath80 the following holds @xmath324 1 . suppose @xmath325 and take @xmath326 to be the valuation for which @xmath91 is true . consider the following set @xmath327 it s easy to find that @xmath328 is a repair and moreover @xmath329 . thus @xmath317 is not a maximally @xmath0-preferred repair . 2 . suppose @xmath330 , i.e. there exists @xmath331 such that @xmath135 and @xmath332 . + first note that @xmath333 . otherwise for @xmath213 to be preferred over @xmath79 the tuple @xmath281 has to be contained in @xmath213 because there is no element dominating @xmath281 except for @xmath44 . since @xmath281 is adjacent with every @xmath270 and @xmath271 then also none of @xmath270 and @xmath271 belongs to @xmath213 . this implies that @xmath334 which is a contradiction . + since @xmath44 is adjacent to every element of @xmath317 and @xmath333 the sets @xmath213 and @xmath317 are disjoint . this implies that for every @xmath335 the set @xmath213 contains either @xmath270 or @xmath271 ( from maximality , independence , and the fact that @xmath135 ) . + take now the following boolean valuation @xmath336 we show that @xmath337 is a valuation for which @xmath91 is true . suppose otherwise , that there exists a conjunct @xmath274 that is not true under @xmath337 . w.l.o.g we can assume that @xmath338 . this implies that @xmath339 . from the construction of @xmath263 we know that there are no elements dominating over @xmath286 except for @xmath340 . and since obviously @xmath341 , we receive @xmath217 which is a contradiction . using the notion of @xmath0-repairs also leads to a significant increase of computational complexity when computing @xmath0-preferred consistent query answers . [ thm : decl_answers_intractable ] there exists a set of four functional dependencies @xmath49 and a quantifier - free ground query @xmath91 ( consisting of one atom only ) such that the set @xmath342 is @xmath343-complete . the membership of @xmath344 in @xmath345 follows from the definition of @xmath0-preferred consistent query answer : query is not @xmath0-consistently true if it is false in some @xmath0-repair , and checking if a given set is a @xmath0-repair is in co - np . we show @xmath345-hardness below . consider a quantified boolean formula @xmath346 of the form @xmath347 where @xmath348 is quantifier - free and is in 3cnf , i.e @xmath348 equals to @xmath349 , and @xmath350 are clauses of three literals @xmath351 . we will construct a database instance @xmath352 ( over the schema @xmath353 ) and a priority relation @xmath354 such that true is a @xmath0-preferred consistent answer to a query @xmath355 if and only if @xmath346 is true ( the value of @xmath213 can be found in table [ tab : tab5 ] ) . the set of integrity constraints is @xmath356 . we define two auxiliary functions @xmath258 and @xmath259 on literals in the following fashion : @xmath357 now , we describe the tuples contained in @xmath352 . @xmath358 the exact values of tuples can be found in table [ tab : tab5 ] . @xmath359 the priority relation @xmath354 is the unique minimal priority relation that satisfies the following conditions : @xmath360 in figure [ fig : fig5 ] we can find a conflict graph of an instance obtained from the reduction of a formula @xmath361 ( 10,-4)(0,2 ) ( 2.5,1 ) ( 3.5,1 ) ( 6.5,1 ) ( 7.5,1 ) ( 5,1 ) ( 5,-3 ) ( 1,-1 ) ( 2,-1 ) ( 4.5,-1 ) ( 5.5,-1 ) ( 8,-1 ) ( 9,-1 ) ( 3.3,-3 ) ( 6.6,-3 ) we partition the set of all repairs of @xmath352 into two ( separate ) classes : 1 . @xmath362-repairs : repairs that contain @xmath213 . @xmath363-repairs : repairs that do nt contain @xmath213 . we will use @xmath363- and @xmath362-repairs to simulate all possible valuations of variables @xmath80 and @xmath364 respectively . because of the functional dependency @xmath365 a repair is @xmath362-repair if and only if it contains any of @xmath366 or @xmath367 . moreover for any @xmath362-repair @xmath57 and for any @xmath368 either @xmath366 or @xmath367 belongs to @xmath57 . therefore there is one - to - one correspondence between @xmath362-repairs and valuations of @xmath369 variables . to easily move from the world of repairs to the world of valuations and vice versa we define the following two operators ( for @xmath57 being a @xmath362-repair and @xmath273 being a valuation of variables in @xmath348 ) : @xmath370(y_j ) = \begin{cases } true & q_j \in r'\\ false & \bar{q}_j \in r ' \end{cases } \qquad r_\mathcal{y}[v ] = \{q_j | v \models y_j\ } \cup \{\bar{q}_j | v \models \neg y_j \ } \cup \{y\}.\ ] ] we will partition further the class of @xmath363-repairs depending on their conformance with @xmath348 . because @xmath363-repairs will correspond only to valuations of @xmath371 we remove any usage of @xmath369 from @xmath346 in the following way : @xmath372 for a given valuation of @xmath373 construct the following set of tuples : @xmath374 = \{p_i | v \models x_i\ } \cup \{\bar{p}_i | v \models \neg x_i \ } \cup \{d_k | v \not \models \tilde{c}_k\ } \cup \{x\}.\ ] ] it s easy to verify that @xmath375 $ ] is a @xmath363-repair . an @xmath363-repair @xmath57 is _ strict _ if and only if there exists a valuation @xmath273 such that @xmath376 $ ] . otherwise the @xmath363-repair is _ non - strict_. it s clear that there is a one - to - one correspondence between strict @xmath363-repairs and valuations of @xmath373 . construction of a valuation of @xmath373 from a strict @xmath363-repair @xmath57 is also straightforward , for technical reasons we extend it to any @xmath363-repair : @xmath377(x_i ) = \begin{cases } true & p_i \in r'\\ false & \bar{p}_i \in r'\\ false & \text{otherwise } \end{cases}\ ] ] note that @xmath363-repairs can be characterized in a alternative way : a repair of @xmath352 is an @xmath363-repair if and only if it contains @xmath79 . in the main proof we use only strict @xmath79-repairs . the following observation will allow us to remove non - strict repairs from consideration . [ cl : strict_are_maximal ] strict @xmath363-repairs are @xmath89-maximal @xmath363-repairs . first we show how for any non - strict @xmath363-repair @xmath57 we construct a ( strict ) @xmath363-repair @xmath378 such that @xmath379 . take the valuation @xmath380 $ ] and let @xmath381 $ ] . the repair @xmath378 is strict and therefore @xmath382 . we show that @xmath383 , i.e. @xmath384 there are three cases of values of @xmath44 to consider : 1 . implies that @xmath378 is not an @xmath363-repair , a contradiction . 2 . for some @xmath335 we have @xmath386 or @xmath387 . w.l.o.g assume that @xmath388 . this implies that @xmath389 . from construction of @xmath375 $ ] this implies that @xmath390 , a contradiction . 3 . for some @xmath254 we have @xmath391 . assume that @xmath392 and @xmath393 . then @xmath394 and @xmath395 ( it s the neighborhood of @xmath396 ) . from the construction of @xmath378 we have that @xmath397 and both @xmath398 and @xmath399 dominate over @xmath396 . now , suppose that there exists a strict @xmath363-repair @xmath57 such that there exists an @xmath363-repair @xmath378 preferred over @xmath57 . we show that @xmath400 . note that @xmath57 and @xmath378 must agree on the tuples corresponding to the valuation of variables @xmath80 , i.e. @xmath401 since @xmath57 is strict , its content is determined by the corresponding valuation of variables @xmath80 . therefore @xmath402 $ ] . we showed in the previous part of the proof that @xmath403 . since @xmath354 is acyclic this implies that @xmath404 . for any valuation @xmath273 of @xmath373 and @xmath369 we have @xmath375 \ll r_\mathcal{y}[v]$ ] if and only if @xmath405 . we prove implication in two directions : * by contradiction . suppose @xmath406 and there exists a tuple @xmath44 of @xmath375 $ ] which is not dominated by any tuple from @xmath407 $ ] . obviously ( from dependency @xmath365 ) @xmath44 can be only one of @xmath408 . assume that @xmath392 and @xmath409 . by construction of @xmath352 this implies that @xmath410 $ ] , @xmath411 $ ] , and @xmath412 $ ] . from the definition of @xmath375 $ ] and @xmath407 $ ] we receive that @xmath413 , @xmath414 , and @xmath415 . this gives us @xmath416 which is a contradiction . * by contradiction . r_\mathcal{y}[v]$ ] and there exists conjunct @xmath350 such that @xmath417 . assume that @xmath392 and @xmath409 . then @xmath413 , @xmath414 , and @xmath415 . consider @xmath396 and note that it belongs to @xmath375 $ ] ( by definition of @xmath418 ) . from the construction of @xmath419 we know that only @xmath398 , @xmath399 , and @xmath420 dominate over @xmath396 . @xmath421 $ ] does nt contain any of those and this gives us a contradiction . qbf @xmath346 is true if and only if for any strict @xmath363-repair @xmath57 there exists a @xmath362-repair @xmath378 such that @xmath379 . by claim [ cl : strict_are_maximal ] we have that only a @xmath362-repair can be more preferred than a strict @xmath363-repair and for any non - strict @xmath363-repair there always exists a more preferred repair . qbf @xmath346 is true if and only if for any @xmath363-repair @xmath57 there exists a different repair @xmath378 such that @xmath379 . from the partition of repairs we know that @xmath363-repairs can be characterized with a formula @xmath422 . @xmath423 . \iff\\ \forall r ' \in \operatorname{rep}_f(r_\psi ) . [ \neg \exists r '' \in \operatorname{rep}_f(r_\psi ) . r'\neq r '' \land r ' \ll r '' ] \rightarrow r ' \models r(y ) \iff\\ \forall r ' \in \operatorname{grep}_f^{\prec_\psi}(r_\psi ) . r'\models r(y ) \iff\\ ( r_\psi,\prec_\psi ) \in d_{f , r(y)}^g\end{gathered}\ ] ] qbf @xmath346 is true if and only if true is @xmath0-preferred consistent answer to @xmath355 in @xmath352 w.r.t . @xmath49 and @xmath354 . if we use as characterization of @xmath363-repairs the formula @xmath424 then we can reduce qbf to answering to a query with one negated atom . qbf @xmath346 is true if and only if true is @xmath0-preferred consistent answer to @xmath425 in @xmath352 w.r.t . @xmath49 and @xmath354 . the postulate [ eq : p5 ] allows us to think of a total acyclic priority as a cleaning program an exact specification of how to resolve all conflicts . to run this program we simply use algorithm [ alg : alg1 ] and obtain a unique - preferred repair . thanks to theorem [ thm : proc_equals_decl ] , this is also the unique @xmath0-repair . given a a total acyclic priority @xmath64 , the unique - repair ( which if also the unique @xmath0-repair ) can be computed in time polynomial in the size of the database . we limit our discussion to work on using priorities to maintain consistency and facilitate resolution of conflicts . the first to notice the importance of priorities in information systems is @xcite . the authors study there the problem of updates of databases containing propositional sentences . the priority is expressed by storing a natural number with each clause ( the integrity constraints should be tagged with the highest priority @xmath426 ) . if an update ( inserting or deleting a sentence ) leads to inconsistency , among all consistent and realizing the update databases the minimally _ different _ are selected . a database @xmath427 is less different than a database @xmath49 w.r.t . @xmath37 if either for some @xmath428 @xmath429 where @xmath253 is the lowest priority in @xmath37 and @xmath430 consists of all sentences from @xmath37 with priority less or equal to @xmath254 . although this framework does not define a notion of a conflict , we note that more than two facts can create a conflict w.r.t some constraint . for sake of the comparison , assume that the conflicts are generated only by pairs of facts ( together with one of the constraints ) . then , the selected minimally different consistent databases are equivalent to @xmath0-repairs ( and because the considered class of priorities has only acyclic extensions it is equivalent to - repairs ) . we note , however , that the chosen representation of priorities imposes a significant restriction on the class of considered priorities . in particular it assumes transitivity of the priority on conflicting facts i.e. if facts @xmath431 , @xmath247 , and @xmath432 are pair - wise conflicting and @xmath431 has a higher priority than @xmath247 and @xmath247 has a higher priority than @xmath432 , then the priority of @xmath431 is higher than @xmath432 . this assumption can not be always fulfilled in the context of inconsistent databases . for example the conflicts between @xmath431 and @xmath247 , and between @xmath247 and @xmath432 may be caused by violation of one integrity constraints while the conflict between @xmath431 and @xmath432 is introduced by a different constraint . while the user may supply us with a rule assigning priorities to conflicts created by the first integrity constraint , the user may not wish to put any priorities on any conflicts created by the other constraint . a similar representation of priorities used to resolve inconsistency in first - order theories is studied in @xcite , where the inconsistent set of clauses is stratified ( again the lowest strata has the highest priority ) . then preferred maximal consistent subtheories are constructed in a manner analogous to - repairs . furthermore , this approach is generalized to priorities being a partial orders , by considering all extensions to weak orders . again , however , this approach assumes transitivity of priority on conflicts , which as we explained previously may be considered a significant restriction . in @xcite priorities are studied to facilitate the process of _ belief revision_. a belief state is represented as an ordered list of propositional formulae and the revision operation simply adds the given sentence at the end of the given belief state . this representation of belief state allows to keep track of revision history , which is later used to impose a preference order on the possible interpretations of the belief state . only maximally preferred interpretations are used when defining the entailment relation . in the context of logic programs , priorities among rules can be used to handle inconsistent logic programs ( where rules imply contradictory facts ) . more preferred rules are satisfied , possibly at the cost of violating less important ones . in a manner analogous to @xmath89 , @xcite lifts a total order on rules to a preference on ( extended ) answers sets . when computing answers only maximally preferred answers sets are considered . @xcite investigate disjunctive logic programs with priorities on facts . the authors use a transitive and reflexive closure ( denoted here @xmath433 ) of a user supplied set of priorities on facts . the preference on answer sets @xmath434 is defined as follows : * @xmath435 for every answer set @xmath79 * @xmath436 if @xmath437,\ ] ] where @xmath161 stands for @xmath438 . * if @xmath436 and @xmath439 , then @xmath440 . the answer to a program in the extended framework consists of all maximally preferred answer sets . the main shortcoming of using this framework is it s computational infeasibility ( which is specific to decision problems involving general disjunctive programs ) : computing answers to ground queries to disjunctive prioritized logic programs under cautious ( brave ) semantics is @xmath441-complete ( resp . @xmath442-complete ) . a simpler approach to the problem of inconsistent logic programs is presented in @xcite . there conflicting facts are removed from the model unless the priority specifies how to resolve the conflict . because only programs without disjunction are considered , this approach always returns exactly one model of the input program . constructing preferred repairs in a corresponding fashion ( by removing all conflicts unless the priority indicates a resolution ) would similarly return exactly one database instance ( fulfillment of [ eq : p1 ] and [ eq : p5 ] ) . however , if the priority does not specify how to resolve every conflict , the returned instance is not a maximal set of tuples and therefore it is not a repair . such an approach leads to a loss of ( disjunctive ) information and violates postulates [ eq : p3 ] and [ eq : p4 ] . @xcite proposes a framework of _ conditioned active integrity constraints _ , which allows the user to specify the way some of the conflicts can be resolved . this notion syntactically extends the notion of embedded dependency @xmath443 $ ] , where @xmath79 and @xmath213 are sets of variables , @xmath348 and @xmath346 are two conjunctions of literals , and each of existential variables @xmath213 is used only once . a conditioned active integrity constraint is obtained by adding a disjunctive list of update atoms ( @xmath444 for adding , and @xmath445 for deletion ) together with conditions @xmath446 specifying when a corresponding update atom can be used . such an extended constraint is denoted as @xmath447\ ] ] a constraint ( or rather its grounded version ) is said to be _ applied _ to by a repair if the original integrity constraint ( @xmath448 ) is satisfied in the database and the repair is obtained by performing updates satisfying the conditional update atom lists ( one of the atoms @xmath449 has been added and the corresponding condition @xmath450 is satisfied , or one of the atoms @xmath451 has been removed and the corresponding condition @xmath452 is satisfied ) . on all repairs , which are obtained in the standard way by taking as integrity constraints only the heads of the conditioned action integrity constraints , we define relation of preference : a repair @xmath86 is preferred over @xmath85 if every ( ground ) constraint applied in @xmath86 is also applied in @xmath85 . we note here that when restricted to functional dependencies the set of preferred repairs is a superset of - repairs . inclusion in the other direction does nt always hold , which is illustrated on the following example . consider a database @xmath453 consisting of three tuples @xmath454 and suppose we work in the presence of two functional dependencies @xmath455 and @xmath456 . suppose also , that the user specifies that if two tuples are conflicting w.r.t . the fd @xmath455 , then the tuple with higher value of the field @xmath457 should be preferred when repairing the database . a similar wish is expressed for conflicts generated by the second functional dependency . this can be expressed using the following two conditioned active integrity constraints @xmath458 , \end{split } \\ \begin{split } \forall x_1,x_2,y_1,y_2,z , s_1,s_2 . [ ( r(x_1,y_1,z , s_1 ) \land r(x_2,y_2,z , s_2 ) \supset s_1 \neq s_2 ) \supset \\ s_1>s_2 : -r(x_2,y_2,z , s_2 ) ] . \end{split}\end{gathered}\ ] ] after grounding we remove constraints with their head equal to false and we obtain the following set @xmath459 the corresponding priority relation is @xmath460 . note that in the context of the database @xmath45 , the user has provided information sufficient to solve all the conflicts , i.e. among the repairs @xmath461 the repair @xmath86 is the unique repair selected by @xmath462 . at the same time only is applied to @xmath86 and only is applied to @xmath85 , what makes both repairs incomparable in terms of the framework of @xcite . this example also shows that the discussed framework violates the postulate [ eq : p4 ] . note also that removing preference information on how to resolve the conflict between @xmath463 and @xmath464 will yield only one repair @xmath86 . this shows that this framework violates the postulate [ eq : p5 ] . at the same time this framework fulfills the property of conservativeness ( the preferred repairs are a subset of standard repairs ) and non - emptiness ( there is always at least one preferred repair ) . @xcite also describes how to translate conditioned active integrity constraints into a prioritized logic program @xcite , whose preferred models correspond to maximally preferred repairs . note that the framework of prioritized logic programming is computationally more powerful ( answering answers under the brave semantics is @xmath442-complete ) than required by the problem of finding if an atom is present in any repair ( @xmath465-complete ) . it is yet to be seen if less powerful programming environment ( like general disjunctive logic programs ) can be used to compute preferred answers . @xcite uses ranking functions on tuples to resolve conflicts by taking only the tuple with highest rank and removing others . this approach constructs a unique repair under the assumption that no two different tuples are of equal rank ( postulates [ eq : p1 ] and [ eq : p5 ] ) . if this assumption is not satisfied and the tuples contain numeric values , a new value , called the fusion , can be calculated from the conflicting tuples ( then , however , the constructed instance is not a repair in the sense of definition [ def : repair ] ) . a different approach based on ranking is studied in @xcite . the authors consider polynomial functions that are used to rank repairs . when computing preferred consistent query answers , only repairs with the highest rank are considered . the postulates [ eq : p1 ] and [ eq : p3 ] are trivially satisfied , but because this form of preference information does not have natural notions of extensions and maximality , it is hard to discuss postulates [ eq : p4 ] and [ eq : p5 ] . also , the preference among repairs in this method is not based on the way in which the conflicts are resolved . an approach where the user has a certain degree of control over the way the conflicts are resolved is presented in @xcite . using repair constraints the user can restrict considered repairs to those where tuples from one relation have been removed only if similar tuples have been removed from some other relation . this approach is monotonic , but not necessarily non - empty . the authors propose method of weakening the repair constraints to restore non - emptiness , however this comes at the price of losing monotonicity . in this paper we proposed a general framework of preferred repairs and preferred consistent query answers by formulating a set of intuitive postulates . we proposed two instantiations of the framework and studied their semantic and computational properties . table [ tab : tab3 ] summarizes the computational complexity results ; its first row is taken from @xcite . we envision several directions for further work . the postulates [ eq : p1][eq : p5 ] can be refined , so that only non - trivial instantiations are captured . for example , the following instantiation fulfills the postulates : we ignore any priority which is not total and return all repairs in this case ; when the priority is total we return the unique - repair . this approach , however , is trivial and obviously does not increase the computational complexity of any of considered problems . also , the computational consequences of further refining the postulates should be examined . the last is generalization of our framework to broader class of constraints . conflict graphs can be generalized to hypergraphs @xcite , which allow to handle broader class of denial constraints . then , more than two tuples can be involved in a single conflict and the current notion of priority does not have a clear meaning .

we present the results of keck telescope laser adaptive optics integral field spectroscopy with osiris of the innermost regions of the ngc 1333 svs 13 outflow that forms the system of herbig - haro objects 7 - 11 . we find a bright 0 long microjet traced by the emission of shock - excited [ feii ] . beyond the extent of this jet , we find a series of bubbles and fragments of bubbles that are traced in the lower excitation h 10 s(1 ) line . while the most recent outflow activity is directed almost precisely ( pa 145 ) to the south - east of svs 13 , there is clear indication that prior bubble ejections were pointed in different directions . within these variations , a clear connection of the newly observed bubble ejection events to the well - known , poorly collimated hh 7 - 11 system of herbig - haro objects is established . astrometry of the youngest of the expanding shock fronts at 3 epochs covering a time span of over two years gives kinematic ages for two of these . the kinematic age of the youngest bubble is slightly older than the historically observed last photometric outburst of svs 13 in 1990 , consistent with that event launching the bubble and some deceleration of its expansion . a re - evaluation of historic infrared photometry and new data show that svs 13 has not yet returned to its brightness before that outburst and thus shows a behavior similar to fuor outbursts , albeit with a smaller amplitude . we postulate that the creation of a series of bubbles and the changes in outflow direction are indicative of a precessing disk and accretion events triggered by a repetitive phenomenon possibly linked to the orbit of a close binary companion . however , our high - resolution images in the h and k bands do not directly detect any companion object . we have tried , but failed to detect , the kinematic signature of rotation of the microjet in the [ feii ] emission line at 1.644 m .

1408.5940

the accretion of matter onto a forming star is inextricably associated with mass outflow . this manifests itself as well - collimated jets typically observed in higher excitation shock - excited forbidden lines such as [ sii ] or [ feii ] , or lower excitation emission of shock - exited h@xmath1 . larger - scale and older outflows are typically detected in co emission of entrained ambient material . while the precise mechanism of jet launching is actively being debated , steady progress has been made in observational studies of the launch regions of jets . such studies require the highest possible spatial resolutions , since all theories postulate that jets are launched from the inner regions of the protostellar disks on spatial scales of a few au , or from the protostar s magnetosphere , on scales of a stellar diameter . at present , the highest spatial resolutions for studies of the jet launch regions are achieved at optical wavelengths with the hubble space telescope ( hst ) , and at near - infrared wavelengths with adaptive optics on large ground - based telescopes . the near - infrared techniques have the advantage of being able to better penetrate dust extinction , so that the jet launch regions of more deeply embedded , generally younger stars can be observed . however , near - infrared laser - guide - star adaptive optics observations today are still limited by the requirement to have a fairy bright , optically visible ( r @xmath5 16 ) tip - tilt reference star close to the object , or that the object itself can serve this purpose . as a consequence , even at near - infrared wavelengths , the best candidates for high spatial resolution studies are objects near the end of their class i phase or early in the classical t tauri star phase because in nearby molecular ( dark ) clouds , the only available optical tip - tilt reference star is often the young star itself . in this paper we present detailed adaptive - optics corrected integral field spectroscopy of ngc 1333 svs 13 , the driving source of the famous chain of herbig - haro objects hh 7 - 11 @xcite . other commonly used names for svs 13 are v512 per and 2mass j03290375 + 3116039 . the svs 13 outflow appears relatively poorly collimated and is comprised of a number of individual shock fronts , as can be seen in fig . 1 that presents a progression from seeing - limited to diffraction - limited images of the svs 13 outflow . there is an anti - parallel counter outflow visible , but it is displaced from the axis of the hh 7 - 11 outflow . for a general overview of ngc 1333 and a review of the literature on the svs 13 subcluster , the reader is referred to the review article by @xcite . the two micron all sky survey ( 2mass ) position of svs 13 is @xmath6 at epoch 1999 nov . 26 with an accuracy of @xmath7 @xcite . this position lies between the vla configuration a radio position of vla 4b at 3.6 cm reported by @xcite at @xmath8 with an effective epoch of 1997.9 and estimated errors of @xmath9 and the 7 mm vla configuration b position of the same radio source , which @xcite give as @xmath10 with an epoch of 2001 may 4 . the epoch of the 2mass observations is between the epochs of the vla observations , and the reported near - infrared coordinates lie between those vla coordinates . within the errors of @xmath9 for the vla data and of @xmath7 for 2mass , and a tentatively indicated small proper motion of vla 4b , these data show that the near - infrared source svs 13 is identical to the vla mm and cm - wavelength source vla 4b . it should be noted that @xcite and @xcite , on the basis of older optical astrometry , had originally identified svs 13 with a different cm - radio source : vla 4a . m ) image based on the same original data as those used by @xcite for their proper motion study . the image shows the full extent of the shocked molecular hydrogen emission . the second panel from the top is a small portion of a h@xmath1 s(1 ) 2.122 @xmath4 m image obtained in 1996 with the quirc camera at the uh 2.2 m telescope that shows the emission features closest to svs 13 . the third panel shows the integrated h@xmath1 s(1 ) line intensity observed with osiris on keck ii in 2011 with the 100 mas spaxel scale , while the fourth panel was taken with osiris on keck i in 2012 with the 20 mas spaxel scale . ] we use the distance to ngc 1333 svs 13 determined by @xcite from vlbi parallax measurements of masers associated with this object : 235 @xmath11 18 pc . the study of the dynamics of the ngc 1333 region by @xcite has determined the systemic velocity of the gas near svs 13 to be 8 km @xmath12 relative to the local standard of rest . throughout this paper , we will give velocities relative to this systemic velocity of the ngc 1333 molecular core that svs 13 is embedded in . young , still accreting stars often experience eruptive changes in brightness that traditionally get classified into either fu orionis ( fuor ) or ex lupi ( exor ) type outbursts , depending on the duration of the outburst and its spectrum . the first review of these phenomena has been given by @xcite . fuor outbursts exhibit time scales of decades to centuries and the spectral characteristics of an optically thick luminous disk with absorption line spectra , while the less substantive exor outbursts have timescales of years , have been observed to actually return to pre - outburst brightness , and show optically thin emission line spectra . jets emanating from young stars are often comprised of a series of individual shock fronts that have been postulated to arise from changes in the jet velocity as a result of repetitive eruptive events , as described in the review by @xcite and references therein . this gives , in principle , a way to study the history of such eruptive accretion instabilities from the `` fossil '' record in the jet shock fronts . our target object , svs 13 , is certainly in this class of multi - shock outflow sources and has been observed to undergo an outburst around 1990 , but the light curve does not match either the classical fuor or exor curves . the relationship of this last observed outburst to the structure of the svs 13 outflow will be studied here . the proper motions of the shock fronts of the hh 7 - 11 system associated with svs 13 have been studied at optical wavelengths by @xcite and later by @xcite , and in the infrared in the h@xmath1 10 s(1 ) line by @xcite using the near - infrared camera and multi - object spectrometer ( nicmos ) , by @xcite using the united kingdom infrared telescope ( ukirt ) , and with the spitzer space telescope by @xcite . these studies all gave proper motions of the shock fronts in the range of 33 mas yr@xmath13 that establish a kinematic expansion age of 2100 yrs for the most distant shock front studied there , the hh 7 bow shock . in general , the jets seen in optical and near - infrared forbidden lines , mostly in [ sii ] and [ feii ] , offer the most direct view of the material ejected from an accretion disk or from a protostar s magnetosphere . there are differences in the details of theoretical models of this process , e.g. , in @xcite , @xcite , @xcite , or @xcite , but they all agree on the main point : jets originating from the magnetosphere of a rotating star , or from a rotating disk , and detected in high - excitation shock - excited lines are expected to carry away excess angular momentum and thereby enable the mass accretion process . however , to date , the most convincing detection of a rotational signature of jets from very young and still deeply embedded objects in sed classes 0 and i have been obtained with radio interferometric observations of outflows in co and sio emission . these observations typically achieve spatial resolutions of a few arcseconds but make up for this disadvantage by exquisite velocity resolution . carbon - monoxide high - velocity emission is mostly associated with the entrainment of ambient molecular material into a jet originally emitted at even higher velocity . any signature of the original jet rotation is expected to be highly confused at this turbulent interface between the jet and the ambient material . nevertheless , @xcite detected the kinematic signature of rotation in the jet of cb 26 in the co(2 - 1 ) line , very close to the source of the jet . @xcite found some evidence for rotational components in the ori - s6 outflow using various higher transitions of co and so , while @xcite recently reported a rotational signature in the hh 797 outflow in ic 348 using co(2 - 1 ) . in these cases , the strongest rotational signatures were found at some distance ( @xmath14 ) from the driving source . even farther away from the driving source , using sio data from the vla , @xcite found a rotation signature of the ngc 1333 iras 4a2 protostellar jet , consistent with the disk rotation in that object . the jet could only be observed at distances of more than 5@xmath15 from its source , and the rotation signature appears most pronounced at distances of 20@xmath15 . these results are far from universal , however . in a series of papers on hh 211 culminating in @xcite they obtained only a tentative detection of rotation . @xcite studied the kinematics of sio emission in the hh 212 jet and did not detect a credible rotational signature . in this paper we have tried to detect the kinematic signature of jet rotation in our [ feii ] data and are presenting our velocity data , but in the end failed to detect this effect . there are two strong systems of near - infrared emission lines that are commonly used for the study of protostellar jets : the ro - vibrational lines of h@xmath1 and the forbidden [ feii ] lines . as was already discussed specifically in the case of svs 13 by @xcite , the h@xmath1 lines , the brightest being the 10 s(1 ) line at 2.122 @xmath4 m , trace low - velocity shocks , either internal shocks within the jet , or the turbulent interface to the ambient medium around the jet . this line is therefore well suited for the study of internal shocks and entrainment of ambient material by the jet , but is less suited for a detection of the jet itself . in contrast , [ feii ] traces higher temperatures and excitations and is usually confined to the densest parts of a jet near its launch region , and to its terminal shocks against ambient material . the [ feii ] lines in the near - infrared are therefore particularly suited for a study of the kinematics of the jet itself . emission from atomic hydrogen , in the near infrared specifically the brackett series of lines , is observed in many young stars , and typically originates in the spatially unresolved accretion disk around the young star itself , and not in the jet . the adaptive optics data reported here were obtained at the keck i and ii telescopes , using laser guide star adaptive optics in conjunction with the oh - suppression infrared imaging spectrograph ( osiris ) built by @xcite . osiris is a lenslet integral field spectrograph where each spatial element ( spaxel ) forms an image of the telescope pupil before being dispersed into a spectrum . this technique separates spatial flux gradients in each spaxel from any velocity shifts and is therefore very suitable for measuring small shifts in line centroid over extended objects with spatial scales of the order of a spaxel . the spectral lines mentioned above can be observed in two setting of osiris . in the 3@xmath16 grating order and through the kn2 filter , the h@xmath1 s(1 ) line at 2.122 @xmath4 m is covered . in the 4@xmath17 grading order and through the hn3 filter , two [ feii ] lines and two brackett - series lines are included in the spectral bandpass . the first observations were carried out in the night of 2011 august 21 , ( ut ) ( mjd 55794.5345 = 2011.638 ) with osiris on the keck ii telescope , using the relatively coarse 100 mas per spaxel and 50 mas per spaxel scales and the kn2 and kbb filters . a second , higher quality set of data was obtained in the night of 2012 november 4 , ( ut ) ( mjd 56235.4853 = 2012.844 ) with osiris now at the keck i telescope . the finest of the spaxel scales of osiris , with lenslets subtending 20 mas was used under conditions of exceptionally good seeing , with the canada - france - hawaii telescope ( cfht ) seeing monitor reporting average seeing at optical wavelengths of @xmath2 0@xmath03 . a third set of data was obtained on 2013 november 22 and 23 ( mjd 56618.5 and 56619.5 , midpoint=2013.893 ) with osiris at the keck i telescope . the seeing was not quite as good as in the year before , but still , very good data were obtained that are well suited for an astrometric comparison with the 2012 data . .log of the keck osiris observations [ cols="^,^,^,^,^,^,^",options="header " , ] on 2012 nov . 04 and 2013 nov . 22 , we obtained data sets with 600 s individual exposure time in the hn3 filter , covering the 1.644002 @xmath4 m a@xmath18 a@xmath19 emission of [ feii ] ( referred in the following as the 1.644 @xmath4 m line ) and also the fainter [ feii ] @xmath20 @xmath21 line at 1.599915 @xmath4 m ( in the following called the 1.600 @xmath4 m line ) . all wavelengths given in this paper refer to vacuum and the line wavelengths and identifications are based on @xcite . 2 shows a raw spectrum of svs 13 in the hn3 filter , averaged over a large 0@xmath02@xmath220@xmath02 box centered on svs 13 , to include the radiation from the object , but also the oh airglow sky flux used for wavelength calibration . the [ feii ] 1.600 @xmath4 m line was evaluated and generally corroborates the conclusions drawn from the 1.644 @xmath4 m line , but due to its lower signal - to - noise ratio , these data are not presented here nor did we try to extract information on excitation conditions from a comparison of those two [ feii ] lines . @xcite have already presented a detailed study of the excitation conditions of h@xmath1 and [ feii ] emission in svs 13 . in addition to these shock - excited forbidden lines , osiris in 4@xmath17 grating order with the hn3 filter also covers two atomic hydrogen emission lines , the 13 - 4 ( 1.611373 @xmath4 m ) and 12 - 4 ( 1.641170 @xmath4 m ) recombination lines of the brackett series @xcite . in the kn2 filter , 4 data sets with 600 s exposure time each were obtained that cover the h@xmath1 10 s(1 ) line at 2.121833 @xmath4 m @xcite . 2@xmath220@xmath02 box containing svs 13 , but also substantial sky oh airglow flux . this figure serves to illustrate the hydrogen br 12 and 13 lines , the [ feii ] lines , and the night - sky oh airglow lines used for wavelength calibration . ] for the monitoring of svs 13 at infrared wavelengths , the brightness of the object around k=8 is actually a problem . most archival infrared images obtained with large telescopes and modern infrared cameras are saturated on svs 13 . from 2012 to 2014 , we have therefore monitored the ngc 1333 region , including svs 13 , with the infrared imaging survey ( iris ) system specifically to study the variability of that object at the present epoch . iris is a 0.8 m telescope and 1024@xmath221024 infrared camera dedicated to the monitoring of infrared variability . it has been described in detail by @xcite . the iris infrared camera , which is a refurbished version of the uh quick infrared camera ( quirc ) described by @xcite , is operated in a mode where after the first non - destructive read of the detector array we immediately take a second read , and compute a double - correlated image with only 2 s exposure time by differencing these first two reads . after this second read - out , the integration continues up to a third read of the detector array , which gives the full integration time of 20 s usually used for our monitoring projects . the first ( 2 s effective integration time ) images are far from saturating on svs 13 and were used for the photometry presented here . the iris camera raw data are processed in a reduction pipeline based on the image reduction and analysis facility ( iraf ) software @xcite . the astrometric solution of the co - added images is calculated based on sextractor @xcite and scamp @xcite . photometry is obtained with the iraf task phot and calibrated against the 2mass catalog @xcite . we chose to correct dark current and other signal - independent detector artifacts in osiris by subtracting the median of 10 dark frames of 600 s exposure time from each object exposure , in order not to waste observing time on frequent observations of an empty sky field for sky background subtraction . the 2d detector frames were then processed into 3d spectral data cubes with the osiris final data reduction pipeline ( frp ) . this process of extracting the spectra from the raw detector format is obviously critical for the wavelength calibration of the spectra and therefore for the extraction of velocity information . in 2013 , the keck observatory released a new version of the data reduction pipeline ( drp v3.2 ) with newly calibrated `` rectification matrices '' that contain the parameters for the extraction of individual spaxel spectra from the raw detector format into wavelength calibrated spectra . a first inspection of the newly reduced 2013 data showed better consistency of the wavelength calibration across all spaxel and therefore the 2012 data were also re - reduced with the new software and calibration data . these newly reduced data did indeed change the velocity structure observed in the svs 13 jet to the point of forcing a different conclusions . the individual data cubes in the hn3 filter were cleaned of spurious bright spaxels using the iraf cosmicray task on individual planes of the data cube . all data cubes in the hn3 filter were spatially mosaicked together with offsets determined by the centroid of the bright continuum image of the svs 13 star . for the shorter wavelengths , this method resulted in superior image quality compared to using offset data supplied by the ao system . for the longer wavelength spectra in the kn2 band , where we only took one dither pattern , relying on the offsets provided by the laser guide star system proved sufficient . for both the hn3 and kn2 data cubes , the individual planes ( wavelengths ) of the resulting combined spectral data cubes were adjusted for zero sky background using the background zero adjustment in the iraf task imcombine . the primary purpose of this procedure was to remove the oh(5,3)r@xmath23(2 ) night sky line at 1.644216 @xmath4 m that partly overlaps with the [ feii ] 1.644002 @xmath4 m line , as was discussed by @xcite . the integral field spectrograph osiris records all wavelength planes simultaneously , so that the psf at continuum wavelengths is recorded under identical atmospheric conditions than the structure at emission line wavelengths . over small wavelength intervals where the wavelength - dependence of diffraction is negligible , very precise continuum subtraction can be achieved . the continuum adjacent to a given line was computed as the average of approximately 10 wavelength planes outside of the line profile and centered on the line . between the 2011 and 2012 runs , the osiris instrument was moved from the keck ii telescope to keck i. both telescopes and their adaptive optics systems have nominally identical focal lengths , but for an astrometric comparison of data from these two telescopes , the spaxel scale in osiris should be observationally verified . the best data available for this purpose were kindly provided by t. do and j. lu from their work on astrometry of stars near the galactic center where they calibrated osiris astrometry relative to astrometry with nirc2 , an instrument that stayed at the keck ii telescope . these data indicate that at keck i , the spaxel scale in mas pixel@xmath13 of osiris is 1.28% larger than it had been at keck ii . while installed on keck i , between our 2012 and 2013 observing runs , the grating of osirs was also changed , resulting in a new wavelength calibration . also , as was discussed above , the data reduction pipeline used at keck for osiris data was upgraded . finally , in moving from keck ii to keck i , the optical path leading to osiris now contains one less mirror reflection , leading to a change in parity of the images . we used the osiris camera with its wider field to obtain a few images tying the osiris spectral data cubes astrometrically to conventional images of svs 13 and one detectable star in its vicinity . in addition , we observed other objects to ascertain that our data are presented in the correct parity and orientation on the sky . the precise scale ratio of the nominal 50 mas ( used in 2011 ) and 20 mas ( used in 2012 and 2013 ) spaxel scales of osiris was measured on setup data cubes obtained in 2008 while observing the @xmath2 0@xmath025 separation binary @xmath24 orionis @xcite . by averaging the component separation measured in each continuum data cube plane within @xmath11 0.1 @xmath4 m of 2.122 @xmath4 m , a ratio of 2.4534 @xmath11 0.0005 between the two spaxel scales was measured . this value deviates by about 1 % from the nominal scale ratio of 2.5 . it should be noted that these deviations from the nominal spaxel scale ratios and the , even smaller , errors of this measurement are much smaller than the astrometric effects of shock - front motion and expansion discussed later in this paper . the continuum images near the emission line features show nothing but the unresolved stellar object svs 13 . they are thus suitable as a psf kernel for deconvolution of the individual ( wavelength ) cube planes . the deconvolution was done with the lucy - richardson algorithm @xcite and @xcite , as implemented in the iraf stsdas package . how fast the lucy - richardson algorithm converges depends on the signal - to - noise ratio in the frame . we have tried out different numbers of iterations and chi - square criteria and have chosen a combination that leads to a significant improvement in the spatial resolution without creating obvious artifacts . working with the original data without continuum subtraction led to strong artifacts in the wings of the overwhelmingly bright stellar source . for the deconvolution of faint emission features well separated from svs 13 we have therefore worked from the continuum - subtracted data . for the purpose of the grey - scale and rgb representations of the images , we have computed a `` high - dynamic - range '' version of the image by adding the deconvolved continuum - subtracted image and a small fraction of the image flux of the original deconvolved image , effectively creating an image where the continuum source is strongly suppressed , but still indicated in the image . 3 shows the s(1 ) image from 2011 in greyscale , and false color representations of the deconvolved `` high - dynamic - range '' 2012 and 2013 data in the h@xmath1 s(1 ) line ( red ) , [ feii ] 1.644@xmath4 m ( green ) and continuum ( blue ) . fiducial marks are included to illustrate the expansion of the s(1 ) bubble over the course of those 2 years . s(1 ) image ( red channel ) and the continuum subtracted [ feii ] emission line image from 2012 nov . 4 . all three images are aligned to better than 1 mas , even though the deconvolution residuals in the continuum subtracted images do not align precisely . to mark the position of the continuum source svs 13 , the blue channel shows 10% of the continuum near the [ feii ] line . a few fiducial marks indicate positions of the bubble rim in 2012 . the comparison of these marks to the images taken earlier and later clearly indicated the expansion motion of the bubble . ] the osiris final data reduction pipeline ( frp ) extracts the data from the raw 2-dimensional detector frame and forms a spectral data cube with 0.2 nm spacing in the h band ( 4@xmath17 grating order ) and 0.25 nm in the k band ( 3@xmath16 grating order ) . the oh airglow lines in the bandpass of the hn3 and kn2 filters give an opportunity to check the wavelength calibration and to measure the spectral resolution in particular , the oh(5,3)r@xmath23(2 ) line at 1.644216 @xmath4 m is very close to the wavelength of the [ feii ] 1.644 @xmath4 m line . for the analysis of the [ feii ] line emission , the uniformly distributed oh emission was eliminated by adjusting the spatial median of the flux in the image to zero . in contrast , for the wavelength calibration , the original data were used and the flux in the oh line was measured . to account for continuum stray light , a continuum subtraction using the same continuum as for the [ feii ] line extraction was used . the individual oh airglow lines were measured by summing up the flux of 400 pixels without detectable object flux , and measuring the position and width of the oh airglow lines in the resulting one - dimensional spectrum using the iraf task splot . we used the vacuum wavelengths of oh lines given by @xcite and found that the wavelength calibration provided by the osiris reduction pipeline required a slight correction in zero point and dispersion , of order 0.01 nm . after this correction , the individual oh line positions have residual wavelength errors of 0.008 nm , about a factor of two larger than the fit residuals given by our wavelength reference , @xcite . the oh airglow lines are , in reality , close doublets with line spacing well below the resolution of osiris . the measured fwhm of the oh lines was 0.44 @xmath11 0.06 nm in the hn3 filter , and 0.61 @xmath11 0.06 nm in the kn2 filter , a little more than two wavelength planes in the osiris data cubes . near the [ feii ] line at 1.644 @xmath4 m , one wavelength interval of the data cube corresponds to 37 @xmath25 in radial velocity . the spectral resolution near that line therefore corresponds to 81 @xmath25 . near the h@xmath1 10 s(1 ) line at 2.122 @xmath4 m , one wavelength interval of 0.25 nm corresponds to a radial velocity difference of 35 @xmath25 , and the measured fwhm of the spectral lines corresponds to 85 @xmath25 width of the spectral profile . most of the velocity effects discussed in this paper are therefore smaller than one spectral resolution element . the accuracy of velocity measurements depends on the signal - to - noise ratio of the emission in question . in our discussion of the velocity structure of the [ feii ] microjet in section 4.5 , we will present a detailed analysis of the measurement errors in that particular case . we have taken data on an a0v star during the observations , but in the end have chosen not to use these for telluric absorption correction . our data analysis does not rely on spectrophotometric correction of the data cubes . in the hn3 filter , the study of the jet in the 1.644 @xmath4 m [ feii ] line and of the hi13 - 4 and hi12 - 4 brackett lines are potentially all affected by the broad atomic hydrogen absorption in the standard star , and the data reduction pipeline would therefore interpolate through the hydrogen lines , making the telluric absorption correction meaningless at these wavelengths . we have also not been able to use those standard star measurements for a flux calibration of our spectra . a consistency check of data obtained with different exposure times showed a problem with the way osiris computes the effective integration times for multi - sampled exposures . we do not have the data to fully calibrate this effect , and given that the flux calibration is not essential for our data analysis , prefer not to give a flux calibration for our data . figure 1 shows the h@xmath1 shock - excited line emission associated with svs 13 on four different spatial scales : the top image is an archival spitzer space telescope infrared array camera ( irac ) @xcite channel 2 ( 4.5 @xmath4 m ) image covering most of the outflow emission . the second panel is a small cutout from a ground - based , seeing limited image in the h@xmath1 10 s(1 ) line obtained at the uh 2.2 m telescope , showing just the emission in the immediate vicinity of svs 13 . the third panel from the top is the s(1 ) integrated line image obtained with the 100 mas scale of osiris in 2011 and the bottom panel shows the 20 mas scale s(1 ) image from osiris obtained in 2012 , without deconvolution . we show this figure to demonstrate the relationship between the bubbles in the svs 13 outflow very close to the source and the larger scale structure of the flow further downwind . this relationship is not entirely trivial , since not all the bubbles have propagated in the same direction from the svs 13 star . the seeing - limited ground - based image shows the brightest of these bubbles in the glare of the psf . the intermediate , 100 mas scale , osiris line image shows three distinct partial bubbles near the source of the outflow , the outer two of which were also observed , but not further discussed , by @xcite , on hst / nicmos images and had also been detected through ukirt narrow - band fabry - perot imaging by @xcite . the best prior detection of these bubbles was by @xcite who used adaptive - optics - assisted long - slit spectroscopy and detected all the bubbles as a series of emission maxima along the slit . the most recent emission , the 0@xmath02 long [ feii ] jet , its faint extension into the s(1 ) bubble , and that youngest s(1 ) bubble are oriented at p.a . 145@xmath3 , as seen in fig . the older two bubbles , visible in the 100 mas image of fig . 1 , are lying more to the south at p.a . 159@xmath3 than those aforementioned features , indicating some variations in the jet emission direction . in contrast the main chain of herbig - haro knots hh 7 - 11 lies at p.a . 123@xmath3 @xcite . while at optical wavelengths , the svs 13 outflow manifests itself only in the herbig - haro chain hh 7 - 11 @xcite , longer wavelengths increasingly reveal a counterflow that appears displaced from the hh 7 - 11 axis , and appears in general fainter and less organized ( fig . 1 , top panel ) . based on spitzer telescope images over a time span of 7 years , @xcite have obtained one proper motion data point for a relatively well defined knot in that counterflow , and this one point is consistent with the counterflow originating from svs 13 . also , no other embedded protostar has been detected by @xcite to the north of svs 13 that might explain this flow as being independent from svs 13 . the displacement in the outflow axis , in combination with the recent changes in outflow direction reported in this paper , suggest that the source of the svs 13 outflow is changing direction , probably due to some precessing motion , as will be discussed in more detail in section 4.2 . as an explanation for the differences between the hh 7 - 11 flow and the counterflow , @xcite have suggested that the northern counterflow enters into the central cavity of ngc 1333 , resulting in different ambient pressure and excitation conditions than the hh 7 - 11 flow . we have obtained astrometry of the bubble expansion on two different spatial scales . relatively wide field osiris data cubes with the 100 mas spaxel@xmath13 scale were obtained on keck ii on 2011 august 21 and keck i on 2013 november 23 . a difference image of the h@xmath1 s(1 ) emission at these two epochs is shown in fig . 4 and illustrates that all three of the closest shock fronts to svs 13 are showing noticeable motion . the farthest of these three shock fronts was defined enough to allow a cross - correlation measurement of its expansion age in the box indicated in the figure . astrometry of the smallest and youngest expanding bubble was done on the 2012 november 4 and 2013 november 22/23 data that were taken at the keck i telescope with the same adaptive optics system and the same spaxel scale ( 20 mas spaxel@xmath13 ) and are therefore suitable for a precise astrometric measurement . in this case , the individual planes of the data cube were deconvolved using the lucy - richardson algorithm , to improve the definition of the bubble edge . figure 3 illustrates the expansion and motion of the youngest , smallest bubble between the three epochs by superposing fiducial marks that outline features in the middle ( 2012 ) image . for fine registration of the images , the iraf task xregister was used to measure the relative alignment of all the frames at the position of the svs 13 stellar object and a magnified and optimally registered version of the images was produced . to measure the expansion of the bubble , we worked under the assumption that its motion relative to the star can be described as a simple constant velocity expansion with the origin at the position of the star . we computed magnified versions of the individual 2012 november 4 images with magnification factors in the range from 0.95 to 1.10 and the center of the magnification on the svs 13 star . we then computed the product of these magnified images and each of the individual 2013 november 22 images . the average of this product in a box centered on the s(1 ) bubble feature image varies smoothly with expansion factor for each of these pairs and the maximum of this one - dimensional correlation function was simply read from the table of cross - correlation values . we show the cross - correlation functions of the 2012 and 2013 high resolution images in fig . 5 to illustrate the variations due to noise and deconvolution noise amplification , and to document how the errors of this measurement were obtained . error of the mean scaling factor . the mean value corresponds to a kinematic start of the expansion in 1980 , an early limit to the true starting date since the expanding bubble will realistically be decelerated . ] the expansion factor so determined was then converted to a kinematic expansion age of 32 yrs ( prior to 2012 ) , i.e. , kinematic starting time of the expansion in 1980 , which gives the upper limit to the true age of the bubble assuming that the bubble has been expanding at constant velocity . since , realistically , the bubble is expanding into the dense environment of a molecular core , the true age of this youngest bubble will be smaller than the kinematic expansion age and is therefore consistent with this expanding bubble having been generated in the 1990 photometric outburst of svs 13 that will be discussed in secion 4.4 . in 2012 , the apex of the bubble was located 654 mas from the star . the front end of the bubble therefore has moved with an average projected angular velocity of 20 mas yr@xmath13 ( 4.7 au yr@xmath13 = 22.3 km s@xmath13 at 235 pc distance ) away from the star . similarly , in 2012 , the bubble had a radius of 212 mas and an average radial proper motion of 6.6 mas yr@xmath13 ( 1.55 au yr@xmath13 = 7.35 km s@xmath13 ) . the proper motion of the bubble center is therefore the proper motion of the leading shock front minus the radial expansion : 13.4 mas yr@xmath13 @xmath2 15 km s@xmath13 . in order to also obtain an estimate of the kinematic age of the next two more distant bubbles , we used data cubes obtained in 2011 august 21 on keck ii with the 100 mas scale , and similar measurements obtained on 2013 november 22 on keck i. the difference of these two images is shown in fig . 4 . the only shock front where a clear maximum of this correlation function was detected in these 100 mas data was the most distant of the three , indicated by a box in fig . 4 , separated from the star by 2.87@xmath26 in 2011 . for this bow shock front , we derive a kinematic formation time of 1919 @xmath11 7 and a linear projected proper motion of 31 mas yr@xmath13 . we have not obtained a reliable kinematic age for the middle shock - excited feature at @xmath2 1@xmath05 from the star , between those two features discussed above ( fig . the fact that this rather poorly defined system of shock fronts lies pretty precisely in the middle of the 1980 and 1919 ( kinematic formation time ) features suggests that this feature must have formed , again in the kinematic sense without accounting for deceleration , around 1950 . if there is a regular pattern to the formation of these bubbles , which with only three examples can not be convincingly established yet , the next outburst could be expected within the next decade . it should be noted , as was already pointed out by @xcite that the proper motion of the major shock fronts in the older parts of the svs 13 outflow indicate a much longer time interval between shock front generating events : about 500 yrs . it is not clear whether svs 13 exhibits multiple periods , or whether the frequency of ejection events has recently increased . an argument for the latter point of view may be that the outflow direction has apparently changed in the past century , as we will discuss now . as demonstrated by fig . 1 , the two previous bubbles are located more toward the se of svs 13 and , for example , @xcite chose p.a . 159@xmath3 as the best slit orientation to cover them . the larger chain of hh objects 7 to 11 is oriented along a position angle of 123@xmath3 @xcite . the larger scale proper motion study by @xcite based on spitzer 4.5 @xmath4 m images shows some emission knots to the se of svs 13 with a proper motion vector generally to the se , in particular the herbig - haro knots 7 , 8 , and 10 . north of svs 13 , in the counter - flow , they find an emission knot with a generally northern proper motion vector ( p.a . -10@xmath3 ) that they ascribe to a chance superposition of another outflow far to the south of svs 13 and with generally northern outflow direction . the [ feii ] jet originating from svs 13 ( fig . 3 ) is oriented along p.a . 145@xmath3 and the h@xmath1 bubble center is displaced from svs 13 along the same angle . with the exception of this most recent h@xmath1 bubble , the other recent mass ejection events from svs 13 have therefore ejected material initially in a more southerly direction ( p.a . @xmath2 155@xmath3 - 159@xmath3 ) , as seen in figs . 1 , and 4 . with outflows generally being bipolar , there must also be mass ejected into a northerly direction . we therefore believe that , contrary to the assertion by @xcite , the 4.5 @xmath4 m emission knot north of svs 13 and with northerly proper motion is part of the counter jet to the bubbles reported here . the more distant parts of the s(1 ) emission nw of svs 13 are anti - parallel , but laterally displaced , from the hh 7 - 11 system of emission knots . we suggest that the hh 7 - 11 chain , the system of bubbles immediately south of svs 13 , the emission knots north of svs 13 , and the more distant emission knots further to the nw ( fig . 1 ) are all part of the same bipolar outflow originating in svs 13 . this outflow has the s - shaped morphology indicative of a precessing or otherwise unstable jet source . corroborating this , radio interferometry mapping of molecular emission near svs 13 by @xcite has similarly found an orientation of high - velocity material south of svs 13 different than the hh 7 - 11 herbig - haro chain . they had already concluded that the differences in the alignment of features of different age indicate a precessing source of the outflow . a prominent other example of such s - shaped morphology of a molecular hydrogen jet , iras 03256 + 3055 , is located just south of svs 13 in ngc 1333 and has been studied in detail by @xcite . the relatively low proper motion and spatial velocity measured for the h@xmath1 bubble studied here is consistent with proper motion measurements of the more distant hh 7 - 11 chain of shock fronts by @xcite , @xcite , and @xcite , but is inconsistent with the much higher proper motions reported by @xcite in the first near - infrared proper motion study of hh 7 - 11 . while a precessing accretion disk provides an explanation for the rapid changes in the outflow direction observed in svs 13 , our data do not show any indication for the presence of a companion object that would be close enough to cause the disk precession on the timescales discussed here . from the size and location of the h@xmath1 bubbles , we can conclude that bubble ejection events happen with a period of several decades . in the model where such events are triggered by periastron passages of a companion object , the orbital semimajor axis must be of order of tens of au , or several of the original 20 mas spaxels of our data . the fact that we do nt see a companion object implies that such an object , if it existed , is intrinsically too faint and/or too deeply embedded to be visible in the h and k atmospheric windows . the only other case of a young stellar object with a pronounced bubble structure of its outflow is xz tau that was studied in detail by @xcite on the basis of optical multi - epoch hst imaging . very similar to the case of svs 13 discussed here , their images show a series of bubbles with a collimated jet propagating inside of the bubbles . @xcite reported initial results of numerical simulations of a very young pulsed jet in close proximity to its driving source . their simulations were specifically tuned to reproduce the observations of the xz tau a chain of bubbles and therefore modeled a faster , more rapidly pulsing jet resulting in more rapidly expanding , overlapping bubbles . the general scenario underlying their model is however applicable to our case of svs 13 : the fuor - like photometric outburst in 1990 created a short - lived pulse of jet activity . this newly created , relatively fast jet ran into slower moving material ejected prior to the outburst event , and this internal shock created an expanding `` fireball '' that subsequently expanded ballistically into a bubble carried away from the star by the outflow . some time after the formation of the bubble , a fast continuous jet then emerges to catch up with the bubble , pierce it , and partially destroy it . in svs 13 this process currently repeats itself about every 30 years , creating the series of bubble fragments that forms the string of herbig - haro objects . in distinction from xl tau , the case of svs 13 also involves a significant change in the direction of the jet and bubble ejection leading to the s - shaped overall morphology of the herbig - haro chain . the individual velocity channels of the h@xmath1 s(1 ) line emission are shown in fig . 6 . in the blue - shifted wing of the velocity distribution , emission projected on the center of the bubble is visible , which is the expected characteristic of an expanding 3-dimensional bubble rather than a 2-dimensional ring . in fig . 7 , a color - coded velocity map of the h@xmath1 10 s(1 ) line emission is presented . the s(1 ) emission shows three distinguishable velocity features . emission near the intersection with the jet ( traced in [ feii ] ) shows the smallest blueshifted velocities and is coded red . the rim of the s(1 ) bubble where the line of sight is tangential to the bubble shows intermediate velocities , coded yellow in the figure . the highest velocities towards the observer are measured in the filamentary features projected against the center of the bubble and are coded in blue . in the brightly visible rim , the velocity centroid varies between -40 and -55 km s@xmath13 relative to the systemic velocity of the molecular material around svs 13 , while in the simple model of an expanding shell , those velocities should be constant and representative of the center motion of the bubble . we take -47 @xmath11 7 km s@xmath13 as the typical radial velocity of the bubble center . 10 s(1 ) line images of the svs 13 jet . the velocities indicated in each panel are relative to the systemic velocity of the svs 13 core . the bubble feature is blueshifted relative to the systemic velocity . ] s(1 ) line at 2.122 @xmath4 m . three distinct velocity features can be distinguished : the emission at the intersection of the jet with the bubble has the lowest ( blueshifted ) radial velocity , the rim of the bubble has intermediate velocities , and the features projected against the center of the bubble have the highest blueshifted velocities . overlayed on the velocity map are contours of the high - dynamic range lucy - richardson deconvolved flux maps that indicate the position of svs 13 itself . a color version of this figure is available in the electronic version of this paper . ] with the proper motion of the bubble center of 15 @xmath11 2 km s@xmath13 , this suggests an inclination angle of 18@xmath3 @xmath11 3@xmath3 against the line of sight . @xcite had given an inclination angle of 20@xmath3 to 40@xmath3 for emission close to svs 13 while @xcite has given 40@xmath3 for more distant emission knots . all the measurements of the inclination angle were done on different shock features . their spread is therefore a combination of measurement uncertainties and the true variations in the motion of these shock fronts . irrespective of which of the shock fronts are measures , all data indicate that the outflow emerging from svs 13 is pointed strongly towards the observer . this explains why the counterjet , which moves away from the observer and into the molecular core around svs 13 , is not detectable at optical wavelengths . shock fronts of this velocity running into stationary ambient molecular hydrogen are certainly capable of exciting v=10 s(1 ) line emission and are not in danger of dissociating the h@xmath1 , see , for example , @xcite . the infrared source svs 13 was discovered by @xcite at a k - band magnitude of 9.08 in a 36@xmath26 aperture . soon after the discovery , @xcite reported k = 8.48 in a 30@xmath26 aperture observed in 1978 . @xcite reported an observation by g. olofsson from 1980 at k=8.7 in a 14@xmath26 aperture as a private communication . @xcite measured a pre - outburst brightness on k=9.34 in a 6 - 8@xmath26 aperture at the irtf on 1981 , oct . 11 - 13 and @xcite measured k=9.30 in a 16@xmath26 aperture on 1981 , dec . the object experienced a sudden increase in brightness around 1990 , as first reported by @xcite and further studied by @xcite , @xcite , and @xcite . in the k band , where the best pre - outburst data are available , as listed above , the pre - outburst magnitude showed some variation between 9.0 and 9.5 mag . post - outburst , @xcite documented brightness variation between 8.0 and 8.6 mag . based on the small amplitude of the brightness increase , the post - outburst brightness fluctuation , and the emission lines in its spectrum , both @xcite and @xcite have concluded that svs 13 underwent an exor or similar outburst , but both papers left the possibility open that the outburst may be of a different nature . @xcite has studied the photometric behavior of svs 13 again and concluded that the object had not returned to its pre - outburst brightness at that time . motivated by the uncertain classification of this event , we have re - examined the historical photometry and are discussing new measurements . we have tried to gather the available photometric data on svs 13 from the literature and data archives . while many images of the svs 13 region exist , the star svs 13 is saturated on most of these , and useful data can only be obtained from shallow surveys , usually with small telescopes . in fig . 8 , we show all the available photometric data as a light curve . the 2mass survey list svs 13 as k@xmath27=8.169 and these data had been obtained on nov . 26 , 1999 . recent k@xmath27 data from the iris telescope @xcite show svs 13 varying in the range of k@xmath27=8.46 to 8.70 between 2012 august and 2014 january . the photometric color transformations between the ukirt system and the 2mass system for the k vs. k@xmath27 filters are insignificant @xcite , so a direct comparison can be made between the measurements by @xcite and the most recent data . there is extended emission around svs 13 , so very large photometric apertures tend to overestimate the brightness . the @xcite , @xcite , @xcite and @xcite data were corrected to the aperture diameter of 6@xmath26 used for the iris photometry , while all other photometric data shown here were originally obtained with apertures in the range of 5@xmath15 - 8@xmath26 , close enough to the iris aperture to not require a correction . figure 8 shows these aperture - corrected photometric values and clearly demonstrates that svs 13 has not declined back to its pre - outburst ( @xmath28 1990 ) brightness , but remains at or near its peak post - outburst brightness with some indication that brightness fluctuations have diminished over the course of the past 24 years . similar to the case of the low - luminosity , deeply embedded , cometary nebula oo ser @xcite and @xcite , svs 13 defies a clear classification as either fuor or exor . while fuors generally show late - type low gravity absorption spectra in the infrared @xcite indicative of a luminous , optically dense disk , the prototypical exor ex lupi shows the co bandheads and many other optical lines in emission during extreme exor outbursts @xcite . during minor outbursts , optical emission lines are observed , but co may be in photospheric absorption @xcite . svs 13 shows the co bandheads in emission @xcite . in the context of this paper , the important conclusion is that the elevated post - outburst level of accretion and therefore jet activity in svs 13 has persisted for the past 2 decades to the present , so that the light curve of svs 13 shares the long duration maximum with fuors , while spectroscopically , it resembles an exor , and the small outburst amplitude resembles neither . this suggests that at least for the younger , more deeply embedded accretion instability events , the traditional two classes may not be appropriate , and that a continous range of outburst characteristics may be a better way to understand this phenomenon . the light curve also implies that the next outburst of svs 13 , if indeed these outbursts occur repetitively about every 30 years , will start from a brighter state of svs 13 than the previous one . this would mean that in addition to the repetitive outbursts , we are also observing cumulative changes in svs 13 . the seeing - limited integral field spectroscopy of @xcite had only marginally resolved the [ feii ] emission around svs 13 . our adaptive - optics corrected osiris data show that [ feii ] traces a high - velocity microjet that extends from the source svs 13 into the area of the most recent molecular hydrogen bubble . the results are summarized in the false - color images ( fig . 3 ) . here , the wavelength - integrated continuum - subtracted flux of the h@xmath1 10 s(1 ) line was lucy - richardson deconvolved and is displayed in the red channel . to show the location of the stellar central object in svs 13 without introducing the artifacts from imperfect deconvolution of a dominant bright source , we have added 10% of the deconvolved flux of two continuum wavelength channels to the continuum - subtracted image . this produces , in effect , a high - dynamic - range version of the svs 13 image for the purpose of showing the relationship of features at different wavelengths . note that the wavelength channels used here were different from those used as the psf . the resulting deconvolved continuum image is therefore not the trivial solution of deconvolving the psf with itself . the green channel of fig . 3 shows the deconvolved integral over the [ feii ] line at 1.644 @xmath4 m . in the same way as described above , a fraction of the continuum wavelengths immediately adjacent to the line was added to the data to mark the location of the stellar source . finally , to adjust the color balance of the stellar object to white , the same fraction of the continuum wavelength channels on either side of the [ feii ] line was assigned to the blue channel . our lucy - richardson deconvolved images of the high - excitation [ feii ] shocks associated with the jets show that the jet is very narrow , about 20 - 40 mas wide . the velocity diagram in fig . 9 shows that the [ feii ] emission of the jet is blueshifted by -140 to -150 km s@xmath13 relative to the systemic velocity of the molecular material around svs 13 @xcite . a comparison of figs . 7 and 9 demonstrates that the [ feii ] emission is more blueshifted than the h@xmath1 s(1 ) emission . the direct superposition of the h@xmath1 and [ feii ] emission line images shows that the bright portions of the microjet [ feii ] emission extend up to the rim of the h@xmath1 bubble , and that faint traces of [ feii ] emission can be detected up to about the center of the bubble . the jet [ feii ] intensity drops by a factor of @xmath2 20 at the bubble surface . in fact , the brightest [ feii ] along the jet axis is seen directly upwind from the h@xmath1 bubble surface . the total 1.644 @xmath4 m [ feii ] flux is dominated by the emission outside of the expanding h@xmath1 bubble . since the front edge of the bright jet component near the bubble rim is not sharply defined , and subject to the degree of lucy - richardson deconvolution and the different quality of the adaptive optics correction achieved at the two epochs , we could not directly measure the proper motion of the jet front edge . from fig . 3 it seems clear that the [ feii ] shock emission from the jet is surrounded by an envelope of entrained ambient material radiating in the low - excitation shocked h@xmath1 lines , and that at least the bright portions of the jet [ feii ] emission terminate at the bubble rim . from the proper motion of the bubble front side , and the rate of bubble expansion ( section 4.2 ) , the expected , but not directly measured , proper motion of the side of the bubble facing svs 13 ( the back side in the direction of motion ) is 6.6 mas yr@xmath13 , which we also take as the proper motion of the front end of the bright jet . the length of the [ feii ] jet is therefore changing only very slowly , at a rate of @xmath2 7 mas year@xmath13 , and the centroid of the [ feii ] flux therefore moves at less than half this speed . this is responsible for the impression noted already by @xcite that the [ feii ] emission looked stationary . what we see from the jet in the [ feii ] emission line are either internal shocks or the shocks resulting from interaction with ambient molecular material that then gets entrained by the jet . the bulk of the jet material propagates into the area of the h@xmath1 bubble , but the excitation conditions are , apparently , less favorable to the formation of such shocks radiating in [ feii ] . 1.644 @xmath4 m wavelength data cubes of the svs 13 jet , deconvolved using the lucy - richardson algorithm , for the 2012 and 2013 data , respectively . the bottom panels show the corresponding velocity sigma maps . the white contours outline the line integrated [ feii ] flux distribution . it clearly shows that all the line emission from the jet is blue - shifted relative to the systemic velocity of svs 13 . a color version of this figure is available in the electronic version of this paper . ] we have tried , both on the 2012 and 2013 data , to detect the signature of jet rotation . initially , using only the 2012 data , such rotation appeared to be indicated @xcite . however , the 2013 data , and the re - reduced 2012 data with the new calibrations , did not confirm this ( fig . the velocity pattern measured in 2013 was more confused and , if anything , a faint indication of the opposite rotation direction was found . in fig . 9 ( lower panels ) we show the rms variations of the velocity measurements on the individual data cubes that were coadded to form the velocity maps in the top panels . these indicate that the errors of the velocity maps are below 10 km s@xmath13 in most parts of the [ feii ] jet . we conclude that the excellent spatial resolution and moderate spectral resolution of osiris are not sufficient to resolve the kinematic signature of jet rotation in svs 13 . the permitted atomic hydrogen line emission from svs 13 was found to be broad with line widths of @xmath2 180 @xmath11 10 km s@xmath13 for br@xmath29 by @xcite and for br-12 by @xcite . the emission is centered on the position of the continuum source , but spatially unresolved . our data cube in the hn3 filter contains the br-12 and br-13 hydrogen recombination lines that trace the same hot hydrogen recombination regions as the more frequently used br-@xmath29 does . in fig . 10 , we show the deconvolved images of svs 13 across the br-13 emission line after subtraction of the continuum emission . this figure confirms that emission in the atomic hydrogen recombination lines is spatially centered on the young star , and has the same flux profile as the continuum , i.e. we see no indication that the br-13 emission is extended . figure 10 has 10 mas pixels ( 2.35 au ) , and any systematic differences between the line and continuum psf are well below that angular scale . we conclude that the zone of atomic hydrogen emission , presumably the accretion disk itself , is less that 2 au in extent . it is actually expected to be only of order of the dimensions of the star itself , i.e. , about 2 orders of magnitude smaller than this limit . our velocity data are consistent with the higher spectral resolution data of @xcite who found that the br@xmath29 line is centered at -25 ( @xmath30 ) km s@xmath13 relative to the systemic velocity of the svs 13 core , essentially at the same velocity as the star . 2 au ) . we have presented adaptive optics corrected integral field spectroscopy of the young outflow source svs 13 in ngc 1333 . the h@xmath1 10 s(1 ) line at 2.122 @xmath4 m , indicating low - velocity shocks , the higher excitation [ feii ] line of the micro - jet at 1.644 @xmath4 m , and atomic hydrogen emission in the hi12 - 4 and hi13 - 4 lines were analyzed and lead us to the following main conclusions : 1 . the hh 7 - 11 outflow originates in svs 13 , which is identical to vla 4b . 2 . the outflow , at present , originates as a micro - jet of @xmath2 0@xmath02 length , detectable in [ feii ] and oriented at p.a . the formation of the youngest partly formed bubble visible in h@xmath1 emission can be traced back to the @xmath2 1990 outburst . 4 . the bright parts of the [ feii ] microjet reach up to the boundary of the h@xmath1 s(1 ) bubble , but fainter [ feii ] emission can be traced another @xmath2 0@xmath02 to near the center of the bubble . 5 . beyond that , h@xmath1 s(1 ) emission outlines a curved path of the jet . 6 . the orientation of the next two bubbles at p.a . @xmath2 159@xmath3 is roughly point - symmetric to the orientation of outflowing material in the counter - jet found by @xcite in co and @xcite at 4.5 @xmath4 m . the chain of bubble fragments and their proper motions suggest that bubble - generating events are occurring repetitively , roughly every 30 years , at the present time . the formation of a series of expanding bubbles within the outflow by a series of eruptive events provides an explanation for the widening of the outflow cavity . atomic hydrogen emission in the hi12 - 4 and hi13 - 4 ( brackett series ) lines is detected around the continuum position of svs 13 , indicating ongoing accretion onto the star . the outflow source svs 13 remains at or near the peak brightness reached during the 1990 outburst . the light curve therefore resembles that of fuor type objects , while the emission line spectrum matches the characteristics of exors . svs 13 therefore represents an object somewhere between those classical classes . most of the data presented herein were obtained at the w.m . keck observatory , which is operated as a scientific partnership among the california institute of technology , the university of california and nasa . the observatory was made possible by the generous financial support of the w.m . keck foundation . some photometric data on svs 13 were obtained at the iris telescope on cerro armazones , which is operated under a cooperative agreement between the `` astronomisches institut , ruhr universitt bochum '' , germany , the `` universidad catolica del norte '' in antofagasta , chile , and the institute for astronomy , university of hawaii , usa . construction of the iris infrared camera was supported by the national science foundation under grant ast07 - 04954 . the operation of the iris telescope is supported by the `` nordrhein - westflische akademie der wissenschaften und der knste '' in the framework of the academy program by the federal republic of germany and the state of nordrhein - westfalen . we wish to thank angie barr dominguez , thomas dembsky , holger drass , lena kaderhandt , michael ramolla and christian westhues for operating the iris telescope for the acquisition of the data used in this paper , ramon watermann for writing the data reduction pipeline , and roland lemke for technical support . we thank tuan do and jessica lu for kindly providing the information about the osiris spaxel scales on the keck i and keck ii telescopes . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center/ california institute of technology , funded by the national aeronautics and space administration and the national science foundation . this publication also uses 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 . skrutskie , m. f. , cutri , r. m. , stiening , r. , weinberg , m. d. , schneider , s. , carpenter , j. m. , beichman , c. , capps , r. , chester , t. , elias , j. , huchra , j. , liebert , j. , lonsdale , c. , monet , d. g. , price , s. , seitzer , p. , jarrett , t. , kirkpatrick , j. d. , gizis , j. , howard , e. , evans , t. , fowler , j. , fullmer , l. , hurt , r. , light , r. , kopan , e. l. , marsh , k. a. , mccallon , h. l. , tam , r. , van dyk , s. , & wheelock , s. 2006 , , 131 , 1163

i generalize the theory of phonon topological band structures of isostatic lattices to frustrated antiferromagnets . i achieve this with a discovery of a many - body supersymmetry ( susy ) in the phonon problem of balls and springs and its connection to local constraints satisfied by ground states . the witten index of the susy model demands the maxwell - calladine index of mechanical structures . `` spontaneous supersymmetry breaking '' is identified as the need to gap all modes in the bulk to create the topological isostatic lattice state . since ground states of magnetic systems also satisfy local constraint conditions ( such as the vanishing of the total spin on a triangle ) i identify a similar susy structure for many common models of antiferromagnets including the square , triangluar , kagome , pyrochlore nearest neighbor antiferromagnets , and the square lattice antiferromagnet . remarkably , the kagome family of antiferromagnets is the analog of topological isostatic lattices among this collections of models . thus , a solid state realization of the theory of phonon topological band structure may be found in frustrated magnetic materials . recently , kane and lubensky ( kl ) identified topological properties of isostatic lattice phonons . they achieved this by discovering a topological index governing mechanical structures by building on calladine s work and further utilizing a dirac - like square - root of the phonon equations of motion , a problem to which they could apply the theory of topological insulators . remarkably , they showed the existence of lattices with gapped phonons for periodic boundary conditions that must have gapless phonons with open boundary conditions . it is natural to wonder if this striking effect is more general and they conclude their study with `` finally , it will be interesting to explore connections with theories of frustrated magnetism . '' where reference 48 is my study identifying topological gauge dynamics of the zero modes of classical kagome antiferromagnets . this paper s goal is to make this connection and use it to generalize kl theory beyond phonons . the kl theory of isostatic lattices is a different branch of topological phases from the theory of topological band insulators . following the original discovery of topological insulators , topological properties of boson band structures have been studied for a wide variety of systems including phonons , photons , acoustic phononic crystals and magnons . these systems achieve their topological properties in the presence of time reversal symmetry breaking and are built directly from the physics of the integer quantum hall effect . in contrast , kl s theory of isostatic phonons is time reversal symmetric and `` purely geometric in nature'' . a connection to the integer quantum hall effect is made only after a dirac - like square - rooting procedure of the equations of motion . it therefore presents a new direction in the theory of topological phases . remarkably , though its may not apply directly to solid state phonons because they are mechanically stable , the kl theory has already seen a variety of applications due to its insight into the general phenomena of mechanical collapse . these include some topological aspects of the jamming and rigidity percolation transitions , metamaterials made from beams and pins , and , remarkably , origami . ref . has also taken it beyond the linearized level and discovered solitons that can propagate freely in the bulk of the isostatic state . so , given the fundamental insight it provides , any extension of kl theory to a new class of systems , including extensions that go beyond the linearized limit , is likely to shed new light on those systems . in this light , frustrated magnets and/or highly frustrated magnets are a prime target for an extension of kl theory . they are magnets not only `` on the verge of collapse '' but also those that have already `` collapsed '' . here collapsing is the analog of destabilizing the magnetic ordered state into a paramagnetic state such as a quantum spin liquid or valence bond solid . a variety of materials including the organics , kagome family and pyrochlores are heavily studied for this reason . in addition , highly geometrically frustrated magnets have a form of accidental degeneracy that results from a special feature of the spin hamiltoinan . this frustration is toy - like ( fine tuned ) perhaps in a similar way that balls and springs are toy - like versions of a general theory of phonons . so if kl theory were applicable to frustrated magnets , this generalization might apply to many already realized solid state materials . in this paper , we show that the key to generalizing kl theory to other systems lies in connecting it to an abstract theoretical framework that takes the form of a many - body supersymmetric ( susy ) structure that extends the description of balls and springs . the new fermionic degrees of freedom , that i dub `` phoninos '' , are superpartners to phonons and are governed by the kl theory s square - rooted equations of motion . for linearized phonons , the two sets of degrees of freedom are decoupled . the phoninos therefore need not be real degrees of freedom but just reflect the specialness of balls and springs compared to a more general theory of phonons . i then show that the topological index identified in kl theory is demanded by the witten index of susy . a similar conclusion was reached by ref . in the continuum limit . remarkably , spontaneous susy breaking here is just the need to gap all modes in the bulk to create the topological state ( as could be the case for topological superconductors ) . finally , we discuss what protects this topology including the role of quantum effects and non - linearities . i then apply the same many - body susy construction to the case of magnons . it turns out susy is found for both unfrustrated and frustrated magnons for either case has ground states that satisfy local constraints and the susy construction is built on these constraints . remarkably , kagome magninos are governed by a hamiltonian which is the dirac s constraint matrix studied in ref . and is related through susy to the ordinary kagome magnon problem . the witten index in this case is then shown to vanish for periodic boundary conditions but not for open boundary conditions demonstrating that kagome magnons are the analog of the isostatic lattice of kl theory if we can gap all their modes without changing topology such as in a distorted kagome crystal . such a phase we call isostatic magnetism . we conclude by identifying some of the new physical phenomena predicted by these formal development : the design of magnets by exploiting weaknesses in isostatic magnetism such as through a rich set of magnetic field induced spin flip transitions that we call `` self field modes '' ; and an application of the ideas here to the origin of the industrial strength of mnir ( why it is a robust antiferromagnet ) ; the potential application of topological band theory of electrons to magninos in the isostatic magnetic phase and via susy to magnons ; the study of the loss of magnetic order via isostatic magnetism which can be viewed as a classical critical point between magnetic order and a classical spin liquid / cooperative paramagnet .

1510.03697

to generalize kl theory to magnetism , we need to first identify the theoretical constructs that underpin the theory of phonons and then seek analogs of these constructs in the magnon problem . the simplest description of phonons , that of the vibrations of balls connected to springs ( in the classical limit ) , is endowed with a number of theoretical constructs that shed much light on their behavior . to begin with , this simplicity demands a specific form of their hamiltonian . if we define the extension of spring @xmath3 to be @xmath4 , then the most general form for an ideal balls and springs classical hamiltonian is @xmath5 here @xmath6 is the @xmath7 component of the momentum of the ball labeled by @xmath8 , @xmath9 is the matrix inverse of the `` mass tensor '' @xmath10 , @xmath11 is the spring constant matrix and repeated indices are summed over . in the simplest setting , @xmath12 and @xmath13 are proportional to identity matrices . here we leave them in the general form to aid our study of the structure of the theory and not its application . the restriction of the hamiltonian to that of balls and springs therefore introduces two ( inverse ) metrics @xmath11 and @xmath9 in configuration space and momentum space respectively and identifies the ground state as satisfying constraints @xmath14 , @xmath15 . balls and springs are a toy - like simplification of a full phonon problem . we will return to the justification of their use and the validity of the following results in the next section ( section [ sec : topologicalprotections ] ) . in the linearized phonon limit , there is also a matrix @xmath16 and an associated topological invariant . this matrix relates @xmath4 to the displacements of each atom @xmath17 from their equilibrium positions via @xmath18 . this matrix also relates the forces @xmath19 to the tensions @xmath20 in each spring via @xmath21 . the topological invariant associated with this matrix ( which we will call the maxwell - calladine index@xcite though it was discovered by kane and lubensky ) is @xmath22 where @xmath23 is the number of components @xmath24 , @xmath25 the number of sites , @xmath26 the number of springs ( bonds ) , @xmath27 denotes the dimension of the null space of @xmath28 and we have used @xmath29 by the fundamental theorem of linear algebra . this quantity relates the topology data of the system , @xmath23 , @xmath25 and @xmath26 to the number of zero modes @xmath30 and number of states of self stress @xmath31 . here a zero mode is a vector @xmath32 that satisfies @xmath33 and a state of self stress is a vector of tensions @xmath34 that satisfies @xmath35 . this quantity is topological in the sense that it does nt depend on the metrics @xmath36 and @xmath9 and remains unchanged under any changes of the matrix @xmath16 . so linearized balls and springs also have a natural topological invariant . kane and lubensky argued that a dirac - like square - rooted equations of motion enables the study of phonons following the theory of topological insulators . these equations of motion are defined by the matrix @xmath37 here i have taken the liberty to add a minus sign to kane and lubensky s matrix and re - inserted @xmath36 and @xmath9 which they set to identity matrices . this is so that squaring this matrix produces @xmath38 where the upper left block defines the second order differential equation for the displacements @xmath39 the minus sign can be removed by multiplying @xmath40 by @xmath41 from the right , something we are free to do since @xmath42 commutes with the @xmath43 ( a feature noted by kane and lubensky ) . this model helps us understand the importance of the topological index . if , for periodic boundary conditions , @xmath44 then there must always be either a zero mode ( if @xmath45 ) or a self stress mode ( if @xmath46 ) . in the spectrum of @xmath40 there is no gap to the excitations . however , if @xmath47 then we can always find an @xmath16 such that the gap disappears . in such a case , if we open the boundary conditions we will invariably find that @xmath44 since we change the counting of the number of sites and bonds and so either special self stresses exist due to boundary conditions or special zero modes exist . in either case , there is an interplay between a gap , topology and boundary conditions as found in other topological systems such as the quantum hall effect or topological insulators . finally , there is one more mathematical object we will find useful for the study of phonons . to study the zero modes of this problem directly , we can follow ref . and view the hamiltonian as energetically imposing constraints on the degrees of freedom . here these constraints are simply @xmath48 and @xmath15 as noted above for if these conditions are met the hamiltonian vanishes . in his development of constrained hamiltonian mechanics , dirac pointed out that an important object in the study of constraints in phase space is the constraint matrix which here takes the form : @xmath49 where @xmath50 is the usual poisson bracket of classical mechanics . a vector in the null space of this matrix is then either associated with a redundant constraint ( when there are more constraint functions than necessary to constrain the variables ) or with a zero mode coordinate that has no conjugate variable in the space of zero modes ( gauge coordinate ) . if we expand the spring extensions @xmath4 in terms of the displacements of the atoms from their equilibrium positions @xmath17 as @xmath51 then the constraint matrix takes the form @xmath52 remarkably , in this form , @xmath53 is similar to the `` square - rooted '' hamiltonian @xmath40 discussed above . if we work in kane and lubensky s units where @xmath54 and @xmath55 they are actually the same . we also see that the redundant constraints are associated with self stress modes living in the null space of @xmath56 and the gauge coordinates with the zero modes living in the null space of @xmath16 . so the eigenvalue problem associated with dirac s constraint matrix here is closely related to the dirac - like square - rooted phonon hamiltonian ! now the relationship between all of these mathematical objects , the metrics , the hamiltonians , the @xmath16 matrix , and the constraint matrix are not so easily understood . they all clearly derive from the balls and springs construction . but how is the topological index dependent on the metrics ? how is the constraint matrix @xmath57 related to the square - rooted hamiltonian @xmath40 ? in what way is the topology protected ? can these constructions be applied to other systems that are not balls and springs ? to address these questions , we need an abstract theoretical framework that identifies all of these mathematical object from a simpler set of constructs . remarkably , such a framework is provided by extending the model s phase space to include a set of fermion degrees of freedom @xmath58 and @xmath59 that we will call phonino modes and imposing supersymmetry . from this vantage point we will then look down on all the mathematical objects discussed previously and see their relationships to each other . the formal nature of the supersymmetry will then also allow us to connect two seemingly different problems : balls and springs and quadratic spin models . we then just need to determine if the connection is merely formal or whether it predicts new physical phenomena for either balls and springs or magnetic systems . a hamiltonian that is a quadratic form in constraint functions that define its ground state has a natural supersymmetry . to construct this supersymmetric model , we begin by defining the supersymmetric charge by the product of a constraint function and a new degree of freedom : @xmath60 this looks like a lagrange multiplier term in an action meant to impose the constraints @xmath48 and @xmath15 where @xmath59 and @xmath58 are the lagrange multipliers . however , here we view these as degrees of freedom that satisfy their own poisson bracket relations . indeed we can construct the supersymmetric hamiltonian using the poisson bracket @xmath61 where @xmath62 denotes a poisson bracket . for this equation to be true , it is necessary for @xmath58 and @xmath59 to be grassmann numbers so that @xmath63 is grassmann odd and @xmath64 does not vanish by the normal antisymmetry of poisson brackets . we also used @xmath65 . if we choose the remaining poisson brackets to be @xmath66 then we obtain the simple relation @xmath67 where @xmath68 is the first term and the second and third terms make up @xmath69 of eq . [ eq : hphonon ] . this defines the poisson bracket to be@xcite @xmath70 which is symmetric if @xmath71 and @xmath72 are both grassmann odd and antisymmetric otherwise . finally , we see that @xmath73 so the two observables form a closed superalgebra . so we can promote the phonon problem to a supersymmetric problem of bosonic phonons and fermionic phoninos where the phoninos satisfy a clifford algebra with metrics @xmath11 and @xmath9 and the phonino hamiltonian is determined by the constraint matrix @xmath74 where @xmath53 is given by eq . [ eq : phononcmatrix ] . note , here we have chosen to study phonons classically using grassmann numbers . in this sense we follow one of the original motivations to introduce grassmann numbers into classical mechanics : to construct simpler - than - field - theory supersymmetry models@xcite . the use of classical mechanics here , however , is just to connect with the history of theories of balls and springs and is not necessary as we will discuss section [ sec : topologicalprotection ] . furthermore , our analysis of the model presented below will be informed by the use of supersymmetry in quantum mechanics ( see ref . for a review ) , because it is another simpler - than - field - theory supersymmetry model . however , unlike quantum mechanics , the above model is a many - body supersymmetric lattice model , one not so different from the lattice model of a supersymmetric topological superconductor studied in ref . . in the linearized limit the two hamiltonians become @xmath75 and @xmath76 where we used @xmath77 . in this limit then , the phonons and the phoninos are decoupled and the hamiltonians are quadratic . we therefore have a model we can apply to real systems since it contains the correct phonon eigenvalue problem . it is not so obvious , however , what we can learn by introducing the phoninos . to settle this , we need to directly study the different eigenvalue problems . to study the eigenvalue problems of @xmath78 , we need to solve the corresponding equations of motion . the first order in time differential equations for the phonons are @xmath79 and for the phoninos are @xmath80 remarkably , the phonino equations of motion are the square - rooted hamiltonian matrix @xmath40 introduced by kane and lubensky . the constraint matrix @xmath53 and @xmath40 are therefore related in the supersymmetric model by one entering the hamiltonian quadratic form and the other by the corresponding equations of motion . we will also find it useful to work with the second derivative in time equations of motion : @xmath81 from these , we see that , @xmath17 , @xmath82 and @xmath59 all obey the same equations of motion . @xmath58 obeys a distinctly different equation . we also can determine that each equation has two types of eigenmodes . for the @xmath17-type equations , all modes either have zero eigenvalue and are in the null space of @xmath16 or they have finite eigenvalues . there are no modes with zero eigenvalue that satisfy @xmath83 . this is because such a mode would satisfy @xmath84 with @xmath85 and violate the assumption that @xmath36 is a positive definite matrix . for a similar reason , the @xmath58 equation has two modes types , those with zero eigenvalue in the nullspace of @xmath56 and those with finite eigenvalues . we can of course proceed by solving these equations separately , but this would not give us any insight into how the phoninos can help us understand the phonons . instead , lets proceed by studying supersymmetry transformations and reduce the problem to solving only the parts not related by supersymmetry . the supersymmetric charge provides us with a map between the phase space observables with an even number of phonino coordinates ( grassmann even ) and those with an odd number ( grassmann odd ) . the single particle phonon modes which are linear combinations of @xmath17 and @xmath6 are in the even group while the single particle phonino modes which are linear combinations of @xmath59 and @xmath58 are in the odd group . the map @xmath86 we can construct from @xmath63 sends an observable from the even sector to the odd sector or vice versa by inserting it into the location of the @xmath87 . this map has a special relationship with the equations of motion : time evolution of the mapped observable is the same as the unmapped observable . for example , @xmath88 is a phonino observable that obeys the equations of motion @xmath89 which is the equations of motion for @xmath17 . similarly , the phonon observables @xmath90 , @xmath91 , @xmath92 obey the @xmath6 , @xmath59 and @xmath58 equations respectively . computing these poisson brackets explicitly , we see that they are @xmath93 so it would seem that supersymmetry is too powerful a symmetry , that the phonino problem is equivalent to the phonon problem ( and therefore uninteresting ) . to study these relationships in more detail , lets now take linear combinations of phase space observables . if @xmath94 is an eigenmode of the @xmath6 equation then it obeys @xmath95 now if @xmath96 so the @xmath97 is in the ( left ) nullspace of @xmath56 , then it is an eigenvector with frequency @xmath98 . if this is not the case , it is an eigenvector with a finite frequency as discussed above . now lets map this to a linear combination of @xmath58 s using @xmath63 : @xmath99 if @xmath97 therefore corresponds to a zero mode , it gets annihilated by this map . only the finite frequency modes therefore actually pass from the @xmath6 eigenmode problem to the @xmath58 eigenmode problem . however , for the finite mode case , notice that the mapped observable is a linear combination whose coefficients obey @xmath100 so it is a ( left ) eigenvector of the @xmath58 differential equation with the same eigenvalue it had in the @xmath6 differential equation before it was mapped . mapped observables therefore carry their eigenvalue with them . extending this study to each of the four differential equations and breaking their set of eigenvectors into two groups , the finite and the zero eigenvalue modes , we have constructed fig . [ fig : phononqmap ] that graphically depicts all the different relationships . in it an arrow represents a function mapping the set of eigenmodes at its tail to the set of eigenmodes at its head . if two different paths exist between any two sets , one going from one set to the other and the other going the reverse direction , then the two sets are isomorphic and in particular have the same number of eigenmodes . in this way , we see that the number of finite energy eigenmodes are the same for all differential equations , just their zero modes are different . '' is the supersymmetry map . notice @xmath63 does not map zero mode eigenstates from @xmath6 to @xmath58 and from @xmath58 to @xmath17 . since all arrows represent maps of unique elements to unique elements , if a path of arrows exists from one set of modes to another and another path exists from that set of modes back to the first then the two sets of modes are isomorphic . for finite modes , any two sets of eigenstates have such paths relating them so they are all isomorphic to each other . for zero modes , only the @xmath17 , @xmath59 and @xmath6 sets of eigenstates are isomorphic . the @xmath58 set is different and hence the phonino problem has a different set of zero modes from the phonon problem . note this figure follows the rules of an olog@xcite.,title="fig : " ] '' is the supersymmetry map . notice @xmath63 does not map zero mode eigenstates from @xmath6 to @xmath58 and from @xmath58 to @xmath17 . since all arrows represent maps of unique elements to unique elements , if a path of arrows exists from one set of modes to another and another path exists from that set of modes back to the first then the two sets of modes are isomorphic . for finite modes , any two sets of eigenstates have such paths relating them so they are all isomorphic to each other . for zero modes , only the @xmath17 , @xmath59 and @xmath6 sets of eigenstates are isomorphic . the @xmath58 set is different and hence the phonino problem has a different set of zero modes from the phonon problem . note this figure follows the rules of an olog@xcite.,title="fig : " ] now , with the above analysis of the eigenmodes , we are in a position to understand the topological index associated with a supersymmetric theory . as pointed out by witten ( and often called the `` witten index '' ) the total number of boson modes minus the total number of fermion modes ( @xmath101 where @xmath102 for a boson mode , and @xmath103 for a fermion mode ) is a topological invariant . applying this to the single particle sector we have @xmath104 but if we break down the count of zero modes into the four groups of two types of modes , we have @xmath105 where we used fig . [ fig : phononqmap ] to establish that the number of finite modes @xmath106 is the same between all modes types while @xmath107 is the same only between @xmath17 , @xmath59 and @xmath6 and is in general different from the number of @xmath58 zero modes , @xmath108 . to derive these results , we heavily relied on the positive definiteness of @xmath9 and @xmath36 for if this were not the case , more modes could fail to be mapped by @xmath63 . so the witten index of supersymmetry , when applied to the single particle sector valid for the linearized theory , is identical to the maxwell - calladine index but its derivation is directly built on eigenvalue problems rather than directly on the properties of the matrix @xmath16 . the witten index is useful in a supersymmetric theory because it dictates whether there must be a zero mode . if @xmath109 there must be @xmath110 bosonic ( phonon ) zero modes . if @xmath111 there must be @xmath112 fermionic ( phonino ) zero modes which here must be self stress modes . so long as @xmath113 then , supersymmetry can exist in the `` ground state '' for there is then a zero mode that satisfies @xmath114 ( in the quantum language ) . however , if @xmath47 then without explicitly breaking supersymmetry one can remove all zero modes . then there are no states satisfying @xmath114 and supersymmetry is spontaneously broken in the ground state . remarkably , an isostatic lattice has both @xmath47 for periodic boundary conditions and all zero modes gapped ( @xmath115 ) . so isostatic lattices are an example spontaneous susy breaking . finally , we need to comment on the question of how the above supersymmetry is related to the quantum mechanics - like supersymmetry of the square rooted equations of motion ( eq . [ eq : sqrtedeom ] ) pointed out by kane and lubensky . this symmetry is a supersymmetry between @xmath59 modes and @xmath58 modes ( both of which are fermionic here ) and has an associated topological index which is numerically the same index as the single particle witten index discussed above and the maxwell - calladine index . so the phonon problem appears to have two different supersymmetries , one at the full phase space level that we discuss above and the other at the level of the equations of motion of the linearized fermions that related different fermion modes to each other ( quantum mechanics - like susy need not always relate bosonic modes to fermionic modes ) . in summary , the supersymmetry model has brought all mathematical objects mentioned so far under one umbrella . the supersymmetric theory has the phonon hamiltoinan , the two metrics @xmath9 and @xmath36 play a central role as they determine the commutation relations of the fermions , the phonino hamiltonian is determined by the constraint matrix @xmath53 and its corresponding equations of motion is the square - rooted hamiltonian @xmath40 . finally , the topological index is determined by the supersymmetry relations between the various eigenmodes . lastly , it suggests that the specialness of the balls and springs description is the requirement that it is supersymmetric . if we break this specialness , if we explicitly break the supersymmetry , then most of the mathematical objects contained in the theory loose their meaning . in this way , i argue that the topological structure of isostatic lattices identified by kane and lubensky is susy protected . the previous results show that models of phonons built on constraint functions like the extension @xmath4 of a spring from its equilibrium length have a topological index tied to their equations of motion by a supersymmetric relationship with phoninos . this index guarantees the existence of zero modes from topology ( even in the absence of a continuous symmetry ) much like goldstone s theorem guarantees the existence of zero modes from spontaneous symmetry breaking . here we discuss the question of to what extent this topological property is robust . for the most part , we worked in the classical limit where phonons are actually lattice vibrations . if we quantize it , then we convert poisson brackets to commutators or anticommutators such as @xmath116 = i\delta_{i\alpha}^{j\beta}$ ] and @xmath117 . quantizing the supersymmetric charge @xmath118 then gives the supersymmetric hamiltonian @xmath119 this is identical in form to the classical case . a similar correspondence between the non - linear forms of the classical and quantum model also holds . so the main results of this paper hold in both the classical and quantum forms and are not restricted to semi - classical or classical regimes . since the metrics @xmath36 and @xmath9 entering the phonon problem are not present in the respective topological indices , the topological protection holds if these undergo a distortion so long as they remain positive definite matrices . in particular , the topological index holds if we stretch and distort the lattice as was used extensively in the discovery of weyl phonon modes in ref . . let us first address this question for the non - linear supersymmetry models before returning to the problem of purely non - linear phonons in the absence of phoninos . in the presence of non - linearities , the eigenmodes of the system are no longer linear combinations of single particle observables . instead , they become a linear combination of all multiparticle observables . as a result , the witten index restricted to the single particle sector looses its meaning but the full index @xmath120 still holds because @xmath63 still maps multiparticle bosonic observables to a multiparticle fermionic observables . to get a sence of this full set of observables , we have considered the two - particle sector . this consists of bosonic observables of types @xmath121 , @xmath122 , @xmath123 , @xmath124 , @xmath125 and @xmath126 and fermionic observables of types @xmath127 , @xmath128 , @xmath129 and @xmath130 . counting all of these two different two particle observables , noting that squares of grassmann observables vanish ( i.e. @xmath131 ) i obtain after simplifying @xmath132 if @xmath133 as in the isostatic lattice case , we see this also vanishes . since the index for the one - particle and two - particle sectors vanishes , likely the total sum over all sectors relevant for the many body non - linear problem also vanishes . so the topological protection remains in effect for the full non - linear supersymmetric problem . with this technical extension , we see that the non - linear models are also topological . since they consist now of majorana fermions coupled to phonons , they could in principle be realized in a carefully engineered superconductor . in practice , however , they may prove to be more useful as a means of gaining theoretical insight into `` symmetry protected topological order '' in the presence of interacting fermions which does not make use of k - theory@xcite , chern - simons quantum field theory@xcite or supercohomology@xcite . unfortunately , the salient feature of the non - linear problem is that the phonons are now coupled to the phoninos . we can no longer think about the added fermionic degrees of freedom as a device to construct a topological index . so , if the fermions are absent the supersymmetry is lost and with it the topological protection . this statement , however , is about the microscopic physics . it may very well be that the full non - linear phonon problem has the same low energy long distance physics as the corresponding supersymmetric non - linear phonon . after all , it is the physics close to the ground state which we used to build the model of balls and springs from constraint functions . it is therefore possible that the susy breaking terms are irrelevant in the renormalization group sense . if so , the topological protection may nevertheless emerge in the low energy limit . the models discussed in this paper are all `` toy '' models in that they correspond to useful simplifications of the actual phonon microscopic physics but in any real system , there will be perturbations that are not of the simple form . for example , the general quadratic phonon potential @xmath134 is not that of balls and springs ( @xmath135 ) . the supersymmetry can therefore be thought of as a symmetry emerging from the ` toy - ness ' of the model . we can speculate on what this might mean . the extent to which the toy model captures the full physics of a real material is likely the extent to which the supersymmetry will be obeyed . in particular , if supersymmetry demands gappless phonon edge modes in a system with gapped bulk phonon modes , these edge modes would likely be gapped by perturbations violating the toy - like property of balls and springs and the size of the gap is likely related to the strength of these non - toy - like perturbations . we can extend the above phonon discussion to magnons ( or other similar systems ) by recognizing its essential ingredient : constraints . the phonon problem is built on the local - in - configuration space condition that the extensions @xmath4 vanish in the ground state . since a ground state also requires a vanishing momenta @xmath6 , we see that the phonon problem is a quadratic form in the set of local - in - phase space constraints on the ground state . to extend to magnons , we therefore need models built around local - in - phase space constraints that define a ground state magnetic ordering pattern . perhaps the simplest example of such a constraint in a magnetic system is the vanishing total ( classical ) spin of a nearest neighbor bond in the nel state of the square lattice antiferromagnet . each nearest neighbor bond @xmath3 has a pair of sites @xmath136 that satisfies @xmath137 . here each @xmath138 is a dynamical spin vector satisfying angular momentum poisson brackets @xmath139 and lives in a spherical @xmath140-dimensional phase space . such a condition is directly analogous to a motionless unstreatched spring with vanishing extension @xmath15 and momenta @xmath14 , @xmath141 . so we can follow our phonon discussion and use the condition @xmath142 to write down the susy charge @xmath143 and susy hamiltonian @xmath144 choosing @xmath145 a positive definite symmetric matrix makes the first term a typical quadratic spin model with @xmath142 for a ground state ( e.g. the case @xmath146 is just the heisenberg model ) . this makes the @xmath147 grassmann variables obeying a clifford algebra just like we found for the phonon problem . additionally , @xmath148 follows from @xmath149 above and defines the fermion hamiltonian again as dirac s constraint matrix . we can then expand about a ground state to linear order using @xmath150 with @xmath151 the phase space coordinates representing deviations of the spin on site @xmath8 from its ground state value . this renders @xmath152 constant and decouples the two terms in @xmath78 . linearizing allows us to demote the magninos described by @xmath147 to fictitious particles whose use is to reveal the implications of the supersymmetry just like the case of the phoninos . however , the equations of motion of these magninos does not give us the `` square - rooted '' hamiltonain for it is @xmath153 which , when squared , is not the magnon equations of motion . so the structure of the square - rooted hamiltonian and its associated quantum mechanics - like supersymmetry does nt apply here . thus , by starting from the ground state condition @xmath142 we obtain the same structure as the phonon problem which presumably has a witten index @xmath110 that can tell us about its zero modes without solving the equations of motion . the above nel state bond constraint @xmath154 is an application of the susy phonon problem to unfrustrated antiferromagnets . it readily generalizes to the ferromagnetic case which has a vanishing difference of spin vectors on each bond ( @xmath155 ) provided we draw arrows on each bond to specify which of the two sites of nearest neighbor bond @xmath3 enters @xmath154 with a negative sign . it also readily generalizes to other bi - partite lattices such as the cubic lattice and the honeycombe lattice . writing down the topological index expected for these generalizations will therefore allow us to determine if an analog of isostatic phonons exists and also how we may use it to study the loss of magnetic order . for each case , we can guess the form of the topological index @xmath156 the last line here is not a guess , it follows from the previous line by the fundamental theorem of linear algebra as before . presumably , if @xmath157 , there must be a finite number of magnon zero modes and if it is negative a finite number of magnino modes . we will discuss the meaning and correctness of this statement below . but assuming it is correct for the present discussion , we see that essentially all unfrustrated magnets have @xmath158 . only dimerized systems which have two sites for each bond or trimerized systems with _ two bonds _ for each trimer have non - negative @xmath159 ( see fig . [ fig : unfrustratednu ] ) . the best case with connected sites is the one dimensional chain with @xmath160 and @xmath161 for periodic boundary conditions . a similar best case holds for bond percolation on the square lattice where at threshold @xmath162@xcite . so for unfrustrated magnets , the topological structure imposed by supersymmetry does nt demand any magnon zero modes . becomes increasingly negative with increasing chain length going negative for four or more spins , essentially all unfrustrated magnetic systems will have a negative topological index . this explains the general robustness of magnetic order in many magnetic materials . ] the above derivation of the magnetic maxwell - calladine index was mostly a guess . to see that it is correct , we need to work out the supersymmetry relations between eigenmodes and the witten index as we did for the phonon problem . this is a more complicated problem than the phonon case due to the existence of two types of zero modes@xcite ( canonical and gauge ) . but this just requires a little more organization and the following proceeds just like the phonon case . the equations of motion for both bosonic and fermionic problems are @xmath163 where @xmath164 is the poisson bracket tensor with @xmath165 the two dimensional levi - civita tensor and we expanded the constraint matrix to leading order ( where @xmath166 ) . lets identify first the types of bosonic eigenmodes , then the types of fermionic eigenmodes and then see how they map into each other by sypersymmetry . for the @xmath167 equations of motion , we can break up the set of eigenmode observables @xmath168 into four types of modes . 1 . finite frequency modes obeying @xmath169 with @xmath170 ; 2 . gauge zero modes that are eigenmodes with @xmath98 but whose observable @xmath171 commutes with all other observables corresponding to other @xmath98 eigenmodes . these obey @xmath172 and @xmath173 where @xmath174 is any vector satisfying @xmath175 . 3 . canonical zero modes that are eigenmodes with @xmath98 that are not gauge zero modes . these obey @xmath172 and @xmath176 for some @xmath174 satisfying @xmath175 . non - eigenmodes of the equations of motion that can arise because the eigenvalue problem here is a non - symmetric matrix that does not guarantee a complete set of eigenvectors . these turn out to be the modes conjugate to the gauge zero modes . we will therefore call them `` conjugate gauge modes '' . these modes are best understood by choosing a canonical basis that pairs them up into position @xmath177 and momentum @xmath178 observables . then each pair has hamiltonian @xmath179 in this light , finite frequency modes have @xmath180 , @xmath181 , canonical zero modes have @xmath182 , @xmath183 and @xmath184 and gauge modes have @xmath185 , @xmath183 . for the @xmath147 equations of motion , we can break up the set of eigenmodes @xmath186 into : * finite frequency modes obeying @xmath187 ; * self field modes obeying @xmath188 ; * magnino gauge modes that do not satisfy @xmath189 but are eigenmodes with @xmath98 . self field modes are modes that correspond to a choice of local magnetic fields on the bonds @xmath190 that contribute no zeeman energy @xmath191 they identify a redundancy in the set of functions @xmath154 when viewing them as constraints on the ground state configuration . the `` magnino gauge modes '' are not gauge modes in the sense of zero modes whose conjugate is not a zero mode . their name here comes from their use in counting gauge modes : dirac found@xcite there is a gauge mode for every eigenvector in the nullspace of the constraint matrix that does not correspond to a redundant constraint ( i.e. is not self field mode ) . now with all these modes identified , we can pass them through the supersymmetry map @xmath86 and see how they are related by supersymmetry . the result is presented in fig . [ fig : fmagnonsqmap ] . since no two eigenmodes mapped by @xmath86 maps to the same eigenmode , we must have : * the number of finite frequency magnon modes @xmath192 is equal to the number of finite frequency magnino modes ; * the number of magnon gauge modes @xmath193 is equal to the number of magnino `` gauge '' modes . in addition , we see that the supersymmetry map gives us a direct explanation of dirac mode counting : magnino gauge modes map to magnon gauge modes under supersymmetry and the conjugate to this magnon gauge mode maps back to the magnino gauge modes . this defines an isomorphism between magnon and magnino gauge modes and so the number of each must be the same . the supersymmetry map also verifies our guess for the single particle witten index / magnetic maxwell - calladine index for the above conditions on the number of modes in each set implies the difference between the number of magnon and magnino modes is @xmath194 where @xmath195 is half the number of zero modes with a conjugate pair that is also a zero mode and @xmath196 is the number of self field modes . and so our guess of eq . [ eq : aindex ] is indeed correct with a positive @xmath110 demanding @xmath197 and a negative @xmath110 demanding @xmath198 . we can also determine the topological another way . we see from the above discussion that @xmath199 ( dimension of the constraint surface @xmath200 in phase space ) and @xmath201 since these null spaces define the modes that do not get mapped under supersymmetry . hence , the witten index in the magnon problem also obeys @xmath202 . , this figure is also an olog@xcite . ] so our guess for the topological index is correct but this in turn implies that unfrustrated magnons are always far away from the analog of mechanical collapse expected for @xmath45 ( i.e. they all have a negative @xmath110 ) . so , to explore the physics of the kane - lubensky theory in magnetic systems , we need to generalize beyond unfrustrated magnetic systems . but before doing so , we turn to an illustrative example of the above magnon and magnino eigenvalue problems : the simple ferromagnet on the square lattice . we can write the hamiltonian for the square lattice ferromagnet heisenberg model as @xmath203 where @xmath146 , @xmath3 labels nearest neighbor bonds @xmath204 and @xmath24 the spin vector components . the ground state therefore satisfies @xmath205 on each horizontal bond and @xmath206 on each vertical bond . up to global spin rotations , this uniquely selects the ground state to be the uniform state . to linearize around a ground state , let s choose the case @xmath207 , @xmath208 . expressing the spin vectors using canonical coordinates @xmath209 with @xmath210 so that @xmath211 we can linearize by expanding in @xmath212 and @xmath213 . this gives @xmath214 . similarly , we can expand the constraint @xmath215 and identify @xmath216 and @xmath217 with @xmath218 indexing to which canonical variable @xmath167 refers . so we find @xmath219 where the sign depends on which spin @xmath220 refers to in the expression @xmath221 . now since the ground state is periodic with the lattice , we may fourier transform . to do this , let @xmath222 with @xmath223 or @xmath224 denoting the center of each bond and @xmath225 denoting the site of the lattice . then @xmath226 and we see its translational symmetry by just shifting both @xmath227 and @xmath228 by @xmath229 and for some lattice vector @xmath229 . we can then expand @xmath230 in fourier components by @xmath231 where @xmath232 is the number of unit cells . this introduces the 6x2 matrix @xmath233 and allows us to express the magnon hamiltonian as @xmath234 multiplying the matrices out for the nearest neighbor heisenberg model ( @xmath235 a 6x6 identity matrix ) we find @xmath236 where @xmath237 is the 2x2 identity matrix giving a quadratic dispersion for small @xmath238 as expected . now we can identify mode types as finite frequency , magnon gauge , conjugate to magnon gauge and magnon canonical zero mode . at @xmath239 we find a zero mode with @xmath240 . it is therefore of the form of eq . [ eq : modeform ] with @xmath241 , @xmath242 and @xmath243 . hence it is a canonical zero mode . for finite @xmath238 , we see that hamiltonian is of the form of eq . [ eq : modeform ] with @xmath244 and @xmath245 finite . hence these are finite frequency magnon modes . there are unfortunately no magnon gauge modes are conjugate gauge modes to illustrate this point . the magnino hamiltonian takes a similar form . it is @xmath246 where @xmath247 is the antisymmetric poisson bracket tensor . now @xmath248 is a @xmath249 antihermetian matrix : @xmath250 choosing @xmath235 a 6x6 identity matrix so that @xmath251 obey canonical commutation relations then allows us to diagonalize this matrix . the result is four zero eigenvalues , @xmath252 and @xmath253 . hence the finite frequency modes have exactly the same eigenvalues . again we can distinguish the different mode types though here we havent expressed a simple canonical form for the diagonalized fermion hamiltonian . we do nt have eq . [ eq : modeform ] to help guide us . but we can work directly with the definitions of the modes . clearly finite frequency modes magnino modes correspond to the eigenvalues @xmath252 and @xmath253 for @xmath254 and there are exactly the same number of these as finite frequency magnon modes . the four zero eigenvalues and the case @xmath255 fall into the null space of @xmath256 and are `` self field modes '' . there are no zero frequency modes not in the null space of @xmath257 so there are no `` magnino gauge modes '' consistent with the absence of magnon gauge modes discuss above . perhaps one way to overcome the ubiquitous negative @xmath110 of unfrustrated magnets is to generalize to other degrees of freedom . for this to improve upon the unfrustrated case , we need to increase the number of degrees of freedom on each site compared to the number of constraints on each bond . this seems difficult for magnetic systems for each site naturally has its own constraints ( such as unit length spin vectors ) in addition to the bond constraints . for example , consider @xmath258 quantum rotors each characterized by @xmath258 spin unit vector @xmath259 , ( @xmath260 , @xmath261 ) and angular momenta @xmath262 conjugate to @xmath263 ( see sachdev s book@xcite ) . in this case @xmath264 and @xmath265 in the ground state so we can define @xmath266 . such a @xmath63 leads to @xmath267 with fermions again decoupling in the linearized limit . but now our guess for the topological index is @xmath268 which is typically even more negative than @xmath159 ( and the situation does nt improve for @xmath269 rotors ) . but notice that here we kept the same number of phase space degrees of freedom on each site and just transferred one of the bond constraints to a site constraint . thus it is even more difficult to introduce magnon zero modes for quantum rotors than for quantum spins and they do not seem to be a viable solution . let us then turn to changing the constraint functions entirely . lets replace the vanishing bond constraint @xmath142 with a vanishing total spin on a plaquette constraint @xmath270 . such a constraint arises in the nearest neighbor triangular , kagome , pyrochlore antiferromagnets and the @xmath271-@xmath272 square lattice antiferromagnet with @xmath0 . all of these cases fall in to the category of frustrated magnets . for example , the triangular lattice heisenberg model can be written @xmath273 where @xmath274 labels each triangle of the triangular lattice and the @xmath271-@xmath272 square lattice antiferromagnet at @xmath275 can be written @xmath276 we can again guess the topological index . here it is @xmath277 where @xmath278 is the number of plaquette constraints ( and fermion grassmann numbers ) . this gives for @xmath232 unit cells and periodic boundary conditions @xmath279 for the triangular lattice ( 2 triangles , 1 site in unit cell ) , @xmath280 for the @xmath281 square lattice case ( 1 square , 1 site in unit cell ) , @xmath282 for the kagome case ( 2 triangles , 3 sites in unit cell ) and @xmath283 for the pyrochlore lattice ( 2 tetrahedrons , 4 sites in unit cell ) . so frustration enables a wide variety of models to explore the impact of the full range of @xmath110 values on the ordering tendencies of the spins . we next work out the special @xmath47 kagome case as a concrete example and use it to clarify a number of the above general statements . on the kagome lattice ( and indeed many highly geometrically frustrated magnets@xcite ) the hamiltonian can be written @xmath284 where @xmath285 is the total spin on triangle @xmath286 . for the case , @xmath287 , this model reduces to the nearest neighbor heisenberg model on a lattice of corner sharing triangles such as the kagome lattice and hyperkagome lattice . we will therefore assume @xmath288 has positive definite eigenvalues and defines a metric . this model has the feature that the ground states all satisfy the local condition @xmath200 . using this constraint , we can again write down a supersymmetric charge @xmath289 and compute the corresponding supersymmetric hamiltonian @xmath290 so we need to choose @xmath291 to make the first term @xmath292 . the second term is the magnino hamiltonian and like the phonino hamiltonian is determined by the constraint matrix ( @xmath293 ) . hence , we obtain the same structure as we did for the unfrustrated case . we just replace the index @xmath3 with @xmath274 . we can now compute the topological index for kagome antiferromagnets . after linearizing as before the analog of the magnetic maxwell - calladine index is now @xmath294 this is different from the unfrustrated case only in replacing @xmath26 the number of bonds with @xmath295 the number of triangles . for periodic boundary conditions , we see that @xmath296 where @xmath232 is the number of unit cells considered , @xmath297 is the number of sites within a unit cell and @xmath298 is the number of triangles within a unit cell . we additionally see this holds for other kagome lattices like the hyperkagome lattice of na@xmath299ir@xmath2o@xmath300@xcite in three dimensions which has @xmath301 , @xmath302 . so the kagome family of lattices all have @xmath47 . given that @xmath47 for periodic boundary conditions , lets see what the index demands for various open boundary conditions . for the two finite kagome clusters shown in fig . [ fig : kagomeedgestates ] , we have @xmath303 and @xmath304 . so either type of boundary condition will have @xmath305 and demand zero modes . presumably , this holds for any boundary condition that still allows us to write the hamiltonian in the general form of eq . [ eq : hfrustratedmagnons ] . however , other boundary conditions may include a bond constraint function @xmath154 in place of a triangle constraint function @xmath306 and we would need to rethink the problem and be sure these constraints really do define ground state configurations ( i.e. it could be that there are no spin configurations that satisfy all the constraints ) . in these cases , it may be that @xmath111 for such boundaries given that we found bond constraints negatively contribute to @xmath110 so that @xmath110 does nt demand any zero modes . hence , it may depend on the nature of the boundary conditions as to whether zero modes are expected . however , there exist boundary conditions where the topological index will demand the existence of zero modes even in the absence of any spin rotational symmetry breaking ( as it does in systems with @xmath45 for periodic boundary conditions such as the nearest neighbor pyrochlore antiferromagnet ) . . a ) has dangling triangles and @xmath307 , b ) has no dangling triangles and @xmath308 . it appears that @xmath45 for any open boundary conditions and only reaches zero for periodic boundary conditions . thus we can always gap out all bulk modes with a suitable choice of the matrix @xmath309 but not so at the edge . a magnon gauge or canonical zero mode must exist on the edge so long as the magnon hamiltonian is of the form eq . [ eq : hfrustratedmagnons].,title="fig : " ] . a ) has dangling triangles and @xmath307 , b ) has no dangling triangles and @xmath308 . it appears that @xmath45 for any open boundary conditions and only reaches zero for periodic boundary conditions . thus we can always gap out all bulk modes with a suitable choice of the matrix @xmath309 but not so at the edge . a magnon gauge or canonical zero mode must exist on the edge so long as the magnon hamiltonian is of the form eq . [ eq : hfrustratedmagnons].,title="fig : " ] the above results show that indeed we can extend the supersymmetry of the linearized phonon problem to both unfrustrated and frustrated magnons . in addition , we find that the kagome family is an analog of `` maxwell lattices''@xcite . it has @xmath47 for periodic boundary conditions and @xmath45 at least for certain open boundary conditions . if we gap out all modes , with supersymmetry allowed perturbations , we would then have a magnetic analog of an isostatic lattice ( a maxwell lattice with no zero modes ) . let us then turn to illuminating some of the new physics that is transported from the theory of phonons into the theory of magnetism by our formal developments . perhaps most striking is the existence of a new phase of magnetism which is on the verge of magnetic collapse that we will here refer to as isostatic magnetism . we can understand the existence of this phase in simple terms as shown in fig . [ fig : isomagnetism ] . first consider the three different mechanical structures , the square , the square with one diagonal beam and the square with two diagonal beams we see three different forms of rigidity : floppy , fragile , and firm . the square is floppy because it has one zero mode . the square with one diagonal brace is fragile because though it is rigid , if you cut / break any of the braces , it becomes floppy . the square with two diagonal braces is firm because it remains rigid if you cut / break any one of the braces . now consider the three magnetic structures , four spins on a square , three spins on a triangle and four spins on a tetrahedron in the figure . in the square case , we subject the spins to @xmath310 for @xmath8 , @xmath311 nearest neighbor , @xmath312 indexes bonds and @xmath313 . in the triangle and tetrahedron , we merely impose that the total spin of the object is zero . the square is the analog of the firm mechanical structure : after imposing @xmath314 on three of its four sides , the fourth side is redundant . removing one bond would not destroy the nel order of its ground state . the triangle is the example of the fragile mechanical structure . all three constraints @xmath315 that define its ground state are needed to set the spins to be 120@xmath316 apart ( up to global spin rotations ) . if we drop one of these three constraints , it gains a zero mode . finally , the tetrahedron is the example of the floppy mechanical structure . we can choose @xmath317 and @xmath318 independently and still @xmath319 . in this light , the new isostatic magnetic order that we can transport from the theory of phonons , is an order like the triangle case : if we were to drop any one term in the hamiltonian that imposes a constraint on the ground state , there is no redundancy and the system becomes floppy . isostatic magnetism is fragile . remarkably , since we have already the other two cases , `` floppy '' magnetism in the form of highly frustrated magnets and `` firm '' magnetism for most other magnetic systems , it may just be a matter of time to identify isostatic magnetism at least within some temperature range . one group of materials that may contain this novel magnetism is the family of distorted kagome antiferromagnets . an undistorted ideal kagome antiferromagnet has @xmath47 but many zero modes . adding distortions could then lift the ground state degeneracy without changing @xmath110 since distortions do not change the topology of the lattice network . since distortioned kagome lattices are much more common than ideal kagome lattices , there may already be a number of such systems that are isostatic magnets . applied to each neighboring spins as shown with black arrows ( equivalent to a uniform field on each spins ) . ] another physical phenomena we can transport from the phonon theory is the existence of self field modes that are the analog of self stress modes . the square with two diagonal beams of fig . [ fig : isomagnetism ] has such a self stress mode because of its redundancy of constraints . if one puts the diagonal beams in tension @xmath320 , then one can place the horizontal and vertical beams in tension @xmath321 and no force will be exerted on the pins at is vertices . similarly , one can add a `` bond magnetic field '' @xmath322 via a term in the hamiltonian @xmath323 and there must exist @xmath322 such that no net field is experienced by a spin to linear order in the displacements of the spins from their ground states . we know actually that there are at least @xmath324 of these from the topological index which here is @xmath325 . because we include spin rotation invariance , there are two zero modes and so @xmath326 demands @xmath327 . if @xmath3 labels bonds cyclically around the square , three of these six modes span @xmath328 . these produce @xmath329 non - linearly ( i.e. independent of setting @xmath330 ) . of the remaining three , one is the case @xmath331 assuming we choose a neel ground state in the @xmath332-direction . this case produces @xmath333 only after we linearize the spins @xmath334 around their magnetic ground state by setting @xmath330 . to non - linear order in the deviations of the spins from their ground state , this case will produce a non - zero energy @xmath335 . the remaining two self field modes behave similarly . so self field modes exist and extend our understanding of the theory of magnetism . actually , we can understand what an experimentalist could achieve with a knowledge of self field modes in analogy with the self stress modes . self stress modes have been studied at least as far back as eshelby s 1957 paper@xcite who emphasized the engineering importance of self stresses in an elastic medium with inhomogeneities . more recently they have been understood as a means of selective mechanical failure of metamaterials@xcite . in a @xmath47 lattice with one self stress mode and one zero mode located in different regions , a load was shown to cause mechanical failure exactly at the self stress mode bonds . similarly , an applied field that corresponds to a self field mode should cause a spin flip . indeed , in the case above for the square with a field @xmath336 applied , we see that the spins will flip due to the non - zero energy @xmath337 caused by the non - linearities . the other two self field modes not discussed in detail similarly cause a spin flip , just not all spins flip in those cases . so such self field modes correspond to spin flip transitions and their identification in the constraint matrix eigenvalue problem may enable the discovery of a variety of such spin flip transitions . in this paper , we have recognized a supersymmetry in the structure of hamiltonians that energetically impose local constraints on their ground states . this allowed us to formally connect recent development in the theory of phonons to a wider range of systems including frustrated and unfrustrated magnets . the connection ultimately arose from the similar role of the vanishing extension of springs @xmath15 of phonon ground states to the vanishing total spin on a bond in unfrustrated magnets and vanishing total spin on a triangle , square , tetrahedron or other local simplex in frustrated magnetic systems . it also highlights the role of dirac s constraint matrix as a central pillar of these recent developments as well as proves that a topological index exists in all these systems . lets end our discussion of these results with some final comments . remarkably , this supersymmetry identifies maxwell s counting of degrees of freedom @xmath338 , where @xmath339 is the total number of degrees of freedom and @xmath340 is the total number of local constraints imposed by the hamiltonian ( i.e. possibly a redundant set ) , to the witten index @xmath110 associated with the supersymmetry . this maxwell counting , though long used for balls and springs and other mechanical structures , was proposed by moessner and chalker as a means of characterizing highly frustrated antiferromagnets@xcite . here we have connected their work with the actual number of zero modes through recognizing it as the topological index . lattice where dark blue are the mn atoms with magnetic moments and yellow atoms are the ir atoms without magnetic moments . one view of the lattice is through the 111 planes for the blue magnetic atoms . they form a kagome lattice . another view is achieved by considering how each triangle shares a side with another triangle in the octahedron . this view suggests irmn@xmath2 is composed of side sharing triangles in 3 dimensions and is therefore more similar to the triangular lattice antiferromagnet than the kagome antiferromagnet . ] the use of analogs of maxwell constraints in magnetism and the associated topological index provides a simple picture of the strength of magnetism in a system . to see this in practice , consider the question of the origin of the industrial strength of antiferromagnetism in irmn@xmath2 . this material magnetically orders in a 120@xmath341 coplanar state at 960 k@xcite but is composed of kagome planes with a dominant isotropic heisenberg spin exchange@xcite that would appear to frustrated the magnetism@xcite . how then can the maxwell - like constraints advanced by this paper explain such industrial strength magnetism ? the answer lies in the many different kagome planes . they all interconnect as shown in fig . [ fig : irmn3 ] . indeed , the lattice is better thought of as side - sharing triangles in three dimensions instead of layers of kagome planes . with eight triangles in the unit cell , the dominant heisenberg term in its hamiltonian can be written @xmath342 where @xmath343 is the total spin on each triangle . we therefore have a topological index @xmath344 ( since there are 8/3 triangles per spin ) . we can contrast this to say an unfrustrated antiferromagnet on the cubic lattice with three bonds in the unit cell and @xmath345 . but upon closer inspection , we should be careful with such a comparison . in irmn@xmath2 , nearest neighbor heisenberg exchange still has a sub - extensive number of zero modes because the side sharing triangles form corner sharing octehedra . so its @xmath346 is likely lower than the unfrustrated antiferromagnet with the same @xmath347 . nevertheless , this effect is subextensive so it is unlikely to be a big effect . therefore , the energetic constraints imposed by the dominant heisenberg exchange in irmn@xmath2 appear to be nearly as redundant as an unfrustrated antiferromagnet on a cubic lattice suggesting great industrial strength even though it is composed of kagome planes . but not only firm magnetism may prove useful for practical applications . fragile isostatic magnetism may prove to be easily manipulated for special design purposes . it may play a role for magnetic systems similar to the role isostatic phonons have played in the design of metameterials . due to the absence of redundant constraints , isostatic lattices can have particular weaknesses that can enable their manipulation . perhaps even metamaterials designs that exploit this could carry over to magnetic materials . for example , one lattice was designed to have selective mechanical failure in ref . where just a local set of bonds associated with a state of self stress broke down under a load . the magnetic analog of this could correspond to a material design with a selective spin flip of a local subset of spins achieved with a global magnetic field . in this way , isostatic magnetism could prove to have an important role in the design of magnetic materials for applications . and given packing fraction @xmath348 . here the isostatic point highlighted by the green dot at the top right occurs for frictionless grains . b ) the phase diagram for quantum disordering a frustrated antiferromagnet obtained in a semi - classical field theory@xcite . the x - axis is a dimensionless temperature and the y - axis a coupling measuring the strength of quantum fluctuations . c ) classical phase diagram of the loss of magnetic order based on the results of this paper . here again the green dot represents the isostatic point . notice in all three cases , a special point exists in the phase diagram that sheds light on the rest of the phase diagram . it would be interesting to understand the relationship between these three points.,title="fig : " ] and given packing fraction @xmath348 . here the isostatic point highlighted by the green dot at the top right occurs for frictionless grains . b ) the phase diagram for quantum disordering a frustrated antiferromagnet obtained in a semi - classical field theory@xcite . the x - axis is a dimensionless temperature and the y - axis a coupling measuring the strength of quantum fluctuations . c ) classical phase diagram of the loss of magnetic order based on the results of this paper . here again the green dot represents the isostatic point . notice in all three cases , a special point exists in the phase diagram that sheds light on the rest of the phase diagram . it would be interesting to understand the relationship between these three points.,title="fig : " ] and given packing fraction @xmath348 . here the isostatic point highlighted by the green dot at the top right occurs for frictionless grains . b ) the phase diagram for quantum disordering a frustrated antiferromagnet obtained in a semi - classical field theory@xcite . the x - axis is a dimensionless temperature and the y - axis a coupling measuring the strength of quantum fluctuations . c ) classical phase diagram of the loss of magnetic order based on the results of this paper . here again the green dot represents the isostatic point . notice in all three cases , a special point exists in the phase diagram that sheds light on the rest of the phase diagram . it would be interesting to understand the relationship between these three points.,title="fig : " ] but isostatic magnetism may also extend our theoretical understanding of magnetic materials . there is likely a separate topological property not discussed in this paper . since the magninos are free fermions , and the ideal isostatic magnetism state is a gapped state , it must fit into our categorization of topological band theory@xcite much like the phoninos where shown to have a topological band theory by kane and lubensky ( though they did nt call them phoninos ) . it is the supersymmetry established by this paper that enables physics that applies to free fermions ( including majorana fermions ) to also apply to phonons and magnons . this is one of the benefits of generalizing kane and lubensky s theory to a broader class of systems . the results of this paper may also shed light on a long standing problem in the field of magnetism beyond band theory : how to melt magnetic order . the connection established by our formal developments shows explicitly that the loss of rigidity in a solid@xcite , the jamming transition@xcite and the loss of magnetic order due to frustration[ramirez1994,moessner1998a , moessner1998b ] may all be different aspects of the same problem ( see fig . [ fig : jamming ] ) . each of these systems are built around constraints , have a topological index @xmath110 , and have isostatic order with @xmath47 that we curiously identify here with spontaneous susy breaking . so from this perspective , isostatic magnetism is at a critical point between magnetic order and a classical spin liquid . it would be interesting to understand how this critical point is related to the quantum melting of antiferromagnetic order through a quantum critical point@xcite particularly given theories of this quantum phase transition that are semi - classical in nature . * practical use in the design of magnetic materials via the topological index , zero modes and self field modes * a novel application of topological band theory of fermions to magnons * insight into the mechanism of quantum spin liquids and other exotic phases of magnetism . 47ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty http://dx.doi.org/10.1038/nphys2835 10.1038/nphys2835 http://www.nature.com/nphys/journal/v10/n1/abs/nphys2835.html{#}supplementary-information [ * * , ( ) ] link:\doibase 10.1016/0020 - 7683(78)90052 - 5 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/15/4/043043 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.103.248101 [ * * , ( ) ] link:\doibase 10.1103/physrevx.5.031011 [ * * , ( ) ] link:\doibase 10.1038/nphoton.2014.248 [ * * , ( ) ] , link:\doibase 10.1038/nphys3228 [ * * , ( ) ] link:\doibase 10.1103/physrevb.87.144101 [ * * , ( ) ] link:\doibase 10.1103/physrevb.89.134409 [ * * , ( ) ] link:\doibase 10.1038/nphys3232 [ * * , ( ) ] http://arxiv.org/abs/1503.01324 [ , ( ) ] , link:\doibase 10.1073/pnas.1405969111 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.116.135501 [ * * , ( ) ] link:\doibase 10.1038/nature08917 [ * * , ( ) ] http://article.pubs.nrc-cnrc.gc.ca/ppv/rpviewdoc?issn=1208-6045{&}volume=79{&}issue=11-12{&}startpage=1283{&}ab=y file:///users / mlawler / dropbox / documents / papers/2001/moessner / canadian journal of physics 2001 moessner.pdf [ * * , ( ) ] http://projecteuclid.org/dpubs?service=ui{&}version=1.0{&}verb=display{&}handle=euclid.jdg/1214437492$\backslash$nhttp://www.math.toronto.edu/mgualt/morse theory / witten morse theory and supersymmetry.pdf [ * * , ( ) ] http://arxiv.org/abs/1407.2890 [ , ( ) ] , link:\doibase 10.1126/science.1248253 [ * * , ( ) ] @noop link:\doibase 10.1098/rspa.1958.0141 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.80.2929 papers://ab7c6739 - 3c9b-4bbe-8ce8 - 067dfd022c15/paper / p6318 file:///users / mlawler / dropbox / documents / papers/1998/moessner / physical review letters 1998 moessner.pdf [ * * , ( ) ] link:\doibase 10.1007/bf01325811 [ * * , ( ) ] link:\doibase 10.1007/bf02729860 [ * * , ( ) ] link:\doibase 10.1016/0003 - 4916(77)90335 - 9 [ * * , ( ) ] link:\doibase 10.1143/ptp.63.599 [ * * , ( ) ] link:\doibase 10.1016/0370 - 1573(94)00080-m [ * * , ( ) ] , @noop _ _ ( , ) link:\doibase 10.1103/physrevlett.116.135503 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.115131 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.125119 [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.115141 [ * * , ( ) ] link:\doibase 10.1007/bf01197577 [ * * , ] @noop _ _ ( , ) link:\doibase 10.1103/physrevb.58.12049 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.137207 [ * * , ( ) ] link:\doibase 10.1098/rspa.1957.0133 [ * * , ( ) ] link:\doibase 10.1038/nphys3185 [ * * , ( ) ] @noop * * ( ) link:\doibase 10.1103/physrevb.79.020403 [ * * , ( ) ] link:\doibase 10.1038/nature06981 [ * * , ( ) ] link:\doibase 10.1103/physrevb.39.2344 [ * * , ( ) ] @noop * * ( ) \doibase http://dx.doi.org/10.1016/s0022-5096(02)00107-2 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.jmps.2005.10.008 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.1841 [ * * , ( ) ] link:\doibase 10.1038/23819 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.61.1029 [ * * , ( ) ]

the distinctive features of equilibrium vortex structures in thin films of anisotropic superconductors in tilted magnetic fields are studied for the limits of moderate and strong anisotropy . the energetically favorable shape of isolated vortex lines is found in the framework of two particular models describing these limiting cases : london theory with an anisotropic mass tensor and london - type model for a stack of josephson decoupled superconducting layers . the increase of the field tilting is shown to result in qualitative changes in the vortex vortex interaction potential : the balance between long range attractive and repulsive forces occurs to be responsible for a formation of a minimum of the interaction potential vs the intervortex distance . this minimum appears to exist only for a certain restricted range of the vortex tilting angles which shrinks with the decrease of the system anisotropy parameter . tilted vortices with such unusual interaction potential form clusters with the size depending on the field tilting angle and film thickness or / and can arrange into multiquanta flux lattice . the magnetic flux through the unit cells of the corresponding flux line lattices equals to an integer number of flux quanta . thus , the increase in the field tilting should be accompanied by the series of the phase transitions between the vortex lattices with different .

1009.0350

according to a standard picture of the mixed state in bulk type - ii superconductors the abrikosov vortices penetrating the homogeneous sample form a periodic arrangement called a flux lattice @xcite . the magnetic flux through the unit cell of such flux line lattice equals to the flux quantum @xmath1 : we have one vortex per unit cell . there are a few examples of rather exotic superconducting systems which may provide a possibility to observe a different vortex lattice periodicity , namely the structures with more than one vortices per unit cell . in particular , the phase transitions to such multiquanta flux lattices can occur , e.g. , for superconductors with unconventional pairing @xcite or 2d fulde - ferrell - larkin - ovchinnikov superconductors @xcite . the goal of this work is to suggest an alternative scenario of the phase transitions between the flux structures with different number of vortices per unit cell which can be realized in thin films of anisotropic superconductors . the underlying physical mechanism for this scenario arises from the interplay between the long range attraction and repulsion between tilted vortex lines in thin films discussed recently in ref . . the unusual attractive part of the vortex vortex interaction potential is known to be a distinctive feature of anisotropic superconductors and the value of the attractive force is controlled by the tilting angle of the vortex line with respect to the anisotropy axis @xcite . the origin of the long range intervortex repulsion in thin films has been analyzed in the pioneering work @xcite by pearl in 1964 . this repulsion force always overcomes the attraction at rather large distances because of the different power decay laws of these contributions . note that , of course , the short range interaction between vortices is also repulsive . finally , this balance between the repulsion and attraction can result in the formation of the nonmonotonic interaction potential @xmath2 vs the intervortex distance @xmath3 . increasing the vortex tilting angle we first strengthen the attraction force between vortices and , thus , the minimum in the vortex interaction potential can appear only for rather large tilting angles when the attraction overcomes the pearl s repulsion . this minimum shifts towards the larger intervortex distances with the further increase in the tilting angle and , finally , at rather large distances the attraction appears to be suppressed due to the exponential screening effect . as a consequence , the minimum in the interaction potential exists only for a certain restricted range of the vortex tilting angles which shrinks with the decrease of the system anisotropy parameter . the appearance of a minimum in the interaction potential points to the possibility to get a bound vortex pair ( or even the clusters with higher vorticities ) for a certain range of vortex tilting angles . for a flux line lattice such vortex vortex interaction potential can cause an instability with respect to the unit cell doubling , i.e. the phase transition to the multiquanta vortex lattices . in this paper we use two theoretical approaches to describe the peculiarities of the intervortex interaction and resulting formation of clusters and multiquanta lattices . one of them is a standard london model accounting for an anisotropic mass tensor which is adequate for the superconductors with moderate anisotropy . this approach assumes that the superconducting coherence length in all directions exceeds the distance between the atomic layers and obviously breaks down in the limit of strong anisotropy , i.e. , for josephson coupled layered structures . in the latter case we choose to apply another phenomenological model , namely the so called lowrence doniach theory @xcite . for rather small intervortex distances this theory can be simplified neglecting the effects of weak interlayer josephson coupling . this approach of josephson decoupled superconducting layers is known to be useful in studies of the vortex lattice structure at low fields @xcite . considering thin film samples in tilted magnetic fields we do not restrict ourselves by the case of only straight vortex lines and study the problem of the energetically favorable vortex line shape in the presence of the inhomogeneous supercurrent screening the field component @xmath4 parallel to the film plane . previously this problem has been addressed in ref . for rather small deviations of the vortex line from the direction normal to the film plane . such approximation is obviously valid only for the @xmath5 values much smaller than the critical field @xmath6 of the penetration of vortices parallel to the film plane . for anisotropic london model this analysis of ref . has been previously generalized for the case of a strongly distorted vortex line ( see ref . ) . for the sake of completeness we present here the calculations of the shape of an isolated vortex line for arbitrary fields @xmath7 within both theoretical models describing the limits of strong and moderate anisotropy . as a next step , we calculate the vortex - vortex interaction potential for such strongly deformed vortex lines . further analysis in the paper includes the calculations of energy of finite size vortex clusters as well as the energy of vortex lattices with different number of vortices per unit cell . experimentally the visualization of unconventional vortex arrangements could be carried out by a number of methods which provided convincing evidence for the existence of vortex chains in bulk anisotropic superconductors caused by the intervortex attraction phenomenon ( such as the decoration technique in @xmath8 @xcite , scanning - tunneling microscopy in @xmath9 @xcite , scanning hall - probe @xcite and lorentz microscopy measurements in @xmath8 @xcite ) . the paper is organized as follows . in sec . ii we find the energetically favorable shape of an isolated vortex line . in sec . iii we calculate the vortex vortex interaction potential and prove the existence of a potential minimum for a certain range of field tilting angles and parameters . the sec . iv is devoted to the calculation of energy of vortex clusters . finally , in sec . v we present our analysis of the phase transition between the vortex lattices with one and two flux quanta per unit cell . the results are summarized in sec . vi . some of the calculation details are presented in the appendices [ apx - a ] and [ apx - b ] . we start our study of the distinctive features of equilibrium vortex structures in thin films of anisotropic superconductors with the consideration of the vortex line shape in the layered systems . let us consider a finite stack of @xmath10 superconducting ( sc ) layers . vortex line of an arbitrary shape pierces the film and can be viewed as a string of 2d pancake vortices : each of these pancakes is centered at the point @xmath11 in the @xmath12-th layer . within the model of the stack of josephson decoupled sc layers , pancakes can interact with each other only via magnetic fields . we denote the interlayer spacing as @xmath13 and consider each of the @xmath10 layers as a thin film with the thickness @xmath14 much less than the london penetration depth @xmath15 . general equation for the vector potential @xmath16 distribution in such system reads @xmath17 where @xmath18 is the effective penetration depth in a superconducting film of a vanishing thickness @xmath14 , each @xmath19th sc layer coincides with the plane @xmath20 ( @xmath21 ) , the sheet current at the @xmath19th layer created by the pancake at @xmath22th layer takes the form @xmath23\ , \ ] ] @xmath24 is the vector potential induced by the only pancake vortex located in the @xmath22th layer ( fig . 1 ) . the vector @xmath25 in the eq . ( [ eq:2 ] ) is given by the expression @xmath26 } { \mathbf{r}^2 } \,,\ ] ] and @xmath27 is the flux quantum . for the layered system without josephson coupling a general expression for the free energy can be written in the form : @xmath28\,.\ ] ] where the total vector potential @xmath29 and the sheet current in the @xmath19th layer @xmath30 , produced by an arbitrary vortex line are the sum of the contributions induced by all 2d pancakes : @xmath31 to find the magnetic vector potential @xmath32 we adopt an approach similar to that in refs . . between the sc layers the vector potential @xmath33 is described by the laplace equation @xmath34 for the gauge @xmath35 the vector potential has only the in - plane components @xmath36 , where @xmath37 and the function @xmath38 can be written as @xmath39 / \sinh(q s)\ , , \\ \qquad z_n < z < z_{n+1},\ ; n=1\ldots n-1 \ , , \\ \\ \alpha_n^m\,\exp\left(-q(z - z_n)\right)\,,\quad z \ge z_n \ , , \\ \\ \alpha_1^m\,\exp\left(q(z - z_1)\right)\,,\quad z \le z_1 \ , . \end{array } \right.\ ] ] taking the fourier transform of eq . ( [ eq:2 ] ) we find : @xmath40,\ ] ] where @xmath41}{q^2 } \ .\ ] ] the sheet current density @xmath42 results in the discontinuity of the in - plane component of the magnetic field @xmath43 across the @xmath12 layer : @xmath44 = \mathbf{z}_0 \times \left.\left [ \mathbf{z}_0 \times \frac{\partial \mathbf{a}^m}{\partial z } \right]\right|_{z_n-0}^{z_n+0 } \,.\ ] ] substituting the expressions ( [ eq:5 ] ) , ( [ eq:6 ] ) , ( [ eq:7 ] ) into above condition ( [ eq:9 ] ) we obtain the system of linear equations for the coefficients @xmath45 : @xmath46 here we introduce two new functions which depend on the wave number @xmath47 : @xmath48 the solution of the linear system ( [ eq:10 ] ) and the eqs . ( [ eq:5 ] ) , ( [ eq:6 ] ) define the distribution of the vector potential @xmath24 which is created by a single pancake vortex positioned in the @xmath22th layer . without the in - plane external magnetic field @xmath49 the relative displacement of the pancakes in the different layers is absent : @xmath50 ( i.e. the pancakes form a vertical stack ) . a rather small magnetic field @xmath51 induces a screening meissner current @xmath52 in each @xmath19th layer . lorentz forces arising from these currents will move the pancakes from their initial positions . taking into account the eq . ( [ eq:9 ] ) , we find the following system of linear equations describing the distribution of the magnetic field screened by the layered structure : @xmath53 here @xmath54 is the magnetic field value between the @xmath19th and @xmath55th layers . the distribution of the meissner screening currents in the layers takes the form : @xmath56 the resulting lorentz forces @xmath57 acting on the pancakes can be written as follows : @xmath58 = ( \phi_0 / \ , c ) j_n^m \mathbf{y}_0\,.\ ] ] the interaction forces between the pancakes can be found using the expression ( [ eq:7 ] ) for the sheet current @xmath59 generated by the pancake positioned in the @xmath22th layer : @xmath60 = \frac{\phi_0 ^ 2}{8\pi^2\lambda\lambda_{ab } } \left\{\ , % \frac{1}{r_{mk}}\,\delta_{mk } - \int\limits_0^\infty d q\ , j_1(q r_{mk})\ , \frac { \alpha_k^m(q)\,g(q)}{z(q)}\ , \right\}\ , \frac{\mathbf{r}_{mk}}{r_{mk}}\,,\ ] ] where @xmath61 is the first - order bessel function of the first kind , @xmath62 is the penetration depth for the in - plane currents , and @xmath63 in order to find the equilibrium form of the vortex line in a finite stack of @xmath10 superconducting layers under the influence of the in - plane external magnetic field @xmath49 , we consider the relaxation of the set of the pancakes towards the equilibrium positions within the simplest version of the dynamic theory : @xmath64 where @xmath65 is the viscous drag coefficient . considering the vortex line consisting of @xmath66 pancakes we start from the initial configuration of pancakes arranged in a straight vortex line parallel to the @xmath67 direction ( see fig . 2 ) . as the system approaches its final force - balanced ( equilibrium ) configuration , the velocities of all pancake motions should vanish : @xmath68 in fig . 2 we illustrate the evolution of the pancake configurations for two values of the applied in plane magnetic field @xmath49 and for two different numbers of layers : @xmath66 ( fig . 2a , 2b ) and @xmath69 ( fig . 2c , 2d ) . the forces @xmath57 caused by the meissner currents rotate and bend the vortex line . for rather small applied field values this rotation and bending result in the formation of a certain stable configuration ( see fig . 2a , 2c ) . for the fields exceeding a certain critical value @xmath70 we do not find such stable pancake arrangement . the vortex line splits into two segments which move in opposite directions ( see fig . 2b , 2d ) . to define the critical value @xmath70 for the breakup of the vortex line we have carried out the calculations of the pancake arrangements increasing the in - plane magnetic field with the step @xmath71 ( @xmath72 ) . the stationary vortex arrangement disappears above a certain threshold field value which is taken as a critical field @xmath70 . the pancake configurations for the both cases @xmath66 and @xmath69 are qualitatively similar but the values of the critical field @xmath70 for @xmath66 and @xmath69 are different . with a decrease in the number of layers @xmath10 ( film thickness ) the value of the critical field @xmath70 grows : @xmath73 for @xmath66 and @xmath74 for @xmath69 . in fact in layered superconductors with very weak interlayer coupling the josephson vortices will appear at much lower field @xmath75 . as a result , at tilted magnetic field , crossing lattice of pancakes , forming abrikosov vortices , and josephson vortices exist rather than a lattice of tilted vortex stacks @xcite . the interaction between pancakes and in - plane field in the form of josephson vortices produces zigzag deformation of the stack of the pancakes @xcite . this deformation is responsible for a long range attraction between such stacks @xcite which is quite similar to the case of considered in the present work . we proceed now with the consideration of the vortex line shape in an anisotropic film which is characterized by the london penetration depths @xmath76 and @xmath77 for currents flowing parallel and perpendicular to the @xmath78 plane , respectively . we consider the case of uniaxial anisotropy which can be described by a dimensionless anisotropic mass tensor @xmath79 , where @xmath80 is the anisotropy parameter and @xmath81 is the anisotropy axis . we choose the @xmath67 axis of the coordinate system perpendicular to the film surface . in the parallel to the film plane direction we apply a certain magnetic field @xmath82 which is screened inside the superconducting film . we consider first a typical geometry when the @xmath83 axis is chosen along the direction normal to the film plane . in such geometry the vortex line is parallel to the plane @xmath84 and can be parameterized by a single valued function @xmath85 . an appropriate thermodynamic potential for determination of the energetically favorable form of the vortex line takes the form @xmath86 where @xmath87 is the film thickness . the first term , @xmath88 , is the ginzburg - landau free energy of the curved vortex line , and the second term corresponds to the work of lorentz force acting on the flux line and distorting this line in the presence of screening currents induced by the external magnetic - field component @xmath89 parallel to the film plane . to simplify the @xmath88 expression we consider a strong type - ii superconducting material with a large ratio of the london penetration depths and coherence lengths . in this case the main contribution to the vortex line energy is determined by the energy of supercurrents @xmath90 flowing around the vortex core @xmath91 where @xmath92 , and @xmath93 is the order parameter phase distribution around the vortex line . the above expression for the free energy reveals a logarithmic divergence which should be cut off at both the small and large spatial length scales . the lower cut off length is naturally equal to the characteristic size @xmath94 of the vortex core which is of the order of the coherence length . of course , in anisotropic case one should introduce two different coherence lengths @xmath95 and @xmath96 in the @xmath78 plane and along the @xmath97 axis , respectively . the resulting core size and lower cut off length @xmath94 for a certain element of the tilted flux line will , thus , depend on both @xmath95 and @xmath96 lengths as well as on the local tilting angle of the vortex line . the upper cut off length strongly depends on the ratio of the film thickness to the london penetration depth . for rather thick films @xmath98 this cut off length @xmath99 is determined by a certain combination of the screening lengths @xmath76 and @xmath77 ( see , e.g. , ref . ) . for a thin film with @xmath100 one can separate two energy contributions : ( i ) the contribution coming from the region of the size @xmath101 around the curved vortex line and providing the logarithmic term in the free energy with the upper cut off length @xmath102 ; ( ii ) the logarithmic contribution @xmath103 coming from the larger distances @xmath104 which weakly depends on the vortex line shape . to sum up , the part of the vortex line energy which depends on its shape can be approximately written in the form : @xmath105 note that we neglect here the weak dependence of the logarithmic factor on the vortex line curvature and local orientation . within such approximation we consider the vortex line as a thin elastic string which is , of course , valid provided the characteristic scale of the string bending is larger than the upper cut off length @xmath99 . the condition of the zero first variation of the gibbs functional gives us the following equation @xmath106 where @xmath107 the equation is valid for magnetic fields which do not exceed the critical field of the penetration of vortices parallel to the film plane @xmath108 thus , analogously to the case of a stack of decoupled layers the stable curved vortex lines can exist only for rather small magnetic fields below the critical field @xmath6 which corresponds to the penetration of a vortex parallel to the film plane . note that in the limit @xmath109 one can obtain the result found previously in ref . : @xmath110 typical shape of a bent vortex line calculated from eq . ( [ anis forma d - arb ] ) is shown in fig . 3a . the above expressions can be easily generalized for an arbitrary angle @xmath111 between the anisotropy axis @xmath81 and the direction normal to the film plane . we restrict ourselves to the case when the axis @xmath81 is parallel to the plane @xmath84 and vortex line can be parameterized by a function @xmath85 as before . in this case the part of the free energy depending on the vortex line shape takes the form : @xmath112 where @xmath113=y'(z)$ ] . thus we find the following equation describing the vortex line shape : @xmath114 where @xmath115 note that the equation is valid in the field range @xmath116 the critical field @xmath117 corresponds to the penetration of a vortex parallel to the film plane . typical plots illustrating the numerical solution of the equation are shown in the fig . 3b for different orientations of the applied magnetic field . note an important difference between the opposite directions of the magnetic field @xmath89 : for @xmath118 the vortex line consists of segments tilted in opposite directions with respect to the @xmath67 axis . in this section we derive general expressions for the interaction energy between two vortices in a thin film of an anisotropic superconductor taking into account both long range attraction and repulsion phenomena . we study both the limits of strong and moderate anisotropy for a wide range of vortex tilting angles . the shape of the interacting vortex lines is assumed to be fixed and not affected by the vortex vortex interaction potential . certainly , such assumption is valid only in the limit of rather larger distances between the vortex lines when the effect of interaction on the vortex shape can be viewed as a small perturbation . in this section we consider the interaction between two vortex lines consisting of pancake vortices . for each vortex the pancake centers are assumed to be positioned along the straight line tilted at the angle @xmath119 with respect to the anisotropy axis @xmath120 ( @xmath67 axis ) . we restrict ourselves to the case of parallel vortex lines shifted by a certain vector @xmath121 in the plane of the layers . using the gauge @xmath122 and the fourier transform @xmath123 @xmath124 one can rewrite the basic equation ( [ eq:1 ] ) in the momentum representation as follows : @xmath125 where @xmath126 . taking account of the relation @xmath127 we obtain from ( [ eq:2c ] ) the following equations for the fourier components of the vector potential @xmath128 : @xmath129 these equations can be reduced to the scalar form @xmath130 where we introduce the new functions @xmath131 : @xmath132 the solution of the linear system ( [ eq:4c ] ) for a fixed distribution of pancakes @xmath133 determines the distribution of the vector potential @xmath29 which is created by an arbitrary vortex line in a finite stack of superconducting layers . in the momentum representation the general expression ( [ eq:6c ] ) for the free energy of the layered system without josephson coupling reads @xmath134 for two vortex lines we can write the total vector potential and the total sheet current as superpositions of contributions coming from the first ( @xmath135 , @xmath136 ) and second ( @xmath137 , @xmath138 ) vortices : @xmath139 calculating the interaction energy @xmath140 of vortex lines we should keep in ( [ eq:7c ] ) only the terms which contain the products of fields corresponding to different vortex lines : @xmath141\,.\ ] ] finally , for the particular case of two parallel vortex lines which are shifted at the vector @xmath142 in the @xmath143 plane we get following expression for the interaction energy via the scalar functions @xmath131 : @xmath144 the expression ( [ eq:9c ] ) and equations ( [ eq:4c ] ) determine the interaction energy of two identically bent vortex lines . our further consideration in this subsection is based on two assumptions : ( i ) for each vortex we choose the centers of pancakes to be positioned along the straight line tilted at a certain angle @xmath119 relative to @xmath67 axis , and put @xmath145 ; ( ii ) we restrict ourselves by the continuous limit assuming @xmath146 and @xmath147 . in this case the eqs . ( [ eq:4c ] ) , ( [ eq:9c ] ) can be simplified ( see appendix [ apx - a ] for details ) : @xmath148 } { \sqrt{q^2 + \lambda_{ab}^{-2}}\,(1 + p^2)^2 \left [ 2 k \cosh l + ( 1 + k^2)\sinh l)\right ] } \right\}\ , , \label{eq:11c}\end{aligned}\ ] ] where @xmath149 and @xmath150 is the thickness of the superconducting film . the first term in ( [ eq:11c ] ) describes the interaction in the bulk system , while the second term is responsible for the effect of film boundaries . the minimum energy configuration corresponds to the case @xmath151 . in fig . 4 we present some typical plots of the interaction energy @xmath152 vs the distance @xmath153 for @xmath154 which corresponds to the lorentz microscopy experiments in ybco @xcite and bi-2212 @xcite samples . analyzing the dependence @xmath155 one can separate three contributions to the energy of vortex vortex interaction : ( i ) a short range repulsion which decays exponentially with increasing intervortex distance @xmath3 ( for @xmath156 ) ; ( ii ) an intervortex attraction which is known to be specific for tilted vortices in anisotropic systems ; this attraction energy term decays as @xmath157 and strongly depends on the angle @xmath158 between the vortex axis and the @xmath120 direction ; ( iii ) long range ( pearl ) repulsion which decays as @xmath159 and results from the surface contribution to the energy . note that the third term does exist even for a large sample thickness @xmath87 ( see ref . @xcite ) although in the limit @xmath160 it is certainly masked by the dominant bulk contribution . at @xmath161 the short range interaction term vanishes and the interaction energy vs @xmath162 takes the form @xmath163where @xmath164 is the effective film thickness . one can observe here an interplay between the long - range attractive ( first term in eq.([energy - interact ] ) ) and the repulsive ( second term in eq.(energy - interact ) ) forces . note that the @xmath165 value increases with an increase in temperature , thus , the effective thickness decreases and the long range attraction force appears to be suppressed with increasing temperature . for large @xmath3 the energy is always positive and corresponds to the vortex repulsion similar to the one between the pancakes in a single layer system . with a decrease in the distance @xmath3 the attraction force comes into play resulting in the change of the sign of the energy . such behavior points to the appearance of a minimum in the interaction potential . we now proceed with the derivation of the expression for the intervortex interaction energy in an anisotropic superconducting film . we choose the anisotropy axis @xmath120 ( @xmath166axis ) to be oriented perpendicular to the film plane and consider two curved vortex lines with identical shapes found in sec . ii b. our further calculations are based on general expressions derived in ref . for the energy of an arbitrary vorticity distribution in an anisotropic superconducting film of finite thickness ( see appendix [ apx - b ] for details ) . the resulting interaction energy of two curved vortices shifted from each other in the @xmath167 direction at a certain distance @xmath3 can be presented in the form : @xmath168 where @xmath169 , while @xmath170 , @xmath171 , and @xmath172 are given by the expressions ( [ eq : b7 ] ) , ( [ eq : b8 ] ) , ( [ eq : b9 ] ) . considering the particular case of straight vortex lines parallel to the plane @xmath173 and tilted at a certain angle @xmath158 with respect to the @xmath120 direction we obtain the following expression for the interaction energy of two vortices : @xmath174 } { \sqrt{q^2 + \lambda_{ab}^{-2}}\,(1 + p^2)^2 \left [ 2 k \cosh l + ( 1 + k^2)\sinh l)\right ] } \right\}\ , , \label{straight - tilted2}\end{aligned}\ ] ] where @xmath175 and the parameters @xmath176 , @xmath177 and @xmath178 are described by the expressions ( [ eq:10c1 ] ) . in the limit of strong anisotropy ( @xmath179 ) the spectral function @xmath180 ( [ straight - tilted2 ] ) naturally coincides with the corresponding expressions ( [ eq:11c ] ) obtained for the layered system without josephson coupling . some typical plots of the interaction energy vs the intervortex distance for tilted vortex lines calculated using the eqs . ( [ straight - tilted]),([straight - tilted2 ] ) are shown in figs . [ fig:5],[fig:51 ] . analyzing the dependence @xmath181 one can separate three contributions to the energy of intervortex interaction : ( i ) a short range repulsion ( for @xmath182 ) which decays exponentially with increasing intervortex distance @xmath3 ; ( ii ) an intervortex attraction which comes into play for the region @xmath183 and decays exponentially with the vortex vortex distance @xmath3 for @xmath184 ; ( iii ) long range ( pearl ) repulsion which decays as @xmath159 at large distances and results from the surface contribution to the energy . taking the limit @xmath182 we get @xmath185 in the region @xmath186 the short range interaction term vanishes and the interaction energy vs @xmath3 is given by the sum ( [ energy - interact ] ) of attractive and pearl s contributions . similarly to the case of decoupled layers discussed above the attractive term can result in the appearance of a minimum in the dependence of the vortex vortex interaction potential vs @xmath3 . the position of this minimum can be roughly estimated as the boundary of the region of the short range repulsion : @xmath187 . obviously , the minimum should disappear provided @xmath188 , i.e. , when the region of the attraction between vortices vanishes . this condition gives us the the upper boundary on the tilting angle @xmath119 restricting the interval of the energy minimum existence : @xmath189 the lower boundary of this angular interval can be found comparing the attractive and repulsive terms in the expression ( [ energy - interact ] ) at the distance @xmath190 : @xmath191 these analytical estimates of the angular interval are in a rough qualitative agreement with the numerical calculations ( see figs . [ fig:5],[fig:51 ] ) for two values of the film thickness @xmath192 . indeed , one can see that increasing the tilting angle we first deepen the minimum in the interaction potential and then make it more shallow . the figures confirms the deepening of the minimum with the increase in the anisotropy parameter @xmath193 . our numerical calculations demonstrate that for the film thickness @xmath194 ( fig . [ fig:5 ] ) the minimum of the interaction energy of two straight tilted vortices can appear only for @xmath195 . starting from @xmath196 the bound vortex pair becomes energetically favorable . an increase in the film thickness reduces the relative contribution of pearl repulsion to the energy of intervortex interaction @xmath140 . as a result attraction of vortices takes place for smaller values of the tilting angle @xmath119 and anisotropy parameter @xmath193 . thus , in a film with the thickness @xmath197 ( fig . [ fig:51 ] ) the minimum in the @xmath181 dependence appears for @xmath198 , whereas creation of the bound vortex pair becomes energetically favorable for @xmath199 . as a next step , we check if the above results obtained for straight tilted vortices remain valid for the curved vortex lines . our analysis of the effect of the vortex line curvature is carried out for model vortex profiles found in sec . ii b. the resulting typical dependencies of the intervortex interaction potential vs @xmath3 for different magnetic field values and anisotropy parameters are shown in figs . [ fig:6],[fig:7 ] . one can clearly see that the minimum in the interaction potential vs @xmath3 survives when we take account of the vortex line curvature . moreover the curving of the vortex line even deepens this minimum as it is confirmed by the comparison of energies of straight tilted and curved vortices presented in fig . [ fig:7 ] . for such comparison we choose the straight vortex lines connecting the ends of curved vortices . we find that for curved vortices the energy minimum exists even for smaller anisotropy parameters than for straight vortices ( i.e. , the threshold anisotropy value for @xmath194 becomes less than @xmath200 ) . of course , increasing the film thickness one can weaken the restrictions on the existence of the minimum in the interaction potential : e.g. , for @xmath201 the minimum appears at @xmath199 . the above theoretical analysis demonstrates that vortex vortex attraction and the formation of chains are possible only for the rather large tilting angles and at low vortex concentrations , i.e. , when the magnetic - field component @xmath202 perpendicular to the film plane is very weak . in fields @xmath202 slightly above @xmath203 abrikosov vortices will form chains due to the long range attractive interaction . peculiarities of penetration of such chains of tilted abrikosov vortices into bulk layered ( anisotropic ) superconductor are well known : in the first approximation , the vortex period in chains does not depend on applied magnetic field , while the distance between chains changes as @xmath204 . the presence of vortex chains significantly modifies the magnetization curves with respect to analogous curves for isotropic superconductors . @xcite in the next sections we discuss additional peculiarities of intervortex interaction specific for thin film samples of layered ( anisotropic ) superconductors . the unusual vortex - vortex interaction potential behavior discussed in the previous section can result in unconventional vortex structures . we start our analysis of energetically favorable vortex structures from the problem of stability of a vortex chain . the formation of infinite vortex chains is known to be a signature of the intervortex attraction in bulk anisotropic superconductors . the long range repulsion of vortices in thin films can destroy the infinite vortex chains . indeed , despite of the fact that two vortices attract each other at a certain distance , further increase in the number of vortices arranged in a chain can be energetically unfavorable because of the slower decay of the repulsive force compared to the attractive one . in this case , for rather thin samples , there appears an intriguing possibility to observe vortex chains of finite length , i.e. , vortex molecules or clusters . in this section we present the calculations of energies of such vortex clusters . as we have demonstrated above , the minimum in the interaction potential exists for both the limits of strong and moderate anisotropy . the vortex molecule cohesion energy is given by the expression : @xmath205where @xmath10 is the number of vortices in the molecule , and @xmath206 are the distances between @xmath207th and @xmath208th vortices in the chain molecule . shown in figs . [ fig:8],[fig:9 ] are typical plots of the interaction energy per vortex vs the intervortex distance @xmath3 for equidistant vortex chains with different @xmath10 numbers calculated within the model of decoupled superconducting layers and anisotropic london theory . the energetically favorable number of vortices in a molecule grows as we increase the film thickness and/or the tilting angle because of the increasing attraction term in the pair potential @xmath140 . shown in the insets of figs . [ fig:8 ] are schematic pictures of vortex matter consisting of dimeric and trimeric molecules . finally , for rather thick samples with @xmath209 we get a standard infinite chain structure typical for bulk systems . note that the formation of an infinite vortex chain may be considered in some sense as a polymerization of the vortex molecules . certainly , the crossover from the vortex molecule state to the infinite chain structure is strongly influenced by the increase in the vortex concentration governed by the component @xmath210 of the external magnetic field perpendicular to the film . indeed , one can expect such a cross - over to occur when the mean intervortex spacing approaches the molecule size . thus , the vortex molecule state can appear only in a rather weak perpendicular field when its observation can be complicated , of course , by the pinning effects . considering the effect of a finite magnetic field ( i.e. , a finite concentration of vortex clusters ) we restrict ourselves by the simplest case of regular vortex arrays . for a regular vortex array the formation of clusters corresponds to the transition with a change in the number of vortices in the elementary lattice cell . the mechanism underlying such transition is naturally connected with the appearance of the minimum in the interaction potential for a vortex pair . in this section we present our calculations of energy of vortex lattices with different number of flux quanta per unit cell . the possibility to get the energetically favorable states with a few vortices per unit cell will be illustrated for a particular intervortex interaction potential derived above for a model of decoupled superconducting layers . the generalization of such consideration for anisotropic london theory is straightforward . note that the vortex lattice structure for bulk anisotropic superconductors in tilted field in the framework of london approach has been calculated in ref . . let s consider a vortex lattice characterized by the translation vectors @xmath211 , where @xmath212 are primitive vectors of the lattice . the primitive cell occupies the area @xmath213\,\cdot \mathbf{z}_0 $ ] and is assumed to contain @xmath0 vortices : @xmath214 . positions of vortices in a cell are determined by the vectors @xmath215 ( @xmath216 ) ( see fig . [ fig:10 ] ) . the interaction energy per unit lattice cell can be expressed via the vortex vortex interaction potentials ( [ eq:10c]),([eq:11c ] ) : @xmath217 the interaction energy ( [ eq:1e ] ) depends on both the relative positions @xmath218 of vortices in the primitive cell and the structure of the vortex lattice defined by the translation vectors @xmath219 . the first term in ( [ eq:1e ] ) describes the interaction energy between vortices in the primitive cell ( without the lattice contribution ) , whereas the second sum takes account of the lattice effects . with the help of the poisson formula , one can rewrite the intervortex interaction energy ( [ eq:1e ] ) in terms of the fourier components @xmath220\,,\ ] ] where the function @xmath221 is determined by the eq . ( [ eq:11c ] ) , and @xmath222 are the reciprocal lattice vectors . the sum and the integral in eq . ( [ eq:2e ] ) diverge both at @xmath223 and at large @xmath222 values . the small @xmath222 divergence corresponds to the linear ( in the system size ) increase in the vortex energy because of the slow @xmath224 decay of the vortex - vortex interaction potential . the large @xmath222 divergence is logarithmic and is associated with the vortex self energy . for simplicity , we restrict ourselves to the case of an instability with respect to the unit cell doubling and tripling , i.e. formation of the vortex lattices with two and three flux quanta per unit cell ( @xmath225 and @xmath226 ) . hereafter we consider only the shifts of vortex sublattices along the @xmath167 direction and choose the appropriate reciprocal lattice vectors @xmath227 @xmath228 for @xmath225 and @xmath226 , respectively . here we consider only equidistant vortex chains within the primitive cells . fixing the value of the field @xmath210 we fix the unit cell area area @xmath229 for @xmath230 and @xmath231 for @xmath226 . thus , the interaction energy ( [ eq:2e ] ) depends only on two parameters : ( i ) @xmath232 ratio characterizing the lattice deformation ; ( ii ) relative displacement @xmath233 of vortex sublattices along the @xmath167-axis ( see fig . [ fig:10 ] ) . to exclude the divergence at @xmath234 it is convenient to deal with the energy difference : @xmath235 the results of our numerical calculations of this energy difference are shown in fig . [ fig:11]a . one can clearly observe that changing the vortex tilting angle we obtain the minimum in the function @xmath236 which gives us the evidence for the phase transition in the lattice structure with the unit cell doubling or tripling depending on the vortex tilting angle . the multiplication of the unit cell is accompanied by the strong change in the lattice deformation ratio @xmath237 ( see fig . [ fig:11]b ) . to sum up , we suggest a scenario of the phase transitions between the flux structures with different number of vortices per unit cell which can be realized in thin films of anisotropic superconductors placed in tilted magnetic fields . we demonstrate that the vortex interaction in the films of anisotropic superconductors placed in tilted magnetic fields is very special . the underlying physics arises from the interplay between the long range attraction and repulsion between tilted vortex lines . in consequence , new and very reach types of vortex structures may appear . they are formed from the vortex dimers , trimers , etc . , and the transition between different types of vortex structures may be controlled by tilting of external magnetic field and/or by varying of the temperature . our theoretical findings are based on two theoretical approaches : anisotropic london model and the london type model of decoupled superconducting layers . taking account of the vortex tilt and bending we analyzed the distinctive features of the vortex vortex interaction potential in a wide range of parameters and fields and demonstrated the possibility to obtain a minimum in the vortex interaction potential vs the intervortex distance . further analysis in the paper included the calculations of energy of finite size vortex clusters as well as the energy of regular vortex arrays with different number of vortices per unit cell . the phase transitions accompanied by the multiplication of the primitive lattice cell appear to be possible for dilute vortex arrays , i.e. for rather small magnetic field component @xmath210 . we believe that our theoretical predictions concerning the unusual vortex configurations are experimentally observable using the modern vortex imaging methods such as lorentz microscopy , scanning tunneling microscopy , scanning hall - probe or decoration technique . we are grateful to professor a. tonomura for stimulating discussions . this work was supported , in part , by the russian foundation for basic research , russian academy of sciences under the program quantum physics of condensed matter " , russian agency of education under the federal program scientific and educational personnel of innovative russia in 20092013 " , and dynasty " foundation . let us evaluate the interaction energy ( [ eq:9c ] ) of two tilted parallel vortex lines taking @xmath238 and assuming @xmath146 and @xmath239 . we introduce a continuous coordinate @xmath240 and continuous function @xmath241 . thus , the linear system of equations ( [ eq:4c ] ) reduces to the following integral equation @xmath242 the equation ( [ eq : a1 ] ) can be rewritten as a differential one @xmath243 at the interval @xmath244 with the boundary conditions @xmath245 introducing the notations @xmath246 one can rewrite the equation ( [ eq : a2 ] ) and boundary conditions ( [ eq : a3 ] ) in dimensionless form @xmath247 the solution of the eq . ( [ eq : a4 ] ) has the form @xmath248 where the constants @xmath249 and @xmath250 are defined by the boundary conditions ( [ eq : a5 ] ) : @xmath251 @xmath252 in the continuous limit the expression for the interaction energy ( [ eq:9c ] ) takes the form : @xmath253 here the function @xmath254 can be calculated analytically : @xmath255 } { \sqrt{q^2 + \lambda_{ab}^{-2}}\,(1 + p^2)^2\ , \left[\,2 k \cosh l + ( 1 + k^2)\sinh l\,\right ] } \ .\ ] ] to calculate the vortex vortex interaction within anisotropic london model we use general expressions derived in ref . for the total energy @xmath256 of an arbitrary arrangement of curved vortices in a superconducting film of thickness @xmath87 with the @xmath97-axis perpendicular to the film plane : @xmath257 where @xmath264a\bigg ] \ , \\ \nonumber \phi^+({\bf k}_{\bot})=(-\upsilon / k_{\bot}c ) \bigg[-2k_{\bot}a+\bigg[(k_{\bot}+\upsilon)e^{\upsilon d}+(k_{\bot}-\upsilon)e^{-\upsilon d}\bigg]b\bigg ] \ , \end{aligned}\ ] ] the summation in the above expressions is carried out over @xmath268 and @xmath269 stands for the vector component parallel to the @xmath270 plane . following ref . we introduce here the fourier transform @xmath271 of the vorticity distribution @xmath272 : @xmath276 + j_0\left[\nu\left(\eta(\zeta_2)-\eta(\zeta_1 ) - \tilde r \right)\right]\bigg\ } \nonumber \\ & & \qquad\times\left(\pi_1(\nu,\zeta_1,\zeta_2 ) + \pi_2(\nu,\zeta_1,\zeta_2)\right)\ , \end{aligned}\ ] ] @xmath277 + j_0\left[\nu\left(\eta(\zeta_2)-\eta(\zeta_1 ) - \tilde r \right)\right]\bigg\ } \nonumber \\ & & \qquad\times\frac{\pi_3(\nu,\zeta_1,\zeta_2 ) } { \tau(\nu)\sinh\left[\tau(\nu)\tilde d\right]\bigg\{e^{-\tau(\nu)\tilde d}\left[\nu-\tau(\nu)\right]^2- e^{\tau(\nu)\tilde d}\left[\nu+\tau(\nu)\right]^2\bigg\}^2 } \ , \end{aligned}\ ] ] @xmath278 + j_0\left[\nu\left(\eta(\zeta_2)-\eta(\zeta_1 ) - \tilde r \right)\right]\bigg\ } \nonumber\\ & & \qquad\times\frac{\pi_4(\nu,\zeta_1,\zeta_2 ) } { \nu\sinh^2\left[\tau(\nu)\tilde d\right]\bigg\{e^{-\tau(\nu)\tilde d}\left[\nu-\tau(\nu)\right]^2- e^{\tau(\nu)\tilde d}\left[\nu+\tau(\nu)\right]^2\bigg\}^2 } \,,\end{aligned}\ ] ] @xmath284\ , \cosh\left[\tau(\nu)\left((-\zeta_1-\zeta_2-|\zeta_1-\zeta_2|)/2\right)\right ] } { \nu^2\sinh\left[\tilde d\tau ( \nu)\right ] } \nonumber \\ & -&\frac{\cosh\left[u(\nu)\left(\tilde d+(\zeta_1+\zeta_2-|\zeta_1-\zeta_2|)/2\right)\right ] \cosh\left[u(\nu)\left((-\zeta_1-\zeta_2-|\zeta_1-\zeta_2|)/2\right)\right ] } { \nu^2u(\nu)\sinh\left[\tilde du ( \nu)\right ] } \nonumber \ , \end{aligned}\ ] ] @xmath285+\tau ( \nu)^2\cosh\left[\tau ( \nu)\tilde d\right ] + 2\nu\tau(\nu)\sinh\left[\tau ( \nu)\tilde d\right]\bigg\ } \nonumber\\ & & \times\bigg\{\cosh\left[\tau(\nu)(\tilde d+\zeta_1)\right]\cosh\left[\tau(\tilde d+\zeta_2)\right ] + \cosh\left[\tau(\nu)\zeta_1\right]\cosh\left[\tau(\nu)\zeta_1\right]\bigg\}\nonumber\\ & + & \cosh\left[\tau(\nu)(\tilde d+\zeta_1)]\right)\cosh\left[\tau(\nu)\zeta_2\right ] + \cosh\left[\tau(\nu)z_1\right]\cosh\left[\tau(\nu)(\zeta_2+\tilde d)\right]\nonumber\,\end{aligned}\ ] ] @xmath286+\sinh^2\left[\tau(\nu)\tilde d\right]+ 2\nu\tau(\nu)\sinh\left[\tau(\nu)\tilde d\right]\cosh\left[\tau(\nu)\tilde d\right]\bigg\}\times \\ \nonumber \bigg\{\cosh\left[\tau(\nu)(\tilde d+\zeta_1)\right]\cosh\left[\tau(\nu)(\tilde d+\zeta_2)\right]+ \cosh\left[\tau(\nu)\zeta_1\right]\cosh\left[\tau(\nu)\zeta_1\right]\bigg\}- \\ \nonumber 2\bigg\{\nu^2\cosh\left[\tau(\nu)\tilde d\right]+\nu\tau\sinh\left[\tau(\nu)\tilde d\right]\bigg\}\times \\ \nonumber \bigg\{\cosh\left[\tau(\nu)(\tilde d+\zeta_1)\right]\cosh\left[\tau(\nu)\zeta_2\right]+ \cosh\left[\tau(\nu)\zeta_1\right]\cosh\left[\tau(\nu)(\zeta_2+\tilde d)\right]\bigg\ } \ .\end{aligned}\ ] ] a. tonomura , h. kasai , o. kamimura , t. matsuda , k. harada , t. yoshida , t. akashi , j. shimoyama , k. kishio , t. hanaguri , k. kitazawa , t. masui , s. tajima , n. koshizuka , p. l. gammel , d. bishop , m. sasase , and s. okayasu , phys . rev . lett . * 88 * , 237001 ( 2002 ) . pancake vortex positioned in the @xmath22th layer of a finite layered structure , @xmath14 is a thickness of the superconducting layer , and @xmath13 is the distance between the layers.,scaledwidth=45.0% ] ( panels a , b ) and @xmath69 ( panels c , d ) pancakes in a finite stack in the presence of the applied in - plane magnetic field @xmath89 . ( a ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath289 . ( b ) pancake configurations at sequential time points @xmath290 for @xmath291 . for the structure with @xmath66 we find @xmath292 . ( c ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath293 . ( d ) pancake configurations at sequential time points @xmath294 for @xmath295 . for the structure with @xmath69 we find @xmath296 . here @xmath72 , @xmath297 , and @xmath298.,title="fig:",scaledwidth=40.0% ] ( panels a , b ) and @xmath69 ( panels c , d ) pancakes in a finite stack in the presence of the applied in - plane magnetic field @xmath89 . ( a ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath289 . ( b ) pancake configurations at sequential time points @xmath290 for @xmath291 . for the structure with @xmath66 we find @xmath292 . ( c ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath293 . ( d ) pancake configurations at sequential time points @xmath294 for @xmath295 . for the structure with @xmath69 we find @xmath296 . here @xmath72 , @xmath297 , and @xmath298.,title="fig:",scaledwidth=40.0% ] ( panels a , b ) and @xmath69 ( panels c , d ) pancakes in a finite stack in the presence of the applied in - plane magnetic field @xmath89 . ( a ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath289 . ( b ) pancake configurations at sequential time points @xmath290 for @xmath291 . for the structure with @xmath66 we find @xmath292 . ( c ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath293 . ( d ) pancake configurations at sequential time points @xmath294 for @xmath295 . for the structure with @xmath69 we find @xmath296 . here @xmath72 , @xmath297 , and @xmath298.,title="fig:",scaledwidth=40.0% ] ( panels a , b ) and @xmath69 ( panels c , d ) pancakes in a finite stack in the presence of the applied in - plane magnetic field @xmath89 . ( a ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath289 . ( b ) pancake configurations at sequential time points @xmath290 for @xmath291 . for the structure with @xmath66 we find @xmath292 . ( c ) the force - balanced ( equilibrium ) configuration of pancakes for @xmath293 . ( d ) pancake configurations at sequential time points @xmath294 for @xmath295 . for the structure with @xmath69 we find @xmath296 . here @xmath72 , @xmath297 , and @xmath298.,title="fig:",scaledwidth=37.0% ] for the anisotropy parameter @xmath299 and for different values of in - plane magnetic field @xmath300 . ( a ) the anisotropy axis is perpendicular to the film plane ( @xmath301 , @xmath302 ) . ( b ) the anisotropy axis is tilted with respect to the @xmath67 axis ( @xmath303 , @xmath304 ) the numbers near the curves denote the values of the ratio @xmath305 . the dashed line shows the shape of a vortex line in the absence of the in - plane magnetic field.,title="fig:",scaledwidth=45.0% ] for the anisotropy parameter @xmath299 and for different values of in - plane magnetic field @xmath300 . ( a ) the anisotropy axis is perpendicular to the film plane ( @xmath301 , @xmath302 ) . ( b ) the anisotropy axis is tilted with respect to the @xmath67 axis ( @xmath303 , @xmath304 ) the numbers near the curves denote the values of the ratio @xmath305 . the dashed line shows the shape of a vortex line in the absence of the in - plane magnetic field.,title="fig:",scaledwidth=45.0% ] ) , ( [ straight - tilted2 ] ) ] vs the distance r between two tilted vortices for an anisotropic film of the thickness @xmath194 . ( a ) interaction energy for the anisotropy parameter @xmath309 and different tilting angles . the numbers near the curves denote the values of tilting angle @xmath119 . ( b ) interaction energy for @xmath310 and different values of anisotropy parameter . the numbers near the curves denote the values of @xmath193.,title="fig:",scaledwidth=40.0% ] ) , ( [ straight - tilted2 ] ) ] vs the distance r between two tilted vortices for an anisotropic film of the thickness @xmath194 . ( a ) interaction energy for the anisotropy parameter @xmath309 and different tilting angles . the numbers near the curves denote the values of tilting angle @xmath119 . ( b ) interaction energy for @xmath310 and different values of anisotropy parameter . the numbers near the curves denote the values of @xmath193.,title="fig:",scaledwidth=40.0% ] ) , ( [ straight - tilted2 ] ) ] vs the distance r between two tilted vortices for an anisotropic film of the thickness @xmath197 . ( a ) interaction energy for the anisotropy parameter @xmath311 and different tilting angles . the numbers near the curves denote the values of tilting angle @xmath119 . ( b ) interaction energy for @xmath312 and different values of anisotropy parameter . the numbers near the curves denote the values of @xmath193.,title="fig:",scaledwidth=40.0% ] ) , ( [ straight - tilted2 ] ) ] vs the distance r between two tilted vortices for an anisotropic film of the thickness @xmath197 . ( a ) interaction energy for the anisotropy parameter @xmath311 and different tilting angles . the numbers near the curves denote the values of tilting angle @xmath119 . ( b ) interaction energy for @xmath312 and different values of anisotropy parameter . the numbers near the curves denote the values of @xmath193.,title="fig:",scaledwidth=40.0% ] ) vs the distance @xmath3 between two curved vortices for an anisotropic film of the thickness @xmath194 : ( a ) @xmath313 ; ( b ) @xmath309 . the numbers near the curves denote the values of the ratio @xmath314 . the shape of vortex lines is schematically shown in the insets.,title="fig:",scaledwidth=45.0% ] ) vs the distance @xmath3 between two curved vortices for an anisotropic film of the thickness @xmath194 : ( a ) @xmath313 ; ( b ) @xmath309 . the numbers near the curves denote the values of the ratio @xmath314 . the shape of vortex lines is schematically shown in the insets.,title="fig:",scaledwidth=45.0% ] ) ] ( solid lines ) and straight tilted [ eqs . ( [ straight - tilted ] ) , ( [ straight - tilted2 ] ) ] ( dashed lines ) vortices for an anisotropic film of the thickness @xmath194 with different anisotropy parameters : ( a ) @xmath313 , @xmath315 ( @xmath316 ; ( b ) @xmath309 , @xmath317 ( @xmath318 . the shape of vortex lines is schematically shown in the insets.,title="fig:",scaledwidth=45.0% ] ) ] ( solid lines ) and straight tilted [ eqs . ( [ straight - tilted ] ) , ( [ straight - tilted2 ] ) ] ( dashed lines ) vortices for an anisotropic film of the thickness @xmath194 with different anisotropy parameters : ( a ) @xmath313 , @xmath315 ( @xmath316 ; ( b ) @xmath309 , @xmath317 ( @xmath318 . the shape of vortex lines is schematically shown in the insets.,title="fig:",scaledwidth=45.0% ] ) , ( [ eq:11c]),([eq:1d ] ) ] vs the intervortex distance @xmath3 in an equidistant chain of n vortices in a stack of decoupled superconducting layers ( @xmath194 ) : ( a ) @xmath319 ; ( b ) @xmath320 . the numbers near the curves denote the number @xmath10 of vortices in molecule . inserts show schematic pictures of vortex matter consisting of dimeric ( a ) and trimeric ( b ) molecules.,title="fig:",scaledwidth=45.0% ] ) , ( [ eq:11c]),([eq:1d ] ) ] vs the intervortex distance @xmath3 in an equidistant chain of n vortices in a stack of decoupled superconducting layers ( @xmath194 ) : ( a ) @xmath319 ; ( b ) @xmath320 . the numbers near the curves denote the number @xmath10 of vortices in molecule . inserts show schematic pictures of vortex matter consisting of dimeric ( a ) and trimeric ( b ) molecules.,title="fig:",scaledwidth=45.0% ] , @xmath309 , @xmath321(@xmath320 ) . the numbers near the curves denote the number n of vortices in a molecule . the shape of vortex line and effective tilting angle @xmath119 are schematically shown in the inset.,scaledwidth=50.0% ] vs the relative displacement @xmath233 of vortex sublattices for different tilting angles @xmath322 ( solid line ) and @xmath323 ( dashed line ) and different number of flux quanta per unit cell @xmath324 . ( b ) lattice deformation ratio @xmath232 vs the relative displacement @xmath233 of vortex sublattices for different tilting angles @xmath322 ( solid line ) and @xmath323 ( dashed line ) and different number of flux quanta per unit cell @xmath324 . here we put @xmath325 . the numbers near the curves denote the number @xmath0 of vortices per unit cell.,title="fig:",scaledwidth=45.0% ] vs the relative displacement @xmath233 of vortex sublattices for different tilting angles @xmath322 ( solid line ) and @xmath323 ( dashed line ) and different number of flux quanta per unit cell @xmath324 . ( b ) lattice deformation ratio @xmath232 vs the relative displacement @xmath233 of vortex sublattices for different tilting angles @xmath322 ( solid line ) and @xmath323 ( dashed line ) and different number of flux quanta per unit cell @xmath324 . here we put @xmath325 . the numbers near the curves denote the number @xmath0 of vortices per unit cell.,title="fig:",scaledwidth=45.0% ]

we present observations and models of the behaviour of the hi and hei lines between 1.6 and 2.2 m in a small sample of compact hii regions . as in our previous papers on planetary nebulae , we find that the ` pure ' 1.7007 m 4d3p and 2.16475 m 7g4f hei recombination lines behave approximately as expected as the effective temperature of the central exciting star(s ) increases . however , the 2.058 m 2p2s hei line does not behave as the model predicts , or as seen in planetary nebulae . both models and planetary nebulae showed a decrease in the hei 2p2s / hi br ratio above an effective temperature of 40000k . the compact hii regions do not show any such decrease . the problem with this line ratio is probably due to the fact that the photoionisation model does not account correctly for the high densities seen in these hii regions , and that we are therefore seeing more collisional excitation of the 2p level than the model predicts . it may also reflect some deeper problem in the assumed model stellar atmospheres . in any event , although the normal hei recombination lines can be used to place constraints on the temperature of the hottest star present , the hei 2p2s / hi br ratio should not be used for this purpose in either galactic hii regions or in starburst galaxies , and conclusions from previous work using this ratio should be regarded with extreme caution . we also show that the combination of the near infrared ` pure ' recombination line ratios with mid - infrared forbidden line data provides a good discriminant of the form of the far ultraviolet spectral energy distribution of the exciting star(s ) . from this we conclude that costar models are a poor match to the available data for our sources , though the more recent wm - basic models are a better fit . [ firstpage ] infrared : ism : lines and bands ism : compact hii regions : general

astro-ph0212135

this is the last paper in a series discussing the near infrared hydrogen and helium recombination lines in photoionised nebulae . in the previous two papers ( lumsden , puxley & hoare 2001a , b ; hereafter paper 1 and 2 ) we discussed the properties as measured in planetary nebulae . we showed that the ratio of pure hydrogen and helium recombination lines are a good measure of the hardness of the exciting radiation field from the central star . this is as expected from simple photoionisation models , since the volume of the he@xmath5 zone relative to the volume of the h@xmath5 zone is simply related to the hardness of the radiation field in the 504912 region of the spectrum of the exciting star . in turn this depends on the stellar effective temperature as well as more specific details of the stellar spectrum as related to stellar evolution ( ie stellar winds , luminosity , surface gravity etc ) . for young stars , this ratio is also therefore clearly a constraint on the upper end of the stellar initial mass function ( imf ) . the use of near infrared hydrogen and helium lines as probes of the underlying radiation field was suggested by thompson & tokunaga ( 1980 ) in the study of heavily embedded compact hii regions . with early infrared instruments the near infrared pure recombination hei lines were too weak to be useful . considerable attention was given instead to the hei 2@xmath3p2@xmath3s line at 2.058@xmath0 m . unfortunately this line is largely pumped by the resonance in the hei 2@xmath3p1@xmath3s transition , and can be influenced by collisional excitation from the metastable 2@xmath3s level as well as from the triplet series . at first it was believed that the complexity of this transition could be modelled in a simple fashion . doyon , puxley & joseph ( 1992 ) suggested that the hei 2@xmath3p2@xmath3s to hi br@xmath4 ratio was ideal for studying the initial mass function in starburst galaxies . they tried to take account of the competing factors that led to emission in the 2.058@xmath0 m line using an analytical approach . however , shields ( 1993 ) showed from a full photoionisation treatment that this approach was a poor approximation to the actual predicted line strength . more recent updates to the atomic data used in the prediction of the hei 2@xmath3p2@xmath3s line strength have also been presented by ferland ( 1999 ) . in the previous two papers in this series we were able to show that the observed near infrared hei and hi line strengths did largely follow the predictions of the photoionisation models for planetary nebulae . this is encouraging since the planetary nebulae are at least simple in the sense that the objects we observed only had a single central exciting star , and could be modelled reasonably as expanding spheres . if the model predictions for such simple systems deviated significantly from the observations , then the method of using near infrared hei and hi line strengths as a constraint on the hardness of the stellar radiation field would have been invalidated almost completely . the aim of this paper is to study how well the line ratios compare between model and observation for compact hii regions . hii regions provide a more challenging test of these models for several reasons . they probably contain a central star cluster rather than a single star , so the cumulative effect of the stellar output needs to be estimated . this in turn requires knowledge of the imf . of course , this is the parameter that doyon et al . ( 1992 ) sought to estimate from the hei to hi line ratios . however , in our case , where we are testing the models , we are forced to make some simple assumptions regarding the form of the initial mass function . compact hii regions are also rather more diverse morphologically than the simple planetary nebulae we were considering . finally , there is considerable uncertainty about the form of the stellar radiation field in massive stars . models diverge markedly depending on whether line blanketing due to metals is included or not , whether the effects of a stellar wind are included , and whether a non - lte treatment is applied . martins , schaerer & hillier ( 2002 ) give a useful summary of the properties of the currently available models for dwarfs . in addition to this overall uncertainty , there is also some doubt as to whether very young massive stars have the same effective radiation field as they would if they were on the main sequence . we know that massive young stellar objects tend to have very dense , slow moving , stellar winds ( eg . bunn , drew & hoare 1995 ) . these winds may reprocess much of the ionising radiation from the young stars into non - ionising radiation . the objects studied by bunn et al . do not have substantial hii regions despite having the bolometric luminosity equivalent to ob stars . the aim of this paper therefore is to determine how well we can reproduce the hei / hi line ratios used in papers 1 and 2 using realistic models of compact hii regions . all of our target hii regions are drawn from the compilation of ultracompact hii regions presented by wood & churchwell ( 1989 ) and kurtz , churchwell & wood ( 1994 ) . the objects were chosen to be sufficiently bright in the near infrared to allow high signal - to - noise low resolution spectroscopy to be obtained . the sources have a variety of morphologies , at least as seen in the radio continuum images . the line ratios presented in this paper were taken from spectroscopic data acquired for other purposes . most of the data were obtained using the facility near infrared spectrograph cgs4 at the united kingdom infrared telescope . the high spectral resolution data from which we draw the 2.16475@xmath0 m hei 7@xmath2g4@xmath2f / hi br@xmath4 line ratio comes from a project to map the kinematics of compact hii regions ( eg . lumsden & hoare 1996 , 1999 ) . the original data have a spectral resolution of @xmath6 , sufficient to easily distinguish the satellite hei lines from br@xmath4 itself . full details of the observations and data reduction can be found in the cited papers . the lower resolution data comes from a project to study the near infrared extinction law ( ridge et al . in preparation ) . here observations of the 1.62.2@xmath0 m region were acquired at low spectral resolution ( @xmath7400 ) . all of the data were acquired with cgs4 with the exception of that for g45.45 + 0.06 , where we have used data acquired with the near infrared imager / spectrograph iris on the aat . the raw data were processed , corrected for atmospheric absorption and flux calibrated in the usual fashion ( see , eg , paper 2 which discusses data acquired on planetary nebulae using cgs4 , or lumsden & puxley 1996 which deals with similar iris data ) . we measured the fluxes of the hi br series lines , in order to determine the extinction , as well as 2.058@xmath0 m hei 2@xmath3p2@xmath3s and 1.7007@xmath0 m hei 4@xmath1d3@xmath1p from the low resolution data . unfortunately , the h band data for g45.07 + 0.13 are of insufficient signal - to - noise to detect the hi br12 or hei 4@xmath1d3@xmath1p lines . the limits that can be placed are consistent with other data for this object , without actually adding any extra constraints . for both high and low resolution data , we measured the line fluxes from the spectrum summed over the whole object . there was no evidence for a variation in line ratio along the slit in any of our sources . since we are only interested in line ratios with hi br@xmath4 , we have combined the low and high resolution data in our analysis . the line ratios from the high resolution data are not corrected for extinction . the small wavelength range involved implies a correction of less than 1% for all our sources . the low resolution data were corrected by comparing the hi lines of br@xmath4 , br10 , br11 and br12 ( the latter 3 lines lie at 1.737@xmath0 m , 1.681@xmath0 m and 1.644@xmath0 m ) . these line ratios vary by less than 5% over the expected range of electron temperatures and densities ( eg wood & churchwell 1989 ) . we assumed a power law dependence for the extinction so that @xmath8 ( landini et al . the error in the hei 1.7007@xmath0m / hi br@xmath4 ratio is then largely set by the derived error in this extinction correction . the hei 2@xmath3p2@xmath3s line must also be corrected for a high frequency component in the atmospheric absorption that is not removed by standard techniques ( see paper 1 for full details ) . the adopted correction values are given to allow the raw line ratios to be estimated . table 1 presents these values , all the observed line ratios after correction for extinction and the adopted values of @xmath9 . finally we also use data on the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio previously published by doherty et al . ( 1994 ) for g5.971.17 and g43.890.78 . the extinction corrections applied there are derived from a comparison of radio continuum and br@xmath4 flux density . we give the values as taken from the original source . we used the data given in wood & churchwell ( 1989 ) and kurtz et al . ( 1994 ) in order to constrain the imf for the exciting stars in these sources . table 2 gives the basic observational properties of our sample . the radio data gives a measure of the flux of hydrogen ionising photons , whereas iras data gives a measure of the total luminosity . the comparison of radio and far infrared data therefore gives us information as to whether the exciting source in an hii region is a single star or a cluster of stars since the two observable properties depend on stellar mass ( and the imf ) in a slightly different way ( see section 3.2 ) . the radio data actually provides a lower limit to the flux of hydrogen ionising photons . the interferometric data presented in wood & churchwell and kurtz et al . are not sensitive to emission on scales much larger than 1030 arcseconds . therefore such extended flux is missed . this is clearly true for g19.610.23a where we have instead used the fluxes presented by garay et al.(1998 ) . kurtz et al . ( 1999 ) demonstrate that such extended emission is commonplace in compact hii regions . such ` compact ' hii regions tend to show a compact dense core , which is what was detected by wood & churchwell ( 1989 ) and kurtz et al . ( 1994 ) , surrounded by low density extended ionised gas emission . we also do not correct for any dust present in the hii region , which also acts to depress the derived flux of hydrogen ionising photons . the work of hoare , roche & glencross ( 1991 ) suggests this is a reasonable approximation . finally we note that we have used the distance estimates given by wood & churchwell ( 1989 ) and kurtz et al . there is evidence that these are on average too large ( see , eg , araya et al . 2002 , which gives distances for six of our targets ) . however , the effect of a slightly too large distance more or less compensates for any missing radio emission , so the actual net effect is small when considering the radio data . by contrast with the radio data , the observed iras fluxes may well be too large , since the large iras beam can encompass emission from neighbouring sources given the high source density in the galactic plane . inspection of higher ( 18@xmath10 ) resolution 21@xmath0 m images from the msx satellite mission ( see , eg , price et al 2001 for details of this mission ) clearly shows the source multiplicity within the iras beam in many cases . the most discrepant are g5.890.39 , g19.610.23a , g35.201.74 and g45.07 + 0.13 , where the observed msx 21@xmath0 m flux densities are only 53% , 65% , 63% and 65% respectively of the iras 25@xmath0 m flux density . in addition , w51d is only a small part of the w51 hii region ( and merges with other sources even in the msx beam ) . therefore , under the assumption that most of the bolometric luminosity is reprocessed by dust , the far infrared luminosity as seen by iras gives an upper limit to the true bolometric luminosity . in addition , the likely overestimate in the distances should be borne in mind since this also inflates the estimated iras luminosity ( by a factor of @xmath11 typically if the distances in araya et al . 2002 are correct ) . any photoionisation model of a hii region requires some assumption as to the form of the spectral energy distribution of the exciting star(s ) . we used the costar model stellar atmospheres as our main input ( schaerer et al . 1996a , b ) . these have the advantage of being widely available and used in such photoionisation modelling , and therefore provide a benchmark against which we can test the observations . the costar models were placed on a physical luminosity - effective temperature scale by schaerer & de koter ( 1997 ) using the calibrations of vacca , garmany & shull ( 1996 ) who presented the fundamental relations between observational properties of o stars . we used the predicted values for @xmath12 from schaerer & de koter to normalise our model set . the vacca et al . calibration may be inaccurate however ( see , eg , the discussion in herrero 2003 , martins , schaerer & hillier 2002 , or smith , norris & crowther 2002 ) . the vacca et al . calibration relied on results from pure hydrogen and helium non - lte models and is likely to overestimate the effective temperature for a given spectral type by as much as 1015% ( see , eg , herrero 2003 for an overview ) . martins et al . ( 2002 ) show the likely effect for the models of most interest to us , the dwarfs , with the effective temperature being too hot by at least 1500k for all o types . crowther et al . ( 2002 ) and herrero ( 2003 ) show that the same effect holds for supergiants as well . there is also observational evidence that the assumed absolute v band magnitude ( and hence luminosity ) derived by vacca et al . is overestimated slightly for the hotter dwarfs ( compare figure 1 in smith et al . 2002 with figure 6 in vacca et al ) . in order to estimate the extent of the effect that the use of the costar grid and the vacca et al . calibration has on our results we also used data from a smaller grid of models presented by smith et al . these are based on the wm - basic models of pauldrach , hoffmann & lennon ( 2001 ) , but use the smith et al . luminosity calibration . we adopted these models since they span the range of effective temperature we are interested in as well as being readily available , though other similar models are now available ( see , eg , herrero 2003 for a summary ) . smith et al . ( 2002 ) show that the hard ultraviolet flux in the costar models is enhanced compared to the wm - basic models ( eg see their figure 4 ) . this is likely due to a fuller treatment of the line blanketing in the wm - basic models ( see also the discussion in section 4.2 of giveon et al.2002 regarding the comparison of wm - basic models with simple lte models ) . we have also included an extension to the effective temperature sequence for stars with masses @xmath13 . these are taken from puxley ( 1988 ) and approximate the behaviour of lower mass stars . the stellar atmospheres used in these models are the lte models from kurucz ( 1979 ) . these models are known to be incorrect to within a factor of a few . therefore we only use these models to demonstrate why we can largely ignore the lower mass stars in our analysis , and do not use them in any detailed photoionisation modelling . if there is a cluster present , then we need to know the form of the stellar imf before we can compare models with observations . scalo ( 1998 ) reviews our current understanding of the imf in depth . for our purposes however we adopted the relatively simple imf proposed by salpeter ( 1955 ) , where the imf is described solely by a power law , @xmath14 this is sufficiently close to our present knowledge of the upper end of the imf that it is still a reasonable approximation . the fact that it is a poor approximation to the lower end of the imf is unimportant because the steepness of the dependence of the ionising photon flux on stellar mass means we need only consider massive stars in any event . the number of hydrogen ionising photons for a given stellar cluster is then @xmath15 where @xmath16 is the number of hydrogen ionising photons at a given stellar mass . a similar relation holds for the total luminosity , @xmath17 . figure 1 shows how the combinations @xmath18 and @xmath19 behave as a function of stellar mass for the costar models . this clearly demonstrates why we can use the comparison of iras and radio data to constrain the form of the imf . the former has a much steeper dependence on @xmath20 than the latter , so that @xmath17 is much more sensitive to the presence of lower mass stars . although the details of the functional form differ slightly , the same is true for the wm - basic models presented in smith et al . ( 2002 ) as well . in figure 2 we plot the actual comparison of @xmath12 and @xmath17 for our sample . the lower solid curve represents the track followed by a single star of increasing mass . the points marked are for ( a ) the dwarf costar models listed in table 3 of schaerer & de koter ( 1997 ) and ( b ) the hottest 9 dwarf models listed in table 1 of smith et al . the upper curve presents the track of a stellar cluster . the points marked here represent the same spectral types as for the single stars , but here these are the most massive stars present in the cluster . we have normalised these curves so that there is exactly one star with that spectral type . other possible combinations such as a steeper imf ( as perhaps indicated by the review of scalo ) , or a truncated imf at both upper and lower mass limits lie somewhere between the two curves shown . the wm - basic models presented by smith et al . ( 2002 ) , using their luminosity calibration , largely overlie the tracks shown for the costar models , but with the points offset approximately 1.5 sub - types towards lower luminosity than the costar models . we have also plotted the direction in which the observational errors described in section 2.2 move the observed luminosity and @xmath12 points in figure 2 . the plotted tracks show the effect of a decrease in the bolometric luminosity by a factor of two ( the extreme suggested by the comparison of the iras and msx data above ) , an increase in @xmath12 by a factor of three ( roughly the case presented by the example of g19.610.34a ) , and a decrease in distance by a factor of two as suggested by araya et al . the distance errors move the observed data approximately parallel with the model tracks , so any large discrepancy between the observed data and the models is unlikely to be due to this source . most of the observed data cluster around or above either the cluster line or single star line in figure 2 . errors of the magnitude shown are certainly sufficient to move all of the sources onto either cluster or single star tracks . we have estimated two characteristic effective temperatures for each of our sources for both costar and wm - basic stellar models . these are estimated simply by locating the nearest model that lies directly below the observed data point . we have not interpolated these temperatures . the results are given in table 2 . the first value given in table 2 is that for the single star required to match the radio data . the second is the hottest star in the cluster that provides a best match to the observed radio and far infrared data . since none of the errors move the objects from significantly below the cluster track to above it , we assume that sources below the cluster track are powered by single stars only . for these cases therefore we do not give a type for the most massive star in a cluster . the likely observational errors imply that a given source may actually be up to two sub - types hotter than this and as many as five sub - types cooler . note that although the costar and wm - basic model sequences in figure 2 do not overlie each other in terms of the actual o sub - type , the same effective temperatures tend to lie in the same place in these diagrams . this explains the rough equivalence between the temperatures of the two model sets given in table 2 . any differences are likely due to the coarseness of the model grids , especially for the smith et al . models . in the actual plots of the observed data we therefore use only the effective temperatures derived from the costar models because the sampling is better . we used cloudy version 94.0 ( ferland 2000 ) to calculate the emission line properties for our model hii regions . we ran a grid of models covering the full range of available stellar effective temperature ( @xmath21 ) for the costar model grid , for all three luminosity classes . although we are interested in young massive stars , there have been suggestions previously that supergiant models are actually a better match to the observational data . this is probably in line with the known properties of some massive young stellar objects , since there is evidence that their spectra bear a resemblance to hypergiants ( eg hamann & persson 1989 ) . in addition , we also ran a smaller set of dwarf and supergiant models using the wm - basic set from smith et al.(2002 ) . we adopt a ` standard ' model where we compare the different stellar inputs , with a flat electron density of @xmath22@xmath23 , a turbulent velocity of 5kms@xmath24 , inner and outer radii to the nebulae of 0.01 pc and 0.5 pc respectively ( see below ) and input stellar models appropriate to solar metallicity . where we vary a given parameter , we kept the other parameters fixed at the values for this standard unless otherwise noted . we assume throughout that the helium abundance is 10% by number when computing the models , but allow the metal abundance to vary in line with the stellar model for the wm - basic model set . we also varied three other physical parameters in the models we considered in order to determine which factors were important in influencing the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio . these were the electron density ( and density structure ) , the turbulent velocity within the ionised gas and the metallicity of the exciting star . we used the the costar model set when considering the effect of density and the microscopic velocity field , and the wm - basic set when considering metallicity . the first two parameters are independent of the nature of the stellar model , whilst only the wm - basic model set spans sufficient metallicity range to be worth considering . we constructed models spanning all five of the metallicities considered by smith et al.(2002 ) . we considered a range of turbulent velocities , from 0 to 15kms@xmath24 . the hei 2@xmath3p2@xmath3s line is expected to be sensitive to local velocity broadening ( hence turbulence ) because of the resonance pumping from the 2@xmath3p1@xmath3s transition ( see paper 1 for more details ) . the range chosen matches the typical range seen in compact hii regions . for the wm - basic models we only used a value of 5kms@xmath24 throughout . we considered two possible density models for the hii regions in the costar models : a flat constant density model , with @xmath25 varying between 1000 and 50000@xmath23 , and a model with a flat baseline density @xmath26@xmath23 with a superposed radial gaussian density profile , with peak density again varying up to 50000@xmath23 . in both cases the inner surface of the cloud was 0.01 pc from the exciting source , and a nominal 0.5 pc outer radius was adopted . the latter matched the observed radii of our sources to within a factor of a few in all cases , though the actual model results are relatively insensitive to this parameter . we adopted a full width at half maximum of 0.3 pc for the gaussian component . the radii were the same in the wm - basic models , but we only considered the flat density law with @xmath22@xmath23 . we also considered two sets of models with a larger inner radius . in the first , we set the inner radius to the cloud from the exciting star of 0.3 pc , but otherwise used the same set of parameters as the flat density models . this radius is actually larger than the total size of many of the objects being considered , and so this is not a realistic physical model of these sources . it can be thought of as a crude way to approximate the effect of using stellar models with a lower ionisation parameter since the net effect is to reduce the ionising flux at the hii region . we define this parameter as the dimensionless ratio of the photon flux to the hydrogen density at the inner surface of the hii region ( note , not the stromgren radius as in the usual definition : our definition agrees with that used in cloudy ) , so that @xmath27 where @xmath28 is the inner radius of the hii region , and @xmath29 the total hydrogen density . trials showed that we had to increase this radius to @xmath30 pc before an effect was seen in the results ( models with inner radius of 0.1 pc are similar to the models with inner radius of 0.01 pc ) . this corresponds to a drop of @xmath31 in the ionisation parameter . the outer radius for these models was increased to 2 pc . the second large inner radius set we considered used costar models with an inner radius of 1 pc and an outer radius of 5 pc . this model assumed a flat electron density profile , and only a model with @xmath32@xmath23 and turbulent velocity of 5kms@xmath24 was calculated . figure 3 and 4 show the results from our cloudy simulations for the different sets of stellar inputs ( luminosity class and model type ) for the key hei 2@xmath3p2@xmath3s ( figure 3 ) and hei 4@xmath1d3@xmath1p ( figure 4 ) to hi br@xmath4 line ratios . in both figure 3(a ) and 4(a ) the results are presented for our standard model for dwarf , giant and supergiant costar model atmospheres , and for dwarf and supergiant wm - basic model atmospheres . the sampling of the wm - basic models is coarser than for the costar models which may partly explain the shift in the peak between the costar and wm - basic dwarf models . the costar models as a whole show greater variance with type than the wm - basic models for both line ratios however . as noted in section 3.1 this is probably a reflection of the fact that the costar models have insufficient line blanketing , which is of greater importance in stars with larger mass loss rate ( eg supergiants as opposed to dwarfs ) . the net result is a spectrum that is ` harder ' at a given effective temperature than it should be , particularly in the far ultraviolet . by comparison , the dwarf and supergiant wm - basic models give very similar predictions ( identical in figure 4(a ) , so only the dwarf models are actually plotted ) . in figures 3(b ) and 4(b ) we consider the effect of having a cluster of stars rather than a single star . we again used the standard model and the costar dwarf model atmospheres . the cluster was constructed so that one star with the spectral type of the maximum effective temperature was present , and the normalisation for the other stars was then set by matching the observed @xmath12 and @xmath33 for a cluster as in figure 2 . there is clearly little difference in the hei / hi ratios between the single - star and the cluster case . this is in line with figure 1 , which clearly shows that the number of helium ionising photons drops sharply as the stellar mass decreases below 40(equivalent to @xmath34k for the costar dwarf models ) . these results suggest a prior understanding of the imf is not required in order to place a constraint on the hottest star present using these ratios . in figure 5 we show the results from our standard model for a different line ratio to show that large differences do exist between the dwarf and supergiant cloudy models and the wm - basic models . the line ratio plotted is that of 15.6@xmath0 m [ neiii ] with 12.8@xmath0 m [ neii ] . this ratio is commonly obtained from spectroscopy obtained by the iso satellite of compact hii regions ( eg morisset et al . 2002 , giveon et al . these model results clearly show much wider variation in behaviour than any of the hei / hi line ratios . we have also shown how metallicity can effect such line ratios as well by examining the variation with metallicity for the wm - basic dwarf models . the result is largely as expected if higher metallicity leads to more line blanketing in the far ultraviolet , and hence a softening of the spectral energy distribution ( the difference occurs at energies larger than 2.5 rydberg , due to the change in the species causing the line blanketing : see the extended discussion of this issue in giveon et al . 2002 for fuller details ) . these results highlight both a strength and weakness of the hei / hi ratios in placing constraints on the effective temperatures of the stars present . the main strength is that the results you would obtain are largely independent of the stellar model chosen and metallicity ( since the helium fraction alters with metallicity by a much smaller amount than any of the metals ) . the main weakness is that it can place only weak constraints on the spectral energy distribution of the stellar atmosphere itself . figure 6 shows how the metallicity of the underlying star effects the results . again we show both the hei 2@xmath3p2@xmath3s ( a ) and hei 4@xmath1d3@xmath1p ( b ) to hi br@xmath4 line ratios . there is a weak metallicity dependence in the former , probably as a result of the increase of the neutral helium fraction in the nebula due to the softening of the ultraviolet flux with increasing metallicity . the resonance of the 2@xmath3p1@xmath3s transition is crucially dependent on this neutral helium fraction as first noted by shields ( 1993 ) . by contrast the changes are relatively small in the pure recombination line ratio . figure 7 shows how the ratio varies as a function of the turbulent velocity . increasing the turbulent velocity depresses the ratio , as was seen for the models for the pn presented in paper 1 . this is simply due to the fact that the resonance scattering of the 2@xmath3p1@xmath3s hei line is more likely to lead to capture by dust or atomic hydrogen if the line is much broader than the thermal value . the differences seen are of the same order of magnitude here as in paper 1 , as expected given that this is a test of the physics in cloudy and not really of the form of the input stellar models . figure 8 shows the effect on the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio of changing the electron density ( the pure recombination line ratio is independent of density ) . figure 8(a ) shows the data for a flat density distribution , and figure 8(b ) the data for the gaussian density distribution . there is clearly little difference between the two , and little apparent dependence on density . this is perhaps surprising since it is well known that the 2@xmath3s level of helium can be pumped by collisional excitation from the triplet 2@xmath35s ` ground - state ' ( see , eg , doherty et al . 1994 ) . in practice these cloudy models predict the 2@xmath35s level is primarily depopulated by photoionisation instead ( see also the discussion in clegg & harrington 1989 ) however . figure 9 shows the results for a flat model with a larger inner nebular radius . the solid lines show the result of increasing the radius to 0.3 pc , with the same densities as those shown in figure 8(a ) , whilst the dashed line shows the result for an inner radius of 1 pc for a single electron density . it must be stressed that these models should only be appropriate for older more extended hii regions . however , it is clear that a reduction in the ionisation parameter at the inner edge of the nebula does have a striking effect on the line ratio . the most obvious feature is the enhanced density dependence of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio , clearly indicating the growing importance of collisional excitation of the 2@xmath3s level . this is a reflection of the much higher ultraviolet photon density at the inner surface of the hii region in our other models . the effect seen in the models with a larger inner radius should in principle be seen in our ` normal ' models if we increased the density significantly beyond 10@xmath36@xmath23 since the collisional rate may then dominate the photoionisation rate . such densities are not in agreement with the published estimates from radio data however . it is also worth noting the fact that the very extended hii region model shows a much shallower decline of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio with increasing effective temperature . the overall trends seen in the models can be summarised fairly simply . there are weak variations in the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio with almost all the parameters considered , but only the increase of the inner radius combined with changes in the electron density have a significant impact . by contrast , the pure recombination line ratios are largely independent of most of the factors we have considered . figure 10 shows the observed hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio together with a representative sample of the costar models from section 3 . we have not plotted the wm - basic model here since it is very close in appearance to the costar dwarf model ( figure 3 ) . the behaviour seen is almost completely at variance with the models . one of the two lowest temperature sources plotted here ( g5.971.17 ) lie well above the predicted curve . the nature of this source is not entirely clear , since it lies very near the exciting star of m8 , herschel 36 . stecklum et al . ( 1998 ) have speculated that the emission actually arises in material near a young , but much less massive , star which is being photo - evaporated by the radiation of herschel 36 . if true , the actual effective temperature of the exciting star is that of herschel 36 ( an o7v star ) , and the point should be shifted to higher @xmath21 on the figure . the other key point to note from figure 10 is that the observed ratio at @xmath37k is always higher than the theoretical prediction unless we adopt the lowest ionisation parameter model . there is also no downturn in the observed ratio as @xmath21 increases beyond 40000k , as predicted in all the models . it must be noted however that the largest inner radius model declines more slowly than the others and is hence a better , though still poor , match to the data . the ` pure ' recombination line ratios are shown in figure 11 . we have also shown a representative spread of the stellar models in figures 11(a ) and ( b ) , with dwarf costar models in ( a ) and dwarf wm - basic models in ( b ) . since the 7@xmath2g4@xmath2f/4@xmath1d3@xmath1p ratio is effectively constant , the wm - basic model for ( a ) would look the same as in ( b ) but with the plateau value the same as for the costar models ( and vice - versa for the costar models not shown in ( b ) ) . the low ionisation parameter / large inner radius models shown in figure 10 are not shown in figure 11 since they given essentially the same results as the dwarf models plotted . in general , these models agree reasonably well with the dwarf models at @xmath38k . there is some discrepancy between dwarf models and observation at lower effective temperature . again , in part , this is due to g5.971.17 . there is some evidence from the data however that the ` plateau ' in these ratios arrives at a slightly cooler effective temperature than the models predict , though the difference is certainly within the errors on the adopted object effective temperatures given in table 2 . the other truly discrepant point is that for g5.890.39 , which lies well below the predicted model curve for the hei 7@xmath2g4@xmath2f / hi br@xmath4 ratio . it may also lie below the model hei 4@xmath1d3@xmath1p / hi br@xmath4 ratio but the observational error for this ratio is too large to draw any conclusions . we discuss this source in more detail below . overall , however , our results are in accord with the study of a large sample of optically visible , and hence more evolved hii regions , by kennicutt et al . they found that the ratio of optical hei and hi lines were in general agreement with the predictions from cloudy using main sequence models for the exciting stars . our results tend to indicate that the same is true regardless of the stellar model used . it is worth briefly considering whether the data might match one of the costar supergiant models shown in figures 3 and 4 . there have been suggestions that such models sometimes give a better match between theory and observations to other line ratios in compact hii regions ( see section 4.2 below for an example ) . note from figures 3 and 4 there is no difference between dwarf and supergiant wm - basic models , so we do not consider these here . of course the observational data as plotted in figure 10 and 11 assume the exciting stars have effective temperatures compatible with the costar dwarf models . in practice , if we define the spectral type using costar supergiant models in the same fashion as figure 2 , then all of the sources fitted by a single exciting star actually lie well below the end of the costar model sequence . the others have types ranging from o7i / o8.5i downwards depending on whether a single star model or a cluster is used . therefore the observational data should all be shifted to the left by @xmath39k when comparing the observational data in figures 10 and 11 with the supergiant models in figures 3 and 4 . this clearly makes most of the observational data disagree completely with the model . the problem is actually exacerbated if we consider a cluster of exciting stars . it seems clear that supergiant costar models are not a sensible match to the hei / hi line ratio data in general . figure 12 shows the ratio of the hei 2@xmath3p2@xmath3s line with the two ` pure ' hei recombination lines . we showed in paper 2 that this ratio helped remove any abundance effects when comparing observational data with the predictions from cloudy assuming a single helium abundance . the data here however are actually more consistent with a flat line than the predicted curves . the results therefore show that the cause of the high hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio is not an abundance effect , but is intrinsic to the strength of the hei 2@xmath3p2@xmath3s line . even the model with a 1 pc inner radius and high electron density fails to explain the observed data . these results are completely different from those for the planetary nebulae , where these hei / hei ratios generally showed reasonable agreement between model and observation . we must stress however that these hii regions have electron densities that are generally a factor of @xmath4010 higher than in the planetary nebulae . it is worth noting that there is a small discrepancy between the observed and model hei 2@xmath3p2@xmath3s / hi br@xmath4 line ratio even in the planetary nebulae ( see figure 4 and 5 of paper 2 ) , in the sense that the hei 2@xmath3p2@xmath3s line is stronger than expected by 1020% from the mean model trend . as noted above , we expect the hei 2@xmath3p2@xmath3s to be stronger if the collisional excitation rate from 2@xmath35s exceeds the photoionisation rate . if the cloudy estimation of these two rates is inaccurate ( or equally if our assumed model inputs for inner cloud radius and electron density are too small ) , then it is possible that the collisional excitation rate actually dominates in these compact hii regions in reality . however , even allowing for this , it is notable that the hei 2@xmath3p2@xmath3s / hi br@xmath4 line ratio does still decrease at higher effective temperatures for the planetary nebulae , and we do not see that effect here . it seems clear that at least one other factor must play a role in explaining the poor match between model and observed hei 2@xmath3p2@xmath3s / hi br@xmath4 line ratios for the compact hii regions . the most important caveat in these general conclusions is that we rely on the data shown in figure 2 to derive @xmath21 . we have used the data as presented in wood & churchwell ( 1989 ) and kurtz et al . as noted in section 2 , there are reasons for believing that at least some of the distance estimates they use are too large . a lower distance estimate may help to explain some of the problem for those sources in figure 10 which lie at @xmath37k , but which lie well above the model hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio , since it would move the observed points to lower @xmath21 . however , it can not be too much lower , since figure 11 shows reasonable agreement between model and data . there are several parameters that we have not discussed so far that may affect our conclusions . these are best illustrated by considering specific examples . a useful example is g29.960.02 . this object has been discussed at length in the literature . there are good arguments for believing that it may be at a distance of 6 kpc rather than the 9 kpc given in wood & churchwell ( 1989 ) , or even the 7.4 kpc estimated by churchwell , walmsley & cesaroni ( 1990 ) ( see pratap , megeath & bergin 1999 for a fuller discussion regarding the distance to this object ) . g29.960.02 is more extended in the radio than indicated by the radio map of wood & churchwell ( 1989 ) . fey et al . ( 1995 ) derive a value of @xmath12 which is @xmath4050% higher than wood & churchwell , largely compensating for any reduction in distance with regard to this indicator ( see also the discussion in section 2.2 ) . using the fey et al . data and a distance of 6 kpc @xmath41s@xmath24 . the bolometric luminosity however does decrease to @xmath42 . the net results is a drop of half a sub - type in the best fitting stellar dwarf models , with a cluster still being preferred over a single star ( see figure 2 ) . the expected pure hei / hi recombination line ratio for such a model is still on the plateau seen at high @xmath21 in figure 11 , in agreement with observation . there have been many previous indirect estimates of the spectral type of the exciting star(s ) in this hii region , in the same spirit as this paper but primarily using mid - infrared forbidden line data . the most recent of these by morisset et al . ( 2002 ) , based on iso data , concluded that the best matched exciting star was actually a relatively cool supergiant ( @xmath40o9.5i , or @xmath43k using costar models ) . this actually disagrees with our pure recombination line ratios ( figure 11 , and see the discussion in section 4.1 ) , and hence we can rule that model out . such a model can also probably be ruled out on the basis of other evidence . the presence of young hot cores ( eg . cesaroni et al . 1998 , de buizer et al . 2002 ) near the hii region strongly suggest that g29.960.02 is part of a complex of star formation taking place on a larger scale ( also consistent with the complexity seen on larger scales at radio wavelengths by kim & koo 2001 ) . furthermore our own work on the kinematics of this object ( lumsden and hoare 1996 , 1999 ) indicate an evolved star travelling through a molecular cloud is not a good match to the data . it is worth briefly considering why morisset et al . reached the conclusion they did . there are several line ratios in the mid and far - infrared that are abundance independent . these include , limiting ourselves to the better studied mid - infrared ratios only , 6.99@xmath0 m [ arii]/8.99@xmath0 m [ ariii ] , 18.7@xmath0 m [ siii]/10.5@xmath0 m [ siv ] , and 12.8@xmath0 m [ neii]/15.6@xmath0 m [ neiii ] . the latter two are of particular interest since s@xmath44 and ne@xmath5 have approximately the same ionisation potential as he@xmath45 . therefore , in principle , these ratios add extra constraints on the form of the stellar spectral energy distribution beyond 504 . here we will only consider the 12.8@xmath0 m [ neii]/15.6@xmath0 m [ neiii ] ratio in detail . morisset et al . give an extinction corrected observed ratio of 3.75@xmath46 ( the observed ratio is close to 2 ) . as can be seen from the inspection of figures 4 and 5 a ratio this low , together with the observed hei / hi ratio actually rules out all costar models . morisset et al . adopted a low value for the hei / hi ratio , based on weak helium radio recombination line data from kim and koo ( 2001 ) , allowing a fit to the data with cool costar supergiant models . our result is certainly more reliable , and hence we can exclude this model as previously noted . indeed , it is not just g29.960.02 where the combination of our near infrared data and published mid - infrared data leads to problems for the costar models . if we accept the best matching costar models to the mid - infrared data all of g5.890.39 , g29.960.02 , g43.890.78 , g45.07 + 0.13 , g45.12 + 0.13 , g45.45 + 0.06 and w51d would have temperatures inconsistent with the hei / hi ratios presented in figure 11 ( see the iso data presented in giveon et al . , or the iras lrs data in simpson & rubin 1990 ) . there is a caveat here , since all of these data were acquired with a large beam and sample most if not all of any extended emission present , whereas our results are largely sensitive only to the dense cores of the hii regions observed . however , this overall trend does indicate a problem with the costar models , in the sense that the far ultraviolet spectral energy distribution is harder than observed . by comparison , the wm - basic models allow a solution to both near and mid - infrared line ratios for @xmath47k ( the exact temperature is somewhat metallicity dependent ) . this range gives reasonable agreement between model and data for all three of the mid - infrared line ratios listed above for the case of g29.960.02 . this is perhaps surprising considering we have not run a detailed physical model such as the one morisset et al . use . however , the agreement is basically a reflection of the fact that at a given effective temperature the wm - basic models are softer in the far ultraviolet region of the spectrum than the costar models ( see also section 4.2 of giveon et al . 2002 ) , and appear to be a better match to the data for most hii regions ( again see giveon et al ) . the most useful direct constraint for g29.960.02 lies in a spectrum of the main exciting star ( watson & hanson 1997 , hanson 1998 , kaper et al . the published spectral type for the star varies significantly between these references ( as hot as o3 in kaper et al . , as cool as o8 in watson & hanson ) , but the most likely range is o4o6 from the strengths of the civ , hei and heii lines present ( hanson , private communication ) . this range is consistent with the photometry of watson et al . ( 1997 ) , especially if we use a lower value for the extinction at @xmath48 of 1.6 ( as proposed by morisset et al . 2002 , and as determined from a preliminary calibration of our own data for studying the extinction law ) . although this would have an absolute magnitude at the limit for a main sequence o4 star ( eg . vacca et al . 1996 or hanson , howarth & conti 1997 ) , it is still just consistent with that type if the distance is only 6kpc ( the cooler types are all consistent ) . it is worth noting that this range of spectral type is still hotter than expected from the photoionisation models . this is not surprising given the uncertainty over the effective temperature / spectral type relation however ( and the implication from all of the more recent work that this scale needs to shift to cooler temperatures ) . it does however emphasise that we should still treat the existing stellar models with some caution . another useful individual example to consider is g5.890.39 . the distance to this source may be slightly less than reported in wood & churchwell ( 1989 ) but the effect is small , and the accuracy of the distance measure for this particular source is good ( see acord , churchwell & wood 1998 for a fuller discussion of the distance ) . for either distance however , there is a deficit of helium ionising photons in this source , as is clear from figure 11(a ) , where the observed hei 7@xmath2g4@xmath2f / hi br@xmath4 ratio is well below both any model prediction and the observed values for other sources with the same expected effective temperature . this discrepancy can not be explained by errors in the assignment of a spectral type , since the distance is well known , and the radio emission compact . it is perhaps noteworthy that g5.890.39 may be the youngest source in this sample . the hii region has an age of only a 600 years from estimates of its rate of expansion ( acord et al . there is also a substantial molecular outflow ( acord , walmsley & churchwell 1997 ) and an infrared reflection nebulosity associated with it ( lumsden & hoare in preparation ) , symptomatic of a rather young source . for the case of g5.890.39 it is tempting to consider the possibility that the stellar models used are actually inappropriate , and its young age may be the clue here . as noted previously , there is some evidence that massive young stellar objects appear to share spectral characteristics with hypergiants ( hamann & persson 1989 , bunn et al . both classes of objects have dense slow moving stellar winds , giving rise to this similarity . it is also true that most massive young stellar objects do not show significant hii regions , although they are luminous enough to generate them . it is assumed that the winds act as sinks for much of the ultraviolet flux from the star , suppressing the ability to form an hii region , and probably also softening the spectral energy distribution over that seen in main sequence stars . if the same behaviour was also shown by the central stars of very young hii regions ( perhaps because the atmosphere of the star has not fully attained its main sequence structure ) then it might explain a case such as this . g5.890.39 is actually more luminous than the stars studied by bunn et al . , so even some leakage of ultraviolet photons could lead to a substantial hii region . of course , we can not prove this speculation from our current data . we have presented models for the infrared helium and hydrogen recombination line ratios in compact hii regions . we fail to find any model that can accurately reproduce the hei 2@xmath3p2@xmath3s line strength as a function of effective temperature . we therefore would discourage anyone from using this ratio as a constraint on the imf , and any evidence derived from the literature about the imf in a particular system that relies on this ratio should largely be discounted . our results do show reasonable agreement between model and observation for pure recombination lines such as hei 4@xmath1d3@xmath1p or 7@xmath2g4@xmath2f when ratioed with hi br@xmath4 as long as we use dwarf costar or wm - basic models . this indicates that these are in principle a useful measure of the stellar effective temperature of the hottest star present . the main drawback with these ratios is the relatively small range of effective temperature over which they are useful ( approximately over the range b0 to o8 for main sequence costar models for example ) . however , at least some of the galaxies in which the 1.7007@xmath0 m hei 4@xmath1d3@xmath1p line has been detected do have values of the ratio within this range ( see , eg , table 4 of vanzi & rieke 1997 ) . it should certainly help to address the issue of whether there is evidence for an anomalous imf from infrared data ( see , eg , the review by scalo 1998 ) . we have outlined possible reasons for the discrepancies seen between the observations and the models for the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio . the first , and perhaps most likely , is that the relative importance of photoionisation and collisional excitation from the @xmath49s hei state may be incorrectly estimated by our hii region models . clearly , from figure 9 , models in which the collisional depopulation of @xmath49s is dominant come closer to matching the observed data . however , it is still clear from figure 12 that the hei / hei ratios are closer to a flat function of effective temperature than the declining function predicted by cloudy ( in agreement with the observations of the planetary nebulae in papers 1 and 2 ) . it seems likely that errors in the 2@xmath35s de - population rate may give rise to part but not all of the discrepancy seen . perhaps , instead , the hei 2@xmath3p2@xmath3s line is telling us something about the nature of the ultraviolet spectral energy distribution of the exciting star ? the current uncertainty in the model stellar atmospheres for massive stars ( eg martins et al . 2002 ) is unlikely to resolve this problem entirely however , since we have shown that the hei / hi line ratios have a relatively weak dependence on model when comparing costar and wm - basic dwarf models . the effective reduction in luminosity given by any re - calibration of the spectral type / effective temperature relation to cooler temperatures may help to lessen the differences between the model hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio and the observed data however . models with a more realistic treatment of the stellar wind ( such as the wm - basic set ) also help to reduce the discrepancy between the pure hei / hi recombination ratios and the mid - infrared forbidden line ratios , since the overall shape of the far ultraviolet continuum is clearly crucial in that regard . they also help in closing the gap between the direct spectral type seen in g29.960.02 and the inferred type from the photoionisation models . similarly , consideration of changes in metallicity of the stellar model do not effect the hei / hi ratios significantly , but can effect forbidden line ratios ( eg figure 5 ) . there is evidence for a gradient of metallicity with galactocentric radius , with metallicity increasing inwards and excitation correspondingly decreasing ( eg giveon et al . many of our sources lie within the solar circle and are likely to have exciting stars with higher than solar metallicity . the most discrepant data point for all three helium lines is probably the youngest of the hii regions in the sample used here . we have outlined the possibility for this case that the exciting star may actually resemble those massive young stellar objects studied by bunn et al . ( 1995 ) . in this case a considerable fraction of the emitted ultraviolet flux is actually reprocessed into non - ionising radiation by a dense stellar wind reducing the excitation in the nebular gas . for the moment this must remain purely speculative however . the easiest way of testing whether our results indicate some residual problems within the photoionisation model that are not otherwise apparent from papers 1 and 2 or a deficiency on the assumed stellar models would be to obtain @xmath48 band spectra of older , more evolved hii regions , such as those studied by kennicutt et al . more evolved regions are less dense , and larger , and hence any residual uncertainty in the depopulation route from 2@xmath35s should be removed . if the same effect as seen here were repeated then it would tend to indicate that even the latest stellar models are still not well matched to the actual ultraviolet continuum of ob stars . sll acknowledges the support of pparc through the award of an advanced research fellowship . we would like to thank the referee paul crowther for his helpful comments , margaret hanson for her comments on the likely spectra type of the exciting star in g29.960.02 and richard norris for making available his code to read the wm - basic models into cloudy = 0.5 in = 1 acord , j.m . , walmsley , c.m . , churchwell , e. , 1997 , apj , 475 , 693 * table 1 : * the observed hei / hi line ratios after correction for extinction and atmospheric absorption . ratios with hi br@xmath4 are shown . there are no suitable observational data for entries marked . note that doherty et al.(1994 ) do not quote errors on @xmath50 . the correction factors for the 2.058@xmath0 m line represent the additional correction necessary to recover the ` true ' 2.058@xmath0 m line flux . the quoted ratios should be divided by this factor to derive the observed ratios . * figure 1 : * the dependence of @xmath51 ( left ) and @xmath52 ( right ) as a function of stellar mass @xmath20 for the main sequence costar models . we have plotted curves representing both the helium and hydrogen ionising continuum . the data for masses @xmath53come from puxley ( 1988 ) , using kurucz ( 1979 ) models , since there are no costar models for this range . there is a clear discontinuity at the transition as seen in the right hand panel , but the actual results with regard to the expected total luminosity and ionising photon rate for a cluster of stars are not particularly affected by this . * figure 2 : * the observed properties of our sample as taken from table 2 . the model tracks represent single stars ( lower ) and clusters of stars ( upper ) from ( a ) the costar model set and ( b ) the wm - basic model set . the location of the specific models given in table 3 of schaerer & de koter ( 1997 ) are marked by @xmath54 in ( a ) . these span the range b0.5v to o3v . cluster models were also constructed spanning this range . these were normalised so that at any upper mass cut - off , there was exactly one star of that spectral type . the locations of these cluster models are marked by @xmath55 . the solid line for the single stars shows the extension to lower masses assuming the use of the kurucz ( 1979 ) models as described in the text . the locations of the wm - basic models given in table 1 of smith et al . ( 2002 ) are similarly indicated in ( b ) . note that the smith et al . grid has coarser sampling than the costar set . the arrows indicate the magnitude and direction of the likely error in the observational data . * figure 3 : * models of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio as ( a ) a function of the luminosity class of the exciting star , and of the input stellar model set , and ( b ) a model stellar cluster . in ( a ) the costar models are shown as solid lines , and the wm - basic models as dotted lines . note that the wm - basic grid is coarser than the costar grid , so the significance of the shift in temperature of the peak in the line ratio between costar and wm - basic dwarf models is not clear . the larger line ratio at high effective temperature seen with the wm - basic models may in part be due to the lower luminosity of the dwarf models given in smith et al . ( 2002 ) . in ( b ) the comparison is between the results for the costar dwarf models when considered as a single star ( dotted line ) and a cluster with standard imf . clearly this line ratio depends on the temperature of the hottest star present more than any other feature of the imf . * figure 4 : * the same models as shown in figure 3 but for the hei 4@xmath1d3@xmath1p / hi br@xmath4 ratio . note the relatively small offset between the costar and wm - basic dwarf models in ( a ) , indicating that the comparison of hei / hi line ratios is not a good test of the detailed shape of the spectral energy distribution of the exciting star(s ) . clearly ( b ) shows there are no differences for this ratio between a single star and a cluster . * figure 5 : * a similar comparison to figure 3 and 4 but using the mid - infrared 15.6@xmath0 m [ neiii ] to 12.8@xmath0 m [ neii ] line ratio . this plot shows that there are significant detectable differences between the stellar models with forbidden metal lines . it also shows the role that metallicity plays in determining such a line ratio , with a difference of a factor of two or more in the final ratio . * figure 6 : * the dependence of ( a ) the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio and ( b ) the hei 4@xmath1d3@xmath1p / hi br@xmath4 ratio as a function of metallicity using the wm - basic model set . clearly , metallicity has a weak effect on determining these line ratios . * figure 7 : * variation of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio as a function of turbulent velocity . the models shown use costar dwarf models and the turbulent velocity has values of 0 , 5 , 10 and 15kms@xmath24 . the result is very similar to that obtained for planetary nebulae in paper 1 . * figure 8 : * models of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio for a variety of electron densities . in ( a ) , the density is constant throughout the hii region , with values of @xmath561000 , 3000 , 6000 , 10000 and 50000@xmath23 . in ( b ) , a constant density of @xmath561000@xmath23 is superposed on a gaussian radial distribution with peak density of @xmath561000 , 3000 , 6000 , 10000 and 50000@xmath23 . the gaussian has a full width at half maximum of 0.3 pc . some of the specific models are indicated on the plot . all models use costar dwarf stellar atmospheres . * figure 9 : * models of the hei 2@xmath3p2@xmath3s / hi br@xmath4 ratio for an extended hii region excited by a single costar dwarf stellar atmosphere . we assume a flat electron density . the solid lines show the same densities as in figure 8(a ) with an inner nebular radius of 0.3 pc . the dashed line shows a single density but with an inner radius of 1 pc . these models correspond to a progressive decline in the ionisation parameter at the inner surface of the nebula . the peak in the ratio clearly shifts to higher effective temperature , and the role of collisional de - excitation of the 2@xmath35s level must also increase with decreasing ionisation parameters to explain these results . * figure 10 : * observed ratios of hei 2@xmath3p2@xmath3s and hi br@xmath4 . vertical error bars are actual observed errors . horizontal error bars represent the likely span in the effective temperature of the hottest star present . the extreme right point of the error bar represents @xmath21 for a single exciting star , the extreme left point represents @xmath21 for a cluster as outlined in section 2 . therefore the observed data point should be thought of as lying at either extreme of the error bar , but not the central position . real observational error in our indirect spectral typing from figure 2 can shift these points as much as 4000k . the models are a selection of those from figures 7 , 8 and 9 , showing the range given by varying electron density , turbulent velocity and ionisation parameters for the costar dwarf model set . the solid lines show models with @xmath32@xmath23 and turbulent velocity of 0kms@xmath24 ( upper curves ) and @xmath26@xmath23 and turbulent velocity of 15kms@xmath24 ( lower curves ) . the dotted lines show the same models but with inner nebular radius of 0.3 pc . the dashed line shows a model with inner nebular radius of 1 pc , @xmath32@xmath23 and turbulent velocity of 5kms@xmath24 . the observed ratio is not a good fit to any of these curves , though the models in which collisional de - excitation dominates photoionisation of hei 2@xmath35s are a better match . * figure 11 : * observed ratios of ( a ) hei 7@xmath2g4@xmath2f and ( b ) hei 4@xmath1d3@xmath1p with hi br@xmath4 . vertical error bars are again actual observed errors , and horizontal error bars have the same meaning as in figure 10 . the models shown in ( a ) are for costar dwarfs , and in ( b ) for wm - basic dwarfs . there is essentially no variation of these curves with density , turbulent velocity or radial density profile . the observed data are mostly a reasonable match to both costar and wm - basic dwarf models . the match argues that our indirect spectral typing is also reasonable . * figure 12 : * observed ratios of hei 2@xmath3p2@xmath3s with ( a ) hei 4@xmath1d3@xmath1p and ( b ) hei 7@xmath2g4@xmath2f . the models plotted are the same as in figure 10 , and the error bars have the same meaning as there . again high density low ionisation parameter models are the best match .

gap formation in a gas disk triggered by disk - planet tidal interaction is considered . density waves launched by the planet are assumed to be damped as a result of their nonlinear evolution leading to shock formation and its subsequent dissipation . as a consequence wave angular momentum is transferred to the disk , leading to evolution of its surface density . planetary migration is an important ingredient of the theory ; effects of the planet - induced surface density perturbations on the migration speed are considered . a gap is assumed to form when a stationary solution for the surface density profile is no longer possible in the frame of reference migrating with the planet . an analytical limit on the planetary mass necessary to open a gap in an inviscid disk is derived . the critical mass turns out to be smaller than mass for which planetary hill s radius equals disk scaleheight by a factor of at least ( is the toomre stability parameter ) depending on the strength of the migration feedback . in viscous disks the critical planetary mass could vary from to , depending on the disk viscosity . this implies that a gap could be formed by a planet with mass depending on the disk aspect ratio , viscosity , and planet s location in the nebula .

astro-ph0110540

the recent discovery of extrasolar planets on orbits very close to their parent stars ( mayor & queloz 1995 ; marcy et al . 2000 ; vogt et al . 2000 ; butler 2001 ) has raised a number of questions about the formation mechanisms of such systems . these close - in planetary companions are all presumed to be gas giants , typically with masses of the order of jupiter mass @xmath5 g. it is very unlikely that such planets were formed at their present locations ( boss 1995 ) : current theories ( mizuno 1980 ; bodenheimer & pollack 1986 ) predict that giant planets were formed by gas accretion onto massive ( @xmath6 ) rocky core which themselves are the result of accumulation of a large number of icy planetesimals . the most favorable conditions for this process exist beyond the so - called `` snow line '' ( hayashi 1981 ; sasselov & lecar 2000 ) which is estimated to lie several au from the star , far larger than the actual orbital radii of the known extrasolar planets . the most popular theory explaining this paradox presumes that giant planets were indeed formed far outside their present locations but then migrated inwards as a result of tidal interaction with a gaseous nebula ( ward 1997a ) . if a planet is fully immersed in a gas disk then the migration process works as follows : planetary gravitational torques produce density waves carrying angular momentum in the surrounding gas disk . for the planet to migrate inwards ( outwards ) two conditions must be fulfilled : ( 1 ) interaction with the outer ( inner ) part of the disk should be stronger than with the inner ( outer ) one , and ( 2 ) , the waves must not return to the planet . in a finite disk the only way to fulfill the second condition is for the waves to dissipate so that their angular momentum gets transferred to the disk flow . it was demonstrated by ward ( 1986 ) that in keplerian disks conditions are usually such that the planet tends to migrate inwards . the typical timescale of this process , @xmath7 yr for a jupiter - mass planet , is very short compared with the lifetime of the nebula itself ( @xmath8 yr ) . this short timescale presents a significant problem for the migration scenario because the migrating planet is likely to drift right into its parent star in very short time . various mechanisms to _ stop _ migration have been proposed : formation of a cavity in the inner disk by magnetospheric activity of the central star ( shu et al . 1994 ) , resonant interaction with another planet in the system ( masset & snellgrove 2001 ) , spin - orbital interaction with the central star ( trilling et al . 1998 ) , etc . on the other hand , there is a process almost inevitable in the course of giant planet formation gap formation in the disk near the planet which can effectively _ slow down _ planet migration . when the planet is not very massive its gravity can not strongly affect the disk surface density distribution in its vicinity and strong interaction with the gas disk through lindblad resonances ( goldreich & tremaine 1980 , hereafter gt80 ) leads to the rapid migration described above ( so called `` type i migration '' , ward 1997a ) . as the planetary mass grows the drift speed increases but at some point the strength of the torque exerted on the surrounding gas becomes so large that planet simply repels the gas and a gap forms . this strongly diminishes the tidal interaction because it is usually dominated by high - order lindblad resonances which now lie inside the gap . thus the orbital evolution of the planet becomes tied to the evolution of the disk and it migrates on the viscous timescale of the disk which could be much longer ( @xmath9 yr or even longer , depending on the viscosity in the disk ) . this stage is called `` type ii migration '' and its existence might help to explain the survival of planetary systems in the course of their orbital evolution or at least significantly alleviate this problem ( ward 1997 ) . gap formation also provides a reason for the existence of a maximum mass of extrasolar giant planets ( nelson et al . 2000 ) . in view of all this a very important question arises : how massive should a planet be in order to open a gap ? the answer depends not only on the conditions in the nebula ( surface density , temperature , viscosity ) but also on the dissipation mechanism of planet - induced density waves and the mobility of the planet . if the viscosity in the disk is absent and the planet is not drifting there is no mechanism which could oppose tidal torques and a gap is opened by an arbitrarily small perturber . this conclusion changes if planetary migration is taken into account self - consistently because a low - mass planet is able to drift through the gap before gap fully forms ( ward & hourigan 1989 , hereafter wh89 ) . the planet must slow down its migration somehow for the gap to be opened . in their study wh89 demonstrated that the disk surface density is enhanced in front of the drifting planet and is reduced behind it . this effect _ diminishes _ the difference between the torques produced by the planet in the disk outside and inside its orbit and acts to _ slow down _ the migration . for some mass of the planet its steady drift becomes impossible , the planet halts , and gap formation ensues . thus , in this picture there is a minimum mass for gap opening even in an inviscid disk . assuming that wave dissipation is a very rapid process and that angular momentum is immediately transferred to the disk wh89 estimated this critical mass to be @xmath10 here @xmath11 is a surface density of the disk , @xmath12 is its geometric thickness ( here @xmath13 is the sound speed and @xmath14 is the angular orbital frequency ) , and @xmath15 is the distance from the central star . typical protosolar nebular parameters were summarized by hayashi ( 1980 ) in the form of the minimum mass solar nebula ( mmsn ) model : @xmath16 using ( [ hayashi ] ) we obtain that @xmath17 and @xmath18 g at @xmath19 au which is about the mass of mercury . of course , if the disk possesses nonzero viscosity this mass would increase because viscous diffusion opposes gap formation . lin & papaloizou ( 1993 , hereafter lp93 ) present a different point of view on the planetary mass required to open a gap . they also assume damping of the density waves to be local but require mass of the planet to be high enough for the waves to be strongly nonlinear from the beginning and shock immediately transferring angular momentum to the fluid . they showed that this minimum mass corresponds to the case when the hill s radius of the planet @xmath20 ( @xmath21 is the mass of the planet , @xmath22 is the mass of the central star ) is comparable to the vertical disk scaleheight yielding @xmath23 as the relevant mass for opening a gap [ the numerical estimate in equation ( [ m1 ] ) is made for mmsn parameters given in equation ( [ hayashi ] ) ] . another way to look at this restriction is to notice that at @xmath24 the pressure gradient in the disk in the vicinity of the planet becomes so high that epicyclic frequency becomes imaginary and rayleigh s instability develops . thus , this restriction could be considered as an upper limit on the mass of the planet that does not open a gap . values of the gap opening mass estimated by these two approaches differ by a huge factor . indeed : @xmath25 where @xmath2 is the toomre stability parameter ( binney & tremaine 1987 ) and @xmath26 is the epicyclic frequency ( @xmath27 for keplerian rotation law ) ; using mmsn parameters given by equation ( [ hayashi ] ) we obtain that @xmath28 and @xmath29 at @xmath19 au meaning that @xmath30 . ward & hourigan ( wh89 ) assume that the damping mechanism is independent of the planetary mass and essentially linear , such as viscous dissipation ( takeuchi et al . 1996 ) or radiative damping ( cassen & woolum 1996 ) . however these mechanisms are probably ineffective in cold , weakly ionized and optically thick systems such as protoplanetary disks ( hawley & stone 1998 ; goodman & rafikov 2001 , hereafter gr01 ) ; in this case tides raised by low - mass planets could propagate much further than just a fraction of disk scale length and gap opening requires significantly more massive perturbers ( ward 1997a ) . requiring waves to be strongly nonlinear and damp immediately as a necessary condition for gap formation is also probably too radical . indeed , density waves produced by the planet can still evolve due to weak nonlinearity and are able to form a weak shock even if the planet is less massive than @xmath0 ( gr01 ) . this mechanism can lead to a gap formation by lower mass planets than lp93 assumed . clearly to obtain a reliable estimate of the critical planetary mass one must use a realistic wave damping prescription . goodman & rafikov ( gr01 ) have considered nonlinear evolution of the density waves produced by low - mass planets in two - dimensional disks using the shearing sheet approximation and assuming the background surface density and sound speed to be constant . they have found that a shock is formed quite rapidly ( depending on @xmath31 , typically after travelling several disk scaleheights from the planet but still not immediately ) because the radial wavelength of the perturbation constantly decreases while its amplitude increases ( as a consequence of angular momentum flux conservation ) thus facilitating wavefront breaking . after wave shocks its angular momentum is gradually transferred to the disk fluid leading to surface density evolution ( thus violating the constant surface density assumption ) . rafikov ( 2001 ) generalized this analysis by taking into consideration the effects of spatial variations of the surface density and sound speed in the disk as well as the cylindrical geometry of the problem . realistic prescription of the global angular momentum dissipation was provided under the condition that the surface density and sound speed vary on scales larger than the wavelength of the perturbation . in this paper we study the criterion for gap opening by planetary tides by determining under which conditions a steady - state solution for the disk surface density perturbation is no longer possible ( method used in wh89 ) . planet migration is taken into account self - consistently . tidal perturbations are supposed to be damped nonlocally by weak nonlinearity and their angular momentum is assumed to be transferred to the disk , as described quantitatively in rafikov ( 2001 ) . for the gap opening the assumption of varying surface density in rafikov ( 2001 ) is the most important one because it allows us to solve the problem self - consistently ( angular momentum transfer depends on how the surface density is distributed radially ) . after developing the general analytical apparatus in [ geneq ] we first explore the case of an inviscid disk in [ visc - less ] to highlight the most important physical mechanisms relevant for gap formation . then we generalize consideration to include the disk viscosity in [ viscous ] . finally , in [ disc ] we discuss our results . we study the surface density evolution of a gas disk driven by planetary gravitational perturbations . in our model a planet of mass @xmath21 moves in a disk on a circular orbit with radius @xmath32 . the background or initial surface density and sound speed @xmath33 and @xmath34 are assumed to vary independently on scales @xmath35 . the values of the surface density , sound speed , orbital frequency , and disk vertical scaleheight at planet s location are @xmath36 , and @xmath37 respectively . we have summarized brief descriptions of and references to definitions of the most important variables we use in table 1 . we are only interested in the radial evolution of the surface density and neglect azimuthal variations . then the general time - dependent equations of disk surface density evolution are the continuity and angular momentum equations ( pringle 1981 ) : @xmath38 where @xmath11 is the disk surface density , @xmath39 is a fluid radial velocity , @xmath40 is the ( azimuthally averaged ) angular momentum flux transferred to the disk fluid from the planetary tidal perturbations , and @xmath41 is the usual viscous angular momentum flux ( @xmath42 is a kinematic viscosity ) . from this system one can easily find that @xmath43^{-1 } \frac{\partial}{\partial r}(g+f ) . \label{v_r}\ ] ] substituting this result back into equation ( [ contin ] ) one obtains a self - consistent equation for the surface density evolution : @xmath44^{-1 } \frac{\partial}{\partial r}(g+f)\right\}. \label{evol}\ ] ] now we describe separately all the important ingredients entering this system of equations and all its necessary simplifications . l l l @xmath2 & toomre s @xmath2 & equation ( [ mass_rat ] ) + @xmath45 & viscous @xmath45-parameter & [ evol_eq ] + @xmath46 & disk vertical scaleheight & [ intro ] + @xmath47 & gas polytropic index & [ ang - mom ] + @xmath48 & total angular momentum flux & equation ( [ f_0 ] ) + @xmath49 & function characterizing torque cutoff & equation ( [ f_0 ] ) + @xmath50 & distance - dependent angular momentum flux & equation ( [ f_j ] ) + @xmath51 & distance - like variable for the shock propagation & equation ( [ t ] ) + @xmath52 & angular momentum damping function & equation ( [ t_w ] ) + @xmath53 & mach-@xmath19 length & equation ( [ t ] ) + @xmath0 & critical planetary mass for a strong wave nonlinearity & equation ( [ m1 ] ) + @xmath54 & critical planetary mass for a strong migration feedback & equation ( [ mwh ] ) + @xmath55 & fiducial mass & equations ( [ mwh ] ) & ( [ m_f ] ) + @xmath56 & radial distance from planet scaled by @xmath53 & equation ( [ xdef ] ) + @xmath57 & dimensionless shocking distance & equation ( [ x_sh ] ) + @xmath58 & proportionality constant in definition of @xmath57 & equation ( [ x_sh ] ) + @xmath59 & radial distance from planet scaled by the shocking distance & equation ( [ z ] ) + @xmath60 & dimensionless surface density & equation ( [ sigma_def ] ) + @xmath61 & dimensionless surface density at planet s location & equation ( [ sigma_def ] ) + @xmath62 & drift velocity & [ local ] & equation ( [ finv ] ) + @xmath63 & relative difference in torques leading to migration & equations ( [ finv ] ) & ( [ warddrift ] ) + @xmath64 & correction factor for drift velocity & equation ( [ finv ] ) + @xmath65 & parameter characterizing strength of the feedback & equation ( [ lambda_s ] ) + @xmath66 & cutoff distance & equation ( [ lambda_s ] ) + @xmath67 & characteristic timescale & equation ( [ tau ] ) + @xmath68 & viscous parameter & equation ( [ lam_nu ] ) + @xmath69 & tidal parameter & equation ( [ lam_t ] ) + goldreich & tremaine ( gt80 ) have considered the gravitational interaction of a gas disk with a planet embedded in it . they have demonstrated that a perturber on a circular orbit exerts a torque on the disk only in the immediate vicinity of the lindblad resonances . density waves launched at these locations carry angular momentum away from the satellite in the outer disk and towards it in the inner one . lindblad resonances of order @xmath70 located about one scaleheight @xmath71 from the planet are the strongest contributors to the angular momentum flux : lower order resonances are far from the planet and do not feel such strong tidal forcing while higher order ones are saturated because of the so - called `` torque cutoff '' ( gt80 ) . the total angular momentum flux across a cylinder of radius @xmath32 carried by all density wave harmonics is shown to be given by the expression ( gt80 ) @xmath72 ^ 2\right\}\label{f_0},\ ] ] where @xmath73 denotes the modified bessel function of order @xmath42 and the function @xmath49 describes the strength of the torque cutoff . it depends only on the disk stability parameter @xmath2 mentioned previously in equation ( [ mass_rat ] ) , and @xmath74 for disks with @xmath75 . beyond several @xmath37 from the planet , tidal perturbations behave basically like sound waves ( lin & shu 1968 ) . their nonlinear evolution finally leads to shock formation ( gr01 ) . after this happens the angular momentum flux of the density waves decreases because it gets transferred to the disk fluid . taking this into effect account rafikov ( 2001 ) has demonstrated that the angular momentum flux carried by the tidal perturbations @xmath76 is given by @xmath77 here @xmath78 is oort s @xmath79 constant , @xmath53 is the mach-@xmath19 distance [ the distance from the planet at which shear velocity is equal to sound speed , @xmath80 for keplerian disks ] , @xmath47 is a polytropic index , and @xmath0 is given by equation ( [ m1 ] ) . the variable @xmath81 plays the role of the distance travelled by the wave from the planet , and the dimensionless flux @xmath82 was calculated in gr01 for @xmath83 . in equations ( [ f_j])-([t ] ) the surface density @xmath11 and sound speed @xmath13 can vary arbitrarily with radius @xmath15 . it was also demonstrated in gr01 that the density wave shocks when @xmath51 reaches @xmath84 in this expression we have neglected an additional term equal to @xmath85 , which takes into account the finite distance required for the wake formation in the linear regime . one can do this only if @xmath86 ( which is always true in this paper ) , so that the region of linear wake formation is separated from the region of its weakly nonlinear evolution . otherwise these regions overlap and the wave shocks before it is fully formed , significantly complicating the analysis . for @xmath87 , the total angular momentum flux of the waves @xmath76 is conserved and has to coincide with @xmath48 given by the expression ( [ f_0 ] ) ; the value of @xmath88 for @xmath87 is such that this condition is fulfilled . note also that although the polytropic index @xmath47 enters equation ( [ f_j ] ) the dimensionless flux @xmath76 does not actually depend on it because @xmath89 ( see gr01 ) . thus , we can rewrite equation ( [ f_j ] ) as @xmath90 the function @xmath52 describes the damping after the shock is formed and is such that @xmath91 when @xmath92 , @xmath93 when @xmath94 , and @xmath95 . its behavior is shown in fig . [ fig : varphi ] . in this paper instead of the true angular momentum flux dependence calculated by gr01 we will be using for numerical convenience a simple analytical fit satisfying all of the above mentioned conditions and described in appendix [ ap1 ] . from the conservation of the total angular momentum ( assuming a steady state ) it follows that the corresponding angular momentum flux of the disk fluid is @xmath96 . \label{f - fluid}\ ] ] this expression together with equations ( [ f_j])-([t ] ) is to be substituted into equation ( [ evol ] ) and then the surface density evolution could be calculated self - consistently because the dependence of @xmath40 on @xmath11 is taken explicitly into consideration by equation ( [ t ] ) . using equations ( [ f_j])-([t ] ) one can in principle study the temporal evolution of @xmath11 taking into account the fact that @xmath97 , @xmath98 , @xmath99 , etc . vary on scales @xmath35 . we prefer to study the problem in a simplified setting assuming that although gap formation is a nonlocal process it still occurs sufficiently close to the planet . later we will determine the necessary conditions for this assumption to be true . this simplification allows us to set @xmath100 , @xmath101 , and to use the shearing sheet approximation to represent the background fluid flow . the only disk quantity which is assumed to vary in our consideration is the surface density @xmath11 . by setting @xmath100 we imply that rapid variations of @xmath11 do not violate the thermal balance in the disk that determines @xmath102 . here the disk temperature distribution is supposed to be the result of external irradiation by the central star and our assumption of constant @xmath13 rests upon the absence of geometrical effects such as the shadowing by the gap edge ( lecar & sasselov 1999 ; dullemond et al . 2001 ) . in this `` quasilocal '' approximation , introducing a new variable @xmath103 we obtain from equation ( [ t ] ) that @xmath104^{1/2 } reproduce the expression for @xmath105 obtained in gr01 . the disk - planet interaction leads to the radial migration of the planet ( gt80 ; ward 1997a ) with some drift velocity @xmath62 . we consider the disk evolution in the coordinate system comoving with the planet . in this case @xmath106 usually it is found that migration is inward , i.e. @xmath107 ( ward 1986 ) . using the definitions @xmath108 one can rewrite equation ( [ evol ] ) in the following form : @xmath109.\label{evol1}\end{aligned}\ ] ] here @xmath110 is the surface density of the disk far away from the planet ; it is different from @xmath111 in the case of viscous disks ( @xmath112 , see [ viscous ] ) . then the ratio @xmath113 is not known a priori and has to be found from a self - consistent solution of the problem . we take @xmath110 to be the same both inside and outside of the planet which means that global surface density gradients are neglected here as we have mentioned earlier . gr01 has demonstrated that in the constant surface density disk , the wave shocks after propagating a distance ( in @xmath56 ) @xmath114 away from the planet . in what follows we will always assume for simplicity that @xmath115 thus taking @xmath116 [ see however fridman & gorkavyi ( 1999 ) ] . in a disk with varying surface density @xmath57 is no longer the distance after which the shock forms . still , it is convenient to change @xmath56 to a new spatial coordinate @xmath59 given by @xmath117 in terms of @xmath59 , the surface density in our particular problem varies on scales @xmath118 . thus , the condition for applying our `` quasilocal '' approach ( see [ local ] ) is @xmath119 we check later in [ disc ] if this is fulfilled in realistic protoplanetary systems . equation ( [ evol2 ] ) does not have @xmath21 inside the argument of @xmath121 and this permits one to extract an explicit dependence on the planetary mass . equation ( [ evol2 ] ) can only describe the surface density evolution when it is supplied with the information about the behavior of the migration velocity @xmath62 as a function of @xmath60 . in appendix [ drift ] we demonstrate that @xmath122 is a correction factor for the drift velocity caused by the feedback from the surface density variations to the planetary migration ; also @xmath123 in the expression for @xmath64 the migration feedback is represented by second term it takes into account modification of the contribution from low - order lindblad resonances , which is important because the planetary torque in the outer and inner parts of the disk due to high - order lindblad resonances almost completely cancel . this means that even a small contribution from low - order resonances could produce a significant effect on the migration speed . when the feedback is absent the drift speed is determined only by global gradients of @xmath97 and @xmath98 represented by factor @xmath63 in equation ( [ finv ] ) ( see appendix [ drift ] ) . we will see later that in interesting cases the planet produces variations of @xmath124 and @xmath125 in the direction of migration and @xmath126 in the opposite direction : a surface density increase forms in the disk _ in front _ of the planet , and a depression appears _ behind _ it . thus density perturbations on both sides of the planet work in one direction to enhance the effect of the disk in front of the drifting planet and decrease it behind the planet [ neglecting for a moment the fact that factor @xmath113 in equation ( [ finv ] ) could be different from unity ] . they do not cancel each other ( like in the case of migration due to variations of @xmath97 ) and produce @xmath127 effect . this means that planetary - driven surface density perturbations tend to _ stall _ migration . however , these variations are strongest at distances @xmath128 from the planet , where torque generation is weak , thus they need to be large to compete with usual drift due to weak global gradients ( for which cancellation effects are important ) . we will often use the numerical form of the parameters entering this system , namely @xmath131 where we have used equation ( [ mass_rat ] ) and our calculation assumes that @xmath75 , @xmath132 , and @xmath133 . equation ( [ main ] ) describes the evolution of the disk surface density and its solution depends on @xmath134 dimensionless parameters : @xmath135 , @xmath67 , and @xmath66 which themselves are combinations of @xmath136 , @xmath31 , @xmath45 , and @xmath2 . the parameter @xmath69 describes the effect of the planetary tidal torques on the surface density distribution , while @xmath68 tells us how strong the viscosity is . obviously , for the gap to form the second term in the r.h.s . of ( [ main ] ) must be larger than the viscous one . the feedback which planetary migration receives from the surface density perturbations is described by the parameter @xmath65 defined in equation ( [ lambda_s ] ) and it is strongest for large @xmath65 . feedback also depends on the value of @xmath66 [ equation ( [ lambda_s ] ) ] which represents the effect of the torque cutoff . to understand better the meaning of these parameters let us consider the time @xmath137 which it takes for a planet to drift across a gap with the width @xmath138 . obviously , @xmath139 , where again @xmath62 is a drift velocity . using equations ( [ f_0 ] ) and ( [ finv ] ) one can easily obtain that @xmath140 time needed for viscous diffusion to fill the gap is clearly @xmath141 finally , for the tidal torques to clear out a gap with width @xmath138 an angular momentum @xmath142 must be supplied to the gas ( see gt80 ) . if we assume that angular momentum flux carried by the density waves is damped on the characteristic length @xmath143 , then the timescale for gap opening is @xmath144 in our particular problem the characteristic gap width and the wave damping distance are the same and given by @xmath145 [ see equation ( [ x_sh ] ) ] . then , substituting this into equation ( [ dt_d ] ) we obtain that @xmath146 , i.e. time @xmath67 is a typical timescale for a gas drift through the gap caused by the planetary migration ( in planet s reference frame ) . one can see that @xmath147 for @xmath86 which justifies our use of azimuthally averaged quantities if @xmath148 [ because @xmath149 is a characteristic timescale for the gap opening , see equation ( [ main ] ) ] . also for this choice of @xmath138 one can easily find that @xmath150 , that is drift through the gap can not replenish material repelled by planetary torques when @xmath151 , and gap could be opened if viscosity is absent . we will further confirm this conclusion in [ visc - less ] . likewise , one can see that @xmath152 which means that viscous diffusion is more important for the surface density evolution than effects of the migration when @xmath153 . it is also clear that for tidal torques to overcome viscosity and open a gap in a viscous disk one needs @xmath154 , or @xmath155 ( see [ viscous ] ) . it is possible that for some values of parameters @xmath135 solution for the surface density does not depend on time - like variable @xmath156 . in this case , the profile of @xmath60 is stationary in the reference frame migrating with the planet . although planetary gravitational torques tend to repel the disk fluid , planet is mobile enough to migrate through the forming gap and stop its development . when such a time - independent solution is no longer possible , planet migration stops and surface density evolves with time to form a gap in the disk . wh89 used the same approach when studying gap formation with instantaneous density wave damping . in their case @xmath121 varies on scales @xmath71 ( wake generation requires a couple of disk scale lenghts to complete ) , i.e. @xmath161 in our notation . also , to significantly affect the migration speed one only needs surface density variations @xmath162 to be produced . combined with the definition ( [ lam_t ] ) this gives @xmath163 as a value of their critical mass , in agreement with equation ( [ mwh ] ) . in our case this mass is larger because wave damping occurs at greater distance from the planet . we now explore separately cases of purely inviscid disks and disks with nonzero viscosity . if the viscosity in the disk is absent equation ( [ kin_main ] ) simplifies to @xmath165 where @xmath166 . it follows from the properties of @xmath167 that @xmath158 for @xmath168 , i.e. for @xmath169 , meaning that @xmath170 ( see appendix [ ap1 ] ) . this implies that in inviscid disk nothing happens with the disk fluid until tidal wave shocks . after that , for @xmath171 , @xmath172 suggesting that for inward migration @xmath125 if @xmath173 ( inner disk ) and @xmath126 if @xmath174 ( outer disk ) just like we mentioned before . the effect of this is to slow down the migration . we can rewrite ( [ finv ] ) as @xmath175 since integrals are @xmath127 , one needs @xmath176 for the drift velocity to be significantly affected . the solution of the system ( [ kin_less])-([fin_less ] ) depends on @xmath177 parameters only , @xmath69 and @xmath65 . in fig . [ fig : profiles ] we plot the surface density profiles in the vicinity of the planet for several values of @xmath69 and @xmath65 . surface density increases in front of the migrating planet because there are two fluxes of the disk fluid there one caused by the migration , another by the planetary repulsion and they are converging causing a surface density pileup . behind the moving planet these fluxes are in the same direction away from the planet . for large enough planetary mass repulsion removes matter from the the disk behind the planet so rapidly that a corresponding surface density decrease can not be replenished by the flux due to the planetary migration ( this flux grows with @xmath21 slower than the one due to the repulsion ) . as a result a gap is carved out in the disk behind the moving planet . to describe this process quantitatively let us notice that r.h.s . of equation ( [ kin_less ] ) depends on @xmath60 only through an integral over @xmath59 . for @xmath174 r.h.s . of ( [ kin_less ] ) varies from @xmath178 to @xmath179 , where @xmath180 is the value of @xmath59 for which @xmath181 reaches minimum , and @xmath182 as described in appendix [ ap1 ] . at the same time l.h.s . of ( [ kin_less ] ) varies from @xmath183 to @xmath179 ( it is minimized at @xmath184 ) . thus , the solution of ( [ kin_less ] ) does not exist if @xmath185 or if @xmath186 closer inspection of equation ( [ kin_less ] ) reveals that @xmath60 develops an infinite derivative when this happens . this would lead to the violation of the rayleigh s criterion and instability associated with this will presumably help to clear out a gap on the outer side of the disk . once the gap is formed , density waves launched in the outer disk are not able to propagate to infinity and to transfer all their angular momentum to the disk material . instead they reflect from the edge of a forming gap and return to the planet . here they could interact gravitationally with planet thus canceling part ( or all ) of the wave angular momentum launched in the outer direction . it would diminish the influence of the outer disk on the migration and planetary drift will stall . for this to occur planet needs to absorb only a small fraction of one - sided angular momentum @xmath48 , about @xmath136 , which is obvious from the discussion in [ evol_eq ] and appendix [ drift ] . for @xmath187 it turns out that @xmath188 and @xmath189 ( planet is still rapidly migrating , feedback effects are not strong enough to slow it down ) when steady - state solution for @xmath60 is no longer possible . this means that gap opens when @xmath190 this corresponds to planetary mass @xmath191 for stable disks ( @xmath75 ) this criterion gives quite low mass at which gap forms although larger by @xmath192 than the one suggested by ward & hourigan ( wh89 ) . for @xmath176 the situation is somewhat different . here the feedback from the density perturbations to the drift velocity of the planet becomes important . to study this region of the parameter space let us divide equation ( [ kin_less ] ) by @xmath193 and integrate the result from @xmath194 to @xmath195 and from @xmath19 to @xmath196 , thus forming integrals present in r.h.s . of equation ( [ fin_less ] ) . substituting them into ( [ fin_less ] ) we find that @xmath197 it turns out that function @xmath198 depends on @xmath199 only very weakly [ which is reasonable because r.h.s . of ( [ v_eq ] ) does not depend on the rapidly varying factor @xmath200 . we confirmed this numerically and found that @xmath201 since l.h.s . of equation ( [ v_eq ] ) attains its maximum equal to @xmath202 at @xmath203 , no solution of equation ( [ v_eq ] ) for @xmath64 exists when @xmath204 from the condition ( [ lim_s ] ) and definitions ( [ lambda_s ] ) & ( [ lam_t ] ) we can find that the feedback stops migration when @xmath205 in fig . [ fig : drift ] we show the dependence of the drift velocity on the mass of a perturber in the regime of strong migration feedback . drift velocity is normalized by the drift speed which a body of mass @xmath206 [ equation ( [ mass_s ] ) ] would have if the feedback were absent ( straight dashed line on this figure ) . analytical solution of equation ( [ v_eq ] ) is also displayed and is in good agreement with numerical result . one can easily see that @xmath207 when @xmath208 . this situation is analogous to what wh89 have found in their inertial limit calculations in the case of instantaneous damping . when a solution for the drift velocity does not exist any more planetary torques can not support a constant flux of disk material seen in planet s frame and time dependent evolution commences leading to a gap formation and stalling the migration . lin & papaloizou ( 1986 ) and wh89 followed this process with time - dependent one - dimensional numerical calculations . the behavior of the drift velocity correction factor @xmath64 is shown in fig . [ fig : contour ] . one can see the cutoff of the possible kinematic wave solutions caused by the migration feedback [ predicted by equation ( [ lim_s ] ) ] on top of the plot where @xmath64 rapidly diminishes . for small @xmath65 the criterion given by ( [ lim_t ] ) becomes important . the agreement between our simple analytical considerations and numerical results is very good . we can unite the gap formation conditions ( [ m_t ] ) and ( [ mass_s ] ) in a single criterion on the limiting mass of the perturber : @xmath209.\label{mass_lim}\ ] ] because @xmath75 we can immediately see from equation ( [ mass_lim ] ) that gap could be formed by a planet with a mass significantly smaller than @xmath0 . since our analysis has only exploited the most basic features of the damping function @xmath52 which are well reproduced by our simple fitting expression ( [ match ] ) we conclude that the results obtained here would be basically the same if one were to use the exact damping function calculated in gr01 . in the case of nonzero background viscosity present in the disk one faces a more complicated situation than in inviscid disk . now the solution of kinematic wave equation ( [ kin_main ] ) depends on @xmath210 parameters : @xmath69 , @xmath65 , which we had before , @xmath68 , which determines the effect of the viscosity on the solution , and @xmath57 , which affects drift velocity feedback . parameter @xmath57 did not appear in the inviscid case because @xmath60 was exactly @xmath19 near the planet in that case . viscosity changes this picture because now @xmath60 varies near the planet , and this is important for the feedback , because tidal forcing is very strong in the immediate vicinity of the planet [ see equation ( [ gt ] ) ] . in the case @xmath211 one can integrate equation ( [ kin_main ] ) once to obtain @xmath212\varphi \left(i(z^\prime)\right)dz^\prime-\varphi\left(i(z)\right)\right\},\end{aligned}\ ] ] or @xmath213 \exp(-s)ds.\label{visceq}\end{aligned}\ ] ] if one integrates r.h.s . of ( [ visceq ] ) by parts and takes the limit @xmath214 , equation ( [ kin_less ] ) is easily recovered . as we have mentioned before , factor @xmath113 in equation ( [ visceq ] ) is not equal to unity any more . it should be determined from the self - consistent solution of equation ( [ visceq ] ) instead . to solve this equation we employ an iterative technique : at each step solution obtained from the previous iteration is substituted into the r.h.s . of ( [ visceq ] ) to obtain a better approximation until this process converges . as a first trial we use @xmath215 and @xmath170 . in fig . [ fig : threshold]a we display several surface density profiles for different @xmath68 and fixed @xmath216 ( gap - forming value of @xmath69 in the inviscid disk without the migration feedback ) and @xmath217 ( thus assuming that migration feedback is absent ) . for small @xmath68 profile is almost the same as in the inviscid disk ; one can show that @xmath218 for small @xmath59 . as @xmath68 grows and becomes comparable with @xmath69 perturbations of @xmath60 near @xmath219 become significant which has profound implications for the migration feedback if @xmath220 . however , as @xmath221 viscous diffusion becomes so strong that surface density inhomogeneities in the vicinity of the planet become very small ( as in the case @xmath222 in fig . [ fig : threshold]a ) . it turns out that in the viscous case ( exactly like in the inviscid one ) strong tidal perturbations can open a gap in the disk even without the drift velocity feedback . the presence of the viscosity only changes the threshold value of the parameter @xmath69 so that it increases when @xmath68 ( and , correspondingly , viscosity ) grows . the dependence of this limiting @xmath69 on @xmath68 for @xmath217 ( no feedback ) is shown on fig . [ fig : threshold]b and is easy to understand : to open a gap the planet has to overcome not only the drift of the fresh disk material into a forming gap behind the planet but also the diffusion due to the viscous stresses which tends to smooth any inhomogeneities in @xmath60 and fill the gap . thus , larger @xmath68 require larger @xmath69 to open a gap . when @xmath223 drift velocity feedback from the surface density perturbations could become important before @xmath69 reaches the threshold value mentioned above . in fig . [ fig : lam_regions ] we show the boundary of the region in the @xmath224 parameter space where a planet could migrate without opening a gap for several values of @xmath68 . as we have said before , for small @xmath65 region where the gap can not be cleared expands as @xmath68 increases confirming results for @xmath217 displayed in fig . [ fig : threshold]b . for larger @xmath65 , however , the situation is more complicated : for a fixed strength of planetary torques ( characterized by @xmath69 ) this region first rapidly contracts but then expands again with increasing @xmath68 . it could be explained as follows : when @xmath225 viscosity can not strongly change the surface density near the planet , @xmath226 and drift velocity feedback is only slightly stronger than in the inviscid case . when viscosity grows so that @xmath227 , surface density perturbations are strong and viscosity is significant to sufficiently modify surface density in the planetary vicinity leading to a strong feedback : as it follows from equations ( [ gt ] ) and ( [ finv ] ) even small deviations of @xmath60 from @xmath228 are strongly amplified by a factor @xmath229 ( and @xmath230 ) in the expression for the feedback correction . this effect rapidly lowers the upper boundary of the region where planet could migrate and not open a gap . however , when @xmath231 strong viscosity smooths out all the inhomogeneities in @xmath60 which planet tends to produce and this reduces the role of the feedback for a fixed @xmath69 critical @xmath65 increases . thus , growing viscosity tends to amplify migration feedback but then attenuates it as @xmath68 exceeds @xmath69 . in the viscous disk the absence of the steady - state solution of equation ( [ visceq ] ) caused by the feedback is not enough to open a gap , a fact emphasized in lin & papaloizou ( 1986 ) and wh89 . planetary torques must also exceed viscous diffusion for this to happen . this additional requirement could be roughly described by a condition that @xmath232 where we have used equations ( [ lambdas])-([t_0 ] ) . this is equivalent to saying that the time to open a gap by tidal torques must be shorter than the viscous diffusion time ( see [ evol_eq ] ) . the condition ( [ viscond ] ) coincides with the corresponding criterion used by lin & papaloizou ( 1986 ) and wh89 , but is different from the condition found in gt80 because only gap formation by a torque exerted at a single lindblad resonance was considered in gt80 . in fig . [ fig : pars_regions ] we plot the boundary of allowed region in the space of physical parameters : @xmath31 and @xmath233 for different values of dimensionless viscosity @xmath45 and toomre stability parameter @xmath2 . in depicting the boundary curves we have taken into account the condition ( [ viscond ] ) ( straight line parts of the curves are due to this restriction ) . for a fixed viscosity , the planet can open a gap only if it is located to the right of the corresponding curve in fig . [ fig : pars_regions ] . for small @xmath233 ( thick disks , small migration feedback ) critical @xmath21 is always larger than in the inviscid case and grows with increasing @xmath45 . as @xmath233 increases both @xmath68 and @xmath68 grow according to equations ( [ lambda_s ] ) & ( [ lam_nu ] ) . this rapidly increases the role of the migration feedback so that corresponding limiting @xmath21 could become smaller than in the inviscid case . complicated shape of the boundary curves for different values of @xmath45 is due to the before mentioned complexity of the influence of the disk viscosity on the migration feedback . one can see from the fig . [ fig : pars_regions ] that for values of @xmath233 typical in protoplanetary disks ( @xmath234 ) @xmath86 if @xmath45 is small ( typically when @xmath235 ) . one can also notice that it is a migration feedback which initiates the evolution of the surface density leading to the gap formation once the condition ( [ viscond ] ) is fulfilled . our consideration of the viscous case assumes that the only surface density gradients producing viscous flux are due to the planet - driven perturbations . however , background surface density varying on scales @xmath35 also contributes to the viscous fluid flux . we neglected this effect completely when we employed our `` quasilocal approximation '' and omitted factors @xmath236 leaving only @xmath237 in our equations ( see [ local ] ) . if we were to include these factors in our analysis a picture similar to that described in ward ( 1997a ) would be found : for large viscosity there would not be an infinite derivative of the drift velocity with respect to @xmath21 at some point as in the case shown in fig . [ fig : drift ] . instead @xmath238 would smoothly ( but rapidly ) decrease at some point as @xmath21 grows and the migration would change gradually from type i to type ii ( ward 1997a ) . we are unable to capture this effect working in the framework of our `` quasilocal approximation '' but this limitation can not affect our major conclusions concerning the gap forming planetary mass . also , when the viscosity is small or when the condition ( [ quazi ] ) is satisfied we should not worry about this effect at all . using equations ( [ lambdas])-([t_0 ] ) we find that these additional viscous fluxes could be neglected if @xmath239 which is usually the case in realistic situations . the validity of our consideration is restricted not only by the requirements ( [ quazi ] ) and @xmath86 but also by the applicability of our nonlinear wave damping prescription . its validity was extensively discussed in gr01 and rafikov ( 2001 ) and we will not repeat this discussion here . for small enough viscosity ( @xmath235 ) we found ( see fig . [ fig : pars_regions ] ) that @xmath240 . this means that typically @xmath241 and @xmath242 so that the condition @xmath243 is marginally fulfilled ( note also that @xmath244 which is the offset value of @xmath51 characterizing the wake generation region , see gr01 ) . thus , a wave typically shocks after travelling only several @xmath37 from the planet ; assumption @xmath245 is reasonable if @xmath246 which is the case in mmsn model described by equation ( [ hayashi ] ) ( @xmath247 at @xmath19 au ) . in disks with higher viscosity , the wave will shock even closer to the planet ( see [ viscous ] ) meaning that initially wake is not tightly wound , but this can hardly change our gap formation criterion . after the planet grows so massive that a steady - state surface density distribution is no longer possible , time - dependent evolution commences . we do not study this process here . it is clear however that as soon as the gap opening criterion is fulfilled and disk viscosity is overcome by planetary torques , gap clearing starts behind the planet ( in the part of the disk opposite to the direction of migration ) . as we have mentioned before this makes propagation of the density waves in the outer disk impossible , and planet migration which is the result of the small difference of the torques on both sides of the disk , has to stall . when this happens the planet gradually repels the surrounding material , carving out a gap in both parts of the disk . far from the planet gap expansion can be stopped by the disk viscosity or tidal torques from other planets in the system . this general picture of gap formation ( after the corresponding gap - opening criterion is fulfilled ) was confirmed by the one - dimensional time - dependent numerical calculations in lin & papaloizou ( 1986 ) and wh89 . of course they used a different gap - forming criterion because of their assumed instantaneous damping of the planet - generated density waves . but because the reason for the gap formation in our case is basically the same as in theirs migration feedback and strong tidal torques exceed the viscous spreading we expect the time - dependent transition from type i to type ii migration in our setting to be similar to what these authors have found in the local damping approximation . the only differences would be the critical mass at which this happens and the transition timescale . our results for the gap - opening criterion also seem to be in general agreement with the time - dependent two - dimensional simulations carried out by lp93 , bryden et al . ( 1999 ) , and nelson et al . ( 2000 ) , although it is hard to make a direct quantitative comparison because our criterion requires knowledge of a larger number of parameters than is quoted in these studies . typical timescale for a gap to form is @xmath149 . using equation ( [ tau ] ) or ( [ t_0 ] ) , assuming @xmath248 and @xmath148 , one can find that for mmsn parameters this time is about @xmath249 yr at @xmath19 au and @xmath250 yr at @xmath134 au ( semimajor axis of jupiter s orbit ) . such long timescales can put some restrictions on the numerical schemes which would be able to check the validity of our gap - opening criterion ( they should also have high spatial resolution , typically a fraction of @xmath37 , and be able to follow the formation and evolution of weak shocks in a disk , although numerical viscosity would be a primary concern ) . using the theory developed in [ kin_wave ] we can determine the planetary mass at a particular location in a disk for which gap formation is expected and migration switches to a type ii mode . for the mmsn parameters represented by equation ( [ hayashi ] ) one obtains @xmath247 , @xmath28 , and @xmath251 at @xmath19 au . using fig . [ fig : pars_regions ] we find that for @xmath252 a gap is opened by a planet with @xmath253 . thus , if the earth was immersed in a gaseous disk at the end of its formation , it probably could not open a gap ( nonzero disk viscosity only strengthens this conclusion ) . then type i migration would have caused it to drift to the sun on a timescale @xmath254 yr , but this apparently did not happen . this result supports the usual view according to which the final accumulation of terrestrial planets via the runaway coagulation of planetesimals or giant impacts of moon - sized planetary embryos occurred _ after _ the nebula was dispersed in the inner solar system . then earth would not have migrated at all and the question of its survival would not arise . at @xmath134 au one finds that @xmath255 , @xmath256 , and @xmath257 ; then critical @xmath258 for @xmath252 . thus , even the usual rocky core of jupiter alone ( without gaseous envelope ) would likely be able to open a gap ; this would significantly slow down its migration towards the sun and leave jupiter on its orbit far from the central star . for some reason this clearly did not happen in systems harboring extrasolar giant planets close to their parent stars , and this issue apparently deserves further study . the value of the critical mass also raises another question : how has jupiter managed to acquire its huge gaseous mass ? it might be that the timescale for the gap formation at the jupiter s location is long enough ( see [ time - dep ] ) for the planet to accrete all its mass during the gap - opening stage if the core instability ( mizuno 1980 ) was operating since the very beginning ( core instability and associated planetary mass growth might have actually triggered the gap formation ) . but obviously more detailed consideration of the gas accretion process is needed to definitely answer this question . an interesting result following from this analysis is that when a gap is opened in an inviscid disk , there is still some material remaining at the planet s orbit . this material has the form of a ribbon with radial width equal to @xmath259 because waves launched by the planet can not shock prior to travelling a minimum necessary distance , and , thus , can not transfer their angular momentum to the disk and cause its evolution . the effects of wave action reflection from the edge of the ribbon or the presence of the moderate viscosity in the disk might lead to the dissipation of such a gas torus . but if these effects are not very strong then this ribbon phenomenon might have observable manifestations . using a realistic damping prescription for tidally - induced density waves we have studied the conditions necessary for a gap to be formed in a gas disk in the vicinity of a planet . it was shown that for small enough planetary mass a steady - state solution for the surface density perturbations exists in the reference frame migrating with the planet . then the details of the surface density distribution in the disk depend on only @xmath210 parameters : the aspect ratio in the disk @xmath136 , the viscosity ( represented here by its dimensionless analog @xmath45 ) , the toomre stability parameter of the disk @xmath2 , and the ratio of the planetary mass to a fiducial mass @xmath0 defined in equation ( [ m1 ] ) . only systems with @xmath260 were considered because larger viscosity could violate our damping prescription based on the nonlinear wave dissipation . we have demonstrated that in disks with @xmath2 between @xmath261 and @xmath262 ( which should be typical in passive protoplanetary disks irradiated by their central stars ) a gap is opened when the planetary mass reaches @xmath263 , depending on the disk viscosity and the planet s location in the nebula . planets further away from the central star must be more massive to repel the gas in their vicinity . we obtained an analytical criterion for a gap - forming planetary mass in inviscid disks [ see equation ( [ mass_lim ] ) ] . our result for this critical mass is in between the previous estimates of this quantity obtained by ward & hourigan ( 1989 ) and lin & papaloizou ( 1993 ) because of the different density wave damping function ( wh89 ) and moderate requirements for the nonlinearity of the wave ( lp93 ) . the apparatus developed here for following the disk surface density evolution and studying stationary structures in the disk such as the kinematic wave solutions ( see [ kin_wave ] ) could be generalized to other astrophysical problems ; for example it could be used to study global disk evolution or to extend our gap formation analysis to a system containing several planets . i want to express my gratitude to scott tremaine and jeremy goodman for useful discussions and a lot of thoughtful comments on the manuscript . financial support provided by the nasa grant nag 5 - 10456 is thankfully acknowledged . binney , j. & tremaine , s. 1987 , galactic dynamics , princeton univ . press , princeton , nj bodenheimer , p. & pollack , j.b . , 1986 , icarus , 67 , 391 bryden , g. , chen , x. , lin , d.n.c . , nelson , r.p . , & papaloizou , j.c.b . , 1999 , apj , 514 , 344 butler , r.p . , tinney , c.g . , marcy , g.w . , jones , h.r.a . , penny , a.j . , & apps , k. , 2001 , apj , 555 , 410 cassen , p. & woolum , d.s . 1996 , apj , 472 , 789 dullemond , c.p . , dominik , c. , & natta , a. , 2001 , astro - ph/0106470 fridman , a.m. & gorkavyi , n.n . , physics of planetary rings . celestial mechanics of continuous media , new york : springer , 1999 goldreich , p. & tremaine , s. 1980 , apj , 241 , 425 ( gt80 ) goodman , j. & rafikov , r. 2001 , apj , 552 , 793 ( gr01 ) hawley , j.f . & stone , j.m . , 1998 , apj , 501 , 758 hayashi , c. 1981 , progr . suppl . , 70 , 35 larson , r.b . 1989 , in : the formation and evolution of planetary systems , p.31 , eds . weaver , h.a . & danly , l. , cambridge university press , cambridge lecar , m. & sasselov , d.d . , 1999 , astro - ph/9911384 lin , d.n.c . & papaloizou , j. , 1986 , apj , 309 , 846 lin , d.n.c . & papaloizou , j. 1993 , in protostars and planets iii , ed . levy & j.i . lunine ( tucson : univ . arizona press ) , 749 ( lp93 ) lin , c.c . & shu , f.h . , 1968 , in _ astrophysics and general relativity _ , ed . m. chretien et al . ( new york : gordon & breach ) marcy , g.w . , butler , r.p . , & vogt , s.s . , 2000 , apjl , 536 , 43 masset , f. & snellgrove , m. , 2001 , mnras , 320 , l55 mayor , m. & queloz , d. , 1995 , nature , 378 , 355 mizuno , h. , 1980 , progr . phys . , 96 , 266 nelson , r.p . , papaloizou , j.c.b . , masset , f. , & kley , w. , 2000 , mnras , 318 , 18 pringle , j.e . , 1981 , ara&a , 19 , 137 rafikov , r.r . , 2001 , astro - ph/0110496 sasselov , d.d . & lecar , m. 2000 , apj , 528 , 995 shakura , n. i. & sunyaev , r. a. 1973 , a&a , 24 , 337 shu , f. , najita , j. , ostriker , e. , wilkin , f. , ruden , s. , & lizano , s. , 1994 , apj , 429 , 781 takeuchi , t. , miyama , s.m . , & lin , d.n.c . 1996 , apj,460 , 832 trilling , d.e . , benz , w. , lunine , j.i . , hubbard , w.b . , & burrows , a. , 1998 , apj , 500 , 428 vogt , s.s . , marcy , g.w . , butler , r.p . , & apps , k. , 2000 , apj , 536 , 902 ward , w.r . 1986 , icarus , 67 , 164 ward , w.r . 1997a , icarus , 126 , 261 ward , w.r . 1997b , apjl , 482 , 211 ward , w.r . & hourigan , k. 1989 , apj , 347 , 490 ( wh89 ) the behavior of the true solution and the matching function is shown in fig . [ fig : varphi ] . for a planet - generated wake profile @xmath265 ( neglecting the fact that there is some distance @xmath67 over which the wake profile forms ) . our simple fit never deviates from the true @xmath52 by more than @xmath266 . asymptotically @xmath267 for @xmath268 , as in the numerical solution . also @xmath93 when @xmath94 . to study the surface density profile near the planet in inviscid disks it is important to know the behavior of @xmath269 . for our assumed matching function ( [ match ] ) one can find that @xmath270 the function @xmath167 reaches its minimum value equal to @xmath271 at @xmath272 . from the global conservation of the angular momentum we find the velocity of migration driven by the planetary gravitational torques : @xmath273 where the integral is taken over the whole disk . one can write @xmath274 , where @xmath40 is the angular momentum flux produced by the planet per unit surface density at distance @xmath15 . gt80 have demonstrated that function @xmath275 rapidly falls off like @xmath276 for @xmath277 . ward ( 1986 ) has shown that @xmath278 $ ] , where @xmath279 is an _ even _ function of @xmath280 and terms of order @xmath136 arise because of the asymmetries in the torque generation pattern intrinsic for keplerian disks . the surface density itself may vary on different scales . in our particular case we are interested in the variations on scales of order @xmath32 [ represented by @xmath33 ] which lead to migration in the first place , and variations on scales @xmath282 , which result from the planet - driven surface density evolution . in the vicinity of the planet @xmath11 could be written in the following form : @xmath283 \approx \sigma_p\left[1+\frac{2 h_p}{3 r_p}\frac{d\ln\sigma_0}{d\ln r } x\right ] \times\left[1+\frac{\sigma -\sigma_p}{\sigma_p}\right].\ ] ] the replacement of @xmath33 with @xmath111 in the last bracket assumes that variations due to planetary torques are more rapid than background ones ( because damping length is supposed to be much shorter than @xmath32 ) . substituting this in ( [ drvel ] ) and recalling that @xmath279 falls off beyond several @xmath37 from the planet one finds that @xmath284 . \label{inttorque}\ ] ] here @xmath63 is a constant of the order of unity which depends on the global gradients of the surface density @xmath285 and temperature @xmath286 in the disk . the first term in brackets arises as a result of the asymmetry in torques produced in the inner and outer parts of the disk . ward ( 1986 ) advocates that for @xmath287 ( @xmath2 is a toomre stability parameter ) @xmath288 ^ 2\right\ } = 6.5(1 + 0.06 k+1.2 l),\label{warddrift}\ ] ] and for @xmath289 r.h.s . of ( [ warddrift ] ) changes to @xmath290 . it means that the surface density decreasing from the center of the disk leads to a very weak acceleration of drift in @xmath287 disk and slows it down for moderate values of @xmath2 . further we will assume for simplicity that the expression in parentheses in r.h.s . of ( [ warddrift ] ) is equal to @xmath177 ( which corresponds to @xmath132 ) . for large @xmath292 it could be demonstrated ( gt80 ) that @xmath293 for @xmath294 torque rapidly decreases ( `` torque cutoff '' ) . for simplicity we will assume that the dependence given by equation ( [ gt ] ) holds true even for @xmath295 , then we only need to correct the cutoff value of @xmath56 so that the integral of ( [ gt ] ) gives us the right amount of the angular momentum flux @xmath48 at infinity . it is done by setting @xmath296 for @xmath297 ; for @xmath298 we assume @xmath299 to be given by equation ( [ gt ] ) . using equation ( [ v ] ) we finally obtain equation ( [ finv ] ) and ( [ lambda_s ] ) . ideally , one should also take into account gradient of @xmath11 on short damping scale in the expression for @xmath275 as it was done by ward ( 1986 ) . this could produce some modification because surface density gradients displace the positions of the lindblad resonances and modify the torque cutoff . however , while variations of @xmath11 produce contribution to the drift @xmath300 , presence of the gradients of @xmath11 gives rise to the effects of order @xmath301 . this implies that contribution due to the gradient of the surface density is @xmath302 compared with the last term in ( [ inttorque ] ) and is unimportant for @xmath86 .

firstly , a systematic procedure is derived for obtaining three - dimensional bound - state equations from four - dimensional ones . unlike `` quasi - potential approaches '' this procedure does not involve the use of delta - function constraints on the relative four - momentum . in the absence of negative - energy states , the kernels of the three - dimensional equations derived by this technique may be represented as sums of time - ordered perturbation theory diagrams . consequently , such equations have two major advantages over quasi - potential equations : they may easily be written down in any lorentz frame , and they include the meson - retardation effects present in the original four - dimensional equation . secondly , a simple four - dimensional equation with the correct one - body limit is obtained by a reorganization of the generalized ladder bethe - salpeter kernel . thirdly , our approach to deriving three - dimensional equations is applied to this four - dimensional equation , thus yielding a retarded interaction for use in the three - dimensional bound - state equation of wallace and mandelzweig . the resulting three - dimensional equation has the correct one - body limit and may be systematically improved upon . the quality of the three - dimensional equation , and our general technique for deriving such equations , is then tested by calculating bound - state properties in a scalar field theory using six different bound - state equations . it is found that equations obtained using the method espoused here approximate the wave functions obtained from their parent four - dimensional equations significantly better than the corresponding quasi - potential equations do .

nucl-th9603008

with the advent of experimental facilities such as cebaf , which are capable of probing hadronic systems at energies where relativistic effects become important , the development of theoretical frameworks in which consistent relativistic calculations of few - body hadronic systems can be performed is no longer merely desirable , it is essential . an obvious starting point for such a calculation in a two - body system ( such as the deuteron ) is the bethe - salpeter equation ( bse ) for the four - dimensional , covariant , two - to - two amplitude @xmath0 , @xmath1 where @xmath2 is the free two - particle propagator , which in our convention is @xmath3 with @xmath4 where @xmath5 @xmath6 is a positive infinitessimal and @xmath7 is the bethe - salpeter kernel @xcite . in principle @xmath7 should include all two - particle irreducible two - to - two feynman graphs . solution of ( [ eq : bse ] ) with the full two - particle irreducible kernel is impractical and usually resort is made to the ladder approximation @xcite . some undesirable features attend this approximation ( see @xcite for a full discussion ) , not the least of which is the fact that the ladder bse does not have the correct one - body limit @xcite . ( an equation is said to have the correct one - body limit if , when the mass of one particle is taken to infinity , the equation reduces to the klein - gordon or dirac equation for the light particle moving in the static field of the now infinitely massive source . ) by contrast , the full bse ( [ eq : bse ] ) does possess the correct one - body limit . three - dimensional quasi - potential equations ( qpes ) are realized by rewriting the bethe - salpeter equation for the two - particle amplitude ( [ eq : bse ] ) as a pair of coupled equations : @xmath8 where the propagator @xmath9 is arbitrary . if @xmath9 is chosen to contain a one - dimensional delta function constraining the relative four - momentum , then eq . ( [ eq : qpe ] ) becomes a three - dimensional equation for the amplitude @xmath0 . examples of this approach include the equations of blankenbecler - sugar @xcite and gross @xcite . in principle the driving term @xmath10 should be determined by solving the four - dimensional integral equation ( [ eq : ueqn ] ) . however , given the difficulty of solving such equations , and our lack of knowledge about the best form of @xmath7 for hadronic physics , usually @xmath10 is chosen to be a one - boson exchange interaction : @xmath11 the t - matrix then obeys the three - dimensional integral equation @xmath12 a reasonable description of few - hadron systems is obtained by fitting coupling constants and some mass parameters of @xmath13 to the nucleon - nucleon scattering data . ( examples of this approach include , but are by no means limited to , refs . @xcite . ) the qpe formalism of gross obeys the one - body limit @xcite . another variant of the quasi - potential approach which respects the one - body limit was derived by wallace and mandelzweig in references @xcite . we provide a generalization of that formalism in this paper . as originally derived , the qpe of wallace and mandelzweig contains , for the nucleon - nucleon system , a one - boson exchange potential which is instant in the center - of - mass frame . this involves the use of a constraint on the relative four - momentum , @xmath14 , of the form @xmath15 where @xmath16 is the total two - body four - momentum . in the center - of - mass frame , the equation for the deuteron vertex function takes the form : @xmath17 where @xmath18 refers to the wallace - mandelzweig choice for the propagator @xmath9 . the formalism is covariant since the constraint ( [ eq : constraint ] ) is expressed covariantly and the pair of equations ( [ eq : wdinst ] ) and ( [ eq : ueqn ] ) is equivalent to the bound - state bse : @xmath19 eq . ( [ eq : wdinst ] ) for the cm frame deuteron vertex function has been used in recent work by devine and wallace @xcite . this formalism may be developed in a manifestly covariant fashion following the technique of fuda @xcite . this involves writing all four - vectors in terms of their components parallel and perpendicular to the total four - momentum : @xmath20 where @xmath21 , with @xmath22 . a fundamental flaw exists in quasipotential formalisms that are based on a form for @xmath9 which contains a delta function . it is generally impossible to systematically correct the lowest - order approximation by use of eq . ( [ eq : ueqn ] ) because unphysical singularities arise . in two - body hadronic systems , such as the deuteron , calculations may still be pursued using equations such as ( [ eq : wdinst ] ) which are themselves free of these singularities . on the other hand , in any electromagnetic reaction where the photon momentum @xmath23 is non - zero , the constraint ( [ eq : constraint ] ) can not hold for the initial and final - state relative four - momenta in the cm frame @xcite . therefore a boosted vertex function is required for electromagnetic matrix element calculations , or , for that matter , for any qpe calculation in a three - body problem . a boost equation for the quasipotential may be deduced from eq . ( [ eq : ueqn ] ) , but unphysical singularities arise in the boosted interaction @xcite . these are removed from the theory if the full result for @xmath10 , as defined by ( [ eq : ueqn ] ) , is used in the quasi - potential equation , but no truncation of @xmath10 at any finite order in @xmath24 is free of singularities . therefore , in this paper we seek a general procedure for the reduction of four - dimensional equations which _ does not involve the use of delta functions_. we develop a procedure by which a four - dimensional equation may be approximated by a three - dimensional equation and the interaction in that equation improved systematically . the technique is presented in section [ sec - section2 ] . it has a close connection to the work of klein @xcite on three - dimensional reductions of four - dimensional equations , and to standard time - ordered perturbation theory , as discussed in section [ sec - section3 ] . in general the interaction in the three - dimensional equations we discuss is not covariant . but , since the application of our delta - function - free reduction technique to infinite order produces an equation equivalent to the original four - dimensional equation , the sum of all terms in the three - dimensional formalism must produce covariant results . in nuclear physics it is known that contributions which are of higher - order in the coupling are of increasingly shorter range . such short - range contributions to hadronic interactions must always be treated in an essentially phenomenological manner . therefore , it is expected that a truncation of the interaction at some finite order in the coupling will be useful for applications . it should always be appropriate to absorb the effects of the neglected higher - order graphs into phenomenological parameters , thus ameliorating the non - covariance of the theory . in particular , we show in section [ sec - section3 ] that the leading - order boost corrections to the interaction obtained by forest , pandharipande and friar @xcite are indeed contained within our truncated interaction . in section [ sec - section4 ] we show that the crossed - box graph may be approximately rewritten as an iterate of the ladder kernel : @xmath25 where the form of @xmath26 is derived in section [ sec - section4 ] . in particular , this approximation is exact in the high - energy ( eikonal ) and one - body limits , and may be systematically improved upon in other regimes . this leads us to propose the four - dimensional integral equation : @xmath27 this equation has the correct one - body limit and results from an approximate resummation of the bse kernel @xmath7 . it reduces to the wallace - mandelzweig equation ( [ eq : wdinst ] ) if the dependence of @xmath28 on the time component of the relative four - momentum in the cm frame is neglected . a three - dimensional reduction of eq . ( [ eq:4dwm ] ) using the method developed in section [ sec - section2 ] provides a systematic way to include retardation effects , and so improve on eq . ( [ eq : wdinst ] ) . this reduction to a three - dimensional equation is performed in section [ sec - section5 ] . the dynamical boost of the three - dimensional interaction present in the equation is realized through simple dependence of the interaction on the total three momentum and energy of the system , thus eliminating the boost problem of eq . ( [ eq : wdinst ] ) . in section [ sec - section6 ] the predictions of the three - dimensional integral equation obtained in section [ sec - section5 ] are compared to those of five other bound - state equations : the ladder bse and eq . ( [ eq:4dwm ] ) , both of which are , of course , four - dimensional equations , and the three - dimensional equations of salpeter , klein and wallace & mandelzweig . the results of sections [ sec - section3][sec - section5 ] are presented concurrently for a scalar and a spinor field theory . in particular , in the scalar case we use the @xmath29 field theory , defined by the lagrangian @xmath30 coupling terms include mass factors so that the couplings @xmath31 and @xmath32 are dimensionless , thus ensuring that the limits @xmath33 and @xmath34 correspond to the appropriate one - body limits . it is this field theory in which the numerical calculations of section [ sec - section6 ] are performed . in the spinor case the lagrangian is @xmath35 the arguments of sections [ sec - section3][sec - section5 ] are , in fact , quite general , and , with small modifications , also apply to field theories involving other types of particles . consider the bethe - salpeter equation for the bound - state vertex function @xmath36 , i.e. , @xmath37 here @xmath16 is the total two - body four - momentum , and @xmath14 and @xmath38 are the relative four - momenta . this equation is completely general , and applies to the bound - state vertex function in any field theory . throughout this paper we find it convenient to omit the explicit integration from such equations , abbreviating them as follows @xmath39 suppose that the driving term of this equation , @xmath7 , is separated into two pieces : @xmath40 where @xmath41 is three - dimensional in the sense that it does not depend on the zeroth component of the relative four - momenta @xmath14 and @xmath38 . for instance , if ( [ eq : bsbse ] ) was being solved in the two - body center - of - mass frame and @xmath7 was a one - boson - exchange kernel then @xmath41 could be chosen to be the usual static one - boson - exchange kernel . note that even though the whole driving term @xmath7 is lorentz covariant , because the restriction on @xmath41 is frame dependent , the pieces @xmath41 and @xmath42 are not . the splitting of @xmath7 leads to the following coupled equations for @xmath36 : @xmath43 both these equations are four - dimensional , however , @xmath44 has no dependence on the zeroth component of the relative four - momentum because of the defining condition of @xmath41 . meanwhile , the second equation may be formally solved to obtain : @xmath45^{-1 } k_2 g_0 \gamma _ 1 \nonumber \\ & = & - \gamma _ 1 + \left [ 1 - k_2 g_0 \right]^{-1 } \gamma _ 1 . \label{eq : gamma2}\end{aligned}\ ] ] this last result may then be substituted into eq . ( [ eq : g1.1 ] ) to yield @xmath46 where @xmath47 eq . ( [ eq : g1.2 ] ) then becomes a three - dimensional equation because the implied integrations over time - components of momenta only affect @xmath48 , i.e. , it reduces to @xmath49 where @xmath50 here we have used the same implied integration notation for the three - dimensional integral equation ( [ eq : rtpt ] ) as for a four - dimensional equation . this practice continues below and the context should make it clear whether the equation in question is a three or four - dimensional one . eqs . ( [ eq : rtpt ] ) and ( [ eq : gg ] ) are exactly equivalent to the original bse ( [ eq : bsbse ] ) . in order to reconstruct the bethe - salpeter amplitude @xmath36 from the non - covariant three - dimensional @xmath44 one must use eq.([eq : gamma2 ] ) rearranged into the form @xmath51^{-1 } \gamma _ 1.\ ] ] the green s function , @xmath52 , defined by eq . ( [ eq : reint ] ) is , in fact , the fourier transform of the corresponding equal - time co - ordinate space green s function @xmath53 . this may be seen by defining @xmath54 where @xmath9 is the appropriate four - dimensional two - body green s function , then inserting the integral representation of the delta functions and taking the fourier transform @xcite . if @xmath41 is now chosen to be a cm - frame instant interaction , @xmath55 , then ( [ eq : gg ] ) may be taken to define @xmath48 order by order in @xmath56 . this provides a systematic way to calculate retardation corrections to the salpeter equation @xcite @xmath57 which results from taking @xmath48 at zeroth order in @xmath56 . in particular , at first order in @xmath56 we find that : @xmath58 the use of this three - dimensional propagator in place of @xmath59 in ( [ eq : rtpt ] ) would therefore incorporate leading - order retardation corrections in the three - dimensional equation . there is , however , a general way to account for all the relative - energy integrations in the modified green s function @xmath52 . recall that the choice of @xmath41 is subject only to the constraint that it should not depend on the zeroth component of the relative four - momentum in , say , the cm frame . consider eq . ( [ eq : gg ] ) rewritten as @xmath60 and choose @xmath41 such that @xmath61 . this provides the defining condition , @xmath62 since this choice of @xmath41 means that @xmath63 the full dynamics reduces to a three - dimensional integral equation with the free propagator @xmath59 , i.e. , @xmath64 and all the complexities of the relative - energy integrations are transferred to the interaction , @xmath41 . it follows that @xmath41 is the two - particle irreducible ( 2pi ) interaction , where two - particle irreducibility is defined with respect to the three - dimensional propagator @xmath59 . to show this formally , rearrange @xmath65 as follows , @xmath66^{-1 } - [ 1 - k g_0 ] ^{-1 } \right\ } \rangle \nonumber \\ & = & \langle g_0 t g_0 \rangle - \langle g_0 t g_0 \rangle k_1 \langle { \cal g } \rangle , \label{eq : rearrange } \end{aligned}\ ] ] where @xmath67 is the bethe - salpeter t - matrix defined by eq . ( [ eq : bse ] ) . use of eq . ( [ eq : rearrange ] ) in eq . ( [ eq : k1 ] ) produces @xmath68 where @xmath69 is the three - dimensional t - matrix corresponding to the bethe - salpeter t - matrix @xmath0 . thus @xmath41 is , indeed , the irreducible interaction which produces the t - matrix @xmath70 using the propagator @xmath71 . we therefore expect that the rules for the construction of @xmath41 will be akin to those for the two - particle irreducible interaction in time - ordered perturbation theory . the connection to time - ordered perturbation theory ( topt ) emerges for a simplified dynamics in which only positive - energy states are kept in @xmath2 . for the case of a field theory of nucleons and pions it has been shown that the green s function @xmath72 obeys the rules of topt in the no - anti - nucleon case @xcite . the same derivations suffice to prove the result for the scalar field theories defined in the introduction . it therefore immediately follows that @xmath41 is the 2pi topt amplitude in the case that only positive - energy states are kept in @xmath2 . we now provide a specific example of the derivation of the time - ordered perturbation theory amplitude by this route . firstly , it is true in general that , if in eq . ( [ eq : k1 ] ) @xmath48 is replaced by @xmath2 , then the three - dimensional kernel becomes @xmath73 we note that this formula was given by klein in his work on deriving three - dimensional scattering equations from four - dimensional ones @xcite . equations ( [ eq : k1 ] ) and ( [ eq : gg ] ) may thus be thought of as providing a generalization of the formalism of klein . secondly , suppose that @xmath7 is expanded according to the power of the coupling constant in each of its contributions , i.e. , @xmath74 and for the moment only the second - order , or ladder , contribution to @xmath7 is kept , i.e. , ( [ eq : k12 ] ) is rewritten as : @xmath75 we now calculate this amplitude in the absence of negative - energy states . first , the propagators are split according to eq . ( [ eq : di ] ) , and only the positive - energy pieces retained . second , the inverse fourier representations of all quantities are inserted , so that the time - ordering can be clearly elucidated . third , a change of variables to time differences is made . fourth , the relative - energy integrations are performed , generating delta functions on some of the time differences . last , the integrals over these time differences are performed . this procedure allows the calculation of @xmath76 in any frame . indeed , it is completely general , and can be used to calculate @xmath41 no matter what @xmath7 is chosen . if the total four - momentum of the two - body system in the frame of interest is written @xmath77 , then it is seen that @xmath76 takes the standard form of topt : @xmath78 where @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 ( @xmath84 ) has the argument @xmath85 ( @xmath86 ) in place of @xmath87 ( @xmath88 ) . here , @xmath89 and @xmath90 are the the masses of the two interacting particles , @xmath91 is the mass of the exchanged meson and the factor @xmath92 is defined by : @xmath93 note that to extract the mass of the bound - state , @xmath94 , one must set @xmath95 . the total and relative three - momenta are related to the individual particle momenta in the usual way : @xmath96 where @xmath97 and @xmath98 are any two real numbers which obey @xmath99 . figure [ fig - fig1 ] shows the time - ordered diagrams corresponding to eq . ( [ eq : toptope ] ) . the bound - state equation corresponding to this interaction is then : @xmath100 where we have omitted the factor of @xmath101 from @xmath59 , which integrates out of the equation trivially . here : @xmath102 where @xmath103 and @xmath104 . now suppose that we have a set of free single - particle positive - energy momentum - space wave - functions , @xmath105 , for both the spin - half and scalar theories . in the spin - half case these will be the spinors , and their normalization may be chosen such that : @xmath106 ( where spin indices have been suppressed ) . in the scalar case eq . ( [ eq : projector ] ) may also be enforced , since the @xmath105 s may be chosen to be merely numerical factors , i.e. , we may define @xmath107 expanding : @xmath108 ( with an implicit summation on spin indices in the spinor case ) and manipulating eq . ( [ eq : bse1 ] ) leads to : @xmath109 with : @xmath110 note that the topt interaction which appears in the bound - state equation ( [ eq : bse2 ] ) changes with total three - momentum of the system . this provides the dynamical boost of the interaction and it is straightforward to show that it incorporates the perturbative boost correction discussed in ref . @xcite . that is , in the equal - mass case , to second order in @xmath111 we have ( neglecting @xmath112 and @xmath113 terms ) : @xmath114 where we have written the interaction as a function of the momentum of the exchanged meson , @xmath115 . if it is true , as argued in ref . @xcite , that this leading - order boost correction is sufficiently accurate for studies of three - nucleon systems , then the boost effects incorporated in the topt result ( [ eq : toptope ] ) should be more than adequate . indeed , we believe such a boost to be preferable to that of eq . ( [ eq : fboost ] ) , since it involves operators which are bounded as @xmath116 , whereas , at large values of @xmath111 the perturbative boost correction ( [ eq : fboost ] ) diverges . the identification of the kernel @xmath41 with the usual time - ordered perturbation theory interaction is only correct in the absence of negative - energy states . with the full four - dimensional propagator , the form of @xmath76 obtained from eq . ( [ eq : k13 ] ) in the manner described above eq . ( [ eq : toptope ] ) is : @xmath117 , \label{eq : kleinpot}\end{aligned}\ ] ] where @xmath118 and @xmath119 . note that here , and throughout the rest of this paper , the @xmath120 prescriptions can be generated by understanding the masses @xmath89 , @xmath90 , and @xmath91 to have a small negative imaginary part . now the three - dimensional propagator is @xmath121 ( where once again we have omitted the momentum - space delta - function for notational simplicity ) , which is the salpeter propagator @xcite . if the negative - energy piece is removed , this becomes the blankenbecler - sugar @xcite or logunov - tavkhelidze @xcite propagator . note that in the spin - half case this propagator does not have a unique inverse , consequently in that case the derivation of the bound - state equation analogous to eq . ( [ eq : bse2 ] ) usually assumes that the @xmath122 and @xmath123 pieces of @xmath124 are zero . the graphs corresponding to the terms in eq . ( [ eq : kleinpot ] ) are shown in figure [ fig - fig2 ] . graphs ( a)-(d ) are generated by the first line of eq . ( [ eq : kleinpot ] ) , graphs ( e)-(h ) by the second line , and graphs ( i ) and ( j ) by the third line . the fourth to sixth lines of ( [ eq : kleinpot ] ) generate the same ten graphs , but with particles and anti - particles interchanged . note that the graphs drawn in figure [ fig - fig2 ] do not represent actual physical processes , but rather contributions to the `` potential '' @xmath76 . as observed by klein , if negative - energy particles are present , the rules for the construction of this interaction differ from those of standard topt @xcite . in general , it can be shown that the rules for the construction of the full two - body equal - time green s function , @xmath72 , for the case of positive - energy initial and final - state particles , are , in either of the two field theories given in the introduction : 1 . draw all topologically distinct , two - to - two graphs , which do not contain vacuum - vacuum subdiagrams , remembering that different time - orderings contribute to different graphs . 2 . construct the expression for each individual graph exactly as in topt , save that : 1 . all one - particle energies are to be relativistic . if in some intermediate state of the graph both the initial - state particles are present with exactly the same momenta as in their initial state , i.e. , their state is unaltered from the beginning of the graph , then in the topt denominator corresponding to that state the quantity @xmath125 must be replaced by @xmath126 . 3 . similarly , if both final - state particles are present with exactly the same momenta as in their final state , i.e. , they have undergone their last interaction but are still present in some intermediate state , then in the topt denominator corresponding to that state the quantity @xmath127 must be replaced by @xmath126 . if both these last two conditions are satisfied then the denominator for the relevant state is the topt denominator for that state , but with @xmath128 replaced by @xmath129 . note that these are not the rules for the amplitude obtained from the ladder bse . the green s function defined by these rules is the @xmath130 piece of the _ full _ two - body equal - time green s function of the field theory . the substitution of these @xmath126s in the intermediate - state denominators can be shown to be due to a crucial difference between the way initial and final - states are treated in the calculation of @xmath131 and in topt . in topt the initial and final times in the diagram are taken to minus and plus infinity respectively , thus guaranteeing that no interaction takes place before the beginning ( or after the end ) of the propagation of the two - particle state . however , when @xmath131 is calculated in our work the initial and final times are kept finite . consequently the first ( last ) event in the diagram need not be the creation of the @xmath132 or @xmath133 pair , it may be the creation out of ( destruction into ) the vacuum of a @xmath134 or @xmath135 state . ( see for instance the second graph in figure [ fig - fig2 ] . ) the possibility of such an event happening before or after the @xmath132 or @xmath133 propagation in some contributions to @xmath131 leads to denominators which differ from the topt ones in those pieces of @xmath131 . it follows that , if the pieces of the hamiltonian of the theory which couple @xmath136 or @xmath135 states directly to the vacuum are removed from the hamiltonian , then the modification of the topt denominators is not necessary , and the rules for calculation of the two - particle to two - particle part of the green s function @xmath131 become exactly those for the calculation of the topt green s function . this equivalence of @xmath131 and the time - ordered perturbation theory green s function in the absence of such terms from the hamiltonian was demonstrated for a field theory of nucleons and pions by kvinikhidze and blankleider @xcite . as discussed in section [ sec - section2 ] , the use of a static interaction in the bethe - salpeter equation leads to the salpeter equation ( [ eq : saleq ] ) . this equation is often used for the interactions of two relativistic particles . it is straightforward to show that the static limit of the potential given by eq . ( [ eq : kleinpot ] ) is the usual instant interaction in the positive - energy sector . on the other hand , eq . ( [ eq : kleinpot ] ) predicts that the coupling between the @xmath137 and @xmath138 states is suppressed by retardation effects which cause it to be a factor @xmath139 ( in the equal - mass case ) smaller than the rest of the static interaction . this factor is zero in the static limit . thus , for a causal interaction , the correct static limit of the equation ( [ eq:3deqn ] ) with the interaction ( [ eq : kleinpot ] ) is the breit equation , i.e. , eq . ( [ eq : saleq ] ) with the @xmath138 states omitted from the calculation . in order to recover the results of klein for two - meson exchange interactions , and so make connection with the recent work on two - pion exchange of rijken and stoks @xcite and lahiff and afnan @xcite , we evaluate @xmath48 to first order in @xmath140 in eq . ( [ eq : k1 ] ) . this produces @xmath141 now expanding @xmath7 as in eq . ( [ eq : kexp ] ) and dropping terms of higher than fourth order in the coupling constant leads to @xmath142 when applied to particle - particle scattering eq . ( [ eq : k14 ] ) gives exactly the same results obtained by klein , except that where klein removed the iterated second - order three - dimensional interaction by hand , here the formalism provides the subtraction naturally . the procedure developed in section [ sec - section2 ] thus provides the generalization of klein s method . the interaction and vertex function defined by eqs . ( [ eq : k1 ] ) , ( [ eq : gg ] ) , and ( [ eq:3deqn ] ) have the full relative - energy dynamics of the bethe - salpeter equation ( [ eq : bsbse ] ) included in them . now let us return to the four - dimensional bse ( [ eq : bse ] ) . the simplest bse with the correct one - body limit is @xmath143 where @xmath144 is the sum of all 2pi ladder and crossed - ladder two - to - two graphs @xcite . the bound - state masses predicted by such an equation have recently been obtained by monte carlo integration in the feynman - schwinger representation of both scalar @xmath145 field theory and scalar qed by nieuwenhuis , tjon and simonov @xcite . in general though , the kernel @xmath144 is too complicated for ( [ eq:3.1 ] ) to be solved by standard means . therefore , in this section we shall derive a four - dimensional equation which has a simple kernel , the appropriate meson - production thresholds and the correct one - body and high - energy ( or eikonal ) limits . a form of the following derivation appeared in @xcite . although it is impossible to rewrite a crossed graph exactly as an iterate of the ladder kernel @xmath28 , the leading contributions of these graphs to the high - energy and one - body limits are iterative . this can be easily shown in the case of the fourth - order crossed - box graph . in both of the field theories defined in the introduction , this graph corresponds to an expression : @xmath146 where the lines have been assigned the momenta shown in figure [ fig - xedbox ] , and @xmath16 is the total momentum , which is conserved : @xmath147 the propagators @xmath148 were given in eq . ( [ eq : di ] ) , and the factor @xmath92 was defined in eq . ( [ eq : m ] ) . following wallace and mandelzweig @xcite we define , for any four - vector @xmath23 , quantities @xmath149 and @xmath150 , as follows : @xmath151 where @xmath152 is the unit four - vector in the direction of @xmath16 and @xmath153 consequently in eq . ( [ eq:3.6 ] ) the argument of the function @xmath154 may be rewritten : @xmath155 now suppose that the momentum of particle 2 is large . this may occur because @xmath156 ( one - body limit ) or because particle 2 has very high energy ( eikonal limit ) . in either case its intermediate and final - state momenta will be largely unaffected by the presence of particle 1 , and so we may approximate the perpendicular components as unchanging , @xmath157 indeed , making the replacement ( [ eq : eikapprox ] ) in ( [ eq:3.6 ] ) will not affect the value of @xmath158 in the limit @xmath159 . this argument shows that @xmath158 may be approximately rewritten as : @xmath160 where @xmath28 is the ladder bse driving term : @xmath161 in operator notation : @xmath162 with @xmath163 here @xmath164 and @xmath165 are defined via : @xmath166 and so in the cm frame @xmath167 . the propagator @xmath26 defined by ( [ eq : gc1 ] ) therefore depends on the parallel components of the external momenta . thus the use of operator notation in eq . ( [ eq : kxit ] ) is not strictly correct . in eq . ( [ eq : newgc ] ) we redefine @xmath26 in order to remove this dependence on external momenta . however , that change in @xmath26 modifies the analytic structure of the amplitude defined by the corresponding integral equation . therefore , for the present we persist with @xmath26 defined by ( [ eq : gc1 ] ) , and use an improper operator notation . note that if the particles are on - shell in their initial and final states then : @xmath168 these arguments show that eq . ( [ eq : kxit ] ) will be exact in the infinite - mass and high - energy limits , thus demonstrating that the pieces of @xmath158 which survive in these two limits may , indeed , be written as iterates of @xmath28 . now suppose that @xmath144 is written as @xmath169 in principle this is always possible , as ( [ eq : inteqkx ] ) may be taken as a definition of @xmath170 . at second order in the coupling we clearly have : @xmath171 while the above argument shows that with this @xmath172 @xmath173 in the high - energy and infinite - mass limits . thus , a reasonable choice for @xmath170 is @xmath174 . equation ( [ eq : inteqkx ] ) then defines corrections to this choice . once this @xmath170 is chosen , eq . ( [ eq : inteqkx ] ) and the bse ( [ eq:3.1 ] ) may be combined to yield an `` improved '' ladder bse , which , in the two - body cm frame , after a change of variables to total and relative four - momenta , takes the form @xmath175 \gamma(p_0,{\bf p};s ) . \label{eq:3.19}\end{aligned}\ ] ] we stress that what has been done here is to take certain pieces of the bethe - salpeter kernel @xmath144 and rewrite them in the form @xmath176 , @xmath177 , etc . consequently , eq . ( [ eq:3.19 ] ) is equivalent to a bethe - salpeter equation in which graphs other than one - meson exchange are approximately included in the kernel . thus we expect that the solution of this equation may provide a better description of the dynamics of two - particle systems than the ladder bse amplitude . this will be especially true for systems with one particle much heavier than the other . in these systems the one - body limit constitutes an important piece of the dynamics . unlike the ladder bse , eq . ( [ eq:3.19 ] ) has the correct one - body limit . this may be shown as follows . for the on - shell vertex function @xmath178 . if @xmath178 the sum of the two particle 2 propagators , multiplied by @xmath92 is , in the spin - zero case : @xmath179= \frac{4 m_1 m_2}{(e_2^{\mbox { on } } - p_0)^2 - \epsilon_2 ^ 2 + i\eta } + \frac{4 m_1 m_2}{(e_2^{\mbox { on } } + p_0)^2 - \epsilon_2 ^ 2 + i\eta},\ ] ] while for the spin - half case the same combination becomes : @xmath180 in the @xmath34 limit , @xmath181 , and so this expression reduces to @xmath182 2 m_1 = - 2 \pi i \delta ( p^0 ) 2 m_1,\ ] ] for the spin - zero case and @xmath183 for the spin - half case , where @xmath184 is the positive - energy projection operator for an infinitely massive particle two . thus , in the infinite @xmath90 limit eq . ( [ eq:3.19 ] ) yields the klein - gordon or dirac equation for the wave function of particle 1 moving in the static @xmath185 field generated by particle 2 @xmath186 or @xmath187 where @xmath188 or @xmath189 and @xmath190 . equation ( [ eq:3.19 ] ) also has the appropriate meson - production thresholds , and therefore could be used as a basis for four - dimensional calculations of two - body bound - state properties . however , the equation can not be written in the form @xmath191 and consequently the method of section [ sec - section2 ] can not be applied to it . therefore , we now seek an approximate version of eq . ( [ eq:3.19 ] ) which can be written in the form ( [ eq:3.19b ] ) . in the on - shell vertex function @xmath178 . in what follows we use @xmath178 in the integrand also when the amplitude is not on - shell . this approximation provides a four - dimensional version of the three - dimensional wallace - mandelzweig equation obtained in ref . @xcite . in the center - of - mass frame it is @xmath192 \gamma(p;e_1^{\mbox { on}},e_2^{\mbox { on } } ) . \label{eq:3.20}\end{aligned}\ ] ] redefining @xmath26 to be , in the cm frame , @xmath193 allows eq . ( [ eq:3.20 ] ) to be written : @xmath194 the following points are worth noting : 1 . because the new choice for @xmath26 does not affect the on - shell fourth - order piece of the amplitude eq . ( [ eq : kxit ] ) is still exactly true on - shell in the one - body and eikonal limits . 2 . in the limit @xmath34 eq . ( [ eq:3.20 ] ) has the correct one - body limit , as can be seen by a similar argument to that given for eq . ( [ eq:3.19 ] ) . if a pinch analysis of the singularities of the amplitude @xmath0 defined by the scattering equation corresponding to ( [ eq:3.20 ] ) is performed then it is found that at order @xmath195 there is a production cut in the @xmath196 plane of the half - off - shell amplitude which extends upwards from @xmath197 and lies infinitessimally below the real axis . in the on - shell amplitude at sixth order this cut appears in the integrand and overlaps with the pole of the green s function at @xmath198 , so producing a cut in the @xmath199 plane of @xmath0 which extends upwards from @xmath200 this is not a particle - production threshold which exists in the full bethe - salpeter scattering amplitude . hence its existence in the amplitude @xmath0 defined by the scattering equation corresponding to eq . ( [ eq:3.20 ] ) must be regarded as a deficiency of that equation . it arises because the method used to derive eq . ( [ eq:3.20 ] ) is only guaranteed to produce the correct cut - structure for the second and fourth - order on - shell amplitudes . it should be noted that the vertex function given by eq . ( [ eq:3.19 ] ) does not contain this anomalous threshold . + the value of @xmath126 at which the threshold ( [ eq : e2thresh ] ) occurs in the on - shell amplitude is @xmath201 . in fact , since equation ( [ eq:3.19 ] ) was designed for use in the @xmath202 regime this is actually above @xmath203 . therefore the theory has the correct threshold structure + for all energies @xmath126 such that @xmath204 . furthermore , if @xmath205 then the threshold ( [ eq : e2thresh ] ) actually lies at @xmath203 , which is where the first production threshold of the crossed - box graph should be . consequently , we think of this cut as representing the usual single - pion - production threshold of the crossed - box graph , but somewhat displaced if @xmath202 . this equation was derived for particles of different mass . if we desire an equation for identical particles a symmetric form of the propagator @xmath26 must be used . ( we shall return to this point in the next section . ) the above derivation could equally well be pursued for the exchange of vector particles . however , in , for instance , scalar qed , account must be taken of the seagull graphs . furthermore , in the case of spinor qed more care must be taken , since an additional piece of the interaction may be generated when one attempts to write @xmath206 . ( the interested reader may consult @xcite for details on these points . ) we now apply the relative - energy integration method of section [ sec - section2 ] to the four - dimensional equation ( [ eq:3.20 ] ) . by so doing we first recover wallace and mandelzweig s original quasi - potential equation , and then calculate a three - dimensional kernel which includes first - order retardation corrections to this result . observe that once @xmath26 is defined by eq . ( [ eq : newgc ] ) , eqs . ( [ eq : inteqkx ] ) and ( [ eq:3.1 ] ) are equivalent to : @xmath207 where @xmath170 is regarded as being _ defined _ by ( [ eq : inteqkx ] ) . this provides a method for calculating corrections to eq . ( [ eq : modified4dapp ] ) , which was obtained by assuming that @xmath174 . the result of applying the method of section [ sec - section2 ] to eq . ( [ eq : modified4d ] ) is : @xmath208 if @xmath170 is taken to be @xmath209 , where @xmath209 is the static one - sigma exchange potential , @xmath210 these equations reduce to @xmath211 which is a three - dimensional equation with a free two - body propagator that in the cm frame takes the form : @xmath212 where once again a factor of @xmath101 has been omitted for notational simplicity . note that this propagator is equivalent to that derived by cooper and jennings by a different technique @xcite . note also that in the case @xmath205 this same propagator is referred to as the equal - time propagator by tjon and collaborators @xcite . the desired generalization of eq . ( [ eq:4.3 ] ) which incorporates retardation and boost effects follows from applying the ideas of sections [ sec - section2 ] and [ sec - section3 ] to eq . ( [ eq:4.2 ] ) . if @xmath213 is chosen such that the last term in eq . ( [ eq:4.2 ] ) is zero , the formula @xmath214 is obtained for the general three - dimensional interaction to be used in @xmath215 a first step in the inclusion of retardation and boost effects may be taken by following klein s work and replacing @xmath216 by @xmath217 . this yields what we refer to as the first - order relative - energy integration result : @xmath218 we now take @xmath170 equal to @xmath28 and calculate @xmath219 in the center - of - mass frame . in order to do this we first decompose the cm frame propagators : @xmath220 where @xmath221 may take on the values @xmath222 . it follows that the amplitude @xmath219 may be written as @xmath223 the sixteen contributions @xmath224 may be calculated by the same method used to obtain the klein potential ( [ eq : kleinpot ] ) . they are found to take on a slightly different form depending on whether @xmath225 is equal to plus or minus one . in order to simplify the expressions for the @xmath219 s we define the following quantities : @xmath226 if @xmath227 then : @xmath228\nonumber\\ & + & f \left[\frac{1 } { e_1 ' + e_2 ' } + \frac{1}{e_2 + e_2 ' -\omega}\right ] \frac{1}{e_1 ' + e_2 - \omega } \frac{1}{e_1 + e_2}. \label{eq:4.15}\end{aligned}\ ] ] conversely , if @xmath229 then : @xmath230\nonumber\\ & + & f \left[\frac{1 } { e_1 ' + e_2 ' } + \frac{1}{e_2 + e_2 ' - \omega } \right ] \frac{1}{e_1 ' + e_2 - \omega } \frac{1}{e_1 + e_1 ' - \omega}\nonumber\\ & + & f \frac{1}{e_1 ' + e_2 ' } \left[\frac{1}{e_1 + e_1 ' - \omega } + \frac{1}{e_2 + e_2 ' - \omega } \right ] \frac{1}{e_1 + e_2}. \label{eq:4.16}\end{aligned}\ ] ] we note the following points about the interaction defined by eqs . ( [ eq:4.5 ] ) , ( [ eq:4.7 ] ) , ( [ eq:4.11 ] ) , ( [ eq:4.15 ] ) and ( [ eq:4.16 ] ) . 1 . as expected from the four - dimensional equation ( [ eq:3.20 ] ) it assumes a static form in the limit @xmath231 . in other words , it has the correct one - body limit . ( recall that we assumed that particle 2 was the heavier of the two . ) 2 . in the static limit of the interaction defined by eqs . ( [ eq:4.15 ] ) and ( [ eq:4.16 ] ) all couplings to the @xmath138 states are suppressed . on the other hand , the other couplings tend to the same static limit as the @xmath130 piece of the interaction . thus , in this limit the @xmath232 coupling , which is due to z - graphs , is correctly given by the instantaneous exchange interaction , and only the @xmath138 states need be omitted from a calculation . the role of the @xmath122 and @xmath138 states in this limit is therefore consistent with the one - body limit which motivated eq . ( [ eq : reewm ] ) . 3 . also as expected from ( [ eq:3.20 ] ) @xmath233 has a singularity structure which is different to that of the ordinary time - ordered perturbation theory amplitude : it contains a cut beginning at @xmath234 . the same comments made about the unusual singularity structure of ( [ eq:3.20 ] ) apply to this cut . 4 . because we have used a form of @xmath26 derived by considering the limit @xmath34 , the interaction is _ not _ symmetric under the interchange of particle 1 and 2 labels . points ( 2 ) and ( 3 ) are not a reflection of the underlying physics of the meson - exchange process , but rather of the particular iterative form we used in our attempt to sum some of the higher - order graphs in the kernel of the bse ( [ eq:3.1 ] ) . the choice of @xmath26 used above involves an approximation which is appropriate in the regime @xmath202 . for equally massive particles a more appropriate choice for the propagator @xmath26 is : @xmath235.\end{aligned}\ ] ] note that this choice is label symmetric , and hence the resulting bound - state equation may be used for identical particles . the @xmath26 defined by ( [ eq : newgc ] ) did not have this property . when eqs . ( [ eq:4.7 ] ) and ( [ eq:4.8 ] ) are applied with this symmetrized form of @xmath26 the result is a @xmath213 of the form ( [ eq:4.11 ] ) , but with an @xmath224 which may be written in any @xmath236-spin channel as : @xmath237 , \label{eq : v1symm}\end{aligned}\ ] ] with all symbols defined as above . having derived the interaction which is to be used in eq . ( [ eq : reewm ] ) we may now compare the two - body bound - state properties predicted by this integral equation with those properties obtained from other three and four - dimensional calculations . this is done in the scalar field theory defined in the introduction . the masses @xmath89 , @xmath90 , and @xmath91 are chosen to be @xmath238 and @xmath239 . units are then chosen so that @xmath240 . the three - dimensional equations to be considered are : 1 . the salpeter equation , which may be written : @xmath241 with @xmath209 given by eq . ( [ eq : vsiginst ] ) and @xmath59 by eq . ( [ eq : g0ave ] ) . note that in this scalar field theory eq . ( [ eq : salbs ] ) is exactly equivalent to the blankenbecler - sugar equation obtained from the ladder bse . 2 . the klein equation for a one - sigma - exchange interaction , which is eq . ( [ eq:3deqn ] ) with the interaction defined by eqs . ( [ eq : kleinpot ] ) and ( [ eq : g0ave ] ) . the equation proposed by wallace and mandelzweig , which we refer to as the et equation : @xmath242 where @xmath18 was given in eq . ( [ eq:4.5 ] ) . 4 . the retarded et equation , eq . ( [ eq:4.1 ] ) , with the label - symmetric interaction obtained in the previous section , i.e. , with @xmath213 defined by ( [ eq : v1symm ] ) , ( [ eq:4.11 ] ) , ( [ eq:4.7 ] ) and ( [ eq:4.5 ] ) the four - dimensional equations these four calculations are to be compared to are : 1 . the ladder bethe - salpeter equation . 2 . equation ( [ eq:3.20 ] ) which , unlike the ladder bse , has the correct one - body limit . we consider the six bound - state equations as falling into two groups : the first contains the ladder bse and the two three - dimensional reductions thereof : the salpeter equation and the klein equation . the second group contains the four - dimensional wallace - mandelzweig equation ( [ eq:3.20 ] ) and the two three - dimensional equations derived from it : the equal - time equation ( [ eq:4.3 ] ) and the first - order relative - energy integration equation ( [ eq : reewm ] ) with the interaction @xmath213 defined by eqs . ( [ eq : v1symm ] ) , ( [ eq:4.11 ] ) , ( [ eq:4.7 ] ) and ( [ eq:4.5 ] ) . within these two groups the three equations may then be thought of as being zeroth ( instant potential equations ) , first ( klein - type ) and infinite ( four - dimensional equations ) order relative - energy integration results . at any given bound - state mass the equation : @xmath243 whether it be a three or four - dimensional integral equation , is an eigenvalue problem . the coupling constant @xmath244 which appears in the kernel must be chosen such that @xmath245 has eigenvalue one . if we are searching for the ground - state of the system then we may assume that @xmath36 is an @xmath246-wave state , thus implying that the function @xmath36 has no angular dependence . the integration over @xmath247 may then be easily performed . the resultant one ( or two ) dimensional integral equation may then be discretized , reducing the problem to one of solving for the eigenvalues of a matrix version of @xmath94 . note that in the four - dimensional equations it is necessary to perform a wick rotation in the variable @xmath248 before the kernel @xmath94 is discretized , as otherwise the analytic structure in the kernel makes this procedure numerically unstable . when this approach is implemented for the equations listed above the six bound - state spectra shown in figure [ fig - bsspectra ] are obtained . the following observations may immediately be made : 1 . the two instant formalisms ( salpeter and et ) predict more binding at a given coupling than either their corresponding four - dimensional equation or their corresponding klein - type interaction . the calculations in the second group always predict deeper binding than the corresponding calculations in the first group . in other words , the inclusion of pieces of higher - order graphs in the kernel of the integral equation yields deeper binding at a given coupling . the first - order relative - energy integration ( klein - type ) interactions give predictions for the bound - state spectra which are significantly closer to those of the full four - dimensional calculation than the spectra of the instant equations . all of these results can be understood on physical grounds . point 1 arises because the inclusion of retardation in the scalar field theory always reduces the amount of attraction in the bound - state . an example of this is provided by one - sigma exchange . when the full retardation corrections predicted by time - ordered perturbation theory are included the instant potential ( [ eq : vsiginst ] ) is replaced by the retarded interaction ( [ eq : toptope ] ) ( provided the effects of negative - energy states are ignored ) . if @xmath249 this will always be a negative number that is smaller in magnitude than the instant potential . meanwhile , the inclusion of pieces of additional higher - order graphs leads to deeper binding since in the @xmath29 theory _ all _ time - ordered perturbation theory graphs are attractive in the bound - state region . most significant though , in our opinion , is point 3 , which shows that the first - order relative - energy integration ( klein - type ) potentials do a better job than the instant interactions of approximating the four - dimensional equation which both are derived from . all six of these calculations aim to sum the main contributions to the two - to - two amplitude . however , figure [ fig - bsspectra ] shows that they give different results , and so they can not all be correct . the question therefore arises : what is the binding energy at a given coupling if the full scattering series is summed ? a partial answer to this question was recently provided by nieuwenhuis and tjon , who summed the series of all ladder and crossed - ladder diagrams using monte carlo integration techniques in the feynman - schwinger representation , and so obtained two - body bound - state masses in this scalar theory @xcite . their results , calculated at six different coupling strengths , show that the sum of all such graphs predicts deeper binding than _ any _ of our six integral equation calculations . in particular , the et appears to do the best job of reproducing the bound - state spectrum obtained by nieuwenhuis and tjon . at first sight this appears to recommend the et as the best formalism for doing three - dimensional calculations . however , two points must be remembered before a definite conclusion is drawn . firstly , the reason that the nieuwenhuis - tjon calculation predicts so much more binding than the two four - dimensional integral equation calculations performed here is precisely the fact discussed in connection with point 2 above : in the @xmath29 theory _ all _ time - ordered perturbation theory graphs give attraction in the bound - state region . hence the nieuwenhuis - tjon calculation , which includes many more such graphs than both the ladder bse and four - dimensional wallace - mandelzweig equation calculations , _ must _ predict more binding . therefore , we claim that the salpeter and et equations do a `` better '' job of imitating the full nieuwenhuis - tjon calculation than the ladder bse and four - dimensional wallace - mandelzweig equation because instant calculations such as the salpeter and et ones ignore two competing effects . 1 . higher - order graphs are left out of their kernel . if included these graphs would lead to deeper binding at a given coupling . exchanged - meson retardation is completely ignored . if included this retardation would lead to shallower binding at a given coupling . in other words , we claim that the et formalism s apparent success is due to the cancellation of these two neglected effects and can not necessarily be interpreted as a signal of that approach having the `` right physics '' for the problem . secondly , _ none _ of these calculations have been performed in the way that calculations in few - body hadronic physics are usually performed . when such bound - state calculations are done in hadronic physics one usually acknowledges that the use of _ any _ integral equation necessarily means the neglect of some graphs of the theory , and so one regards the couplings that appear in that equation as effective . these couplings are then usually fixed by fitting to some set of observables , and then certain different observables are predicted . this makes the connection to an underlying field theory somewhat more tenuous , but it does allow calculation to proceed in situations where : * the value of the coupling which appears in the `` fundamental '' lagrangian is not known . * the problem is highly non - perturbative and consequently _ any _ truncation is somewhat suspect . therefore , to test the procedure of section [ sec - section2 ] for deriving three - dimensional integral equations in a way consistent with that in which it would be used in hadronic physics , we fix the mass of the bound - state at @xmath250 and calculate the wave - functions @xmath251 for each of the four different three - dimensional calculations . ( here @xmath9 is the free two - particle green s function for the relevant calculation . ) each three - dimensional wave - function must be normalized according to @xmath252 where @xmath253 is the full two - particle green s function appearing in the three - dimensional equation from which @xmath36 was obtained . these wave - functions are then compared to the four - dimensional wave functions , integrated over the zeroth component of the relative four - momentum : @xmath254 here @xmath9 is the appropriate free two - particle four - dimensional green s function . the four - dimensional vertex function used in ( [ eq:4dwf ] ) is normalized according to : @xmath255 where @xmath253 is the full two - particle green s function appearing in the four - dimensional equation from which @xmath36 was obtained . the left - hand sides of eqs . ( [ eq:3dnormn ] ) and ( [ eq:4dnormn ] ) may be reduced to two ( and four)-dimensional integrals using the fact that the @xmath246-wave wave - function @xmath256 is angle - independent . note that in the four - dimensional case , wick rotation in the zeroth component allows use of the vertex function on the @xmath248 or @xmath196 imaginary axis . when the normalized wave - functions of the equations in group 1 ( group 2 ) are compared the results shown in figure [ fig - wfcomp1 ] ( [ fig - wfcomp2 ] ) are obtained . it is immediately seen that all the wave - functions at this bound - state mass have essentially the same features . however , in both figures it is clear that the three - dimensional equation with the first - order relative - energy integration , or klein - type , interaction reproduces the integrated four - dimensional wave function considerably more accurately than the instant calculation does . this shows that the procedure given in section 2 _ does _ provide a way to systematically derive three - dimensional equations which are closer to their four - dimensional counterparts than those which are obtained from merely adopting an instant interaction . quasi - potential equations ( qpes ) , despite their success as a basis for hadronic phenomenology , can not be improved upon systematically . this is because whenever they are used beyond first - order they predict amplitudes with unphysical singularities . these singularities also appear whenever attempts are made to boost the interaction appearing in the qpe from one frame to another . they arise from these equations use of delta - function constraints on the relative four - momentum . because of this shortcoming of qpes , in section [ sec - section2 ] of this paper we sought , and found , a systematic procedure for deriving three - dimensional equations from four - dimensional ones which does not involve the use of delta functions . as shown in section [ sec - section3 ] , this procedure is akin to the work of klein . it allows the derivation of a three - dimensional kernel which includes , in a systematic expansion which may be pursued to any desired accuracy , the effects of the relative - energy integration present in the four - dimensional integral equation . if the procedure is applied to infinite order , a result equivalent to the original four - dimensional equation is obtained . at any order the kernel derived has a simple interpretation in terms of the diagrams of time - ordered perturbation theory , except that the presence of negative - energy states requires additional rules . the resultant three - dimensional equations therefore include more of the meson - retardation effects than quasi - potential equations , and incorporate the effects of a dynamical boost . since the bse with any kernel which is a finite sum of feynman graphs does not have the correct one - body limit , if the procedure of section [ sec - section2 ] is directly applied to any solvable bse a three - dimensional equation without the correct one - body limit is found . however , as demonstrated by wallace @xcite , and recapitulated here in section [ sec - section4 ] , a bethe - salpeter equation with the correct one - body limit may be reorganized so that the pieces of the kernel which contribute at leading order in the one - body limit take on an iterative form . hence a four - dimensional equation , eq . ( [ eq:3.19 ] ) , with the correct one - body limit and meson - production thresholds is derived . a further approximation leads to a simple four - dimensional equation to which the procedure of section [ sec - section2 ] may be applied , eq . ( [ eq:3.20 ] ) . in making that approximation some of the meson - production thresholds are displaced . nevertheless , the equation still has exactly the correct cut - structure for @xmath257 . in section [ sec - section5 ] the procedure of section [ sec - section2 ] was applied to the four - dimensional equation ( [ eq:3.20 ] ) . the result is a simple three - dimensional equation with the correct one - body limit which has straightforward boost properties . the main shortcoming of this equation is the fact that certain thresholds of the original bethe - salpeter wave function are modified in the three - dimensional wave function . however , this modification is due to the approximation in moving from ( [ eq:3.19 ] ) to ( [ eq:3.20 ] ) , rather than to any deficiencies of our procedure for deriving three - dimensional equations . it could be systematically corrected for , at the price of complicating the three - dimensional kernel . finally , in section [ sec - section6 ] we compared and contrasted the bound - state properties predicted by six different bound - state equations in a scalar field theory . a first group of equations contained the ladder bethe - salpeter equation and two three - dimensional equations based on it : the salpeter equation , and the equation obtained by applying our method to first order which we referred to as the klein equation , since the formula obtained for the three - dimensional interaction appears in klein s work . in the second group were eq . ( [ eq:3.20 ] ) , the so - called equal - time equation and the equation derived in section [ sec - section6 ] . it was found that in each group of three equations the equation derived by the klein - like delta - function - free reduction technique of section [ sec - section2 ] approximated its parent four - dimensional equation better than the corresponding instant equation did . however , the bound - state spectra from the instant et equation lies closest to the bound - state spectrum obtained from the sum of ladder and crossed - ladder graphs , as calculated by nieuwenhuis and tjon @xcite . this can be understood on physical grounds and , we argued , is not necessarily a recommendation for the use of the et equation in physical systems . indeed , when the wave functions of a two - body system in a scalar field theory are examined it is seen that the equation derived in section [ sec - section5 ] does a much better job of reproducing the integrated four - dimensional wave function than the et equation wave function does . the results of section [ sec - section6 ] are in harmony with recent results for phase shifts in scalar - scalar scattering obtained by lahiff and afnan @xcite . they found that the klein method reproduces the phase shifts obtained from the corresponding bethe - salpeter equation much better than a blankenbecler - sugar calculation with an instant interaction . these formal and numerical results indicate that the procedure of section [ sec - section2 ] is successful in its goal of providing a way to systematically obtain three - dimensional equations from four - dimensional ones without the use of delta functions . in particular , when applied to the four - dimensional equation ( [ eq:3.20 ] ) this procedure yields a three - dimensional equation which , unlike the corresponding quasi - potential equation ( [ eq:4.3 ] ) , has a well - defined boost . this would seem to make such an equation a good starting point for calculations in few - hadron systems .

the spitzer survey of stellar structure in galaxies ( s g ) is a volume , magnitude , and size - limited survey of 2352 nearby galaxies with deep imaging at 3.6 and 4.5 . in this paper we describe our surface photometry pipeline and showcase the associated data products that we have released to the community . we also identify the physical mechanisms leading to different levels of central stellar mass concentration for galaxies with the same total stellar mass . finally , we derive the local stellar mass - size relation at 3.6 for galaxies of different morphologies . our radial profiles reach stellar mass surface densities below . given the negligible impact of dust and the almost constant mass - to - light ratio at these wavelengths , these profiles constitute an accurate inventory of the radial distribution of stellar mass in nearby galaxies . from these profiles we have also derived global properties such as asymptotic magnitudes ( and the corresponding stellar masses ) , isophotal sizes and shapes , and concentration indices . these and other data products from our various pipelines ( science - ready mosaics , object masks , 2d image decompositions , and stellar mass maps ) , can be publicly accessed at irsa ( http://irsa.ipac.caltech.edu/data/spitzer/s4g/ ) .

1505.03534

understanding how galaxies acquired their baryons over cosmic time is a key open question in extragalactic astronomy . how and when did galaxies of different types assemble the bulk of their stellar mass ? nearby galaxies are of particular relevance in this context : they constitute the present - day product of billions of years of evolution , so the past assembly history of these galaxies is encoded in the present - day spatial distribution of old stars within them . therefore , an accurate census of the current stellar structure of nearby galaxies provides essential constraints on the physics of galaxy formation and evolution . to address this critical issue , in this paper we present deep mid - infrared radial profiles for the more than 2300 galaxies in the spitzer survey of stellar structure in galaxies ( s@xmath0 g , see @xcite for the full survey description ) . large surveys of nearby galaxies traditionally have been carried out in the optical regime , for example the third reference catalogue of bright galaxies ( rc3 , @xcite ) , the sloan digital sky survey ( sdss , @xcite ) , or the carnegie - irvine galaxy survey ( @xcite ) . however , translating optical measurements into stellar masses is not a straightforward task . on the one hand , the optical mass - to - light ratio ( @xmath3 ) is strongly dependent on the star formation history of the galaxy ( e.g@xmath4 @xcite ) . on the other hand , internal extinction by dust obscures a significant fraction of the optical output of galaxies ( e.g@xmath4 , @xcite ) . observations at near- and mid - ir bands can circumvent these issues . at these long wavelengths , @xmath3 is only a shallow function of the star formation history , and dust extinction plays a minor role , leading to milder variations in @xmath3 compared to optical wavelengths ( @xcite ) . several wide field ir surveys have been carried out over the years , such as the two micron all sky survey ( 2mass , @xcite , see also @xcite ) , the deep near infrared survey ( denis , @xcite ) and more recently the wide - field infrared survey explorer ( wise , @xcite ) . these surveys provide robust number statistics , but at the expense of a shallow image depth , insufficient to map the stellar content in the faint outskirts of galaxies . in the case of wise , the spatial resolution at 3.4 and 4.6@xmath1 is relatively coarse ( @xmath5 ) . conversely , other surveys have obtained much deeper ir images for small samples of several tens or a few hundred objects . projects like these include the ohio state university bright galaxy survey ( osubgs , @xcite ) , the near - infrared atlas of s0-sa galaxies ( nirs0s , @xcite ) , the spitzer infrared nearby galaxies survey ( sings , @xcite ) and the spitzer local volume legacy survey ( lvl , @xcite ) . but a complete understanding of galaxy assembly involves many independent parameters , such as galaxy mass , morphology , environment , bar presence , gas and dark matter content , etc . with so many independent dimensions along which a sample should be sliced , any sample with a few hundreds objects will be broken down into bins too small for a reliable understanding of the impact of any given parameter . s@xmath0 g was specifically designed to answer the need for a deep , large and uniform ir survey of nearby galaxies . the survey contains over 2352 galaxies within 40mpc , away from the galactic plane ( @xmath6 ) , with an extinction - corrected b - band vega magnitude brighter than 15.5 and a b - band diameter larger than 1 . we observed these galaxies at 3.6 and 4.5@xmath1 with the infrared array camera ( irac , @xcite ) onboard spitzer ( @xcite ) . we followed the successful observing strategy of the sings and lvl programs , reaching azimuthally - averaged stellar mass surface densities @xmath7 ; this is a physical regime where the baryonic mass budget is dominated by atomic gas . the s@xmath0 g images are processed through a suite of different pipelines designed to produce a wealth of enhanced data products . the first three pipelines are described in this paper . pipeline 1 ( p1 ) creates science - ready mosaics by combining all the individual exposures of each galaxy . pipeline 2 ( p2 ) masks out foreground stars , background galaxies and artifacts . pipeline 3 ( p3 ) performs surface photometry on the images and derives integrated quantities such as asymptotic magnitudes , isophotal sizes , etc . pipeline 4 ( p4 , @xcite ) uses galfit ( @xcite ) to decompose each galaxy into bulges , disks , bars , etc . finally , pipeline 5 ( p5 , @xcite ) combines the 3.6 and 4.5@xmath1 images to produce stellar mass maps via an independent component analysis , following the methodology developed by @xcite . the goal of this paper is twofold . first , we describe the technical details and inner workings of p1 , p2 and p3 , as well as the resulting data products . then , we discuss two particular scientific applications of our data : the local stellar mass - size relation , and the physics behind the vast diversity of morphologies in galaxies with the same stellar mass . for each galaxy we have obtained radial profiles both with fixed and free ellipticity and position angle . from these profiles we have also derived global measurements such as asymptotic magnitudes and stellar masses , isophotal sizes and ellipticities , and concentration indices . this dataset constitutes a unique tool to address many important issues on stellar structure , including but not limited to : * the radial structure of dark matter in galaxies . our profiles constrain the radial distribution of stellar mass , which is a necessary ingredient when modelling rotation curves to infer the radial distribution of dark matter ( @xcite ) . * the scaling laws of disks ( @xcite ) , in particular the local stellar mass - size relation ( @xcite ) and , when kinematic data are available , the tully - fisher relation ( @xcite ) . * the inside - out assembly of disks , by comparing our profiles of old stars with uv and optical profiles probing younger stellar populations ( @xcite ) . * the local bar fraction and the sizes , strengths and shapes of bars ( @xcite ) . * radial migration of old stars and the assembly of galactic outskirts ( @xcite ) , in particular the study of breaks in radial projected surface density profiles and their links to bar resonances ( @xcite ) . * one - dimensional structural decompositions ( @xcite ) , from which one can derive physical parameters such as disk scale - lengths , bulge effective radii , bulge - to - disk ratios , etc . * the intrinsic face - on circularity of disks and their vertical thickness ( @xcite ) . * quantified galaxy morphology via non - parametric estimators such as concentration indices ( @xcite ) . * the internal structure of elliptical galaxies ( @xcite ) , in particular their intrinsic axial ratios ( @xcite ) , their srsic indices ( @xcite ) and their boxy / disky structure ( @xcite ) . this paper is structured as follows . in section [ sec_sample ] we summarize our sample selection criteria and observations . section [ sec_p1 ] contains a brief outline of the image reduction performed by p1 . a more detailed description can be found in appendix [ app_p1 ] . the object masking procedure is described in section [ sec_p2 ] . then , in section [ sec_p3 ] we explain our ellipse fitting technique and the resulting data products . readers interested in the scientific applications of these products can proceed to section [ sec_results ] , where we discuss the stellar - mass size relation , as well as the variety of morphologies and radial concentration that we find at a fixed stellar mass . then , section [ sec_irsa ] explains how to access and download our data . finally , in section [ sec_summary ] we summarize our main conclusions . to assemble the s@xmath0 g sample we made use of the hyperleda database ( @xcite ) . we selected all galaxies with radio - derived radial velocity @xmath8km / s , which translates into a distance cut of @xmath9mpc for @xmath10kms@xmath11mpc@xmath11 . we only considered galaxies with total corrected blue magnitude @xmath12 , blue light isophotal diameter @xmath13 , and galactic latitude @xmath6 . the final sample after applying all these selection criteria comprises 2352 galaxies . using radio - based velocities biases the sample against gas - poor early - type galaxies . however , our recently approved cycle 10 program ( prog . i d 10043 , pi : k. sheth ) will complete spitzer s legacy with archival and new observations of all @xmath14 early - type galaxies within the s@xmath0 g volume , following the same observing strategy and selection cuts . out of the 2352 s@xmath0 g galaxies , @xmath15% were already present in the spitzer archive as a result of previous programs carried out during the cryogenic phase . we observed the remaining @xmath16% during the post - cryogenic mission . in this regard , from now on we will refer to the galaxies in our sample as either `` archival '' or `` warm '' , respectively . we observed the warm galaxies with a total on - source exposure time of 240s per pixel , and mapped them out to at least @xmath17 . depending on the apparent size of each galaxy , we used either a single dithered map or a mosaic of several pointings . each galaxy was observed in two visits separated by at least 30 days , in order to image the galaxy with two different orientations thanks to the rotation of the telescope . this allowed for a better correction of cosmetic effects , cosmic ray and asteroid removal , and subpixel sampling . most of the archival galaxies had been also mapped with at least 240s per pixel and out to @xmath18 . however , six archival galaxies had exposure times between 90 and 200s per pixel , and 125 were mapped out to less than @xmath19 . they represent a small fraction of the total sample , and only specific science goals are affected by this . we therefore decided not to reobserve these galaxies , as the incremental gain of repeating these observations would not have made up for the required additional observing time ( see @xcite for more details on the sample selection ) . we refer the reader to appendix [ app_p1 ] for a more detailed description of p1 . briefly speaking , p1 creates science - ready mosaics by combining the different exposures of each galaxy . the pipeline first matches the background level of the individual exposures , using the overlapping regions among them . it then combines all frames following standard dither / drizzle procedures ( @xcite ) . the final science - ready mosaics are delivered in units of mjysr@xmath11 , with a pixel scale of 0.75 . the fwhm of the psf at 3.6 and 4.5@xmath1 is 1.7 and 1.6 , respectively . for the farthest galaxies in our sample at @xmath20mpc , this translates into a physical size of @xmath21pc . besides the scientific images themselves , p1 also produces weight - maps showing how many individual frames cover each pixel of the final mosaics . for each galaxy in the sample , p2 creates masks of contaminant sources like background galaxies , foreground stars and artifacts . such masks are necessary for subsequent analysis such as background measurements and surface photometry ( p3 ) , 2d image decompositions ( p4 ) and mass map generation ( p5 ) . the mask creation procedure consists of two steps : ( i ) automatic generation of initial masks , and ( ii ) visual check and editing to create final clean masks . a first set of masks is automatically generated by running sextractor ( @xcite ) on the 3.6 and 4.5@xmath1 images ; the resulting segmentation maps constitute our initial raw masks . masking sources is relatively easy in those areas of an image with little or no emission from the main target galaxy . however , it becomes more complicated on top of the galaxy , where we need to avoid masking regions belonging to the galaxy itself . therefore , for each galaxy and band we create three automatic masks with high , medium , and low detection thresholds in sextractor that control how aggressively different sources are masked on the main body of the galaxy . during the visual quality check step , for each galaxy we first choose the best mask among the three ones with different thresholds . this best mask is then visually inspected and edited by hand , masking additional sources missed by sextractor and unmasking any regions of the galaxy that may have been masked by mistake . in particular , we often have to manually add to the masks the extended haloes and diffraction spikes of bright stars . conversely , certain sources on the target galaxies like bright clumps or the ansae of bars are sometimes picked up by sextractor , and need to be excluded from the masks by hand . this editing process is performed with a custom code described in @xcite . this idl routine displays side by side the original and masked images , and provides several geometric shapes that the user can employ to interactively mask or unmask regions . the editing is first manually done on the 3.6@xmath1 images , and then the editing is automatically transferred to the 4.5@xmath1 ones , checking that artifacts remain properly masked . thus , we end up with two final , edited masks for each galaxy , one for each band . a sample mask is shown in fig . [ sample_mask ] . the masking process is to some extent subjective , especially for the faintest sources , due to the lack of color information . theses masks are mainly intended to remove extraneous sources that would significantly contaminate and distort our radial profiles and integrated magnitudes . therefore , users interested in specific structures such as very faint hii regions or globular clusters around our galaxies should not blindly use our masks without double - checking the nature of these sources with ancillary multi - wavelength data . in this section we describe in detail the inner workings of pipeline 3 . briefly speaking , p3 first determines the background level of the images . it then locates the centroid of the galaxy and performs different sets of ellipse fits . finally , it measures several properties such as asymptotic magnitudes , isophotal sizes and concentration indices . a careful determination of the background level and noise is essential to perform reliable surface photometry . this is often done by placing several `` sky boxes '' around each galaxy and measuring the corresponding background statistics in them . with large samples such as s@xmath0 g it is desirable to implement this process in a way that is automatic yet flexible . pipeline 3 automatically places several sky boxes around each galaxy ( see fig . [ sample_mask ] ) . this is done by defining two concentric and adjacent elliptical annuli that surround the entire galaxy . each ring is then azimuthally subdivided into 45 sectors or boxes . since these boxes will contain in general a certain amount of masked pixels , each box is grown radially outwards until each one contains 1000 unmasked pixels . we then measure the median sky level and local rms inside each box , as well as the large - scale rms between the sky levels of all boxes . while the number of 1000 pixels per box is to some extent arbitrary , it is chosen to provide a reliable measurement of the local rms , while at the same time yielding boxes that are small enough to be easily accommodated within our images . note also that before using the p2 masks , we first grow the masked areas by 2 pixels to make sure that the faint wings of the psf do not contaminate our sky measurements . the sky boxes are initially placed by default at @xmath22 from the galaxy center , but this value is modified as needed for each galaxy in order to ensure a proper background subtraction . to do this , we compare the values between the inner and outer rings to make sure that there are no significant differences , which could be a telltale sign of contamination from the galaxy itself . we double check this by verifying that the growth curve is flat ( see section [ sec_magnitudes ] ) . we also check that the boxes are not too close to the frame edges ( which are noisier as a result of the dithering pattern ) , or that they do not fall in the adjacent frames ( which may have a somewhat different background value ) . in cases with a complicated background structure , the pipeline allows one to manually distribute the sky boxes as deemed appropriate . the distribution of measured background levels in the whole s@xmath0gsample is shown in fig . [ sky_hist ] . the histograms peak at @xmath23 and @xmath24abmagarcsec@xmath25 for the 3.6 and 4.5@xmath1 bands , respectively . it is illustrative to verify whether our measured values of the background level agree with theoretical expectations . the background in the s@xmath0 g images is almost entirely dominated by zodiacal light ( or `` zodi '' ) coming from dust grains in the ecliptic plane . both the thermal emission of the grains and the scattered sunlight contribute to the zodi . at 3.6 and 4.5@xmath1 , both contributions are roughly equal . thermal emission from interstellar cirrus in the milky way amounts to merely 1%-3% of the background in our images , as it peaks at much longer wavelengths ( and the s@xmath0 g sample specifically avoids the galactic plane anyway ) . contamination by the unresolved cosmic infrared background is negligible at these bands compared to the zodi ( @xcite ) . figure [ sky_ecliptic ] shows how the observed sky level varies with the ecliptic latitude of each galaxy . as expected , since the primary source of background emission at these wavelengths is zodi , the distribution clearly peaks in the ecliptic plane . for each galaxy we retrieved from the fits header the background level predicted for its galactic coordinates and epoch of observation.@xmath26 skydrkzb . ] the green curves in fig . [ sky_ecliptic ] are the upper and lower envelopes of these predicted values . we can see that both the dependency with ecliptic latitude and the range in background level at fixed latitude nicely agree with our measured values . interestingly , about 30 warm galaxies exhibit background levels at 3.6@xmath1 well above the predicted values . a visual inspection of these images revealed a recurrent diffuse artifact , both in the main frame where the galaxy is and in the flanking one ( fig . [ backgr_artifacts ] ) . these few galaxies were observed in a narrow time window during august - september 2009 , very shortly after the spitzer s warm phase began . this smooth background artifact may be therefore due to the detector temperature and bias not being settled yet at that time . besides the background level itself , it is also important to characterize the background noise at different spatial scales , as this determines our ability to detect and measure faint structures at the outskirts of our galaxies . figure [ sky_noise ] shows the distribution of the local , pixel - to - pixel noise in our images . the histogram for the warm galaxies is much narrower than that for the archival ones , which is expected given that the warm galaxies were imaged with the same observing strategy , whereas the archival come from a variety of different programs . in general we reach a local surface brightness sensitivity per pixel of @xmath27 and @xmath28abmagarcsec@xmath25 at @xmath29 and @xmath30 , respectively . note , however , that these are values _ for individual pixels_. when measuring the flux of a given extended source , this noise component scales down with the square root of the number of pixels in the region where the measurement is being performed . in particular , when measuring radial profiles at the outskirts of galaxies , this local noise component becomes negligible compared to the large scale background noise , which is given by the rms between the median sky values measured in the different boxes . as shown in fig . [ sky_noise ] , this large scale component peaks at 26 - 26.5abmagarcsec@xmath25 at a @xmath29 level , and @xmath31abmagarcsec@xmath25 at @xmath30 . as a reference , @xmath32abmagarcsec@xmath25 corresponds to a stellar mass surface density of @xmath33 , adopting @xmath34 as measured by @xcite ; see also @xcite . + once the background level and noise have been measured , our pipeline proceeds to perform surface photometry on the images . we first use the iraf task imcentroid to find accurate coordinates for the center of the galaxy . we then run the task ellipse ( @xcite ) to obtain radial profiles of surface brightness ( @xmath35 ) , ellipticity ( @xmath36 ) and position angle ( pa ) . we perform three separate runs of ellipse with different settings : 1 . * fixed center ; free @xmath36 and pa ; radial resolution @xmath37 . * we use these fits with a coarse radial increment to derive @xmath36 and pa in the outer parts of the galaxy , thus defining a global shape and orientation for each object . for each band , we provide values of @xmath36 and pa at two levels of surface brightness , @xmath38 and @xmath39abmagarcsec@xmath25 . after testing how sensitive these values are to variations in the sky subtraction , input fitting parameters , amount of masked objects , etc . , we recommend using the 25.5 values as they are more stable . * fixed center ; @xmath36 and pa fixed to the values at 25.5abmagarcsec@xmath25 ; @xmath40 . * these fits have a finer radial resolution that matches the irac psf at these wavelengths . by keeping @xmath36 and pa fixed and equal to the global outer values , these fixed - fits are ideal to measure disk scale - lengths , disk break radii , to perform 1d bulge - disk decompositions , etc . we also employ these profiles to measure the integrated magnitude of each galaxy from the growth curve ( sect . [ sec_magnitudes ] ) . * fixed center ; free @xmath36 and pa ; @xmath40 . * these free - fits with a fine resolution are well suited to study in detail the structural properties of features such as bars , which leave very characteristic signatures in the radial profiles of @xmath36 and pa . an example of these fits is shown in fig . [ sample_prof ] . [ cols="^,^ " , ] for simplicity and to improve the robustness of the fits , once we measure the central coordinates of galaxy , these are kept fixed during the ellipse fitting . however , the pipeline allows one to leave the center as a free parameter too . we have in fact performed such fits to study individual galaxies with offset bars ( e.g@xmath4 ngc3906 , de swardt et al@xmath4 2014 , in prep . ) . we normally begin the ellipse fitting at an intermediate radius , typically between @xmath41 , using the optical @xmath36 and pa from hyperleda as input guesses for the fitting routine . starting at that initial radius , the ellipse fitting proceeds first outwards and then inwards . it should be noted that the ellipse algorithm was specifically designed for galaxies with a smooth radial brightness distribution . it therefore works correctly for elliptical galaxies , but it does not cope well with features such as spiral arms or rings that can significantly modify the local luminosity gradient . as a result , ellipse can sometimes stop fitting at a given radius in spiral galaxies . tuning the input parameters can sometimes fix this problem and provide a good fit , but in general this is a blind trial - and - error solution with an unpredictable outcome . some authors have in fact circumvented this issue by automatically running ellipse tens or even hundreds of times on each image , varying the initial fitting parameters each time until a good fit is obtained ( @xcite ) . while this approach can be practical in small images of distant galaxies , it is prohibitively time - consuming for the large images in s@xmath0 g , where a single fit can take up to several minutes ( and this is further complicated by the large number of galaxies in our sample ) . we thus decided to implement an optimized procedure that is better tailored to our needs . after each run of ellipse , the pipeline looks for any radial interval where the fit did not converge , and it then refits that particular radial interval alone . the process is then iterated several times to ensure that no radial gap is left unfit . in general most of our galaxies can be properly fit with a single ellipse run , but this automatic iterative process was of great help to handle troublesome cases that would have otherwise required considerable manual work . in any case , for each galaxy we always visually inspect the output of the pipeline : we overlay the fitted ellipses on the galaxy image , and we plot the radial profiles of @xmath35 , @xmath36 and pa , making sure that the results accurately probe the different structures within the galaxy . dwarf galaxies can be particularly challenging for our pipeline . they are often partially resolved into stars in our images , and their clumpy and patchy appearance can sometimes fool the ellipse fitting algorithm . for these galaxies we recommend not to overinterpret any signature in the @xmath36 and pa profiles without inspecting the images themselves . since these galaxies normally lack well - defined large - scale structures such as bars , the profiles with fixed @xmath36 and pa are more suitable in these cases . we have released all these ellipse fits in the form of ascii tables containing the full output of the ellipse task . these tables not only include radial profiles of @xmath35 , @xmath36 and pa , but also other quantities such as the harmonic deviations from a perfect ellipse , which are commonly used to quantify the boxiness / diskiness of the isophotes at different radii ( e.g@xmath4 @xcite ) . apart from the usual output from ellipse , we also include in our ascii profiles additional columns that are specific to the irac data used here . in particular , the irac photometry needs to be corrected for the extended wings of the psf and the diffuse light that is scattered throughout the detector . here we rely on the extended source aperture correction provided in the irac instrument handbook . given an elliptical aperture with major and minor radii @xmath42 and @xmath43 , if @xmath44 is the total observed flux inside such an aperture , the corrected flux is given by : @xmath45 in this expression @xmath46 is the equivalent radius of the elliptical aperture , in arcseconds . the coefficients @xmath47 , @xmath48 and @xmath49 are equal to 0.82 , 0.370 and 0.910 , respectively , at 3.6@xmath1 , and 1.16 , 0.433 and 0.94 at 4.5@xmath1 . similarly , if @xmath50 is the surface brightness along an isophote ( rather than the total flux inside that radius as before ) , the aperture - corrected surface brightness can be obtained by performing a series expansion on the previous equation : @xmath51 our ascii tables include radial profiles with and without these aperture corrections . these corrections are estimated to be uncertain at a 5%-10% level . we have always applied the extended source aperture correction at all radii in our profiles . however , the reader should keep in mind that the point source corrections might be more suitable at very small radii ( @xmath52-@xmath53 ) , especially if the nucleus is bright and compact . finally , all the released profiles have been corrected for foreground extinction , using the color excess map of @xcite and the milky way extinction curve of @xcite . nevertheless , at 3.6 and 4.5@xmath1 this extinction correction is typically around 0.005mag . the pipeline also estimates the uncertainty in @xmath35 , @xmath36 and pa at different radii . for @xmath36 and pa we rely on the errors determined by the ellipse task , which result from the internal errors in the harmonic fit ( @xcite ) . as for @xmath35 , the error provided by ellipse is derived from the rms of the pixel values along each isophote , so it mostly reflects azimuthal variations of the stellar emission at a given radius rather than the true uncertainty of the mean surface brightness at that radius . we therefore opted to measure the error in @xmath35 ourselves . if @xmath54 is the incident pixel intensity , prior to sky subtraction , and @xmath55 is the sky level , then the surface brightness is given by : @xmath56 where @xmath57 includes the magnitude zero - point plus any other global multiplying factor , such as the aperture corrections mentioned before . following the methodology detailed in @xcite and @xcite , the uncertainty in @xmath35 can be estimated as : @xmath58 the error in @xmath57 is dominated by the uncertainty in the aperture corrections , and is of the order of 5 - 10% as explained in the previous section . since any change in @xmath57 will merely translate into a global offset of the radial profiles , the values of @xmath59 in our ascii tables and plots do not explicitly include the contribution of @xmath60 . we computed @xmath61 , which is the error in the incident pixel intensity , by assuming poissonian statistics : @xmath62 where @xmath63 is the effective gain and @xmath64 is the number of pixels within each isophote . the effective gain , in turn , was derived from the nominal detector gain and the exposure time of each pixel according to the weight maps . the term @xmath65 represents the uncertainty in the background level , and was obtained as : @xmath66 where @xmath67 and @xmath68 are the local and large scale background rms , respectively , as described in section [ sec_sky ] . here @xmath64 is , again , the number of pixels along a given isophote , and @xmath69 is the number of pixels inside the boxes used to measure the background level . the first term in eq . [ eq_sky_err ] reflects the contribution of the local , pixel - to - pixel noise to the final error in the surface brightness . this contribution scales down as @xmath70 , so it becomes negligible at large radii , where the flux is averaged over a large number of pixels . it is the second term , the large - scale noise , the one that dominates the error budget at large radii . note that the measured rms between the sky values in the different boxes is partly contaminated by the local noise , in the sense that even if there were no true large - scale fluctuations , the rms between the sky boxes would be @xmath71 , on average . the correction term in eq . [ eq_sky_err ] accounts for this . from the surface photometry we measure the asymptotic magnitude of each galaxy by extrapolating the growth curve to infinity . to do this we use the profiles with fixed @xmath36 and pa and 2 resolution . as an example , the bottom panel of figure [ gcurve ] shows the 3.6@xmath1 growth curve of ngc0936 , which rises fast at the center and then slowly approaches a flat regime . to determine this asymptotic value , we first compute the gradient of the growth curve at all radii . in the outer parts of the galaxy , the gradient at a given radius and the magnitude inside that radius follow a linear trend ( top panel ) . we then apply a linear fit whose y - intercept ( the magnitude for a flat gradient ) is by definition the asymptotic magnitude . note that growth curves are often used in shallow images to recover the flux buried beneath the noise in the outskirts of galaxies . this is not the case in the s@xmath0 g images , which are deep enough to always comfortably reach the flat regime of the growth curve . therefore , in practice there is no extrapolation involved when fitting the growth curves of our galaxies . in this regard , the error bugdet in the surface photometry is dominated by the large - scale rms , and not by any possible non - zero slope in the growth curve . once we have determined the asymptotic apparent magnitudes , we derive the absolute ones using redshift - independent distances from ned whenever available ( 79% of the sample ) , and redshift - dependent ones otherwise , assuming @xmath10kms@xmath11mpc@xmath11 . for those galaxies with more than one redshift - independent distance measurement ( 61% of the whole sample ) , the typical rms between the different measured distances is @xmath72% , which translates into an error of @xmath73mag in the absolute magnitude . while this is not strictly speaking a true formal uncertainty , for such nearby galaxies it provides a more realistic estimate of the actual distance error than the uncertainty in the radial velocity , which is just @xmath73% for our galaxies . in fig . [ morpho_mass ] we plot the 3.6@xmath1 absolute magnitude for all our galaxies as a function of their optical morphological type as given by hyperleda ( the 4.5@xmath1 values follow a very similar trend ) . the conversion to stellar mass in the rightmost vertical axis was done via the @xmath3 value of @xcite . at the late - type end of the sequence , the stellar mass rises monotonically from @xmath74@xmath75 for irregular galaxies to @xmath76@xmath75 for sc ones . the trend then remains remarkably flat out to s0 galaxies , beyond which the stellar mass increases again for elliptical galaxies . the rms in stellar mass for each bin of hubble type is typically between 0.5 - 0.6dex , or a factor of 3 - 4 . central light concentration is one of the parameters that varies most prominently along the hubble sequence , and was in fact used early to modify hubble s original classification scheme ( @xcite ) . light concentration and other non - parametric estimators such as the asymmetry , clumpiness , second - order moment or gini coefficient , have been extensively used over the years to quantify the morphology of nearby and distant galaxies ( @xcite ) . in particular , @xcite presented a detailed study of the quantitative morphology of the s@xmath0 g galaxies through a suite of commonly used non - parametric estimators . here we present the concentration indices derived from the growth curves measured by our pipeline . among the different definitions existing in the literature , we settled on @xmath77 ( @xcite ) and @xmath78 ( @xcite ) , defined as follows : @xmath79 where @xmath80 is the semi - major axis of the ellipse enclosing @xmath81% of the total luminosity of the galaxy . there is some discrepancy in the literature regarding how the total luminosity is measured when computing concentration indices ; here we simply use the asymptotic magnitudes described before . also , unlike concentration indices derived from srsic fitting , our values are truly non - parametric , in the sense that we do not assume any functional form for the radial profiles or the growth curves . for each galaxy and band we computed both concentration indices . the results are presented in sec . [ sec_cindex ] , where we investigate why galaxies with the same stellar mass can exhibit vastly different morphologies and concentration indices . from the growth curves we have also obtained effective radii ( @xmath82 ) containing half of the total asymptotic luminosity , independently at both irac bands . radial gradients in age , metallicity and internal extinction have a negligible impact at these wavelengths ; therefore , our half - light radii can be safely interpreted as half - mass radii . given that we measured @xmath82 along the major axis of elliptical isophotes , our values do not need to be corrected for inclination , as is the case in studies employing circular apertures . also , since we do not perform any sort of srsic or similar fitting , the values that we provide constitute the true effective radius of each galaxy , and not the effective radius of a simple model fitted to that galaxy . in sect . [ sec_mass_size ] we discuss in detail the stellar mass - size relation of the s@xmath0ggalaxies for different morphologies . the effective radii depend on the level of central concentration , in the sense that highly - concentrated galaxies tend to have small values of @xmath82 compared to the overall extent of those galaxies . therefore , it is also desirable to measure the global sizes and shapes of our galaxies , including their outermost parts . here we follow an rc3-like approach and measure the radius , ellipticity and position angle at 25.5 and 26.5abmagarcsec@xmath25 , both at 3.6 and 4.5@xmath1 ( that is , four sets of measurements for each galaxy ) . we recommend using the 25.5 values at 3.6@xmath1 to characterize the overall extent of the s@xmath0 g galaxies . even though there is always plenty of detectable emission beyond that surface brightness level , ellipse fitting becomes less reliable and more sensitive to convergence issues , object masking and background subtraction . the distribution of isophotal sizes as a function of stellar mass and morphology is presented in sect . [ sec_mass_size ] . regarding the outer isophotal ellipticities , they can be used as a photometric proxy for the disk inclination in the case of spiral galaxies . we have in fact used the p3 ellipticities to deproject our galaxies and recover the intrinsic shapes of rings and bars , for instance ( @xcite ) . note , however , that this is complicated by the fact that ( a ) disks may not be perfectly circular when viewed face - on , and ( b ) the intrinsic vertical thickness of disks and bulges / haloes will distort the outermost isophotes at high inclinations ( see , e.g. , @xcite and references therein for a discussion on both issues ) . the influence of the vertical thickness is particularly obvious in our deep irac images , given that thick disks and spheroids are primarily composed of old stellar populations , which show up prominently at these wavelengths . users should therefore keep this in mind when studying highly - inclined galaxies in our sample ( @xcite ) . we will present inclinations derived from multicomponent 2d fitting of the galaxy images in a forthcoming paper ( salo et al@xmath4 2014 , in prep . ) . figure [ cindex_mass ] shows the distribution of the @xmath78 index at 3.6@xmath1 as a function of the stellar mass for all s@xmath0 g galaxies . figure [ cindex_mass_hist ] displays the same information in histogram form . none of these plots change noticeably when using @xmath77 and/or the 4.5@xmath1 data instead . the concentration index distribution exhibits a pronounced narrow peak centered at @xmath83-@xmath84 , and a much broader one peaking at @xmath85-@xmath86 . the first group corresponds to disk - dominated galaxies ; indeed , the theoretical concentration index for a perfectly exponential profile is 2.8 . the second group comprises bulge - dominated galaxies and ellipticals . the `` l''-shaped distribution of concentration vs. stellar mass shown in fig . [ cindex_mass ] nicely describes how the spatial distribution of old stars within galaxies varies across the hubble sequence . for galaxies later than sc , the concentration index remains roughly constant at the value expected for disk - dominated galaxies , with some scatter but no obvious dependence on the stellar mass . but for galaxies earlier than sc the situation changes , and now the concentration index ranges anywhere from a disk - dominated profile to a highly concentrated one . a similar distribution was found by @xcite in their analysis of the @xmath87-band photometry of galaxies in nearby clusters . they found that galaxies with @xmath88 exhibit a small and constant concentration index , as well as blue optical - ir colors . at @xmath89 galaxies were found to span a large range of concentration indices , with redder colors . to better quantify the structural variety of the s@xmath0 g galaxies , in fig . [ median_profs ] we show the median profiles after grouping the galaxies in bins of stellar mass and concentration . based on the fact that @xmath90 for a pure exponential profile and @xmath91 for a de vaucouleurs one , we use these values to define four bins of mass concentration , as shown in fig . [ median_profs ] : blue and green curves are the median profiles of disk - dominated galaxies , with the blue ones being less concentrated than the green ones . on the other hand , orange and red curves correspond to bulge - dominated galaxies , with the latter being the most concentrated ones . in galaxies more massive than @xmath92 , the observed differences in concentration at a fixed mass are driven by variations in the global radial stellar structure over scales of many kpc , all the way from the center of the galaxies to their outer parts . however , for galaxies less massive than @xmath92 the situation is different : differences in concentration between the blue and green curves ( both disk - dominated ) are mostly due to the presence or absence of central bright features smaller than 1kpc ; the disks at @xmath93 are otherwise very similar in terms of surface density and slope . in the following two subsections we investigate in more detail why galaxies with the same stellar mass and morphological type present such varied radial structure and concentration indices . according to fig . [ cindex_mass ] , the structural transition between low- and high - concentration galaxies occurs at stellar masses between @xmath94 and @xmath95 . previous studies based on optical imaging and spectroscopy have found abrupt transitions in the stellar ages and star formation histories in the same mass interval ( @xcite ) , with stellar populations suddenly becoming older when going from low - mass , disk - dominated galaxies to more massive , highly - concentrated ones . to quantify changes in the stellar populations of the s@xmath0 g galaxies we made use of galex measurements in the far and near ultraviolet . bouquin et al@xmath4 ( 2015 , submitted ) measured fuv and nuv surface photometry and asymptotic magnitudes for the s@xmath0 g galaxies following the same methodology described here for the spitzer images . here we use their measurements to derive extinction - corrected star formation rates ( sfr ) . for each galaxy we first estimated the internal dust extinction via the integrated fuv@xmath96nuv color , following the prescription calibrated by @xcite on nearby galaxies . while this recipe is a good proxy for internal extinction in normal star - forming spirals , it overestimates the true extinction in elliptical galaxies , where a significant fraction of the observed uv reddening is due to the intrinsicially old and red stellar populations . it also overestimates the extinction in sdm - irr galaxies , which tend to have somewhat redder uv colors than other low - mass disks with similar dust content , possibly due to differences in the extinction law , dust geometry , or a bursty star formation activity . therefore , we only relied on the fuv@xmath96nuv color to determine the internal extinction in the disk - dominated galaxies of our sample . following @xcite , for elliptical galaxies we adopted a constant extinction in the fuv of 2mags , and 0.5mags in the case of sdm - irr galaxies . the extinction - corrected fuv luminosities were then converted into sfr via the calibration of @xcite . note that in elliptical galaxies most of the uv light is not associated to recent star formation , so their true sfr is lower than the value resulting from the @xcite prescription . the data - points in fig . [ cindex_mass ] are colored according to their specific sfr ( ssfr , the sfr per unit of stellar mass ) . the transition from low- to high - concentration galaxies in the mass regime between @xmath94 and @xmath95 is accompanied by a decrease of one order of magnitude in the ssfr , which drops from @xmath97 in massive , low - concentration galaxies to @xmath98 in high - concentration galaxies with the same stellar mass . the same decrease in ssfr is found in spectroscopic studies , where the ssfr is estimated from spectral measurements such as the break at 4000 or the h@xmath99 equivalent width ( @xcite ) . this drop in ssfr with increasing concentration does not occur all at once for all morphological types , though . figure [ cindex_mass ] demonstrates that it is only in s0/a galaxies and earlier types that the ssfr drops below @xmath100 in highly - concentrated objects . in these galaxies the high - concentration and red colors result from the dominant contribution of large bulges and stellar haloes to the total light . but in the morphological range between sa and sbc , highly - concentrated galaxies still exhibit a relatively high ssfr of @xmath101 , similar to the ssfr of less concentrated galaxies in the same morphological regime . what is the physical nature of these galaxies with a concentrated stellar mass distribution but high ssfr ? a visual inspection of these objects reveals that they are a mixed bag of galaxies with very different physical properties , which can be nevertheless sorted out into three broad categories ( fig . [ high_cindex_ssfr ] ) : 1 . barred galaxies with prominent nuclear rings , such as ngc 7552 , ngc 4593 or ngc 1365 . the strong bars in these galaxies drive gas inwards , but this gas eventually stalls in a nuclear ring , most likely associated with the inner lindblad resonance ( @xcite ) . this shrinks @xmath102 , the radius of the isophote containing 20% of the total luminosity of the galaxy , thus increasing the concentration index @xmath78 in these galaxies . note that given the intense star formation activity in nuclear rings and their elevated gas and dust content , non - stellar sources can locally dominate the 3.6 and 4.5@xmath1 emission in these rings . nevertheless , the clean stellar mass maps resulting from our pipeline 5 ( @xcite ) still reveal a prominent central concentration of stellar mass in these galaxies , indicating that bar - driven gas inflows play a key role in the secular assembly of ( pseudo-)bulges ( @xcite ) . 2 . interacting systems such as ngc 2782 , ngc 7714 or ngc 5534 . these galaxies often exhibit up - bending radial profiles , with a steep inner disk followed by a flatter outer one , normally accompanied by tidal features in the outermost regions . this decreases @xmath102 and increases @xmath103 , yielding high concentration indices . numerical simulations of minor mergers such as those by @xcite can reproduce this radial structure : the interaction leads to gas inflows that contract the inner radial profile , while at the same time the outwards transfer of angular momentum expands the outer disk . also , the ssfr is temporarily enhanced in these galaxies during the interaction , with prominent star forming knots that are clearly visible in the galex images . galaxies with compact bulges and smooth extended disks , often undisturbed as in ngc 3642 . this again leads to a small @xmath102 and large @xmath103 . while these extended outer disks often have low surface brightness in the infrared , they can be much brighter at uv wavelengths , as happens with ngc 3642 . these so - called extended uv disks are typically found in @xmath10410% - 20% of nearby galaxies ( @xcite ) , but the origin of these galaxy - wide star formation events is still under debate . + + in brief , for stellar masses above @xmath92 , galaxies with a centrally concentrated radial distribution of old stars are either ( a ) : quiescent galaxies with a prominent stellar bulge / halo , or ( b ) star - forming galaxies with central mass concentrations due to a variety of mechanisms such as bar- or merger - driven inflows . figure [ cindex_mass ] shows that galaxies with stellar masses below @xmath92 still exhibit a considerable scatter in the concentration index at any given mass . while all these galaxies have concentration indices broadly consistent with a disk - dominated profile , the actual values can range anywhere from @xmath105 to @xmath106 . moreover , this scatter in @xmath78 does not seem to be correlated with changes in the ssfr , as was the case in the more massive galaxies analyzed before . to ascertain the origin of these structural variations in low mass disks , we visually inspected galaxies with the same stellar mass but extreme values of concentration . figure [ sample_cindex ] displays several representative examples . the top two rows show very low - mass dwarf disks , all of them with stellar masses between @xmath107 . the objects in the first row are characterized by very low concentration indices of @xmath108 ( smaller than for an exponential profile ) , whereas those in the second row all have @xmath109 ( larger than for an exponential ) . this higher concentration is due to central aggregations of stars ( usually resolved in our images ) that are absent in the first group of galaxies . galex images of some of these concentrated dwarfs , like ngc 0059 and ngc 3738 , reveal bluer uv colors in the central parts than in the outskirts , possibly hinting to an outside - in formation for these objects . the third and fourth rows in fig . [ sample_cindex ] correspond to more massive disks , with stellar masses between @xmath110 . galaxies in the third row present @xmath111 , while those in the fourth row have @xmath109 . in this case the higher concentration results from bright bars as in ngc 0600 or ngc 6140 . unbarred galaxies can be also concentrated if they host bright and compact ( pseudo-)bulges , as in ngc 3344 . in contrast , low concentrated disks in this mass range either have weakly - defined bars ( ngc 5147 ) or no bars at all ( ngc 5949 ) . moreover , some galaxies such as ngc 4353 have bright inner disks that lead to central plateaus in their radial profiles , leading to concentration indices lower than those expected for an exponential profile . we also find a wealth of asymmetric galaxies which may have recently undergone gravitational interactions . in some of them , like ngc 2793 , the stellar mass is so dislodged that the radial surface density is almost flat , leading to very low concentration indices . on the other hand , galaxies like ngc 1637 exhibit a much milder asymmetry , and therefore have higher concentration values ( @xcite and references therein ) . in summary , even though galaxies with stellar masses below @xmath92 tend to have low concentration , at any given mass variations in concentration are still present , due to the presence of central star clusters , bars , pseudo - bulges , or lack thereof . the mass - size relation provides very stringent constraints on models of galaxy formation and evolution . from a cosmological point of view , the trend between increasing galaxy size and mass reflects , to first order , the physical connection between the mass of the dark matter halo and its angular momentum . indeed , in a @xmath112cdm cosmology the characteristic size of a disk should scale as @xmath113 ( @xcite ) . in practice , though , the stellar mass - size relation also depends on whether the halo gas conserves its angular momentum as it settles onto the disk , how efficiently that gas is later converted into stars , the impact of agn and stellar feedback , and whether internal and external processes rearrange stars within the galaxy . the mass - size ( or luminosity - size ) relation has been extensively explored both in the local universe ( see , e.g@xmath4 , @xcite ) and at high redshift ( e.g@xmath4 @xcite ) . here we take advantage of the little sensitivity of the irac bands to dust extinction and @xmath114 variations to derive a self - consistent stellar mass - size relation in the local universe . in fig . [ iso_size_mass ] we plot @xmath115 , the semi - major radius at @xmath116abmagarcsec@xmath25 , as a function of the total stellar mass for all s@xmath0 g galaxies . we find an obvious and expected monotonic trend , in the sense that galaxies with higher stellar mass are also larger . however , the slope of this trend is not constant , as galaxies of different hubble types populate different areas in this plot , defining clearly distinct sequences . as in figs . [ morpho_mass ] and [ cindex_mass ] , sc galaxies constitute a well - defined transition type in this diagram , in the sense that earlier - type galaxies are up to a factor of @xmath117 more massive than later - type ones for the same outermost size @xmath115 . isophotal radii represent a robust metric of global galaxy sizes in the nearby universe , but their applicability to distant galaxies is hampered by cosmological surface brightness dimming . in order to facilitate the comparison of the s@xmath0 g mass - size relation with observations at higher redshits , we have also measured the effective radius @xmath82 of our galaxies . in fig . [ reff_mass ] we show the trend between the effective radius at 3.6@xmath1 and the total stellar mass of our galaxies , in bins of morphological type . as explained in sect . [ size_shape ] , @xmath82 was measured along the semi - major axis of elliptical apertures , so no inclination corrections are required . the diagonal lines in this plot represent different values of the average stellar mass surface density inside @xmath82 , that is , @xmath118 . points are again color - coded according to their extinction - corrected ssfr . the main difference with respect to the mass - size relation based on isophotal radii ( fig . [ iso_size_mass ] ) is an increased spread in @xmath119 at fixed size , because a higher concentration index will significantly shrink @xmath82 , but not so much the outer isophotal radius . as a result , early - type galaxies can be up to two orders of magnitude more massive than late - type ones with the same @xmath82 . as a comparison , we have overplotted the mass - size relations for early- and late - type galaxies found by @xcite at @xmath120 , based on data from the galaxy and mass assembly survey ( gama , @xcite ) . these authors relied on different criteria to separate early - type galaxies from late - type ones : srsic index , colors and visual identification ; here we plot the latter . we show their mass - size relation in the @xmath121-band ( where their sample selection was performed ) , but shifting @xmath82 down by 0.076dex to account for the slightly smaller sizes of galaxies at 3.6@xmath1 . this offset was derived from the empirical fits of @xmath82 vs@xmath4 wavelength in @xcite . we find an excellent agreement between the s@xmath0 g mass - size relation and the gama one . massive early - type galaxies exhibit a roughly constant average stellar surface density of @xmath122 , whereas in late - type galaxies the surface density decreases monotonically from @xmath122 in massive disks to @xmath123 in low - mass disks ( @xcite ) . the transition from the early- to the late - type sequence occurs at a specific sfr of @xmath124 . in agreement with our findings when discussing the spread in concentration at fixed stellar mass , here we also note that the early - type mass - size sequence is not entirely populated by red and old systems : bars , mergers and xuv - disks can lead to relatively high ssfr in galaxies that have otherwise high infrared concentration indices and stellar mass surface densities ( see fig . [ high_cindex_ssfr ] ) . a similar mass - size relation was previously found by @xcite , using sdss data of nearby galaxies as well . in fig . [ reff_mass ] we have overplotted their trends ( @xmath125 ) for early- and late - type galaxies , which were classified as such based on their concentration and srsic indices . it is worth noting that @xcite used circular apertures instead of elliptical ones , an this typically decreases the effective radius as @xmath126 , where @xmath127 is the axial ratio . we computed the median offset that our data - points would undergo had we used circular apertures , based on the distribution of axial ratios in each bin of morphological type . these offsets are shown as vertical arrows in fig . [ reff_mass ] . offsets are negligible in early - type , spheroid - dominated galaxies , which tend to be round regardless of the observing angle . however , inclination does play a role in late - type , disk - dominated galaxies , where @xmath128 can decrease by @xmath129dex when circular apertures are used . galaxy size measurements , and effective radii in particular , can be biased by the image depth , the observed wavelength ( due to radial color gradients ) , and the fitting methodology . in order to assess the impact of these factors in determining accurate galaxy sizes , here we perform a direct galaxy - to - galaxy comparison of the s@xmath0 g effective radii with published measurements from other surveys that overlap with ours . we first cross - matched our sample with the 2mass extended source catalog ( xsc ; @xcite ) , which yielded 1687 galaxies in common ( unmatched galaxies typically have @xmath130ab mag ) . the xsc provides surface photometry carried out in essentially the same way as in s@xmath0 g : first , 1d radial profiles are measured using concentric elliptical apertures ; second , total magnitudes are obtained by fitting the outer part of the growth curve ; finally , @xmath82 is derived by integrating on the growth curve until reaching the half - light value . the xsc quotes effective radii along the major axis in @xmath131 , @xmath87 and @xmath132 , but the values differ by less than @xmath133% among the three bands , so for each galaxy we simply averaged the three values . in fig . [ 2mass ] we compare the s@xmath0 g and xsc effective radii . given the similar methodologies and wavelengths of both surveys one would expect an excellent agreement between both sets of measurements . while this is generally the case , the xsc sizes are considerably affected by the shallowness of the 2mass images . to illustrate this , we have color - coded the data - points according to @xmath134 , the average surface brightness inside the half - light elliptical aperture at 3.6@xmath1 , uncorrected for inclination . for bright galaxies with an effective surface brightness @xmath135abmagarcsec@xmath25 the 2mass radii agree with the s@xmath0 g ones with a 1@xmath136 scatter of @xmath137% . however , galaxies with fainter surface brightness levels appear to be several times smaller in 2mass . this is further highlighted in fig . [ 2mass_sample ] , where we compare the s@xmath0 g and 2mass images of low surface brightness galaxies whose effective radius is severely underestimated in 2mass . this effect must be also taken into account at high redshift , where cosmological surface brightness dimming hampers the detection of galaxy outskirts . + we can also compare our size measurements with those obtained from sdss data using parametric fits , in order to understand how many components these fits must have in order to accurately recover the global non - parametric sizes of galaxies . the sdss pipeline performs simple exponential or de vaucouleurs fits to the light profiles of galaxies , but as we will show below these basic fits provide only rough ( and potentially biased ) estimates of galaxy sizes . more sophisticated fits , including srsic profiles and/or multi - component fits , have been widely used by other authors on sdss data ( see , e.g. , @xcite ) . here we compare our effective radii with those measured by @xcite on a sample of sdss nearby galaxies , which includes 124 s@xmath0 g objects . these authors fitted the @xmath121-band image of each galaxy with five different 2d models : a single exponential model , a single de vaucouleurs one , a srsic model , an exponential bulge plus an exponential disk , and a de vaucouleurs bulge plus an exponential disk . we compare their effective radii along the major axis disk models , @xcite do not provide global @xmath128 for the whole galaxy , only for each individual component . we therefore added up the bulge and disk 1d profiles along their respective major axes and computed a global @xmath128 . ] with the s@xmath0 g ones in fig . [ lackner ] . panels ( a ) to ( e ) show the five models described above , and panel ( f ) shows the model deemed by @xcite as the most representative for each particular galaxy . the @xmath121-band radii are systematically larger than the 3.6@xmath1 ones by @xmath138dex . this is quantitatively consistent with the wavelength dependence of @xmath128 found by @xcite , due to color gradients . after accounting for this offset , in term of scatter the `` best model '' option in panel ( f ) provides the best agreement with our non - parametric sizes , with a 1@xmath136 scatter of merely @xmath139dex ( @xmath140% ) . the bulge@xmath141disk models yield a similar accuracy . the srsic fit does a somewhat poorer job , recovering the non - parametric sizes within @xmath142% . for the single exponential fit , the scatter around the 1:1 line is broader and more skewed , as such model is only suitable for almost bulgeless systems . finally , the de vaucouleurs profile completely overestimates the sizes except for some compact objects . the enhanced data products from our different pipelines are available through a dedicated webpage at the nasa / ipac infrared science archive . for each galaxy , users can download the p1 images and weight - maps , the p2 masks , the p3 radial profiles , the p4 galfit decompositions , and the p5 stellar mass maps . irsa also hosts an interactive catalog that contains all the global properties measured in p3 for each galaxy : central coordinates , background level and noise , outer isophotal sizes and shapes , apparent and absolute asymptotic magnitudes , stellar masses , and concentration indices . we have also incorporated into this catalog other parameters mined from external sources . in particular , the catalog contains redshift - independent distances from ned , and a variety of ancillary measurements compiled in hyperleda such as optical sizes , magnitudes and colors , radial velocities , gas content and internal kinematics . this allows users to define detailed subsamples within s@xmath0 g according to different selection criteria ; the resulting list of galaxies can be then fed into the irsa query system to retrieve the images , profiles and other data for the selected objects . s@xmath0 g is the largest and most homogeneous inventory of the stellar mass and structure in the nearby universe . our 3.6@xmath1 and 4.5@xmath1 data peer through dust at the old stellar backbone of more than 2300 nearby galaxies of all morphological types in different environments . in this paper we have described the methodology and algorithms of two of the pipelines in our data flow . pipeline 2 creates masks of foreground stars and background galaxies . pipeline 3 measures the background level and noise , performs surface photometry on the images , and derives global quantities such as asymptotic magnitudes , stellar masses , isophotal sizes and shapes , and concentration indices . we find a structural transition between disk - dominated to bulge - dominated galaxies at stellar masses between @xmath94 and @xmath95 , concurrent with a drop in the specific sfr from @xmath97 to @xmath98 with increasing concentration . however , we find that not all galaxies with high stellar concentration indices are quiescent spheroidal systems . bars and mergers can yield central concentrations of stellar mass while at the same time enhancing the overall sfr of the galaxy . extended uv disks can also increase the sfr in the outer disk of galaxies that look otherwise compact in the irac bands . we have also studied the local stellar - mass size relation at 3.6@xmath1 , which is consistent with previous results , but has the advantage of being little influenced by gradients in dust extinction or stellar age and metallicity . early - type galaxies present an approximately constant stellar mass surface density of @xmath122 . in late - type galaxies the surface density decreases with mass from @xmath122 to @xmath123 . given the large ancillary value of s@xmath0 g , we have made our enhanced data products available to the community through a dedicated webpage at irsa . this data release includes not only the products described in the present paper ( masks , radial profiles and derived global quantities ) , but also other enhanced products from our various pipelines , including science - ready mosaics and weight - maps , 2d image decompositions , and stellar mass maps . the authors are grateful to the entire s@xmath0 g team for their collective effort in this project . we also thank the staff at irsa , and in particular justin howell , for implementing the online access to our data . we acknowledge useful suggestions from an anonymous referee , which helped to improve the scientific content of this paper . we thank rebecca lange for sharing the contour data of the gama mass - size relation . jcmm acknowledges the receipt of an eso fellowship . this work was also co - funded by nasa jpl / spitzer grant rsa 1374189 provided for the s@xmath0gproject . jcmm , ks and tk also acknowledge support from the national radio astronomy observatory , which is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . we also acknowledge financial support from the dagal network from the people programme ( marie curie actions ) of the european union s seventh framework programme fp7/2007 - 2013/ under rea grant agreement number pitn - ga-2011 - 289313 . ea and ab also acknowledge financial support from the cnes ( centre national detudes spatiales - france ) . j.h.k . acknowledges financial support from the spanish mineco under grant number aya2013 - 41243-p . acknowledge financial support under grant number aya2012 - 30717 . this work is based on data acquired with the spitzer space telescope , and makes extensive use of the nasa / ipac extragalactic database ( ned ) , both of which are operated by the jet propulsion laboratory , california institute of technology under a contract with the national aeronautics and space administration ( nasa ) . we have also made use of the hyperleda database ( http://leda.univ-lyon1.fr ) and nasa s astrophysics data system bibliographic services . _ facilities : _ the s@xmath0 g pipeline 1 is an evolved version of the irac pipeline used for the sings legacy program ( @xcite ) . a major change to the code from previous versions is the removal of a step to improve the registration of individual frames using cross - correlation , because the spitzer calibration now uses 2mass to improve the astrometry . also , sky subtraction has been moved into pipeline 3 . below we summarize the steps in the pipeline . the first step is optional and only occurs when the irac observations were taken in high dynamic range ( hdr ) mode . in this mode a short exposure is taken before the main exposure at each location . for the vast majority of the observations in the sample , the main observation is 30 seconds and the hdr exposure is 1.2 seconds . for new galaxies that were only observed during the warm mission we did not use hdr mode since very few have bright nuclei . if the basic calibrated data ( bcd ) images have an hdr counterpart and the flux in the pixel is above @xmath143mjysr@xmath11 , the value from the normal bcd is replaced by the hdr one . in addition , we do a rough cosmic ray removal by replacing the value of the pixels in the normal bcd image with the value in the hdr pixel when the difference is above 3mjysr@xmath11 . before we can compare the sky level in overlapping bcd images , we must first correct each bcd for optical distortion and rotate the images to have the same orientation . this is required to get pixels to all have the same projeted size and orientation on the sky . to do this we use the drizzle routine and create a new optically corrected bcd image for each bcd image in the mosaic . once we have all the images with constant size on the sky we can find the matching regions of overlap . for all pairs of overlapping bcd images where at least a minimum number of pixels ( nominally 20,000 ) overlap between the two images we find the background offset . to determine the background offset we find the difference in the 20th percentile pixel flux in both images . using the 20th percentile has several advantages over a mean or median . our goal is to find the change in the background level rather than any changes when pixels receive flux from resolved sources . sub - pixel changes in the telescope position will change the flux level of pixels observing a point source substantially with the irac undersampled optics . these same sub - pixel changes do not affect the observations of the unresolved background . also , some detector artifacts such as mux bleed and column pull down can affect large areas leading to biases in even the median , while using the 20th percentile value minimizes the bias . once we have the background offset between each pair of overlapping bcd images , we used the method of @xcite to find the background offset for each image that minimizes the residuals . this method performs a least square minimization of the residuals using a single offset that is applied to each image . due to their small size , many sample galaxies were not observed in irac mapping mode but instead were observed in dither mode . in this case two ( or even three ) non - overlapping regions of the sky are observed and there is a need of additional information to obtain a solution . therefore , when the regions of the sky do not overlap we add in an offset between a bcd image in each region of the sky . this will force the sky level to be similar between the non - overlapping regions of the sky . the set of measured offsets of overlapping images does not provide a zero point for the solution , which requires having an external paramenter fixed to set the zero point for the relative offsets . to minimize the effect of applying the offsets on the background level , we place the additional constraint that the average offset should have a value of 0 . once we know the correction that needs to be applied , we create a corrected version of each input bcd image . the final mosaics are created using the standard stdas pyraf task drizzle . the resuling images have a plate scale of 0.75 per pixel and are oriented to have north to the top . we have to make a correction to the output of the drizzling task to recover the correct photometry . drizzle assumes the pixels have units of flux instead of surface brightness . we apply a correction to account for the smaller pixel area in the final mosaic . drizzle also returns a weight map that shows the number of seconds of integration time for each final pixel . this image is also included in the s@xmath0 g archive and is used by subsequent s@xmath0 g pipelines . in the final step we update the fits header to include a list of the astronomical observation requests ( aors ) that contributed to the final mosaic and set all the pixels outside of the field of view to have the value of nan . , g. , de vaucouleurs , a. , corwin , jr . , h. g. , et al . 1991 , third reference catalogue of bright galaxies . volume i : explanations and references . volume ii : data for galaxies between 0@xmath144 and 12@xmath144 . volume iii : data for galaxies between 12@xmath144 and 24@xmath144 . , m. w. , & gruendl , r. a. 1995 , in astronomical society of the pacific conference series , vol . 77 , astronomical data analysis software and systems iv , ed . r. a. shaw , h. e. payne , & j. j. e. hayes , 335

aiming at deriving a statistically well - justified galactic center distance , , and reducing any occurrence of publication bias , we compiled the most comprehensive and most complete database of galactic center distances available to date , containing 273 new or revised estimates published since records began in october 1918 until june 2016 . we separate our compilation into direct and indirect distance measurements . the latter include a large body of estimates that rely on centroid determinations for a range of tracer populations as well as measurements based on kinematic observations of objects at the solar circle , combined with a mass and/or rotational model of the milky way . careful assessment of the galactic center distances resulting from orbital modeling and statistical parallax measurements in the galactic nucleus yields our final galactic center distance recommendation of kpc . the centroid - based distances are in good agreement with this recommendation . neither the direct measurements nor the post-1990 centroid - based distance determinations suggest that publication bias may be important . the kinematics - based distance estimates are affected by significantly larger uncertainties , but they can be used to constrain the galaxy s rotation velocity at the solar galactocentric distance , . our results imply that the international astronomical union - recommended galactic center distance ( kpc ) needs a downward adjustment , while its recommendation ( km s ) requires a substantial upward revision .

1610.02457

the distance from the sun to the galactic center , @xmath0 , provides the basic calibration for a wide range of methods used for distance determination , both on galactic and extragalactic scales . calculations of many physical parameters , including of the distances , masses , and luminosities of galactic objects , as well as the galaxy s integrated mass and luminosity , depend directly on @xmath0 . in fact , most luminosity and a large number of mass estimators scale as distance squared , while masses based on total densities or orbital modeling scale as distance cubed . this dependence could involve adoption of a galactic mass and/or rotation model , in which case we also need to know the sun s circular velocity accurately . as @xmath0 estimates are refined , so are the estimated distances , masses , and luminosities of numerous galactic and extragalactic objects , as well as our best estimates of the rate of galactic rotation and the size of the milky way . in addition , a highly accurate direct galactic center distance determination would immediately allow reliable recalibration of the zero points of numerous secondary distance calibrators , including of cepheids , rr lyrae , and mira variable stars , which would consequently reinforce the validity of the extragalactic distance scale ( e.g. , olling 2007 ) . in turn , this would enable better estimates of globular cluster ages , the hubble constant , as well as a lower limit to the age of the universe ( monelli et al . 2015 ) , and place tighter constraints on a range of cosmological scenarios ( e.g. , reid et al . 2009 ) . it is no surprise , therefore , that determinations of the galactic center distance have been the subject of many studies ever since the first attempt by harlow shapley in 19141918 . however , the importance of determining an accurate @xmath0 value , combined with the large number of studies undertaken to achieve this goal , have led to speculations that some degree of publication bias ( also known as ` observation bias ' or a ` bandwagon effect ' ) may have affected subsequent galactic center distance determinations ( e.g. , reid 1989 , 1993 ; nikiforov 2004 ; foster & cooper 2010 ; malkin 2013a , b ; francis & anderson 2014 ) . in de grijs et al . ( 2014 , henceforth paper i ) , de grijs & bono ( 2014 ; paper ii ) , and de grijs & bono ( 2015 ; paper iii ) , we embarked on large - scale data mining of the nasa / astrophysics data system ( ads ) to explore whether distance determinations to , respectively the large magellanic cloud ( lmc ) , the m31 group , and the small magellanic cloud had been polluted by such bandwagon effects . in this paper , we extend our series of papers by adding a similar analysis of distance estimates to the galactic center . our ultimate aim is to provide a self - consistent distance framework that can serve as a benchmark for the structure of the local group of galaxies ( see , e.g. , table 4 in paper ii , combined with the recommended distance to the small magellanic cloud from paper iii ) . to achieve our aim , in section 2 we discuss our approach to mining the nasa / ads database , eventually resulting in the most complete and most comprehensive database of galactic center distances published , starting from shapley s ( 1918 ) distance estimate , published in october 1918 , until early june 2016 . in section 3 , we review three different types of galactic center distance indicators , including direct distance measurements as well as centroid- and kinematics - based galactic center distance determinations . this is followed in section 4 by a detailed discussion of the validity of and the uncertainties affecting the ` best ' measurements , which we also place in the context of the most recent progress in the field . finally , section 5 summarizes and concludes the paper . the distance to the galactic center has long been a subject of intense scrutiny . since the mid-1970s , it has become standard practice in meta - analyses of the galactic center distance to publish compilations of previously published values ( e.g. , harris 1976 ; de vaucouleurs 1983 ; kerr & lynden - bell 1986 ; reid 1993 ; nikiforov 2004 ; perryman 2009 , his table 9.1 ; malkin 2013a , b ) . although the latest such compilation dates only from 2013 , upon close examination it transpired that malkin s ( 2013a , b ) compilation of 53 galactic center distance estimates published between 1992 and 2011 is incomplete : our own perusal of the literature from this same period revealed an additional 32 papers with newly derived or updated galactic center distance estimates . since gaps in the data may mask or , alternatively , artificially suggest the presence of publication bias ( for a discussion , see paper i ) , we decided to compile our own database of galactic center distances that is as complete as possible until the present time . we followed a similar two - pronged approach as employed in papers i , ii , and iii . first , we scanned all 23,516 papers until and including february 2016 tagged with the ` galactic center ' keyword in the nasa / ads for potential new or rederived galactic center distance determinations . at the same time , we carefully followed the reference trail : where a paper referred to the provenance of the galactic center distance adopted by its authors , we made a note of the original reference and double checked that the latter was actually included in our final database . this approach led to a final database containing a total of 273 galactic center distance measurements since shapley s ( 1918 ) first attempt at determining the centroid of the galaxtic globular cluster distribution known at that time . as for papers i iii , the full database is available at http://astro-expat.info/data/pubbias.html , where we provide our compilation of galactic center distances both as a function of publication date and by tracer , supported by full bibliographic information . we compiled the extinction - corrected distance moduli , as well as their statistical and systematic uncertainties , if available . only 25 authors published their systematic uncertainties separately ; in addition , five papers specified that their published error bars include the systematic uncertainties . for the remaining galactic center distance measurements , the uncertainties refer to the statistical errors only . as in papers i iii , instead of combining individual values based on different assumptions or input parameters , we have included all ( final ) galactic center distance measurements published in a given paper . the range spanned by such alternative values provides a valuable estimate of the systematic uncertainty inherent to the distance determined , although we note that these values are often highly correlated . figure [ fig1 ] shows an overview of the data as published in the original articles , i.e. , without having homogenized or recalibrated the various measurements . we show both the full data set in fig . [ fig1]a and subsets selected according to their provenance , i.e. , direct measurements ( statistical parallaxes , galactic center orbital modeling , galactic center cepheids ) , centroid determinations ( @xmath6 scuti , rr lyrae , cepheids , and mira variables , red clump stars ) , and determinations involving adoption of a kinematic disk model ( cepheids , rr lyrae , miras , open clusters ) . we have also included the galactic center distance values recommended by the international astronomical union ( iau ) . at its xii@xmath7 general assembly held in hamburg ( germany ) in 1964 , the iau released its first formal recommendation for the value of @xmath8 kpc . this was subsequently amended to @xmath9 kpc ( kerr & lynden - bell 1986 ) , a value that has been in force since 1986 . nevertheless , even a casual comparison of the galactic center distances in our database ( see fig . [ fig1 ] ) with the current iau recommendation reveals that the latter value is well above the mean of most current and recent determinations of @xmath0 . as such , the community has called for a further adjustment of the iau recommendation , but this would need to be done in tandem with adjustments of the galactic rotation rate at the solar circle as well as of the oort constants , which all depend on one another . our final database is significantly more comprehensive than any previously published compilation of galactic center distance estimates . we will demonstrate this by comparison of our data set with those of three benchmark papers , kerr & lynden - bell ( 1986 ) , reid et al . ( 1993 ) , and malkin ( 2013a ) . the latter paper was selected for this comparison since it is the most recent large compilation . as already addressed earlier in this section , we added 32 papers ( containing 42 new galactic center distance estimates ) to malkin s ( 2013a ) compilation of 53 articles published between 1992 may and 2011 july . the review by reid ( 1993 ) has long been used as the ` gold standard ' in the field of galactic center distance determinations . it includes 37 @xmath0 values published between 1972 june and 1992 april , taken from 38 articles . our compilation , by contrast , includes 20 additional galactic center distance estimate , found in 17 papers not included by reid ( 1993 ) . finally , the current iau recommendation for the value of @xmath0 is based on the compilation of 26 estimates by kerr & lynden - bell ( 1986 ) . again , our own perusal of the literature published in the period spanning from 1973 august to 1986 may revealed nine additional galactic center distance determinations . based on observations taken since 1914 , shapley ( 1918 ) used the light curves and , hence , the period luminosity relation of cepheid variables in 69 galactic globular clusters to visualize the spatial distribution of the galactic globular cluster system . he eventually extended his analysis to include all 93 galactic globular clusters known at that time . he subsequently determined the centroid of the three - dimensional globular cluster distribution to determine the first ever distance to the galactic center , @xmath10 kpc . armed with our current understanding of the structure of the milky way , it is clear that this distance estimate was considerably too large . in fact , the period luminosity relation he applied was @xmath111 mag too faint , while he used population ii cepheids ( w virginis stars)combined with the 25 brightest stars in his seven sample clusters ( since population ii cepheids were only known in a limited number of clusters)instead of the population i cepheids he thought he had observed . population ii cepheids are some 2 mag fainter than their population i counterparts , while it appears that shapley ( 1918 ) may also have made calibration errors regarding the absolute magnitudes of the brightest stars and the associated reddening corrections ( sandage 2004 , pp . 301302 ) , thus offsetting the error introduced by his use of an incorrect period luminosity relation to some extent . as a consequence , shapley s ( 1918 ) distance scale was off by @xmath111 mag , corresponding to a factor of @xmath111.6 . note that the implicatiom of shapley s ( 1918 ) results based on his approach of using both population ii cepheids and his clusters brightest stars may suggest that the difference between classical and population ii cepheids only applies to a limited number of clusters . while shapley s ( 1918 ) galactic center distance determination now seems a mere historical curiosity , the fact remains that this was the first viable ` centroid ' approach to estimate the galactic center s distance . nevertheless , it is instructive to compare shapley s ( 1918 ) result with the most recent @xmath0 determination based on the globular cluster centroid . nikiforov & smirnova ( 2013 ) found that the globular clusters composing the galaxy s metal - rich and metal - poor subsystems separately form bar - like structures that closely resemble the galactic bar as a whole . they conclude that only non - axisymmetric models can provide sufficiently strong constraints on @xmath0 . in addition , these authors found evidence for an extinction component associated with the galactic bar which affects the observational incompleteness of globular clusterss on the far side of the galactic center , a selection effect that must be considered seriously in determining @xmath0 . in this section , we will focus on the ` modern ' period since 1990 to distinguish among three different types of galactic center distance determinations in order of decreasing accuracy , i.e. , direct ( geometric ) methods , centroid - based methods , and distance determinations based on observations of tracer objects at some distance from the galactic center , combined with a kinematic model of the galactic disk . we will also address the uncertainties inherent to each of the methods applied . the current - best , ` direct ' ( geometric ) estimates of @xmath0 are based on astrometric orbit determinations of the so - called s stars in the galactic center region . two competing groups are leading efforts in this field , i.e. , genzel et al . versus ghez and collaborators : see table [ gcdirect.tab ] . figure [ fig1]b shows the run of these determinations as a function of publication date ( black points ) . the latest result from the genzel group is @xmath12 kpc ( gillessen et al . 2013 ) , while the ghez group published @xmath13 kpc ( yelda et al . 2011 ; but see morris et al . 2012 for a rather surprising downgrade ) . we will return to a discussion of these direct measurements in section [ discussion.sec ] . the next best direct galactic center estimates come from statistical parallax measurements of the nuclear star cluster . these measurements are also included in table [ gcdirect.tab ] . taking the straight mean and standard deviation of all such measurements gives @xmath14 kpc . however , this approach does not do full justice to the published results . systematic uncertainties related to the dynamical cluster models affect the resulting distance determinations . assuming uniform , isotropic , and fully phase - mixed systems ( e.g. , eisenhauer et al . 2003 ; do et al . 2013 ) seems to yield systematically smaller @xmath0 values than adopting anisotropic , spherical jeans models ( e.g. , do et al . the current - best results are based on joined - up analyses of both the nuclear star cluster s stellar velocity dispersions and s - star orbital modeling . the latter lead to @xmath15 kpc ( do et al . 2013 , combined with ghez et al . 2008 ) and @xmath16 kpc ( chatzopoulos et al . 2015 , combined with gillessen et al . 2009a ) . 0.5 mm @cccll@ publ . date & @xmath0 & & & + ( mm / yyyy ) & ( kpc ) + + 11/2003 & @xmath17 & & eisenhauer et al . ( 2003 ) & s2 only , 19922001 ; systematic uncertainty 0.16 kpc + 07/2005 & @xmath18 & & eisenhauer et al . ( 2005 ) & s2 only , 19922004 + 07/2005 & @xmath19 & & eisenhauer et al . ( 2005 ) & s2 only , 19922004 , excl . data from 2002 + 03/2006 & @xmath20 & & zucker et al . ( 2006 ) & s2 only , 19922004 ; corrected for relativistic effects + 07/2007 & @xmath21 & & olling ( 2007 ; ghez , priv . commun . ) & bias - free ` orbital parallax method ' ( armstrong et al . 1992 ) + 07/2008 & @xmath22 & & ghez et al . ( 2008 ) & s2 only , 19952007 + 12/2008 & @xmath23 & & ghez et al . ( 2008 ) & s2 only , 19952007 ; black hole freely moving + 12/2008 & @xmath24 & & ghez et al . ( 2008 ) & s2 only , 19952007 ; black hole at rest + 02/2009 & @xmath25 & & gillessen et al . ( 2009b ) & s stars , 19922008 ; incl . systematic uncertainties + 02/2009 & @xmath26 & & gillessen et al . ( 2009b ) & s stars excl . s2 , 19922008 ; incl . errors + 12/2009 & @xmath27 & & gillessen et al . ( 2009b ) & s stars , 19922008 , combined eso / keck data sets ; syst . 0.29 kpc + 12/2009 & @xmath28 & & gillessen et al . ( 2009b ) & s2 only , 19922008 , combined eso / keck data sets ; syst . 0.52 kpc + 05/2011 & @xmath29 & & yelda et al . ( 2011 ) & s2 only , 19952007 ; new distortion corrections + 08/2012 & @xmath30 & & morris et al . ( 2012 ) & s stars , 19952011 + 02/2013 & @xmath31 & & gillessen et al . ( 2013 ) & 5 s stars , 19922012 + + 12 1995 & @xmath32 & & huterer et al . ( 1995 ) & 50 m giants + 09 2000 & @xmath33 & & genzel et al . ( 2000 ) & 104 stars with proper motions ; 71 stars with @xmath34 velocities + 09 2000 & @xmath35 & & genzel et al . ( 2000 ) & corrected for the effects of a central point mass + 11 2003 & @xmath36 & & eisenhauer et al . ( 2003 ) & uniform , isotropic , phase - mixed system + 12 2008 & @xmath37 & & trippe et al . ( 2008 ) & 664 late - type giants + 05 2011 & @xmath38 & & trippe et al . ( 2011 ) & velocity dispersion ; systematic uncertainty 0.13 kpc + 12 2013 & @xmath39 & & do et al . ( 2013 ) & isotropic velocity distribution + 12 2013 & @xmath40 & & do et al . ( 2013 ) & anisotropic spherical jeans models + 12 2013 & @xmath41 & & do et al . ( 2013 ) & combined with ghez et al . ( 2008 ) + 02 2015 & @xmath42 & & chatzopoulos et al . ( 2015 ) & systematic uncertainty 0.1 kpc + 02 2015 & @xmath43 & & chatzopoulos et al . ( 2015 ) & combined with gillessen et al . ( 2009a ) + matsunaga et al . ( 2011 ) used the near - infrared period luminosity relation of van leeuwen et al . ( 2007 ) , adopting solar metallicity , calibrated on the basis of cepheids with parallax - based distances , to determine the distance modulus to three classical cepheids in the galactic nucleus . these three objects were found within 40 pc ( projected distance ) of the central black hole , leading the authors to conclude that @xmath44 kpc ( see also bono et al . 2013 ; matsunaga 2013 ) . the latter value was scaled based on adoption of @xmath45 mag for the distance modulus to the large magellanic cloud . although this is not strictly speaking a ` direct ' galactic center distance measurement , we include it here in a separate subsection rather than in the section where we discuss the cepheid centroid given the close proximity of these nuclear cepheids to the actual galactic center . the direct measurements discussed in section [ direct.sec ] and the nuclear cepheid distance of matsunaga et al . ( 2011 ) are indeed fully consistent with one another , with the cepheid distance somewhat on the short side . ) suggests that any differences are negligible . ] if this small distance differential between the nuclear cepheids and the actual galactic center is real , this may imply that the former objects are seen in projection onto the galactic center . in fact , matsunaga et al . ( 2011 ) do not claim precise coincidence with the galactic center , but merely that their projected positions are consistent with their presence in the thin disk - like structure of the nuclear bulge . these authors point out that their mean distance estimate suffers from an additional , systematic uncertainty of approximately 0.3 kpc , which is predominantly driven by the uncertainties inherent to the extinction law applied ( i.e. , the total - to - selective extinction ratio ) and by scatter among the prevailing period luminosity relations , both corresponding to uncertainties of 0.09 mag in distance modulus . matsunaga et al . ( 2015 ) added a fourth nuclear cepheid to their sample and obtained their kinematics . they concluded that the velocities of these cepheids suggest that the stars orbit within the nuclear stellar disk , i.e. , within @xmath46 pc of the galactic center . this is indeed consistent with their earlier suggestion . finally , we note that some authors have suggested that the distances to different galactic center tracers affected by significant extinction should be adjusted upward , for instance by adopting a non - standard or variable extinction law toward the galactic center ( e.g. , collinge et al . 2006 ; vanhollebeke et al . 2009 ; pietrukowicz et al . 2012 ; nataf et al . 2013 ; nataf et al . 2016 ; but see francis & anderson 2014 ) . in addition , the effects of population differences ( ages and metallicities ) may cause certain tracer objects such as red clump stars to be redder in the galactic bulge than their solar neighborhood equivalents ( e.g. , girardi & salaris 2001 ) , which thus implies that significant systematic uncertainties likely remain in the zero - point calibrations of many ` standard ' secondary distance tracers . since the pioneering effort by shapley ( 1918 ) , many authors have attempted to determine @xmath0 based on the centroids of the distributions of a variety of tracers , including globular clusters , cepheid , rr lyrae , and mira variables , red clumps stars , and even delta scuti stars and planetary nebulae . while fig . [ fig1]c e showed the ` raw , ' uncorrected centroid data collected in our database , fig . [ fig2 ] includes the most recent ( post-1990 ) galactic center measurements based on different centroid tracers as a function of publication date . we have attempted to homogenize the distance calibration where appropriate : the original data points as listed in the online database are shown as black open circles ; the corrected measurements are shown as red bullets with error bars . the calibrations applied are included in the individual panels . we used a common horizontal - branch magnitude of @xmath47 mag for globular clusters , @xmath48 mag for rr lyrae stars , and @xmath49- and @xmath50-band red clump absolute magnitudes from _ hipparcos _ ( stanek & garnavich 1998 ; see also laney et al . 2012 ) for the red clump stars , i.e. , @xmath51 mag and @xmath52 mag , with @xmath53 mag ( grocholski & sarajedini 2002 ) . in addition , we adopted a homogenized distance scale based on @xmath54 mag , thus also ensuring internal consistency with the recommended distance benchmarks derived in papers i iii . 0.5 mm @cclll@ publ . date & @xmath0 & & & + ( mm / yyyy ) & ( kpc ) + + 08/1993 & @xmath55 & maciel ( 1993 ) & @xmath56 mag & 46 gcs , @xmath57 } > -1.2 $ ] dex + 09/1994 & @xmath58 & rastorguev et al . ( 1994 ) & @xmath59 mag & @xmath60 + 00/1998 & @xmath61 & surdin & feoktistov ( unpubl . ) & @xmath60 & quoted in surdin ( 1999 ) as ` submitted ' + 00/1999 & @xmath62 & surdin ( 1999 ) & @xmath63 } + 1.00 $ ] mag & 126 gcs + 03/2006 & @xmath64 & bonatto et al . ( 2009 ) & calibration from bica et al . ( 2006 ) & @xmath60 + 04/2006 & @xmath65 & bica et al . ( 2006 ) & @xmath66 } + 0.84 $ ] mag & 116 gcs + 10/2013 & @xmath67 & nikiforov & smirnova ( 2013 ) & @xmath66 } + 0.84 $ ] mag & all methods + 10/2013 & @xmath68 & nikiforov & smirnova ( 2013 ) & @xmath66 } + 0.84 $ ] mag & all spatial methods + 10/2013 & @xmath69 & nikiforov & smirnova ( 2013 ) & @xmath66 } + 0.84 $ ] mag & shapley s method and related spatial methods + 06/2014 & @xmath21 & francis & anderson ( 2014 ) & @xmath70}$ ] & 154 gcs , syst . kpc + & & & @xmath71}^2 $ ] mag + + 04/2008 & @xmath72 & groenewegen et al . ( 2008 ) & @xmath73 mag & cepheids and rr lyrae ; syst . 0.26 kpc + 04/2008 & @xmath74 & groenewegen et al . ( 2008 ) & @xmath73 mag & type ii cepheids + 06/2008 & @xmath75 & feast et al . ( 2008 ) & @xmath76 mag & type ii cepheids + 06/2008 & @xmath77 & feast et al . ( 2008 ) & @xmath76 mag & type ii cepheids , incl . @xmath78 pav + 09/2009 & @xmath79 & majaess et al . ( 2009 ) & @xmath73 mag & type ii cepheids + 09/2009 & @xmath80 & majaess et al . ( 2009 ) & @xmath73 mag & type ii cepheids + bulge model + 02/2013 & @xmath72 & matsunaga ( 2013 ) & @xmath73 mag & 100 miras , cepheids , rc stars ; syst . 0.26 kpc + + 09/1991 & @xmath81 & walker & terndrup ( 1991 ) & @xmath82 mag & @xmath83 , baade s window ; @xmath84 + 09/1991 & @xmath85 & walker & terndrup ( 1991 ) & @xmath82 mag & @xmath83 , baade s window ; @xmath86 , + & & & & carney et al . ( 1992 ) distance scale + 10/1995 & @xmath87 & carney et al . ( 1995 ) & @xmath82 mag & @xmath88 , baade s window + 10/1995 & @xmath89 & carney et al . ( 1995 ) & @xmath73 mag & @xmath88 , baade s window ( but see feast 1997 ) + 11/1996 & @xmath90 & layden et al . ( 1996 ) & @xmath91 mag & baade s window + & & & @xmath92 } + 0.95 $ ] mag + 11/1996 & @xmath93 & layden et al . ( 1996 ) & @xmath91 mag & baade s window + & & & @xmath94 mag + 11/1996 & @xmath95 & layden et al . ( 1996 ) & combined approach & ` best ' value + 01/1997 & @xmath96 & feast ( 1997 ) & @xmath97 mag & @xmath98 + & & & @xmath99 } + 1.13 $ ] mag + 04/1998 & @xmath100 & udalski ( 1998 ) & @xmath101 mag & @xmath102 + 11/2006 & @xmath103 & collinge et al . ( 2006 ) & @xmath104 mag & @xmath105 , baade s window + 11/2006 & @xmath106 & collinge et al . ( 2006 ) & @xmath107 mag & @xmath105 , baade s window + 11/2006 & @xmath108 & collinge et al . ( 2006 ) & combined approach & @xmath105 , baade s window + 02/2008 & @xmath109 & kunder et al . ( 2008 ) & @xmath104 mag & @xmath60 + 02/2008 & @xmath95 & kunder et al . ( 2008 ) & @xmath107 mag & @xmath60 + 04/2008 & @xmath72 & groenewegen et al . ( 2008 ) & @xmath73 mag & cepheids+rr lyrae + 04/2008 & @xmath110 & groenewegen et al . ( 2008 ) & @xmath73 mag & @xmath111 , sollima et al . ( 2006 ) plc calibration + 04/2008 & @xmath112 & groenewegen et al . ( 2008 ) & @xmath73 mag & @xmath111 , sollima et al . ( 2006 ) plr calibration + 12/2008 & @xmath113 & kunder & chaboyer ( 2008 ) & @xmath114}$ ] & @xmath115 , bulge + & & & @xmath116}^2 $ ] mag + 03/2010 & @xmath117 & majaess ( 2010 ) & @xmath73 mag & @xmath60 + 05/2012 & @xmath118 & pietrukowicz et al . ( 2012 ) & @xmath119 & @xmath120 } - 1.765 $ ] dex + & & & @xmath121 mag + 10/2013 & @xmath122 & dkny et al . ( 2013 ) & catelan et al . ( 2004 ) calibration & systematic uncertainty 0.14 kpc + 10/2015 & @xmath123 & pietrukowicz et al . ( 2014 ) & catelan et al . ( 2004 ) calibration & systematic uncertainty 0.40 kpc + 0.5 mm @cclll@ publ . date & @xmath0 & & & + ( mm / yyyy ) & ( kpc ) + + 06/1991 & @xmath124 & whitelock et al . ( 1991 ) & @xmath125 mag & @xmath60 + 06/1991 & @xmath126 & whitelock et al . ( 1991 ) & @xmath125 mag & @xmath60 + 00/1992 & @xmath127 & whitelock ( 1992 ) & @xmath60 & @xmath60 + 03/1995 & @xmath128 & glass et al . ( 1995 ) & @xmath129 mag & @xmath60 + 03/1995 & @xmath130 & glass et al . ( 1995 ) & @xmath129 mag & if located in the galactic bar + 11/2005 & @xmath131 & groenewegen & blommaert ( 2005 ) & @xmath73 mag & @xmath132 + 11/2005 & @xmath133 & groenewegen & blommaert ( 2005 ) & @xmath134 mag & @xmath132 ; feast ( 2004 ) calibration + 09/2009 & @xmath135 & matsunaga et al . ( 2009 ) & @xmath136 mag & systematic uncertainty 0.42 kpc + 11/2009 & @xmath135 & matsunaga et al . ( 2009 ) & @xmath136 mag & systematic uncertainty 0.42 kpc + 02/2013 & @xmath137 & matsunaga ( 2013 ) & @xmath136 mag & cepheids , miras , rc stars ; syst . 0.26 kpc + 01/2016 & @xmath138 & catchpole et al . ( 2016 ) & @xmath139 mag & @xmath140 } < 2.6 $ ] + 01/2016 & @xmath141 & catchpole et al . ( 2016 ) & @xmath139 mag & @xmath142 } < 2.7 $ ] + 01/2016 & @xmath138 & catchpole et al . ( 2016 ) & @xmath139 mag & combined approach + + 02/1998 & @xmath143 & paczyski & stanek ( 1998 ) & @xmath144 mag & @xmath60 + 04/1998 & @xmath145 & udalski ( 1998 ) & @xmath146}^{\rm rc}$ ] & @xmath57}^{\rm rc } = + 0.2 $ ] dex ; 73 rrab lyrae , rc stars + & & & @xmath147 mag + 08/1998 & @xmath148 & stanek & garnavich ( 1998 ) & @xmath149 mag & @xmath60 + 06/2000 & @xmath150 & stanek et al . ( 2000 ) & @xmath151 mag & @xmath60 + 08/2000 & @xmath152 & alves ( 2000 ) & @xmath153 mag & @xmath60 + 02/2001 & @xmath154 & gould et al . ( 2001 ) & @xmath155 mag & updated zero points + 03/2005 & @xmath156 & nishiyama et al . ( 2005 ) & @xmath155 mag & @xmath60 + 04/2005 & @xmath157 & babusiaux & gilmore ( 2005 ) & @xmath158 mag & @xmath60 + 04/2005 & @xmath159 & babusiaux & gilmore ( 2005 ) & @xmath155 mag & @xmath60 + 08/2006 & @xmath160 & nishiyama et al . ( 2006 ) & @xmath161 mag & @xmath60 + 08/2006 & @xmath162 & nishiyama et al . ( 2006 ) & @xmath158 mag & recalibration of alves ( 2000 ) + 07/2007 & @xmath163 & rattenbury et al . ( 2007 ) & @xmath164 mag & upper limit + 12/2010 & @xmath165 & mcwilliam & zoccali ( 2010 ) & @xmath166 mag & @xmath60 + 08/2011 & @xmath167 & fritz et al . ( 2011 ) & @xmath168 mag & @xmath169 + 08/2011 & @xmath170 & fritz et al . ( 2011 ) & @xmath168 mag & @xmath171 + 08/2011 & @xmath172 & fritz et al . ( 2011 ) & @xmath168 mag & @xmath173 + 08/2011 & @xmath174 & fritz et al . ( 2011 ) & @xmath168 mag & mean + 02/2013 & @xmath175 & matsunaga ( 2013 ) & @xmath161 mag & cepheids , miras , rc stars + 06/2013 & @xmath176 & nataf et al . ( 2013 ) & @xmath177 mag & @xmath60 + 09/2013 & @xmath178 & cao et al . ( 2013 ) & @xmath177 mag & @xmath60 + 11/2013 & @xmath179 & wegg & gerhard ( 2013 ) & @xmath180 mag & @xmath60 + 03/2014 & @xmath181 & gardner et al . ( 2014 ) & @xmath166 mag & @xmath60 + 06/2014 & @xmath182 & francis & anderson ( 2014 ) & @xmath183 mag & _ hipparcos _ recalibration + & & & @xmath184 mag + 06/2014 & @xmath185 & francis & anderson ( 2014 ) & @xmath183 mag & recalibration of alves ( 2000 ) + & & & @xmath184 mag + 06/2014 & @xmath186 & francis & anderson ( 2014 ) & @xmath183 mag & recalibration of babusiaux & gilmore ( 2005 ) + & & & @xmath184 mag + 06/2014 & @xmath187 & francis & anderson ( 2014 ) & @xmath183 mag & recalibration of nishiyama et al . ( 2006 ) + & & & @xmath184 mag + @xmath188 abbreviations used : gcs , globular clusters ; @xmath189 , number of objects ; @xmath190 , period ; plc , period luminosity color relation ; plr , period luminosity relation ; rc , red clump ; syst . unc . , systematic uncertainty . the left - hand side of table [ tracers.tab ] provides an overview of the mean galactic center distances and their standard deviations based on published centroid determinations since 1990 . it is readily apparent that the rr lyrae and mira centroids yield systematically larger distances than the other tracers . we checked whether this might be a signature of publication bias by recalculating the rr lyrae statistics for galactic center distances published since 2000 . their ( mean , @xmath191 ) combinations are ( 8.21 , 0.31 ) kpc and ( 8.31 , 0.39 ) kpc before and after correction , respectively . determination of @xmath0 based on the centroid approach is significantly more uncertain than using direct geometry . uncertainties hindering accurate galactic center distance estimation include ambiguities associated with the prevailing extinction law , a preference for smaller values of @xmath0 because of sampling biases ( where extinction causes tracers on the near side of the bulge to have a greater chance of inclusion than their counterparts on the far side ) , and difficulties in converting the mean distance to one s tracer sample into a value of @xmath0 . 0.5 mm [ cols="<,^,^,^,^,^,^,<,^,^,^,^,^ " , ] a large number of galactic center distance determinations base their estimates on the distances to a variety of tracer populations at or near the solar circle , combined with a kinematic disk ( mass ) model of the milky way galaxy . suitable tracers include cepheid and rr lyrae variable stars , open clusters , and maser sources , as well as the distributions of hi and co. figure [ fig3 ] shows the relevant measurements published since 1990 . as before , we have attempted to homogenize these estimates where possible , particularly the cepheid- , rr lyrae- , and open cluster - based distances . the latter are benchmarked with respect to the distance to the hyades , @xmath192 mag , which also ensures internal consistency with the lmc distance modulus adopted in this series of papers . the mean @xmath0 values and their associated standard deviations based on kinematic disk modeling are included on the right - hand side of table [ tracers.tab ] . note , however , that this method of galactic center distance determination suffers from larger uncertainties than the centroid approach , given the larger number of assumptions one has to make . in addition to the need for an accurate milky way mass model ( e.g. , reid et al . 2009 ; mcmillan 2011 ) , the basic , underlying assumption on which the accuracy of this methodology hinges is that the tracers are stationary with respect to the local standard of rest ( at the solar position ) and in circular motion at the solar circle . since both of these assumptions are usually imprecise , good results can not be expected from individual objects . 0.5 mm @cclccl@ publ . date & @xmath0 & & & & + ( mm / yyyy ) & ( kpc ) & & & ( km s@xmath5 ) + + 02/1992 & @xmath193 & caldwell et al . ( 1992 ) & @xmath129 mag & @xmath194 & @xmath195 + 05/1994 & @xmath196 & pont et al . ( 1994 ) & @xmath125 mag & @xmath197 & @xmath60 + 05/1995 & @xmath198 & dambis et al . ( 1995 ) & @xmath199 mag & @xmath60 & @xmath60 + 11/1997 & @xmath200 & feast & whitelock ( 1997 ) & @xmath125 mag & @xmath60 & @xmath0 from pont et al . ( 1994 ) ; new plr + 11/1997 & @xmath201 & feast & whitelock ( 1997 ) & @xmath125 mag & @xmath60 & @xmath0 from metzger et al . ( 1998 ) ; new plr + 01/1998 & @xmath202 & glushkova et al . ( 1998 ) & @xmath199 mag & @xmath203 & @xmath204 : open clusters , cepheids , rsgs + 02/1998 & @xmath205 & metzger et al . ( 1998 ) & @xmath73 mag & @xmath206 & systematic uncertainty 0.44 kpc + 02/1998 & @xmath207 & metzger et al . ( 1998 ) & @xmath73 mag & @xmath206 & @xmath208 , solar circle ; no orbital ellipticity + 02/1998 & @xmath209 & metzger et al . ( 1998 ) & @xmath73 mag & @xmath206 & @xmath208 , solar circle ; 4% orbital ellipticity + 08/1998 & @xmath210 & feast et al . ( 1998 ) & @xmath211 mag & @xmath60 & @xmath212 + 00/1999 & @xmath213 & glushkova et al . ( 1999 ) & @xmath199 mag & @xmath214 & @xmath204 : open clusters , cepheids , rsgs ; use caution@xmath215 + 01/2010 & @xmath216 & shen & zhang ( 2010 ) & & @xmath217 & @xmath60 + 02/2013 & @xmath218 & bobylev ( 2013 ) & @xmath219 mag & @xmath220 & @xmath221 ; @xmath222 days + 02/2013 & @xmath223 & bobylev ( 2013 ) & @xmath219 mag & @xmath224 & @xmath98 ; proper motions + 02/2013 & @xmath225 & zhu & shen ( 2013 ) & @xmath219 mag & @xmath226 & @xmath227 + + 06/2009 & @xmath228 & dambis ( 2009 ) & @xmath229 mag & @xmath230 & @xmath60 + 02/2010 & @xmath231 & dambis ( 2010 ) & @xmath232 mag & @xmath233 & @xmath60 + 11/2013 & @xmath234 & dambis et al . ( 2013 ) & @xmath235 mag & @xmath236 & recalibration of carney et al . ( 1995 ) + 11/2013 & @xmath237 & dambis et al . ( 2013 ) & @xmath235 mag & @xmath236 & recalibration of groenewegen et al . ( 2008 ) + 11/2013 & @xmath238 & dambis et al . ( 2013 ) & @xmath235 mag & @xmath236 & recalibration of collinge et al . ( 2006 ) + 11/2013 & @xmath239 & dambis et al . ( 2013 ) & @xmath235 mag & @xmath236 & average value + + 01/1998 & @xmath240 & glushkova et al . ( 1998 ) & @xmath199 mag & @xmath203 & @xmath204 : open clusters , cepheids , rsgs + 00/1999 & @xmath241 & glushkova et al . ( 1999 ) & @xmath199 mag & @xmath214 & @xmath204 : open clusters , cepheids , rsgs ; use caution@xmath215 + 02/2004 & @xmath22 & gerasimenko ( 2004 ) & @xmath60 & @xmath60 & @xmath242 ; weaver s method + 02/2004 & @xmath243 & gerasimenko ( 2004 ) & @xmath60 & @xmath60 & @xmath244 ; fokker s method + 02/2007 & @xmath245 & shen & zhu ( 2007 ) & @xmath219 mag & @xmath246 & @xmath247 + 02/2007 & @xmath248 & shen & zhu ( 2007 ) & @xmath219 mag & @xmath246 & mean value ( open clusters , ob stars ) + 11/2007 & @xmath249 & bobylev et al . ( 2007 ) & @xmath219 mag & @xmath250 & @xmath251 + 11/2007 & @xmath252 & bobylev et al . ( 2007 ) & @xmath219 mag & @xmath250 & young open clusters ( @xmath253 myr ) + 11/2007 & @xmath254 & bobylev et al . ( 2007 ) & @xmath219 mag & @xmath250 & old open clusters ( @xmath255 myr ) + 12/2009 & @xmath256 & zhu ( 2009 ) & @xmath219 mag & @xmath257 & @xmath258 + 02/2013 & @xmath256 & zhu & shen ( 2013 ) & @xmath219 mag & @xmath259 & @xmath258 + @xmath188 new abbreviation used : rsgs , red supergiant stars . + @xmath215 note that glushkova et al . ( 1998 ) and glushkova et al . ( 1999 ) are identical peer - reviewed articles , although this duplicate publication is not formally acknowledged . nevertheless , the galactic center distance estimates differ by 0.1 kpc between both papers , which remains unexplained . it is not obvious how one could self - consistently homogenize the galactic center distance estimates based on other kinematic disk tracers . we considered using the ratio of the rotational velocities at the solar circle , @xmath2 , and @xmath0 to transfer all these measurements onto a common scale . mcmillan & binney ( 2010 ) carefully re - evaluated a number of current models describing solar rotation about the galactic center . using data of 18 masers in high - mass star - forming regions at the solar circle from reid et al . ( 2009 ) , they found that the best - fitting models yield galactic center distances in the range from @xmath260 kpc to @xmath261 , and @xmath2 from @xmath262 km s@xmath5 to @xmath263 km s@xmath5 . despite these large ranges in @xmath0 and @xmath2 , which are largely driven by one s choice of galactic rotation curve , these authors found that with one exception the @xmath264 ratio should be narrowly constrained between @xmath265 km s@xmath5 kpc@xmath5 and @xmath266 km s@xmath5 kpc@xmath5 ( see also reid et al . . however , fig . [ fig4 ] shows that , in practice , the @xmath264 ratios adopted by different authors are significantly smaller than the range proposed by mcmillan & binney ( 2010 ; see the horizontal dashed lines in the figure ) . the red data points are three - year averages , aimed at showing the general trend , where the horizontal ` error bars ' indicate the relevant time period sampled and the vertical error bars represent the corresponding standard deviations . between 1990 and 2007 , the majority of @xmath264 ratios adopted by authors who recently determined @xmath0 are found near @xmath267 km s@xmath5 kpc@xmath5 ; since then , the ratio has slowly increased to @xmath268 km s@xmath5 kpc@xmath5 ( 20132015 ) . combined with the current - best galactic center distance estimates based on s - star orbital modeling or statistical parallaxes , @xmath269 kpc , these @xmath264 ratios imply that the galactic rotation speed at the solar circle is best constrained to the range from @xmath270 km s@xmath5 ( 19902007 ) to @xmath271 km s@xmath5 based on our current best understanding of galactic dynamics , i.e. , significantly faster than the iau - recommended value of @xmath272 km s@xmath5 . in fact , this is confirmed statistically : 33 of the 55 post-1990 values of @xmath2 used in this paper attained values of @xmath273 km s@xmath5 ( only 17 out of 55 publications derived @xmath274 km s@xmath5 ) . the equivalent numbers for the period since 2000 are 25 out of 39 high values ( versus 10 out of 39 low rotation speeds ) . an increasing body of evidence now suggests that the galactic rotation speed at the solar circle is indeed well in excess of the iau recommendation ( e.g. , reid et al . 2009 ) , with most authors supporting @xmath275 $ ] km s@xmath5 at the present time ( e.g. , shen & zhu 2007 ; bovy et al . 2009 ; zhu 2009 ; mcmillan & binney 2010 ; sato et al . 2010 ; shen & zhang 2010 ; brunthaler et al . 2011 ; mcmillan 2011 ; honma et al . 2012 ; schnrich 2012 ; dambis et al . 2013 ; reid 2013 ; zhu & shen 2013 ; reid et al . 2014 ) . since the dual basic premises of science are reproducibility and replicability ( e.g. , de grijs 2016 , and references therein ) , as fallible humans scientists may be tempted most likely unknowingly to trust results that are in good agreement with previously published values . such behavior may lead to a ` bandwagon effect , ' more commonly referred to as publication bias ( for background , see paper i ) . in the context of galactic center distance determinations , the possible presence of publication bias has been addressed repeatedly ( e.g. , reid 1989 , 1993 ; nikiforov 2004 ; foster & cooper 2010 ; malkin 2013a , b ; francis & anderson 2014 ) , although their conclusions are not in general agreement . publication bias can be revealed either by a gradual convergence of published values to a generally accepted number ( e.g. , malkin 2013a , b ; paper i ) or by a sudden shift in preferred values following a landmark publication . malkin ( 2013a , b ) explored evidence of the former using his compilation of 53 @xmath0 determinations published over a period of some 20 years ( 19922011 ) , but he did not find any statistically significant trend . ; see also reid 1989 ; 1993 ; surdin 1999 ; nikiforov 2004 ) , which may be the first suggestion of the presence of a bandwagon effect in @xmath0 estimates , although this could also simply be the result of improvements in either our physical understanding or of the methods used , or both . ] francis & anderson ( 2014 ) used an unspecified , larger sample of 150 galactic center distances , cited as `` available online , '' of which they deemed 137 values as broadly distinct , for a more detailed analysis of possible publication bias affecting published galactic center distance estimates . these latter authors claimed to have uncovered a continuing small but significant trend toward decreasing estimates based on 48 values published between 1980 and 2000 . ( this encompasses approximately half the number of distance determinations we recorded during the same period , i.e. , our database includes 100 galactic center distance determinations published between 1980 february and 2000 september . ) a student s @xmath276 test implies that this trend s significance is reduced from 92% prior to 2000 to 73% post-2000 ( francis & anderson 2014 ; see also foster & cooper 2010 ) . note that even if the range of published values becomes narrower over time , this does not necessarily imply that authors ( or reviewers ) are unconsciously affected by a bandwagon effect . on the contrary , as we already addressed in paper i in the context of lmc distance determinations , later publications may benefit from having access to larger tracer samples than earlier papers , the calibration relations are likely better understood , the analysis method applied are improved , and systematic errors might be reduced ( e.g. , malkin 2013a , b ) . objective consideration of progress in the field of galactic center distance determinations singles out one paper in particular that may qualify as a landmark publication . eisenhauer et al . ( 2003 ) were the first to determine a direct , geometric distance to the galactic center based on the orbital motion of s2 , one of the small sample of stars orbiting the galactic center s black hole on roughly keplerian orbits . although the direct galactic center distance determinations by the genzel and ghez groups are arguably the most precise galactic center distance estimates available to date , their derivation is very sensitive to a range of modeling assumptions ( e.g. , gillessen et al . 2009 ; francis & anderson 2014 ) . two potentially important issues the underlying assumption of keplerian motion of the s stars around sgr a*. first , newton s shell theorem , which posits that the gravity inside a spherical shell is unimportant , provides only an approximate solution to characterize the real , three - dimensional distribution of matter in the nuclear disk and central bar of the milky way . second , the initial keplerian approximations to the s - star orbits did not take into account relativistic effects ( zucker et al . 2006 ; gillessen et al . francis & anderson ( 2014 ) pointed out that the corrections for the relativistic geodetic effect ( de sitter precession ) attempted by gillessen et al . ( 2009 ) are based on the assumption that this precession arises in the schwarzschild metric , whereas one should instead adopt the kerr metric associated with a relativistically rotating central black hole . in turn , this leads to ` frame dragging ' ( also known as the lense thirring effect ) , which is predicted to affect both the s stars orbits and the light paths into the line of sight . neither zucker et al . ( 2006 ) nor gillessen et al . ( 2009 ) considered these effects ( for a detailed discussion , see francis & anderson 2014 ) . yet , in 2002 , when the star s2 was near its pericenter and relativistic effects were most significant , both zucker et al . ( 2006 ) and gillessen et al . ( 2009 ) reported difficulties in modeling the star s orbital motion . this thus implies that relativistic effects must be taken into account and certainly can not be ignored . perhaps surprisingly , this appears to be a case where more data actually lead to less accurate results . the potentially significant relativistic effects affecting the s - star orbits result in ever more serious cumulative differences from keplerian motion ( francis & anderson 2014 ) . nevertheless , here we consider the @xmath0 determinations based on the orbital motions of the s stars as a function of publication date . we refer to fig . [ fig1]b and table [ gcdirect.tab ] ( top ) for guidance . the most significant change in the central @xmath0 values since eisenhauer et al . ( 2003 , 2005 ) occurred with the publication of gillessen et al . ( 2009 ) , which resulted in an upward adjustment of approximately 0.6 kpc . gillessen et al . ( 2009 ) point out that their new values for @xmath0 are consistent within the errors with the eisenhauer et al . ( 2003 , 2005 ) results , and that adding more stars to the orbital solutions does not change the results significantly . they assert that their main improvement is owing to the more rigorous treatment of the systematic errors . in fact , it appears that this change is indeed not driven by publication bias , nor did their consideration of relativistic effects have a significant impact . the latter statement is supported by comparing the zucker et al . ( 2006 ) galactic center distance estimate with the earlier determinations of eisenhauer et al . ( 2003 , 2005 ) . instead , the gillessen et al . ( 2009 ) galactic center distance is more robust than their previous results , given that it is based on astrometric observations of much higher quality ( affected by smaller astrometric uncertainties ) and covering a significantly longer time span . the latest results obtained by both genzel s group ( gillessen et al . 2009 , 2013 ) and ghez collaboration ( ghez et al . 2008 ; yelda et al . 2011 ) are indeed fully mutually consistent , as also expected from the @xmath0 values resulting from combining the data of both groups ( gillessen et al . 2009 ) . indeed , the low value of @xmath277 kpc derived by morris et al . ( 2012 ) , members of the ghez group , is an odd one out . these latter authors only commented that this estimate corresponds to their current best fit to the observational data from 19952011 . they did not discuss their discrepant result , other than by pointing out that the uncertainty in their estimate is highly correlated with the uncertainty in the mass of the black hole . indeed , the same degeneracy affects the central values ( e.g. , gillessen et al . 2009 ) . in conclusion , we have not unearthed any evidence of possible publication bias in the galactic center distances resulting from s - star orbital modeling . careful assessment of the @xmath0 values obtained from statistical parallax measurements of the nuclear star cluster also show broad consistency during the 20 years leading up to 2015 ( for a brief discussion , see section [ direct.sec ] ) . therefore , the current - best distance determination to the galactic center is , arguably , based on appropriate combined direct observations of both the s - star orbits and the galactic center s statistical parallax properties . ( 2013 ) and chatzopoulos et al . ( 2015 ) both attempted such a combination , yielding @xmath278 kpc and @xmath279 kpc , respectively . however , other recent direct galactic center distance measurements yield somewhat smaller distances ( e.g. , trippe et al . 2008 , 2011 ; yelda et al . 2011 ; gillessen et al . 2013 ) . at the present time , we therefore advocate a benchmark galactic center distance of @xmath280 kpc , where the uncertainty is the statistical error resulting from combination of the do et al . ( 2013 ) and chatzopoulos et al . ( 2015 ) ` combined ' values . the additional , systematic uncertainty , which should be added in quadrature , is of order @xmath281 kpc . the latter estimate follows from consideration of the full set of updated values of @xmath0 in table [ gcdirect.tab ] , where the variety of assumptions made is reflected in the spread of the central values ( see also the discussion in section [ direct.sec ] ) . the unprecedented volume of our catalog of galactic center distance determinations has allowed us to conclude that the globular cluster , cepheid , and red clump centroids suggest smaller galactic center distances than the direct distances estimates , while the rr lyrae and mira variables result in larger galactic center centroid distances ( see , e.g. , table [ gccentroids.tab ] ) . indeed , it has been suggested that some tracers , and in particular the ( @xmath282 gyr ) old tracers ( i.e. , the rr lyrae stars ) might or might not follow the galactic bar ( e.g. , zoccali & valenti 2016 ) . this implies that distance determinations based on these tracers might depend on the line of sight of the adopted sample . moreover , we do not yet have similarly robust identifications of rr lyrae stars located in the galactic center center as we do for the galactic center cepheid sample . this suggests that rr lyrae variables might be more prone to possible systematic effects ( e.g. , zoccali & valenti 2016 ) . in addition , there are reasons to believe that the red clump results might also be affected by large systematic uncertainties . in fact , if a metallicity gradient is present from the galactic center to the outer bulge ( e.g. , zoccali et al . 2008 ; pietrukowicz et al . 2012 , 2015 ; uttenthaler et al . 2012 ) , this also means that the mean magnitude and color of red clump stars change when approaching the galaxy s inner core . a separate concern regarding systematic effects introduced by our homogenization is the following . if , as we have done for many rr lyrae - based galactic center distance determinations , we adopt any of the prevailing @xmath283 versus [ fe / h ] relations , we have to deal with a strong dependence on the reddening correction and the metallicity distribution of rr lyrae in the central galaxy . specifically , we have adopted @xmath104 mag . if the original papers quoting galactic center distance determinations used any other absolute magnitude calibration ( see table [ gccentroids.tab ] ) , we converted the @xmath0 value to a distance modulus , corrected it for the difference in calibration used , and eventually converted it back to a linear distance . if we were to adopt the @xmath50-band magnitudes for our calibration instead , the dependence on metallicity and on the reddening uncertainties is smaller . however , in this case we would introduce a dependence on the pulsation period , since rr lyrae exhibiting longer periods are brighter . where appropriate , we have also assumed that @xmath57 } = -1.0 $ ] dex for rr lyrae in the bulge ( e.g. , bono et al . if we assume that a typical mean period for a fundamental rr lyrae in the bulge is around 0.550.60 days and the mean metallicity is around [ fe / h]@xmath284 , @xmath285 mag ( marconi et al . 2015 ) . at the present time , we have access to too few @xmath50-band calibrated rr lyrae distance measurements to the galactic center ( only two such calibrations are available : carney et al . 1995 ; dambis 2009 ) . overall , the adjustments thus made take into account the metallicity dependence of the rr lyrae distance scale to the best of our current ability . finally , we explored whether any of the centroid tacers or those used for the kinematics - based galactic center distance estimates exhibited statistically significant trends in the modern , post-1990 era . in the context of exploring possible bandwagon effects , one should , of course , only consider the _ original _ galactic center distance estimates , before homogenization . the significant scatter among the data points pertaining to any of the kinematics tracers precludes us from resolving any such trend . among the centroid tracers , we do not have sufficient numbers of data points for the globular clusters , planetary nebulae , or cepheids spanning a sufficiently long time span to draw statistically justifiable conclusions either ( see fig . [ fig2 ] ) . for all other tracers we only find statistically _ insignificant _ ( @xmath286 ) trends of @xmath0 as a function of publication date , both for the full 19902016 time span and for the more recent period since 2005 . however , we find that in general the uncertainties have been steadily decreasing for all tracers , which largely owing to improvements in the calibration approaches used and access to new and larger samples of tracer objects ( for a related discussion , see paper i ) . in conclusion , neither the centroid- nor the kinematics - based galactic center distance tracers published since 1990 suggest the presence of any significant trends of @xmath0 with publication date . aiming at deriving a statistically well - justified galactic center distance based on a large variety of tracers and reducing any occurrence of publication bias , we embarked on an extensive data - mining effort of the scientific literature , eventually yielding 273 new or revised @xmath0 estimates published since records began in october 1918 until june 2016 . our large database of galactic center distance estimates , a fully linked version of which is made available to the scientific community , allowed us to explore the pros and cons of a variety of different approaches used to determine the distance to the galactic center . we separated our compilation into direct and indirect distance measurements . the former include distances such as those based on orbital modeling of the so - called s stars orbiting sgr a * , the closest visual counterpart of the milky way s central supermassive black hole , as well as those relying on statistical parallaxes of either the nuclear star cluster or the stellar population in the inner galactic core . careful assessment of the body of published galactic center distances based on these methods resulted in our galactic center distance recommendation of @xmath287 kpc . a much larger body of galactic center distance determinations is based on indirect methods , either those relying on centroid determinations for a range of different tracer populations ( e.g. , globular clusters , cepheid , rr lyrae , or mira variables , or red clump stars ) or measurements based on kinematic observations of objects at the solar circle , combined with a mass and/or rotational model of the milky way . the latter approaches are affected by significantly larger uncertainties than the former , while the central , mean galactic center distances based on the kinematic methods are systematically smaller than those based on centroid determinations . most centroid - based distances are in good agreement with those resulting from the direct methods . we did not find any conclusive evidence of the presence of a bandwagon effect in the post-1990 galactic center distance measurements , neither among the direct distance estimates nor among those based on centroid determinations . our set of kinematics - based distance measurements can not be used to explore this issue given the significant uncertainties associated with the latter methods . however , these latter measurement can indeed be used to constrain the galaxy s rotation velocity at the solar galactocentric distance using the @xmath264 ratios employed by their respective authors . we found a gradual increase in the mean value of @xmath2 from @xmath288 km s@xmath5 in the 1990s to @xmath289 km s@xmath5 more recently . ( both values represent the means of the distributions of rotation speeds ; their standard deviations are of order 2030 km s@xmath5 . ) our results thus imply that the iau - recommended galactic center distance ( @xmath290 kpc ) needs a downward adjustment , while the recommendation for the galactic rotation velocity at the solar circle ( @xmath4 km s@xmath5 ) requires an upward revision . finally , in view of the recent _ gaia _ data release 1 ( lindegren et al . 2016 ) , this is an opportune time to consider the impact on galactic center distance determinations of the improved parallax measurements of upwards of a billion stars that the mission will provide by the end of its nominal five - year duration . the final _ gaia _ catalog will map a significant fraction of the galactic volume , with particularly high ( distance - dependent ) parallax precision for stars in the solar neighborhood , gradually decreasing toward the galactic center . because of the high extinction and significant crowding in the galaxy s inner regions , _ gaia _ s optical ( @xmath291-band ) measurements will reach the galaxy s central regions , but not the center itself . it is anticipated that _ gaia _ s direct distance determinations of stars near the galactic center will be accurate to approximately 20% , i.e. , significantly worse than the distance estimates provided by any of the direct methods discussed in section [ direct.sec ] or the cepheid - based result of matsunaga et al . however , _ gaia s _ homogeneous and extensive final catalog will undoubtedly facilitate a significantly improved _ statistical _ determination of @xmath0 based on kinematic measurements at as well as inside the solar circle . in addition , _ gaia _ will provide improved distance measurements to large numbers of standard candles , thus almost certainly improving their zero - point accuracies , which in turn will allow us to test for the effects of population differences associated with the use of secondary distance tracers . _ gaia _ is currently among the best - placed facilities to resolve the remaining uncertainties in the galactic distance scale , which is no longer seriously affected by statistical uncertainties . the prevailing uncertainties preventing us from determining a more accurate distance to the galactic center are systematic in nature . among the latter , the effects of not just variations in the extinction , but of possible variations in the prevailing extinction _ law _ are among the most important stumbling blocks . the data set analyzed in this paper is , unfortunately , unsuitable for systematic studies of the uncertainties introduced by variations in the extinction law ( e.g. , nishiyama et al . 2009 ; nataf et al . 2016 ; and references therein ) . this is particularly so , because the vast majority of the individual data points retrieved from the literature have been extinction - corrected by their respective authors using the most appropriate approaches available at the time of their analyses . this implies the inherent presence of intrinsic inhomogeneities in the extinction corrections , which are impossible to fully correct for at the present time . in addition , we point out that current empirical estimates of reddening laws in the local universe are mainly based on optical and near - infrared photometry . to further constrain possible systematic effects , independent approaches based on either analyses of diffuse interstellar bands ( e.g. , munari & zwitter 1997 ; wallerstein et al . 2007 ; munari et al . 2008 ; kashuba et al . 2016 ) or spectroscopic studies ( e.g. , kudritzki et al . 2012 ) may shed new light on this long - standing problem . indeed , detailed , large - scale studies such as that of nataf et al . ( 2016 ) , perhaps combined with high spatial resolution observations of carefully selected , homogeneous samples of galactic center objects at near- to mid - infrared wavelengths ( such as those anticipated to result from _ wfirst _ operation or from campaigns with the next generation of 30 m - class ground - based telescopes ) , seem most promising to overcome the remaining systematics in the reddening laws . r. d. g. is grateful for research support from the national natural science foundation of china through grants 11373010 , 11633005 , and u1631102 . this work was also partially supported by prin - miur ( 2010ly5n2 t ) , ` chemical and dynamical evolution of the milky way and local group galaxies ' ( pi f. matteucci ) . this research has made extensive use of nasa s astrophysics data system abstract service .

based on the axiomatization of reversible computing racp , we generalize it to quantum reversible computing which is called qracp . by use of the framework of quantum configuration , we show that structural reversibility and quantum state reversibility must be satisfied simultaneously in quantum reversible computation . racp and qracp has the same axiomatization modulo the so - called quantum forward - reverse bisimularity , that is , classical reversible computing and quantum reversible computing are unified . [ firstpage ] reversible computation ; process algebra ; algebra of communicating processes ; axiomatization ; quantum computing

1501.05260

reversible computation has gained more and more attention in many application areas , such as the modeling of biochemical systems , program debugging and testing , and also quantum computing . quantum computing is one of the most important application areas of reversible computation . there are several research works on reversible computation . abramsky maps functional programs into reversible automata @xcite ; danos and krivine s reversible rccs @xcite uses the concept of thread to reverse a ccs @xcite@xcite process ; boudol and castellani @xcite@xcite compare three different non - interleaving models for ccs ; phillips and ulidowski s ccsk @xcite formulates a procedure for converting operators of standard algebraic process calculi such as ccs into reversible operators , while preserving their operational semantics . while racp is a new reversible axiomatical algebra , which is an axiomatical refinement to ccsk . in quantum process algebra , qacp @xcite @xcite is an axiomatization of quantum processes , which is a quantum generalization of process algebra acp . qacp still uses the concept of a quantum process configuration @xmath0 , which is usually consisted of a process term @xmath1 and state information @xmath2 of all ( public ) quantum information variables . in qacp , quantum operations are chosen to describe transformations of quantum states , and behave as the atomic actions of a pure quantum process . quantum measurements are treated as quantum operations , so probabilistic bisimulations are avoided . as qacp is a quantum generalization of acp , we generalize racp in quantum reversible computing in this paper , which is called qracp . racp and qracp has the same axiomatization modulo the so - called quantum forward - reverse bisimularity , that is , classical reversible computing and quantum reversible computing are unified . the paper is organized as follows . in section [ pre ] , some basic concepts about basic linear algebra , basic quantum mechanics , equational logic , structural operational semantics and process algebra acp are introduced . in section [ sosqp ] , we extend classical structural operational semantics to support quantum processes . the brqpa is introduced in section [ brqpa ] , arqcp is introduced in section [ arqcp ] , recursion is introduced in section [ recursion ] , and abstraction is introduced in section [ abstraction ] . classical reversible computing and quantum reversible computing are unified in section [ unifying ] . an application of qracp is introduced in section [ verification ] . the extensions of qracp is discussed in section [ extensions ] . finally , we conclude this paper in section [ conclusions ] . for convenience of the reader , we introduce some basic concepts about basic linear algebra , basic quantum mechanics ( please refer to @xcite for details ) , equational logic , structural operational semantics and process algebra acp ( please refer to @xcite and @xcite for more details ) . * definition [ bla].1 ( hilbert space)*. an isolated physical system is associated with a hilbert space , which is called the state space of the system . a finite - dimensional hilbert space is a complex vector space @xmath3 together with an inner product , which is a mapping @xmath4 satisfying : ( 1)@xmath5 with equality if and only if @xmath6 ; ( 2)@xmath7 ; ( 3 ) @xmath8 , where @xmath9 is the set of complex numbers , and @xmath10 denotes the conjugate of @xmath11 ( @xmath12 ) . * definition [ bla].2 ( orthonormal basis)*. for any vector @xmath13 in @xmath3 , the length @xmath14 . a vector @xmath13 with @xmath15 is called a unit vector in its state space . an orthonormal basis of a hilbert space @xmath3 is a basis @xmath16 with @xmath17 * definition [ bla].3 ( trace of a linear operator)*. the trace of a linear operator @xmath18 on @xmath3 is defined as @xmath19 * definition [ bla].4 ( tensor products)*. the state space of a composite system is the tensor product of the state space of its components . let @xmath20 and @xmath21 be two hilbert spaces , then their tensor product @xmath22 consists of linear vectors @xmath23 , where @xmath24 and @xmath25 . for two linear operator @xmath26 on hilbert space @xmath20 , @xmath27 on hilbert space @xmath21 , @xmath28 is defined as @xmath29 where @xmath30 and @xmath31 . let @xmath32 and @xmath33 . then the inner product of @xmath34 and @xmath13 is defined as follows . @xmath35 * definition [ bqm].1 ( density operator)*. a mixed state of quantum system is represented by a density operator . a density operator in @xmath3 is a linear operator @xmath2 satisfying:(1 ) @xmath2 is positive , that is , @xmath36 for all @xmath13 ; ( 2 ) @xmath37 . let @xmath38 denote the set of all positive operators on @xmath3 . * definition [ bqm].2 ( unitary operator)*. the evolution of a closed quantum system is described by a unitary operator on its state space . a unitary operator is a linear operator @xmath39 on a hilbert space @xmath3 with @xmath40 , where @xmath41 is the identity operator on @xmath3 and @xmath42 is the adjoint of @xmath39 . * definition [ bqm].3 ( quantum measurement)*. a quantum measurement consists of a collection of measurement operators @xmath43 , where @xmath44 is the measurement outcomes and satisfies @xmath45 * definition [ bqm].4 ( quantum operation ( super operator))*. the evolution of an open quantum system can be described by a quantum operation . a quantum operation on a hilbert space @xmath3 is a linear operator @xmath46 from the space of linear operators on @xmath3 into itself satisfying : ( 1 ) @xmath47\leq tr(\varrho)$ ] for each @xmath48 ; ( 2 ) for any extra hilbert space @xmath49 , @xmath50 is positive if @xmath18 is a positive operator on @xmath51 , where @xmath52 is the identity operator in @xmath49 . if @xmath47= tr(\varrho)$ ] for all @xmath48 , then @xmath46 is said to be trace - preserving . * definition [ bqm].5 ( relation between quantum operation and unitary operator)*. let @xmath39 be a unitary operator on the hilbert space @xmath3 , and @xmath53 for any @xmath48 . then @xmath46 is a trace - preserving quantum operation . * definition [ bqm].6 ( relation between quantum operation and measurement operator)*. let @xmath43 be a quantum measurement on the hilbert space @xmath3 . for each @xmath44 , let @xmath54 for any @xmath48 . the @xmath55 is a quantum operation and is not necessarily trace - preserving . in this paper , we refer to quantum operations all as reversible quantum operations . quantum measurements are treated as a special kind of quantum operations , in @xcite , quantum measurements can also be reversible . we introduce some basic concepts about equational logic briefly , including signature , term , substitution , axiomatization , equality relation , model , term rewriting system , rewrite relation , normal form , termination , weak confluence and several conclusions . these concepts are coming from @xcite , and are introduced briefly as follows . about the details , please see @xcite . * definition [ elp].1 ( signature)*. a signature @xmath56 consists of a finite set of function symbols ( or operators ) @xmath57 , where each function symbol @xmath58 has an arity @xmath59 , being its number of arguments . a function symbol @xmath60 of arity _ zero _ is called a constant , a function symbol of arity one is called unary , and a function symbol of arity two is called binary . * definition [ elp].2 ( term)*. let @xmath56 be a signature . the set @xmath61 of ( open ) terms @xmath62 over @xmath56 is defined as the least set satisfying : ( 1)each variable is in @xmath61 ; ( 2 ) if @xmath63 and @xmath64 , then @xmath65 . a term is closed if it does not contain variables . the set of closed terms is denoted by @xmath66 . * definition [ elp].3 ( substitution)*. let @xmath56 be a signature . a substitution is a mapping @xmath67 from variables to the set @xmath61 of open terms . a substitution extends to a mapping from open terms to open terms : the term @xmath68 is obtained by replacing occurrences of variables @xmath69 in t by @xmath70 . a substitution @xmath67 is closed if @xmath71 for all variables @xmath69 . * definition [ elp].4 ( axiomatization)*. an axiomatization over a signature @xmath56 is a finite set of equations , called axioms , of the form @xmath72 with @xmath73 . * definition [ elp].5 ( equality relation)*. an axiomatization over a signature @xmath56 induces a binary equality relation @xmath74 on @xmath61 as follows . ( 1)(substitution ) if @xmath72 is an axiom and @xmath67 a substitution , then @xmath75 . ( 2)(equivalence ) the relation @xmath74 is closed under reflexivity , symmetry , and transitivity . ( 3)(context ) the relation @xmath74 is closed under contexts : if @xmath76 and @xmath58 is a function symbol with @xmath77 , then @xmath78 . * definition [ elp].6 ( model)*. assume an axiomatization @xmath79 over a signature @xmath56 , which induces an equality relation @xmath74 . a model for @xmath79 consists of a set @xmath80 together with a mapping @xmath81 . ( 1)@xmath82 is sound for @xmath79 if @xmath72 implies @xmath83 for @xmath84 ; ( 2)@xmath82 is complete for @xmath79 if @xmath83 implies @xmath72 for @xmath84 . * definition [ elp].7 ( term rewriting system)*. assume a signature @xmath56 . a rewrite rule is an expression @xmath85 with @xmath73 , where : ( 1)the left - hand side @xmath86 is not a single variable ; ( 2)all variables that occur at the right - hand side @xmath87 also occur in the left - hand side @xmath86 . a term rewriting system ( trs ) is a finite set of rewrite rules . * definition [ elp].8 ( rewrite relation)*. a trs over a signature @xmath56 induces a one - step rewrite relation @xmath88 on @xmath61 as follows . ( 1)(substitution ) if @xmath85 is a rewrite rule and @xmath67 a substitution , then @xmath89 . ( 2)(context ) the relation @xmath88 is closed under contexts : if @xmath90 and f is a function symbol with @xmath77 , then @xmath91 . the rewrite relation @xmath92 is the reflexive transitive closure of the one - step rewrite relation @xmath88 : ( 1 ) if @xmath85 , then @xmath93 ; ( 2 ) @xmath94 ; ( 3 ) if @xmath93 and @xmath95 , then @xmath96 . * definition [ elp].9 ( normal form)*. a term is called a normal form for a trs if it can not be reduced by any of the rewrite rules . * definition [ elp].10 ( termination)*. a trs is terminating if it does not induce infinite reductions @xmath97 . * definition [ elp].11 ( weak confluence)*. a trs is weakly confluent if for each pair of one - step reductions @xmath98 and @xmath99 , there is a term @xmath100 such that @xmath101 and @xmath102 . * theorem [ elp].1 ( newman s lemma)*. if a trs is terminating and weakly confluent , then it reduces each term to a unique normal form . * definition [ elp].12 ( commutativity and associativity)*. assume an axiomatization @xmath79 . a binary function symbol @xmath58 is commutative if @xmath79 contains an axiom @xmath103 and associative if @xmath79 contains an axiom @xmath104 . * definition [ elp].13 ( convergence)*. a pair of terms @xmath86 and @xmath87 is said to be convergent if there exists a term @xmath100 such that @xmath96 and @xmath95 . axiomatizations can give rise to trss that are not weakly confluent , which can be remedied by knuth - bendix completion@xcite . it determines overlaps in left hand sides of rewrite rules , and introduces extra rewrite rules to join the resulting right hand sides , witch are called critical pairs . * theorem [ elp].2*. a trs is weakly confluent if and only if all its critical pairs are convergent . the concepts about structural operational semantics include labelled transition system ( lts ) , transition system specification ( tss ) , transition rule and its source , source - dependent , conservative extension , fresh operator , panth format , congruence , bisimulation , etc . these concepts are coming from @xcite , and are introduced briefly as follows . about the details , please see @xcite . also , to support reversible computation , we introduce a new kind of bisimulation called forward - reverse bisimulation ( fr bisimulation ) which first occurred in @xcite . we assume a non - empty set @xmath105 of states , a finite , non - empty set of transition labels @xmath18 and a finite set of predicate symbols . * definition [ sosp].1 ( labeled transition system)*. a transition is a triple @xmath106 with @xmath107 , or a pair ( s , p ) with @xmath108 a predicate , where @xmath109 . a labeled transition system ( lts ) is possibly infinite set of transitions . an lts is finitely branching if each of its states has only finitely many outgoing transitions . * definition [ sosp].2 ( transition system specification)*. a transition rule @xmath110 is an expression of the form @xmath111 , with @xmath112 a set of expressions @xmath113 and @xmath114 with @xmath115 , called the ( positive ) premises of @xmath110 , and @xmath116 an expression @xmath113 or @xmath114 with @xmath115 , called the conclusion of @xmath110 . the left - hand side of @xmath116 is called the source of @xmath110 . a transition rule is closed if it does not contain any variables . a transition system specification ( tss ) is a ( possible infinite ) set of transition rules . * definition [ sosp].3 ( proof)*. a proof from a tss @xmath117 of a closed transition rule @xmath111 consists of an upwardly branching tree in which all upward paths are finite , where the nodes of the tree are labelled by transitions such that : ( 1 ) the root has label @xmath116 ; ( 2 ) if some node has label @xmath118 , and @xmath119 is the set of labels of nodes directly above this node , then ( a ) either @xmath119 is the empty set and @xmath120 , ( b ) or @xmath121 is a closed substitution instance of a transition rule in @xmath117 . * definition [ sosp].4 ( generated lts)*. we define that the lts generated by a tss @xmath117 consists of the transitions @xmath116 such that @xmath122 can be proved from @xmath117 . * definition [ sosp].5*. a set @xmath123 of expressions @xmath124 and @xmath125 ( where @xmath87 ranges over closed terms , @xmath126 over @xmath18 and @xmath108 over predicates ) hold for a set @xmath46 of transitions , denoted by @xmath127 , if : ( 1 ) for each @xmath128 we have that @xmath129 for all @xmath130 ; ( 2 ) for each @xmath131 we have that @xmath132 . * definition [ sosp].6 ( three - valued stable model)*. a pair @xmath133 of disjoint sets of transitions is a three - valued stable model for a tss @xmath117 if it satisfies the following two requirements : ( 1 ) a transition @xmath116 is in @xmath134 if and only if @xmath117 proves a closed transition rule @xmath135 where @xmath123 contains only negative premises and @xmath136 ; ( 2 ) a transition @xmath116 is in @xmath137 if and only if @xmath117 proves a closed transition rule @xmath135 where @xmath123 contains only negative premises and @xmath138 . * definition [ sosp].7 ( ordinal number)*. the ordinal numbers are defined inductively by : ( 1 ) @xmath139 is the smallest ordinal number ; ( 2 ) each ordinal number @xmath140 has a successor @xmath141 ; ( 3 ) each sequence of ordinal number @xmath142 is capped by a limit ordinal @xmath11 . * definition [ sosp].8 ( positive after reduction)*. a tss is positive after reduction if its least three - valued stable model does not contain unknown transitions . * definition [ sosp].9 ( stratification)*. a stratification for a tss is a weight function @xmath143 which maps transitions to ordinal numbers , such that for each transition rule @xmath110 with conclusion @xmath116 and for each closed substitution @xmath67 : ( 1 ) for positive premises @xmath113 and @xmath114 of @xmath110 , @xmath144 and @xmath145 , respectively ; ( 2 ) for negative premise @xmath124 and @xmath125 of @xmath110 , @xmath146 for all closed terms @xmath147 and @xmath148 , respectively . * theorem [ sosp].1*. if a tss allows a stratification , then it is positive after reduction . * definition [ sosp].10 ( process graph)*. a process ( graph ) @xmath1 is an lts in which one state @xmath86 is elected to be the root . if the lts contains a transition @xmath149 , then @xmath150 where @xmath151 has root state @xmath152 . moreover , if the lts contains a transition @xmath153 , then @xmath154 . ( 1 ) a process @xmath155 is finite if there are only finitely many sequences @xmath156 . ( 2 ) a process @xmath155 is regular if there are only finitely many processes @xmath157 such that @xmath156 . * definition [ sosp].11 ( reverse transition)*. there are two processes @xmath1 and @xmath151 , two transitions @xmath158 and @xmath159 } } p$ ] , the transition @xmath159 } } p$ ] is called reverse transition of @xmath158 , and the transition @xmath158 is called forward transition . if @xmath158 then @xmath159 } } p$ ] , the forward transition @xmath158 is reversible . where @xmath160 $ ] is a kind of special action constant @xmath160\in a \times \mathcal{k}$ ] , @xmath161 , called the histories of an action @xmath126 , and @xmath162 . * definition [ sosp].12 ( bisimulation)*. a bisimulation relation @xmath163 is a binary relation on processes such that : ( 1 ) if @xmath164 and @xmath165 then @xmath166 with @xmath167 ; ( 2 ) if @xmath164 and @xmath166 then @xmath165 with @xmath167 ; ( 3 ) if @xmath164 and @xmath154 , then @xmath168 ; ( 4 ) if @xmath164 and @xmath168 , then @xmath154 . two processes @xmath1 and @xmath169 are bisimilar , denoted by @xmath170 , if there is a bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].13 ( forward - reverse bisimulation)*. a forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on processes such that : ( 1 ) if @xmath164 and @xmath165 then @xmath166 with @xmath167 ; ( 2 ) if @xmath164 and @xmath166 then @xmath165 with @xmath167 ; ( 3)if @xmath164 and @xmath171}}p'$ ] then @xmath172}}q'$ ] with @xmath167 ; ( 4 ) if @xmath164 and @xmath172}}q'$ ] then @xmath171}}p'$ ] with @xmath167 ; ( 5 ) if @xmath164 and @xmath154 , then @xmath168 ; ( 6 ) if @xmath164 and @xmath168 , then @xmath154 . two processes @xmath1 and @xmath169 are fr bisimilar , denoted by @xmath173 , if there is a fr bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].14 ( congruence)*. let @xmath56 be a signature . an equivalence relation @xmath163 on @xmath66 is a congruence if for each @xmath174 , if @xmath175 for @xmath176 , then @xmath177 . * definition [ sosp].15 ( panth format)*. a transition rule @xmath110 is in panth format if it satisfies the following three restrictions : ( 1 ) for each positive premise @xmath178 of @xmath110 , the right - hand side @xmath147 is single variable ; ( 2 ) the source of @xmath110 contains no more than one function symbol ; ( 3 ) there are no multiple occurrences of the same variable at the right - hand sides of positive premises and in the source of @xmath110 . a tss is said to be in panth format if it consists of panth rules only . * theorem [ sosp].2*. if a tss is positive after reduction and in panth format , then the bisimulation equivalence that it induces is a congruence . * definition [ sosp].16 ( branching bisimulation)*. a branching bisimulation relation @xmath163 is a binary relation on the collection of processes such that : ( 1 ) if @xmath164 and @xmath165 then either @xmath179 and @xmath180 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath182 such that @xmath183 and @xmath184 with @xmath167 ; ( 2 ) if @xmath164 and @xmath166 then either @xmath179 and @xmath185 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath186 such that @xmath187 and @xmath188 with @xmath167 ; ( 3 ) if @xmath164 and @xmath154 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath182 such that @xmath183 and @xmath189 ; ( 4 ) if @xmath164 and @xmath168 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath186 such that @xmath187 and @xmath190 . two processes @xmath1 and @xmath169 are branching bisimilar , denoted by @xmath191 , if there is a branching bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].17 ( branching forward - reverse bisimulation)*. a branching forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on the collection of processes such that : ( 1 ) if @xmath164 and @xmath165 then either @xmath179 and @xmath180 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath182 such that @xmath183 and @xmath184 with @xmath167 ; ( 2 ) if @xmath164 and @xmath166 then either @xmath179 and @xmath185 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath186 such that @xmath187 and @xmath188 with @xmath167 ; ( 3 ) if @xmath164 and @xmath154 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath182 such that @xmath183 and @xmath189 ; ( 4 ) if @xmath164 and @xmath168 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath186 such that @xmath187 and @xmath190 ; ( 5 ) if @xmath164 and @xmath171}}p'$ ] then either @xmath179 and @xmath180 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath192 such that @xmath183 and @xmath193}}q'$ ] with @xmath167 ; ( 6 ) if @xmath164 and @xmath172}}q'$ ] then either @xmath179 and @xmath185 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath194 such that @xmath187 and @xmath195}}p'$ ] with @xmath167 ; ( 7 ) if @xmath164 and @xmath154 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath192 such that @xmath183 and @xmath189 ; ( 8) if @xmath164 and @xmath168 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath194 such that @xmath187 and @xmath190 . two processes @xmath1 and @xmath169 are branching fr bisimilar , denoted by @xmath196 , if there is a branching fr bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].18 ( rooted branching bisimulation)*. a rooted branching bisimulation relation @xmath163 is a binary relation on processes such that : ( 1 ) if @xmath164 and @xmath165 then @xmath166 with @xmath197 ; ( 2 ) if @xmath164 and @xmath166 then @xmath165 with @xmath197 ; ( 3 ) if @xmath164 and @xmath154 , then @xmath168 ; ( 4 ) if @xmath164 and @xmath168 , then @xmath154 . two processes @xmath1 and @xmath169 are rooted branching bisimilar , denoted by @xmath198 , if there is a rooted branching bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].19 ( rooted branching forward - reverse bisimulation)*. a rooted branching forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on processes such that : ( 1 ) if @xmath164 and @xmath165 then @xmath166 with @xmath199 ; ( 2 ) if @xmath164 and @xmath166 then @xmath165 with @xmath199 ; ( 3 ) if @xmath164 and @xmath171}}p'$ ] then @xmath172}}q'$ ] with @xmath199 ; ( 4 ) if @xmath164 and @xmath172}}q'$ ] then @xmath171}}p'$ ] with @xmath199 ; ( 5 ) if @xmath164 and @xmath154 , then @xmath168 ; ( 6 ) if @xmath164 and @xmath168 , then @xmath154 . two processes @xmath1 and @xmath169 are rooted branching fr bisimilar , denoted by @xmath200 , if there is a rooted branching fr bisimulation relation @xmath163 such that @xmath164 . * definition [ sosp].20 ( lookahead)*. a transition rule contains lookahead if a variable occurs at the left - hand side of a premise and at the right - hand side of a premise of this rule . * definition [ sosp].21 ( patience rule)*. a patience rule for the i - th argument of a function symbol @xmath58 is a panth rule of the form @xmath201 * definition [ sosp].22 ( rbb cool format)*. a tss @xmath117 is in rbb cool format if the following requirements are fulfilled . ( 1 ) @xmath117 consists of panth rules that do not contain lookahead . ( 2 ) suppose a function symbol @xmath58 occurs at the right - hand side the conclusion of some transition rule in @xmath117 . let @xmath202 be a non - patience rule with source @xmath203 . then for @xmath176 , @xmath204 occurs in no more than one premise of @xmath110 , where this premise is of the form @xmath205 or @xmath206 with @xmath207 . moreover , if there is such a premise in @xmath110 , then there is a patience rule for the i - th argument of @xmath58 in @xmath117 . * theorem [ sosp].3*. if a tss is positive after reduction and in rbb cool format , then the rooted branching bisimulation equivalence that it induces is a congruence . * definition [ sosp].23 ( conservative extension)*. let @xmath208 and @xmath209 be tsss over signatures @xmath210 and @xmath211 , respectively . the tss @xmath212 is a conservative extension of @xmath208 if the ltss generated by @xmath208 and @xmath212 contain exactly the same transitions @xmath113 and @xmath114 with @xmath213 . * definition [ sosp].24 ( source - dependency)*. the source - dependent variables in a transition rule of @xmath110 are defined inductively as follows : ( 1 ) all variables in the source of @xmath110 are source - dependent ; ( 2 ) if @xmath113 is a premise of @xmath110 and all variables in @xmath87 are source - dependent , then all variables in @xmath147 are source - dependent . a transition rule is source - dependent if all its variables are . a tss is source - dependent if all its rules are . * definition [ sosp].25 ( freshness)*. let @xmath208 and @xmath209 be tsss over signatures @xmath210 and @xmath211 , respectively . a term in @xmath214 is said to be fresh if it contains a function symbol from @xmath215 . similarly , a transition label or predicate symbol in @xmath209 is fresh if it does not occur in @xmath208 . * theorem [ sosp].4*. let @xmath208 and @xmath209 be tsss over signatures @xmath210 and @xmath211 , respectively , where @xmath208 and @xmath212 are positive after reduction . under the following conditions , @xmath212 is a conservative extension of @xmath208 . ( 1 ) @xmath208 is source - dependent . ( 2 ) for each @xmath216 , either the source of @xmath110 is fresh , or @xmath110 has a premise of the form @xmath113 or @xmath114 , where @xmath217 , all variables in @xmath87 occur in the source of @xmath110 and @xmath147 , @xmath126 or @xmath108 is fresh . acp@xcite is a kind of process algebra which focuses on the specification and manipulation of process terms by use of a collection of operator symbols . in acp , there are several kind of operator symbols , such as basic operators to build finite processes ( called bpa ) , communication operators to express concurrency ( called pap ) , deadlock constants and encapsulation enable us to force actions into communications ( called acp ) , liner recursion to capture infinite behaviors ( called acp with linear recursion ) , the special constant silent step and abstraction operator ( called @xmath218 with guarded linear recursion ) allows us to abstract away from internal computations . bisimulation or rooted branching bisimulation based structural operational semantics is used to formally provide each process term used the above operators and constants with a process graph . the axiomatization of acp ( according the above classification of acp , the axiomatizations are @xmath219 , @xmath220 , @xmath221 , @xmath221 + rdp ( recursive definition principle ) + rsp ( recursive specification principle ) , @xmath222 + rdp + rsp + cfar ( cluster fair abstraction rule ) respectively ) imposes an equation logic on process terms , so two process terms can be equated if and only if their process graphs are equivalent under the semantic model . acp can be used to formally reason about the behaviors , such as processes executed sequentially and concurrently by use of its basic operator , communication mechanism , and recursion , desired external behaviors by its abstraction mechanism , and so on . acp is organized by modules and can be extended with fresh operators to express more properties of the specification for system behaviors . these extensions are required both the equational logic and the structural operational semantics to be extended . then the extension can use the whole outcomes of acp , such as its concurrency , recursion , abstraction , etc . we extend the above classical reversible structural operational semantics to support quantum processes , about the details , please refer to @xcite @xcite . here , we use @xmath223 to denote quantum operations in contrast to classical actions @xmath224 . * definition [ sosqp].1 ( quantum process configuration)*. a quantum process configuration is defined to be a pair @xmath225 , where @xmath1 is a process ( graph ) called structural part of the configuration , and @xmath48 specifies the current state of the environment , which is called its quantum part . * definition [ sosqp].2 ( quantum process graph)*. a quantum process ( graph ) @xmath225 is an lts in which one state @xmath86 is elected to be the root . if the lts contains a transition @xmath226 , then @xmath227 where @xmath228 has root state @xmath152 . moreover , if the lts contains a transition @xmath153 , then @xmath229 . ( 1 ) a quantum process @xmath230 is finite if and only if the process @xmath155 is finite . ( 2 ) a quantum process @xmath230 is regular if and only if the process @xmath155 is regular . * definition [ sosqp].3 ( quantum transition system specification)*. a quantum process transition rule @xmath110 is an expression of the form @xmath111 , with @xmath112 a set of expressions @xmath231 and @xmath232 with @xmath115 and @xmath233 , called the ( positive ) premises of @xmath110 , and @xmath234 , called structural part of @xmath112 and denoted as @xmath235 . and @xmath116 an expression @xmath231 or @xmath232 with @xmath115 and @xmath233 , called the conclusion of @xmath110 , and @xmath234 , called structural part of @xmath116 and denoted as @xmath236 . the left - hand side of @xmath116 is called the source of @xmath110 . @xmath237 is called the structural part of @xmath110 and denoted as @xmath238 . a quantum process transition rule @xmath110 is closed if and only its structural part @xmath238 is closed . a quantum transition system specification ( qtss ) is a ( possible infinite ) set of transition rules . * definition [ sosqp].4 ( quantum reverse transition)*. there are two quantum processes @xmath225 and @xmath228 , two transitions @xmath239 and @xmath240 } } \langle p , \varrho\rangle$ ] , the transition @xmath240 } } \langle p , \varrho\rangle$ ] is called reverse transition of @xmath158 , and the transition @xmath239 is called forward transition . if @xmath239 then @xmath240 } } \langle p , \varrho\rangle$ ] , the forward transition @xmath239 is reversible . where @xmath241 $ ] is a kind of special quantum operation constant @xmath241\in a \times \mathcal{k}$ ] , @xmath161 , called the histories of an quantum operation @xmath140 , and @xmath162 . * definition [ sosqp].5 ( quantum forward - reverse bisimulation)*. a quantum forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on processes such that : ( 1 ) if @xmath242 and @xmath243 then @xmath244 with @xmath245 ; ( 2 ) if @xmath242 and @xmath244 then @xmath243 with @xmath245 ; ( 3 ) if @xmath242 and @xmath246}}\langle p',\varrho'\rangle$ ] then @xmath247}}\langle q',\varsigma'\rangle$ ] with @xmath245 ; ( 4 ) if @xmath242 and @xmath247}}\langle q',\varsigma'\rangle$ ] then @xmath246}}\langle p',\varrho'\rangle$ ] with @xmath245 ; ( 5 ) if @xmath242 and @xmath229 , then @xmath248 ; ( 6 ) if @xmath242 and @xmath248 , then @xmath229 . two quantum process @xmath225 and @xmath249 are fr bisimilar , denoted by @xmath250 , if there is a fr bisimulation relation @xmath163 such that @xmath242 . * definition [ sosqp].6 ( relation between quantum fr bisimulation and classical fr bisimulation)*. for two quantum processes , @xmath250 , with @xmath251 , if and only if @xmath173 and @xmath252 , where @xmath2 evolves into @xmath253 after execution of @xmath1 and @xmath254 evolves into @xmath255 after execution of @xmath169 . * definition [ sosqp].7 ( quantum branching forward - reverse bisimulation)*. a quantum branching forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on the collection of quantum processes such that : ( 1 ) if @xmath242 and @xmath243 then either @xmath256 and @xmath257 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath258 such that @xmath259 and @xmath260 with @xmath245 ; ( 2 ) if @xmath242 and @xmath244 then either @xmath256 and @xmath261 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath262 such that @xmath263 and @xmath264 with @xmath245 ; ( 3 ) if @xmath242 and @xmath229 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath258 such that @xmath259 and @xmath265 ; ( 4 ) if @xmath242 and @xmath248 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath262 such that @xmath263 and @xmath266 ; ( 5 ) if @xmath242 and @xmath246}}\langle p',\varrho'\rangle$ ] then either @xmath256 and @xmath257 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath267 such that @xmath259 and @xmath268}}\langle q',\varsigma'\rangle$ ] with @xmath245 ; ( 6 ) if @xmath242 and @xmath247}}\langle q',\varsigma'\rangle$ ] then either @xmath256 and @xmath261 or there is a sequence of ( zero or more ) @xmath181-transitions @xmath269 such that @xmath263 and @xmath270}}\langle p',\varrho'\rangle$ ] with @xmath245 ; ( 7 ) if @xmath242 and @xmath229 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath267 such that @xmath259 and @xmath265 ; ( 8) if @xmath242 and @xmath248 , then there is a sequence of ( zero or more ) @xmath181-transitions @xmath269 such that @xmath263 and @xmath266 . two quantum processes @xmath225 and @xmath249 are branching fr bisimilar , denoted by @xmath271 , if there is a branching fr bisimulation relation @xmath163 such that @xmath242 . * definition [ sosqp].8 ( relation between quantum branching fr bisimulation and classical branching fr bisimulation)*. for two quantum processes , @xmath271 , with @xmath251 , if and only if @xmath272 and @xmath252 , where @xmath2 evolves into @xmath253 after execution of @xmath1 and @xmath254 evolves into @xmath255 after execution of @xmath169 . * definition [ sosqp].9 ( quantum rooted branching forward - reverse bisimulation)*. a quantum rooted branching forward - reverse ( fr ) bisimulation relation @xmath163 is a binary relation on quantum processes such that : ( 1 ) if @xmath242 and @xmath243 then @xmath244 with @xmath273 ; ( 2 ) if @xmath242 and @xmath244 then @xmath243 with @xmath273 ; ( 3 ) if @xmath242 and @xmath246}}\langle p',\varrho'\rangle$ ] then @xmath247}}\langle q',\varsigma'\rangle$ ] with @xmath273 ; ( 4 ) if @xmath242 and @xmath247}}\langle q',\varsigma'\rangle$ ] then @xmath246}}\langle p',\varrho'\rangle$ ] with @xmath273 ; ( 5 ) if @xmath242 and @xmath229 , then @xmath248 ; ( 6 ) if @xmath242 and @xmath248 , then @xmath229 . two quantum processes @xmath225 and @xmath249 are rooted branching fr bisimilar , denoted by @xmath274 , if there is a rooted branching fr bisimulation relation @xmath163 such that @xmath242 . * definition [ sosqp].10 ( relation between quantum rooted branching fr bisimulation and classical rooted branching fr bisimulation)*. for two quantum processes , @xmath274 , with @xmath251 , if and only if @xmath275 and @xmath252 , where @xmath2 evolves into @xmath253 after execution of @xmath1 and @xmath254 evolves into @xmath255 after execution of @xmath169 . * definition [ sosqp].11 ( congruence)*. let @xmath56 be a signature and @xmath38 be the state space of the environment . an equivalence relation @xmath163 on @xmath276 is a congruence , i.e. , for each @xmath174 , if @xmath277 for @xmath176 , then @xmath278 . an equivalence relation @xmath163 on @xmath276 is a congruence , if for each @xmath174 , @xmath175 for @xmath176 , and @xmath177 . * definition [ sosqp].12 ( quantum conservative extension)*. let @xmath208 and @xmath209 be qtsss over signature @xmath210 and @xmath279 , and @xmath211 and @xmath280 , respectively . the qtss @xmath212 is a conservative extension of @xmath208 if the ltss generated by @xmath208 and @xmath212 contain exactly the same transitions @xmath281 and @xmath282 with @xmath213 and @xmath283 , and @xmath284 . * definition [ sosqp].13 ( relation between quantum conservative extension and classical conservative extension)*. the qtss @xmath212 is a quantum conservative extension of @xmath208 with transitions @xmath281 and @xmath282 , if its corresponding tss @xmath285 is a conservative extension of @xmath286 with transitions @xmath287 and @xmath114 . in the following , the variables @xmath288 range over the collection of process terms , the variables @xmath289 range over the set @xmath18 of quantum operations , @xmath290 , @xmath291 are closed items , @xmath181 is the special constant silent step , @xmath292 is the special constant deadlock . we define a kind of special action constant @xmath241\in a \times \mathcal{k}$ ] where @xmath161 , called the histories of a quantum operation @xmath140 , denoted by @xmath241,\alpha[n],\cdots$ ] where @xmath293 . let @xmath294 . brqpa includes three kind of operators : the execution of quantum operation @xmath140 , the choice composition operator @xmath295 and the sequential composition operator @xmath296 . each finite process can be represented by a closed term that is built from the set @xmath18 of atomic actions or histories of an atomic action , the choice composition operator @xmath295 , and the sequential composition operator @xmath296 . the collection of all basic process terms is called basic reversible quantum process algebra ( brqpa ) , which is abbreviated to brqpa . we give the forward transition rules under transition system specification ( qtss ) for brqpa as follows . @xmath297,\upsilon(\varrho)}\ ] ] @xmath298,\varrho'\rangle \quad \upsilon\notin y}{\langle x+y,\varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m]+y,\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle\xrightarrow{\upsilon}\langle x',\varrho'\rangle \quad \upsilon\notin y}{\langle x+y,\varrho\rangle\xrightarrow{\upsilon}\langle x'+y,\varrho'\rangle}\ ] ] @xmath299,\varrho'\rangle \quad \upsilon\notin x}{\langle x+y,\varrho\rangle\xrightarrow{\upsilon}\langle x+\upsilon[m],\varrho'\rangle } \quad\frac{\langle y,\varrho\rangle\xrightarrow{\upsilon}\langle y',\varrho'\rangle\quad \upsilon\notin x}{\langle x+y,\varrho\rangle\xrightarrow{\upsilon}\langle x+y',\varrho'\rangle}\ ] ] @xmath300,\varrho'\rangle\quad \langle y,\varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m],\varrho'\rangle}{\langle x+y , \varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m],\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle\xrightarrow{\upsilon}\langle x',\varrho'\rangle\quad \langle y,\varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m],\varrho'\rangle}{\langle x+y , \varrho\rangle\xrightarrow{\upsilon}\langle x'+\upsilon[m],\varrho'\rangle}\ ] ] @xmath300,\varrho'\rangle\quad \langle y,\varrho\rangle\xrightarrow{\upsilon}\langle y',\varrho'\rangle}{\langle x+y , \varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m]+y',\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle\xrightarrow{\upsilon}\langle x',\varrho'\rangle\quad \langle y,\varrho\rangle\xrightarrow{\upsilon}\langle y',\varrho'\rangle}{\langle x+y , \varrho\rangle\xrightarrow{\upsilon}\langle x'+y',\varrho'\rangle}\ ] ] @xmath300,\varrho'\rangle}{\langle x\cdot y,\varrho\rangle\xrightarrow{\upsilon}\langle\upsilon[m]\cdot y,\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle\xrightarrow{\upsilon}\langle x',\varrho'\rangle}{\langle x\cdot y,\varrho\rangle\xrightarrow{\upsilon}\langle x'\cdot y,\varrho'\rangle}\ ] ] @xmath301,\varrho'\rangle}{\langle\upsilon[m]\cdot y,\varrho\rangle\xrightarrow{\omega}\langle\upsilon[m]\cdot \omega[n],\varrho'\rangle } , m\neq n \quad\frac{\langle y,\varrho\rangle\xrightarrow{\omega}\langle y',\varrho'\rangle}{\langle \upsilon[m]\cdot y,\varrho\rangle\xrightarrow{\omega}\langle\upsilon[m]\cdot y',\varrho'\rangle}.\ ] ] * the first transition rule says that each quantum operation @xmath302 can execute successfully , and leads to a history @xmath303 $ ] . because in reversible computation , there is no successful termination , the forward transition rule @xmath304,\upsilon(\varrho)\rangle}$ ] implies a successful execution . * the next four transition rules say that @xmath305 can execute only one branch , that is , it can execute either @xmath86 or @xmath87 , but the other branch remains . * the next four transition rules say that @xmath305 can execute both branches , only by executing the same quantum operations . * the last four transition rules say that @xmath306 can execute sequentially , that is , it executes @xmath86 in the first and leads to a successful history , after successful execution of @xmath86 , then execution of @xmath87 follows . we give the reverse transition rules under transition system specification ( qtss ) for brqpa as follows . @xmath307,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon,\upsilon[m](\varrho)\rangle}\ ] ] @xmath308}}\langle\upsilon,\varrho'\rangle\quad \upsilon[m]\notin y}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon+y,\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x',\varrho'\rangle \quad \upsilon[m]\notin y}{\langle x+y , \varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x'+y,\varrho'\rangle}\ ] ] @xmath309}}\langle\upsilon,\varrho'\rangle \quad \upsilon[m]\notin x}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x+\upsilon,\varrho'\rangle } \quad\frac{\langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle y',\varrho'\rangle \quad \upsilon[m]\notin x}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x+y',\varrho'\rangle}\ ] ] @xmath308}}\langle\upsilon,\varrho'\rangle\quad \langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon,\varrho'\rangle}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon,\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x',\varrho'\rangle\quad \langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon,\varrho'\rangle}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x'+\upsilon,\varrho'\rangle}\ ] ] @xmath308}}\langle\upsilon,\varrho'\rangle\quad \langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle y ' , \varrho'\rangle}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon+y',\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x',\varrho'\rangle\quad \langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle y',\varrho'\rangle}{\langle x+y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x'+y',\varrho'\rangle}\ ] ] @xmath308}}\langle\upsilon,\varrho'\rangle}{\langle x\cdot y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon\cdot y,\varrho'\rangle } \quad\frac{\langle x,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x',\varrho'\rangle}{\langle x\cdot y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle x'\cdot y,\varrho'\rangle}\ ] ] @xmath310}}\langle\omega,\varrho'\rangle}{\langle\upsilon[m]\cdot y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\omega[n]}}\langle\upsilon[m]\cdot \omega,\varrho'\rangle } , m\neq n \quad\frac{\langle y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\omega[n]}}\langle y',\varrho'\rangle}{\langle\upsilon[m]\cdot y,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\omega[n]}}\langle\upsilon[m]\cdot y',\varrho'\rangle}.\ ] ] * the first transition rule says that each history of a quantum operation @xmath303 $ ] can reverse successfully , and leads to a quantum operation @xmath302 . similarly , the reverse transition rule @xmath311,\varrho\rangle { \ext@arrow 0055{\rightslashedarrowfillg@}{}{\upsilon[m]}}\langle\upsilon,\upsilon[m](\varrho)\rangle}$ ] implies a successful reverse . * the next four transition rules say that @xmath305 can reverse only one branch , that is , it can reverse either @xmath86 or @xmath87 , but the other branch remains . * the next four transition rules say that @xmath305 can reverse both branches , only by executing the same histories of quantum operations . * the last four transition rules say that @xmath306 can reverse sequentially , that is , it reverses @xmath87 in the first and leads to a successful quantum operation , after successful reverse of @xmath87 , then reverse of @xmath86 follows . we design an axiomatization @xmath312 for brqpa modulo quantum fr bisimulation equivalence as table [ axiomforbrqpa ] shows . .axioms for brqpa [ cols= " < , < " , ] * theorem [ extensions].3*. @xmath313 + rsp + rdp + cfar + qrn1-qrn4 is sound for @xmath314 with guarded linear recursion and renaming operators , modulo quantum rooted branching fr bisimulation equivalence . since quantum rooted branching fr bisimulation is both an equivalence and a congruence for @xmath314 with guarded linear recursion and renaming operators , only the soundness of the first clause in the definition of the relation @xmath74 is needed to be checked . that is , if @xmath72 is an axiom in @xmath313 + rsp + rdp + cfar + qrn1-qrn4 and @xmath67 a closed substitution that maps the variable in @xmath86 and @xmath87 to basic quantum process terms , then we need to check that @xmath315 . since axioms in @xmath313 + rsp + rdp + cfar + qrn1-qrn4 ( same as @xmath316 + rsp + rdp + cfar + qrn1-qrn4 ) are sound for @xmath317 with guarded linear recursion and renaming operators modulo rooted branching fr bisimulation equivalence , according to the definition of quantum rooted branching fr bisimulation , we only need to check if @xmath252 when @xmath251 , where @xmath2 evolves into @xmath253 after execution of @xmath318 and @xmath254 evolves into @xmath255 after execution of @xmath68 . we can find that every axiom in table [ axiomforr ] meets the above condition . * theorem [ extensions].4*. @xmath313 + rsp + rdp + cfar + qrn1-qrn4 is complete for @xmath314 with guarded linear recursion and renaming operators , modulo quantum rooted branching fr bisimulation equivalence . to prove that @xmath313 + rsp + rdp + cfar + qrn1-qrn4 is complete for @xmath314 with guarded linear recursion and renaming operators modulo quantum rooted branching fr bisilumation equivalence , it means that @xmath319 implies @xmath72 . it was already proved that @xmath313 + rsp + rdp + cfar + qrn1-qrn4 ( same as @xmath316 + rsp + rdp + cfar + qrn1-qrn4 ) is complete for @xmath317 with guarded linear recursion and renaming operators modulo rooted branching fr bisimulation equivalence , that is , @xmath320 implies @xmath72 . @xmath319 with @xmath251 means that @xmath320 with @xmath251 and @xmath252 , where @xmath2 evolves into @xmath253 after execution of @xmath86 and @xmath254 evolves into @xmath255 after execution of @xmath87 , according to the definition of quantum rooted branching fr bisimulation equivalence . the completeness of @xmath313 + rsp + rdp + cfar + qrn1-qrn4 for @xmath317 with guarded linear recursion and renaming operators modulo rooted branching fr bisimulation equivalence determines that @xmath313 + rsp + rdp + cfar + qrn1-qrn4 is complete for @xmath314 with guarded linear recursion and renaming operators modulo quantum rooted branching fr bisimulation equivalence . we can see that qracp with renaming operator and racp with renaming operator can also be unified under the framework of quantum process configuration @xmath321 . in this paper , we give a quantum generalization of reversible algebra racp . based on the relationship between quantum fr bisimularity and classical fr bisimularity , we establish a series of axiomatization for reversible quantum processes called qracp . we also unify qracp and classical racp under the framework of quantum process configuration @xmath225 . it makes qracp can adapt to all quantum communication protocols in a reversible computing flavor . v. danos , j. krivine . : _ transactions in rccs . _ proceedings of 16th international conference on concurrency theory , concur 2005 , lecture notes in computer science , vol . 3653 , springer - verlag , 2005 , pp . 398412 . c. h. bennett and g. brassard . quantum cryptography : public - key distribution and coin tossing . _ proceedings of the ieee international conference on computer , systems and signal processing , 1984 , 175179 .

we study the influence of the relative size of the reservoir on the adsorption isotherms of a fluid in disordered or inhomogeneous mesoporous solids . we consider both an atomistic model of a fluid in a simple , yet structured pore , whose adsorption isotherms are computed by molecular simulation , and a coarse - grained model for adsorption in a disordered mesoporous material , studied by a density functional approach in a local mean - field approximation . in both cases , the fluid inside the porous solid exchanges matter with a reservoir of gas that is at the same temperature and chemical potential and whose relative size can be varied , and the control parameter is the total number of molecules present in the porous sample and in the reservoir . varying the relative sizes of the reservoir and the sample may change the shape of the hysteretic isotherms , leading to a `` reentrant '' behavior compared to the grand - canonical isotherm when the latter displays a jump in density . we relate these phenomena to the organization of the metastable states that are accessible for the adsorbed fluid at a given chemical potential or density .

0809.2314

under confinement in disordered mesoporous materials , the characteristic time scale for relaxation of a fluid can become extremely long . as a result , and although not always appreciated , equilibrium is often not attained and , accordingly , _ bona fide _ thermodynamic transitions such as the liquid - gas transition in one - component fluids or macroscopic phase separation in mixtures , are unobservable . `` capillary condensation '' in disordered solids is an out - of - equilibrium phenomenon , as illustrated by the irreversibility and hysteresis effects found in experiments . one typically observes a hysteresis loop that describes the isothermal evolution of the amount of fluid adsorbed in the porous solid as a function of the applied pressure , with branches that differ on adsorption ( filling ) and on desorption ( draining ) . this hysteresis loop appears rate - independent , and its size and shape vary with temperature as well as with the characteristics of the solid ( _ e.g. _ its porosity ) or those of the solid - fluid interaction potential . such a phenomenon is related to the existence of a large number of `` metastable states '' in which the system can be trapped on the experimental time scale ; evolution from one metastable state to another then only occurs as a result of the action of the applied pressure ( or equivalently , chemical potential ) and it proceeds through a sequence of irreversible cooperative condensation ( or evaporation ) events , generically denoted as `` avalanches''.@xcite the fact that the location and the shape of the hysteresis loop are reproducible in experiments indicates that the observation time is smaller than the time to reach the global equilibrium state , but is larger than local equilibration processes by which the system settles in one metastable state . as a result , the behavior of a fluid during filling or draining in disordered mesoporous materials can be rationalized by envisaging the evolution of the system in a free - energy landscape characterized by many local minima , _ i.e. _ metastable states.@xcite the above picture of gas adsorption in disordered porous media brings in a strong analogy with the out - of - equilibrium response of systems driven by an external force in the presence of impurities or other types of quenched disorder . this is for example the case of magnetization cycles in ferromagnetic materials when a magnetic field is ramped up and down and of hysteretic martensitic transformations in alloys;@xcite in both examples , avalanches can be detected through some `` crackling noise'',@xcite magnetic barkhausen noise in the former , acoustic emission in the latter . in such driven disordered systems , one expects the occurrence of out - of - equilibrium phase transitions as one changes , on top of the driving force , some external parameters such as the temperature or the characteristics of the intrinsic disorder ( _ e.g. _ the porosity in a porous solid).@xcite the branches of the hysteresis loop , in particular the adsorption and the desorption isotherms , may then display jumps ( discontinuities ) : indications for such behavior are for instance seen in the adsorption of helium in very light aerogels.@xcite such discontinuities and out - of - equilibrium phase transitions , however , are theoretically predicted on the basis of a grand - canonical set - up ( for gas adsorption ) in which the gas reservoir is infinite . in real experiments , the reservoir has a finite size which may not always be large enough for considering that the fluid inside the porous material is in a grand - canonical ensemble with fixed chemical potential . what should be expected in such situations ? the evolution of the system among metastable states may depend on the specific experimental set - up . as far as we know , there has been no systematic experimental study of the effect of changing the size of the gas reservoir ( relative to that of the porous medium ) and no estimate of the condition under which a grand - canonical situation is approximately reached . answering these questions is the primary goal of the present paper . the question of the dependence of the hysteretic response on the chosen control parameter has been a little more studied in other driven disordered systems . two extreme cases have been considered : the response of an extensive quantity to a change in the ( conjugate ) `` force '' and the response of an intensive quantity ( a force ) to a change in the conjugate extensive quantity ; the former is akin to the grand - canonical situation for a fluid in which the chemical potential ( or actually , the pressure in the gas reservoir ) is controlled and the latter to a canonical situation in which the number of adsorbed fluid molecules is controlled . examples of such studies are found in the context of martensitic transformations in which either stress or strain is controlled@xcite and in that of barkhausen noise in which either magnetic field or , via some feed - back mechanism , magnetization is controlled.@xcite in all cases , the loop obtained with the extensive variable as control parameter appears as reentrant when compared to that obtained with the force as control parameter . ( for a theoretical study , see ref . @xcite ) one anticipates that the influence of the relative size of the reservoir@xcite on the adsorption isotherms of a fluid in a disordered porous material is intimately connected to the organization of the metastable states in the adsorbed - density / chemical - potential plane . indeed , when the isotherms are smooth both on adsorption and on desorption , one may experimentally probe metastable states located inside the main hysteresis loop by studying ascending and descending `` scanning curves '' , which involve partial filling or draining.@xcite these curves lead to a variety of hysteretic subloops that provide direct evidence for the presence of metastable states inside the main loop.@xcite actually , metastable states are expected everywhere inside the latter . the situation is quite different , however , when the adsorption or the desorption branch displays a jump ( in a grand - canonical setting with a very large reservoir ) . a discontinuity prevents the realization of scanning curves in some portion of the isotherm . as a result , part of the interior of the main hysteresis loop is now inaccessible . in the present work we address the above questions , namely the effect of the relative size of the gas reservoir on the adsorption isotherms of a fluid in a disordered or inhomogeous porous solid and the connection to the distribution of metastable states inside the hysteresis loop . we show that even when the grand - canonical isotherms display discontinuous jumps , there are branches of metastable states inside the main hysteresis loop and that many of these states can be reached by varying the relative size of the reservoir , the smaller the reservoir the larger the extent to which the branches of metastable states are probed . in particular , the extent is maximal when the reservoir is so small that the fluid inside the porous solid behaves as in a canonical situation of fixed number of adsorbed molecules . when compared to the loop obtained for an infinite reservoir ( grand - canonical situation ) and plotted as the amount adsorbed versus the chemical potential , the hysteresis loop for a finite reservoir size then appears as reentrant . on the contrary , when the isotherms are smooth ( continuous ) , there is no influence of the size of the reservoir . the rest of the paper is organized as follows : in section ii , we introduce the two models that have been investigated and give some details on the methods used to compute the adsorption isotherms . we have considered : ( i ) an atomistic model of a fluid in simple , yet structured pore , whose adsorption isotherms are computed by molecular simulation , and ( ii ) a coarse - grained model for adsorption in a disordered mesoporous material , studied by a density functional approach in a local mean - field approximation . in both cases , the fluid inside the porous solid exchanges matter with a reservoir of gas that is at the same temperature and chemical potential and whose relative size can be varied . the overall system composed of the sample plus the reservoir is taken in the canonical ensemble , with the total number of molecules as control parameter . in section iii , we present the results for the atomistic model . the simplicity of the system allows one to get a clear interpretation of the physical nature of the branches of metastable states , whose number is limited . the existence of metastable states is a direct consequence of the intrinsic inhomogeneity of the pore space . as the relative size of the gas reservoir is varied , we find that the exploration of the branches changes , the isotherm displaying larger jumps in the adsorbed density as the size increases . section iv is devoted to the coarse - grained model of a disordered porous material . the number of metastable states is now enormous , since it is likely to increase exponentially with the size of the sample . we find a drastic change of behavior between the `` strong - disorder '' regime for which the isotherm is continuous and the `` weak - disorder '' one for which the isotherm , here the adsorption isotherm on which we focus our study , displays a jump corresponding to a macroscopic avalanche . in the former regime , the isotherm does not display any dependence on the relative size of the reservoir whereas in the latter , it is reentrant for finite reservoir sizes , the more so as one decreases the reservoir size , and it shows increasing jumps in adsorbed density with increasing reservoir size . finally , we summarize our main results and give our conclusions in section v. in particular , we discuss the relevance of our study to experimental situations and we stress the important role of the intrinsic inhomogeneity induced by the solid matrix . the principles of the method are the following . one considers a starting situation where @xmath0 particles are placed inside two cells which are in thermal equilibrium with an infinite heat bath at temperature @xmath1 . one of the cells represents a porous solid sample of volume @xmath2 and the other a reservoir of volume @xmath3 . the total volume is then equal to @xmath4 . we allow mass exchange between the cells whose volumes are kept fixed so that at ( exchange ) equilibrium the chemical potential @xmath5 of particles inside the reservoir is equal to the chemical potential @xmath6 of particles inside the porous solid : @xmath7 . we measure the average amount of particles @xmath8 present in the porous solid . varying the total number of particles @xmath0 in small steps , one changes both the chemical potential of the cells , @xmath9 , and the number of adsorbed particles , @xmath10 . we then monitor the adsorption isotherm @xmath11 ( for simplicity , we omit the subscript @xmath12 in the fluid density adsorbed in the porous sample ) as a function of @xmath13 . introducing the ratio @xmath14 and the average reservoir density @xmath15 , where @xmath16 , one has to satisfy for each @xmath13 the constraint @xmath17 with the condition @xmath18 . taking the limit @xmath19 corresponds to the grand - canonical ensemble for the adsorbed fluid ( with a controlled chemical potential ) whereas taking the limit to 1 corresponds to the canonical ensemble for the adsorbed fluid ( with a controlled number of adsorbed particles ) . between these two extremes , one has a mixed ensemble , with a reservoir of variable size compared to the sample , somewhat reminiscent of the mesoscopic canonical ensemble considered by neimark and coworkers.@xcite more details about the algorithms used in our two models are given below . an atomistic model of confined fluid was considered because it allows a realistic description of the various ( metastable ) states the system can adopt and , possibly , of the transition mechanisms at a molecular scale . the chosen system is a simple atomic fluid ( lennard - jones like ) confined in a mesoporous substrate . ( the model more specifically corresponds to argon adsorbed in nanoporous solid carbon dioxide.)@xcite a fully realistic model of the substrate should take into account the surface roughness , pore morphology ( pore size distribution ) , and interconnections between pores . the number of metastable states would however be too large for a systematic study . the model was therefore designed to exhibit only a few metastable states . the main potential sources for generating metastable states are nanometer - scale heterogeneities , due to pore - size distribution , and variations in surface chemistry . in both cases , the prominent effect on the adsorbed fluid comes from the modulation of the effective fluid / wall interaction.@xcite we then chose to investigate a cylindrical pore containing few domains of variable fluid / wall interaction.@xcite the diameter of the nanopore is 8 @xmath20 ( @xmath20 is the fluid - fluid lennard - jones diameter ) . in this way , the smooth - wall approximation can be applied : the external potential seen by a fluid particle in the pore is calculated by integrating the fluid - wall ( @xmath21 ) lennard - jones potential ( @xmath22 = 1.277 @xmath23 , @xmath24 = 1.094 @xmath20 ) over a uniform distribution of substrate interacting sites of density 0.8265 @xmath25 . the calculated reduced external potential @xmath26 in this perfectly cylindrical pore is given in fig . [ figpotr ] . the heterogeneity is introduced by modulating this external potential along the axial direction z : [ eq01 ] ^*_pore(r , z)=^*_cyl(r ) where @xmath27 = 0.3 for @xmath28 and @xmath27 = 0.2 for @xmath29 , and @xmath30 is the simulation box length . this external potential exhibits four domains of spatial extension of a few molecular diameters along the axial direction ( see fig . [ figpotrz ] ) . periodic boundary conditions are applied along the axial direction @xmath31 . the interactions are truncated at half the simulation box size ( minimal image convention ) . the fluid adsorption properties are calculated by monte - carlo simulations . thermalization of the adsorbed fluid is performed by particle displacement trials . chemical equilibration between the adsorbed fluid and the reservoir is done by particle exchange trials . the acceptance probabilities are given by the metropolis algorithm . the fluid in the reservoir is assumed to be ideal , which is a good approximation since the simulations are performed well below the critical point . as a consequence , the gas in the reservoir does not need to be treated explicitly , which speeds up the calculations . in the limit of infinite reservoir size , one recovers the usual grand canonical monte carlo algorithm ( gcmc ) . the adsorption / desorption curves are determined as follows . the initial configuration is the empty pore . few molecules are then added in one step to the system ( pore+reservoir ) . in order to mimic the experimental situation , the molecules are actually added to the reservoir cell , which results in an increase in @xmath32 , and , accordingly , to a lack of equilibrium between the pore and its reservoir . the whole system is then relaxed for a while . molecules exchange between the cells , which produces a small flow from the reservoir to the pore in the first steps . this flow comes to zero when stationarity is reached ( @xmath33 ) . we emphasize that this matter flow does not describe correctly the mass transport , since the monte carlo algorithm allows transfer of matter everywhere in the porous substrate . @xmath34 monte carlo trials per particle are then performed to acquire statistics for computing averaged quantities . this gives the first point of the isotherm . the subsequent points are obtained according to the same procedure , _ i.e. _ a small increase in the number of molecules in the reservoir , followed by a relaxation run , and , finally , by a long run for acquisition . after complete adsorption has been achieved , the desorption isotherm can be calculated by decreasing the amount of particles in the reservoir step by step , as in a real experiment . this algorithm allows us to obtain the scanning curves embedded within the main hysteresis loop . note that for the case of an infinite reservoir , it is the chemical potential which is increased stepwise instead of the amount of particles in the reservoir , which is then infinite . as discussed in previous papers,@xcite our approach to fluid adsorption in disordered mesoporous materials is based on a coarse - grained lattice - gas description which incorporates the essential physical ingredients of the solid - fluid system . we consider a three - dimensional bcc - lattice of linear size @xmath35 in which each of the @xmath36 sites may be occupied by a fluid or by a solid particle ( @xmath35 is measured in units of the lattice spacing @xmath27 and we set @xmath37 ; for a lattice , the volume @xmath2 is then simply equal to the number of sites @xmath38 times the volume of the elementary cell @xmath39 , so that we shall speak of volume and number of sites indistinctly ) . multiple occupancy of a site is forbidden and only nearest - neighbor interactions are taken into account . the fluid particles can equilibrate , as explained below , with a finite - size reservoir at fixed temperature whereas the solid particles are `` quenched '' and distributed according to a specific choice of the porous material structure . for the sake of simplicity , we study a random matrix with a porosity ( i.e. fraction of sites without solid particles ) @xmath40% . the relevant correlation length of the solid is around one lattice spacing so that even the smallest systems studied ( with a linear size @xmath41 ) can correctly describe the collective effects occurring inside the matrix on long length scale ( such as a sharp condensation event in the whole pore space ) . the starting point of our theoretical analysis is the following expression of the free - energy functional in the mean - field approximation : @xmath42=k_bt \sum_i [ \rho_i \ln \rho_i + ( \eta_i -\rho_i ) \ln ( \eta_i -\rho_i ) ] \nonumber \\ & -w_{ff } \sum_{<ij > } \rho_i \rho_j - w_{sf } \sum_{<ij > } [ \rho_i ( 1-\eta_j)+\rho_j ( 1-\eta_i)]\end{aligned}\ ] ] where @xmath43 is the thermally averaged fluid density at site @xmath44 and @xmath45 is a quenched variable describing the occupation of the lattice by solid particles ( @xmath46 if the site @xmath44 is occupied by the solid and @xmath47 otherwise ) ; @xmath48 and @xmath49 denote the fluid - fluid and fluid - solid attractive interactions , respectively , and the double summations run over all distinct pairs of nearest - neighbor sites . we first start with the grand canonical situation ( @xmath50 ) where fluid particles can equilibrate with an infinite reservoir that fixes the chemical potential @xmath13 . minimizing the grand - potential functional @xmath51= f_p[\ { \rho_i \}]-\mu \sum_i \rho_i$ ] with respect to @xmath52 at fixed @xmath1 and @xmath53 for a given realization of the solid yields a set of coupled equations , [ eq03 ] _ i= , where the sums run over the @xmath54 nearest neighbors of site @xmath44 . by using a simple iterative method to solve these equations , one finds solutions that are only minima of the grand potential surface , _ i.e. _ metastable states . for a given realization of the solid , the adsorption isotherm is obtained by increasing the chemical potential in small steps @xmath55 . at each subsequent @xmath53 , the converged solution at @xmath56 is used to start the iterations . what happens in the mixed situation ( @xmath57 ) ? the fluid particles can equilibrate between the solid sample and a finite - reservoir so that their total number @xmath0 is fixed . the isotherm is then obtained by increasing @xmath0 in small steps @xmath58 . in this case , the system tries to minimize its total helmholtz free - energy @xmath59=f_p[\ { \rho_i \},t]+f_r[\rho_r , t]$ ] , where @xmath60 is the helmholtz free - energy of the gas reservoir ( @xmath61=v_r\{k_bt [ \rho_r \ln \rho_r + ( 1 -\rho_r ) \ln ( 1 -\rho_r ) ] -w_{ff}c\rho_r^2/2\}$ ] with @xmath3 the volume of the reservoir , _ i.e. _ , again up to @xmath39 , the number of sites of the reservoir , while satisfying the global constraint @xmath62 . this can be solved in a natural way by the method of lagrange multipliers . we consider the function [ eq04 ] |_t[\ { _ i } , _ r,,t]=f_t[\ { _ i } , _ r , t]+\ { n-_i _ i - v_r _ r } , where @xmath63 is a lagrange multiplier that has the meaning of the chemical potential coupled to the densities . minimizing @xmath64 with the constraint on densities amounts to solve simultaneously the coupled equations @xmath65 , @xmath66 and @xmath67 or equivalently @xmath68-\lambda - w_{ff}\sum_{j / i } \rho_j - w_{sf}\sum_{j / i } ( 1-\eta_j)= 0 , \nonumber \\ & k_bt \ln [ \frac{\rho_r}{1-\rho_r}]-\lambda - w_{ff}\rho_r=0 , \nonumber \\ & n-\sum_i \rho_i - v_r \rho_r=0 , \quad 1 \leq i \leq n_{sp}.\end{aligned}\ ] ] one has then to define an iterative scheme that specifies how the system goes from one converged solution to another as the total number of particles is slowly changed . the details are given in the appendix . we are searching for configurations for the local densities @xmath69 that are local extrema of the grand - potential functional @xmath70 for the special value of the chemical potential @xmath71 . however , it is not fully guaranteed that the above algorithm necessarily converges to a local minimum , nor even to an extremum , since the constraint could in principle stabilize an unstable state . therefore , we regularly ascertained that configurations obtained in the mixed ensemble are stable in the grand - canonical ensemble , _ i.e. _ that they indeed are metastable states , by starting grand - canonical calculations with the converged solution at chemical potential @xmath63 . in addition , it must be emphasized that the global constraint is not satisfied until convergence is reached , in the spirit of the lagrange method . it is therefore doubtful that one can attribute any physical meaning to the intermediate stages of the iterative process . recall also that one does not take into account mass transport : local densities can change everywhere and instantaneously inside the porous sample . moreover , we focus on the adsorption ( filling ) process . as shown in previous papers,@xcite fluid desorption may crucially depend on the presence of an external surface for the porous solid : the system then includes a real interface between the adsorbed fluid and the external vapor , and , during desorption , the vapor domain may penetrate and drain the solid from the outer surface ( the so - called percolation and depinning transitions discussed in ref . these mechanisms are no longer bulk phenomenon and their , _ a priori _ more subtle , analysis will be carried out elsewhere . we have therefore used periodic boundary conditions for the sample and the reservoir ( separately ) . the ratio @xmath72 has been fixed to @xmath73 so that the adsorption isotherms exhibit the interesting range of phenomena as the temperature changes ( see ref.@xcite ) . we first focus on the adsorption / desorption isotherms obtained in the limit of infinite reservoir size ( @xmath74 ) , _ i.e. _ the gcmc data . the results are given in fig . [ figisothmain ] for various temperatures . as can be seen , the highest temperature isotherm ( reduced temperature @xmath75 ) is perfectly reversible , _ i.e. _ the adsorption and desorption curves are superimposed . on the other hand , for lower temperatures , the adsorption and desorption curves differ and exhibit hysteresis with vertical steps . we emphasize that the vertical lines are guides to the eye and do not correspond to gcmc data . the isotherm jumps in one step , and can not be stabilized in between . quite noticeably , the simulation points may be grouped into branches . for instance , for a reduced temperature of @xmath76 , three branches are present : the gas - like branch ( with a fluid layer adsorbed at the wall , the rest of the pore being filled with gas ) , which exists down to very low @xmath13 , the liquid - like branch ( pore filled by dense fluid ) at high @xmath13 , and a branch of intermediate density which exists only for a limited range of @xmath13 . we have found that adsorption and desorption are reversible along any given branch . these branches are associated to deep local minima in the free energy landscape describing the system ( the grand potential ) . the energetic barriers separating these minima are generally large compared to the thermal fluctuations sampled by the monte - carlo algorithm : the system then remains trapped in these local minima , which explains the reversibility of adsorption / desorption along the branches . however , for some particular values of @xmath13 the barriers become sufficiently small for allowing thermal fluctuations to make the system jump into an adjacent local minimum . these values define the limit of stability of the branches ( in the grand canonical ensemble ) . the vertical lines in fig . [ figisothmain ] indicate the new local minimum ( branch ) reached by the system . note that the @xmath13-range of the various branches overlap , which means that for some @xmath13 , the system may be stabilized in gcmc at various degrees of pore filling @xmath77 . this is the main origin of hysteresis . as can be seen , the lower the temperature , the larger the @xmath13-range of existence of the branches , and the wider the hysteresis . the number of branches increases for decreasing temperature , at least down to @xmath78 ( fig . [ figisothmain ] ) . we shall show later that this is actually still true down to @xmath79 if one takes into account metastable states that are hidden in the gcmc . for @xmath80 , the three branches are visited during both adsorption and desorption . they belong to the main adsorption and desorption curves . for @xmath81 , five metastable states belong to the main adsorption curve , and three to the main desorption curve . the branch of intermediate density reached during desorption ( second branch , @xmath82 ) looks like a continuation of the third branch visited during adsorption ( @xmath83 ) . this point was checked by showing the reversibility of the complete branch by gcmc ( by increasing and decreasing @xmath13 ) . the total number of metastable states is then five , and their range of existence was determined in a systematic way by @xmath13-variations . the results are shown in fig . [ figisothall ] . here again , lines are guides to the eye , and the vertical lines show the new metastable state reached by the system beyond the stability limit . the case @xmath79 deserves special attention . the procedure that consists in following simple ascending and descending @xmath13-paths does not allow to reach the two branches with reduced densities around @xmath84 and @xmath85 . these states were obtained from the corresponding states at higher temperature @xmath81 by following a particular ( @xmath53,@xmath1 ) path along which their stability is preserved . this path essentially consists of slowly decreasing the temperature and increasing the chemical potential in order to keep the average number of particles roughly constant . when @xmath86 has been reached , the @xmath13-range stability can be determined for both states ( see fig . [ figisothall ] ) . it is possible that more metastable states actually exist , but the procedures previously explained did not allow finding more than five branches in our simple system . figure [ figisothenlarg ] displays an enlargement of the lowest temperature ( @xmath86 ) gcmc results , without the vertical lines . as previously mentioned , the simulation points group into reversible branches ( solid lines are guides to the eye ) which correspond to local minima in the free - energy landscape . we now give a molecular - level description of these local minima . the lowest density branch corresponds to gas - like fluid filling the pore with adsorbed molecules at the wall . the highest density branch corresponds to liquidlike fluid saturating the pore . the three other states are stabilized by the chemical corrugation . visual inspection of molecular configurations ( see lower panel of fig . [ figisothenlarg ] ) shows that the three branches correspond respectively to : one liquid - like meniscus in the most attractive domain of the pore ; two liquid - like menisci in the two attractive regions of the pore ; one single large meniscus bridging the two attractive domains . we now focus on the adsorption / desorption results obtained for a finite - size reservoir . two ratios @xmath87 and @xmath88 have been considered ( @xmath89 and @xmath90 ) . the procedure previously used to obtain all metastable states will be applied here in its simplest form , _ i.e. _ isothermal @xmath13-paths , because it corresponds to what is actually feasible for realistic mesoporous substrates and in real experiments ( one essentially measures the main adsorption / desorption hysteresis and scanning curves ) . the results for the main hysteresis loops are given in fig . [ figisothalpha ] , where the grand canonical case ( @xmath91 ) has also been plotted for comparison . in all cases the symbols are the simulation data , and the color ( thick ) lines are guides to the eye that connect the simulation points in the same order as they were obtained by slowly increasing ( adsorption ) or decreasing ( desorption ) the total number of particles in the system . the thin ( black ) lines are the five branches previously obtained by extensive gcmc study ( fig . [ figisothenlarg ] ) . in the limit of infinite reservoir size ( @xmath91 ) , starting from the empty system and gradually increasing @xmath53 results in the continuous filling of the system up to @xmath92 where the system jumps into the second local minimum . increasing further the chemical potential makes the system jump directly to the fifth local minimum . upon desorption , the system remains in the saturated state for a while ( hysteresis ) and finally jumps ( for @xmath93 ) directly into the first local minimum ( gas - like branch ) without visiting any intermediate state . the vertical lines represent the constraint imposed by the reservoir ( @xmath94= cte ) but not necessarily the path actually followed by the system during the transition ( transient phenomenon are not properly described by the monte - carlo algorithm ) . in the cases of finite reservoir , it is the total number of particles in the system plus its reservoir which is increased by small steps as in a real experiment . as can be seen from fig . [ figisothalpha ] , the amount of adsorbed fluid initially increases along the gas - like branch in a continous manner until it reaches its stability limit . ( note that this limit occurs at slightly larger @xmath53 for smaller reservoir : @xmath95= -9.43 for @xmath74 , @xmath96 for @xmath97 and @xmath98 for @xmath89 . ) at this stability limit , a small addition of extra molecules in the reservoir destabilizes the adsorbed fluid which then jumps into the second state . during the transition , particles leave the reservoir and adsorb into the pore in the most attractive region to form a meniscus ( second metastable state ) , which results in a decrease of the chemical potential . here again , our algorithm is not meant to describe this transition . the path actually followed by the system is not known . the ( curved ) line shown in the figure corresponds to the constraint of conservation of the total number of particles in the system plus its reservoir . after equilibration , the system reaches the second branch . further increase in the total number of particles makes the system follow this branch until it reaches a new stability limit . ( as previously , this limit slightly increases for a finite reservoir : @xmath99 for @xmath100 , @xmath101 = -9.23 for @xmath97 and @xmath102 for @xmath89 . ) the system then jumps onto the fifth branch for @xmath100 , onto the fourth branch for @xmath97 , and onto the third branch for @xmath89 . the above results illustrate one of the main effects of the reservoir : the system does not necessarily visit the same metastable states as one changes the size of its reservoir . from fig . [ figisothalpha ] , this effect may be tentatively interpreted as follows : the reservoir imposes a constraint on the relation between the chemical potential and the amount adsorbed , materialized by the vertical or inclined curves . after a branch stability limit has been reached , the system approximately follows this line of constraint during the transition , and finally meets a new branch at a location that depends on the ratio between the reservoir and the system size . to illustrate this point , starting from the highest @xmath13 gcmc point of the second branch ( limit of stability ) , we have determined the new metastable state reached by the system after a small increase in the total number of molecules for various reservoir sizes . the results are given in fig . [ figjumpalpha ] . this picture is however somewhat oversimplified and does not apply in all situations since it does not take into account the complexity of the underlying free - energy ( grand - potential ) landscape that determines which path is actually followed by the system ( being in configurational space , the free energy landscape is an object of very high dimension ) . at a molecular level , this means that complex nucleation and fluid adsorption processes determine the path actually followed by the system . one of the consequences is that the system does not necessarily stop on the first branch crossed by the constraint line . this is illustrated in our simple system for @xmath89 during desorption from the fourth branch : as can be seen , the constraint line crosses the third metastable branch , but the system avoids this state and actually reaches the second branch . in the grand - potential landscape picture , this means that , starting from the fourth local minimum , and following the steepest slope , the system has not crossed the basin of attraction associated to the third minimum and has followed its way until being finally trapped by the basin associated to the second local minimum . in the molecular - level description , the fourth branch corresponds to a large meniscus bridging the two attractive regions . upon desorption , the system would have to nucleate a bubble in between the attractive regions in order to reach the third branch corresponding to two menisci in the attractive regions ( see fig . [ figmechanism ] , upper path ) . this nucleation barrier is probably too high , and it is more favorable for the system to desorb by recession of the meniscus in the less attractive regions until it is left with one single meniscus in the most attractive region corresponding to the second state ( see fig . [ figmechanism ] , lower path ) . as for the grand - canonical case , it is possible to perform a systematic search for all possible metastable states in the system in contact with a finite reservoir . the five branches already described are found in all cases . figure [ figallmetasalpha ] shows the superimposition of all simulation points ( stable over long runs ) obtained for the various ratios between the reservoir and system sizes already presented . as can be seen , the points group onto branches corresponding to the five metastable states found in the system . it is however reminded that the limits of stability of the branches slightly vary with the reservoir size . this may be related to the fact that transitions proceed via energetic barriers and that the amplitude of the fluctuations allowed by the reservoir depends on its size . to summarize the results , reducing the size of the reservoir allows one to explore metastable states upon adsorption / desorption that are not accessible in the grand - canonical situation . it is also found that the number of metastable states visited by the fluid increases with decreasing reservoir size . typical results for adsorption isotherms obtained at various temperatures in the grand - canonical ensemble are shown in fig . [ figek1 ] . when the temperature decreases , the isotherm shape changes from smooth to steep . as discussed in previous papers,@xcite this corresponds to a true out - of - equilibrium phase transition , the so - called `` avalanche transition'',@xcite with the sudden appearance of a macroscopic , connected liquid domain in the whole porous sample . a previous extensive scaling study@xcite showed that it exists a critical value of the temperature for which the isotherm changes from continuous to discontinuous in the thermodynamic limit . at higher temperature , isotherms look gradual but at reduced temperatures @xmath103 , they consist of little steps of varying sizes ( see insert in fig . [ figek2 ] ) . the above results are related to the characteristics of the grand - potential landscape . at low temperatures , this multidimensional landscape ( recall that it is a function of the local fluid densities , with here @xmath104 ) is characterized by a large number of local minima , the metastable states in which the system may be trapped . since thermally activated processes are neglected in the lmft ( an approximation that , as already discussed , finds its justification in the experimental reproducibility of the adsorption isotherms on the time scale of most experiments ) , the evolution of the system is only due to a variation of the external driving ( here the chemical potential ) . as @xmath53 varies , the system either follows the minimum in which it is trapped as this minimum deforms gradually ( the flat portions in the insert of fig . [ figek2 ] ) , or it falls instantaneously into another minimum when the former reaches its stability limit . this later move is a discontinuous and irreversible process , an avalanche , which is at the origin of the history - dependent behavior of the system , _ e.g. _ the hysteresis . the avalanche corresponds to some collective condensation event inside the porous sample that manifests itself by a jump in the adsorption isotherm . the size of the avalanche may be macroscopic as for @xmath105 , and the adsorption isotherm is discontinuous in the thermodynamic limit ; or it may be microscopic as for @xmath103 , and the adsorption isotherm remains smooth in the thermodynamic limit . ( more calculations should be necessary to conclude on what happens at @xmath106 . ) the hysteresis loop encloses all the metastable states of the system : this is illustrated in fig . [ figek2 ] by the scanning curves obtained by performing incomplete filling of the matrix and then decreasing the chemical potential to drain the adsorbed fluid . of course , the number of metastable states is very large , probably exponential in the system size , and only few of them are revealed with this simple procedure . what happens when the porous sample is coupled to a finite reservoir ? first , we begin with a vanishingly small reservoir , which is the canonical situation where one controls the number of adsorbed particles . the main results are summarized in fig . [ figek5 ] where we compare the adsorption isotherms obtained with the `` @xmath13-driven '' ( grand - canonical ) and `` @xmath77-driven '' ( canonical ) procedures . the behavior is quite different in the low and the high temperature regimes , respectively characterized by the absence and the presence of a macroscopic @xmath13-driven avalanche . at the highest temperature studied ( @xmath107 , not shown in the figure ) , there exists only one stable state , the equilibrium one , and controlling either the adsorbed density or the chemical potential yields the same result . this is also true for all temperatures at very low adsorbed densities ( not shown in the figure ) . at lower temperatures , avalanches appear in the grand - canonical isotherm . collective localized events , which we define as `` avalanches '' irrespective of the control variable , also show up in the canonical isotherm in the form of small jumps in the chemical potential toward a lower value : with the disappearance of the initial minimum , the system has to find another metastable state with the required adsorbed density ; as this state can not be found at a higher value of the chemical potential since the grand - canonical adsorption / desorption isotherms have been shown to represent the extremal curves that encompass all the metastable states of the system,@xcite the chemical potential must decrease . ( note that in a grand - canonical setting `` avalanches '' appear as jumps in the adsorbed density @xmath77 at constant chemical potential @xmath13 whereas in a canonical one , they appear as jumps in @xmath13 at constant @xmath77 . ) when a metastable state is found and fluid is further added , the system smoothly follows this state ( see the quasi - linear portions in fig . [ figek4 ] ) until it reaches the corresponding stability limit . then , there is a new jump in the chemical potential to a smaller value , and the evolution proceeds in this way until the porous sample is completely filled with liquid . the contrast between @xmath13-driven and @xmath77-driven isotherms that is illustrated in fig . [ figek5 ] is very reminiscent of what was found in ferromagnetic systems when comparing magnetization - driven and magnetic - field - driven protocols.@xcite as in the latter case , the canonical ( @xmath77-driven ) isotherms are closely related to the distribution of metastable states inside the ( grand - canonical ) hysteresis loops . we now discuss in more detail the canonical ( @xmath77-driven ) isotherms . at @xmath103 with a sample of linear size @xmath108 , the jumps in @xmath109 are very small and one needs to zoom in on the isotherm to see small differences between the canonical and grand - canonical protocols , as illustrated in fig . [ figek4 ] . this smallness is related to the presence of many metastable states in the close vicinity of the grand - canonical isotherm . since both the grand - canonical and the canonical avalanches remain of microscopic size , we expect that the small jumps in either @xmath77 or @xmath13 become infinitesimally small in the thermodynamic limit . this is indeed what is observed in fig . [ figek4 ] where isotherms obtained for different system sizes are compared . therefore , we predict that the canonical and grand - canonical curves should become identical in the thermodynamic limit . a similar behavior is observed at @xmath106 , as illustrated in fig . [ figek6 ] . however , the jumps in @xmath77 are larger than at @xmath103 and are clearly associated with a much more fluctuating chemical potential @xmath109 : each jump in @xmath77 generates a re - entrant behavior in @xmath109 . it is difficult to conclude on how the isotherm will evolve in the thermodynamic limit : as the size of the system increases ( see fig . [ figek6 ] ) , there are more and more intermediate points in the steepest part of the grand - canonical isotherm and , at the same time , the canonical isotherm is less and less fluctuating . the temperature @xmath106 is probably barely above the critical temperature of the ( grand - canonical ) avalanche transition and we expect that both the canonical and the grand - canonical isotherms would become continuous and would coincide in the thermodynamic limit . on the other hand , @xmath110 is undoubtedly below the critical temperature of the avalanche transition : all samples studied , whatever their size , display a large jump in their grand - canonical isotherms ( see fig . [ figek7 ] ) . ( in a previous work on a random matrix,@xcite we concluded through an extensive finite - size study that a macroscopic jump exists in the thermodynamic limit at @xmath105 for @xmath111 ; since the value @xmath112 used in this paper corresponds to a weaker disorder than @xmath111 , this is necessarily true here as well . ) as shown in fig . [ figek7 ] , the canonical curves show a pronounced reentrant behavior , with a large difference with the grand - canonical ones : a whole region void of metastable states then appears . there are always finite - size effects , which affect both the canonical and the grand - canonical isotherms . in the latter case , the rare events that trigger the macroscopic avalanches are very sensitive on details about the structure of the matrix , and the corresponding chemical potentials vary strongly with the system size at small @xmath35 . we have nonetheless checked by performing grand - canonical isotherms on very large systems of linear size @xmath113 that the position of the jump in the thermodynamic limit is very close to the jump found in the isotherm for @xmath108 that is displayed in fig . [ figek7 ] . on the other hand , even if the chemical potentials of the canonical isotherms fluctuate less and less as the size of the system grows ( especially at the end of the adsorption process , when the porous sample is nearly filled with liquid ) , the overall location of the isotherm in the @xmath77-@xmath13 plane does not change significantly with @xmath35 . to discuss the fluctuations of the chemical potential with the density in the canonical protocol , we have performed finite - size studies of the first moments of their distribution , after averaging over disorder ( _ i.e. _ porous sample ) realizations . more precisely , we compute the mean @xmath114 ( hereafter an overbar denotes the average over disorder ) and the variance @xmath115 for three sizes of the system . as explained in ref . @xcite , the standard argument concerning self - averaging quantities , whose value in a macroscopic sample is equal to the average over all disorder realizations , can not be applied : as a consequence , one can not be sure that @xmath109 is a self - averaging quantity . to delve more into its behavior , it is interesting to investigate the dependence on system size of the standard deviation @xmath116 . the result is shown in fig . [ figek9 ] , where @xmath117 is plotted as a function of the adsorbed density ( computed in bins of width 0.01 ) . except at low and high density , the standard deviation remains nearly constant with @xmath77 ( typically , between @xmath118 and @xmath119 ) . it decreases with @xmath35 but quite slowly : we find that in the region between @xmath118 and @xmath119 , @xmath120 with a finite - size scaling exponent @xmath121 around 1.4 , _ i.e. _ significantly smaller than 3 . this indicates that @xmath109 is self - averaging , but only weakly so . as in ref . @xcite , we have checked that the corresponding histograms are roughly gaussian ( see fig . [ figek9 ] ) . although the situation with respect to self - averaging remains somewhat unclear as far as the whole isotherm is concerned,@xcite it is nonetheless instructive to study the mean @xmath114 as a function of system size . this is displayed in fig . [ figek8 ] , where one can see that the @xmath114-isotherms vary significantly with the size of the system , becoming steeper as @xmath35 increases . we can not conclude whether or not the isotherm becomes vertical in the thermodynamic limit , but the data strongly suggest that it will be distinct in this limit from the grand - canonical isotherm : the @xmath114-isotherms for all studied system sizes intersect near the point @xmath122 and @xmath123 , at a significantly lower chemical potential than those found for the grand - canonical jumps ( @xmath124 ) for @xmath113 ( see fig . [ figek8 ] ) . we now comment on the two types of behavior found along the @xmath77-driven adsorption isotherm . for increasing adsorbed density @xmath77 , the chemical potential may either increase continuously or decrease by discontinuous jumps . in the former case , there are only slight swellings of the liquid domain , with the liquid - gas interfaces retaining the same shapes locally and all the @xmath125 s increasing . in the latter case , we find that each jump corresponds to a single condensation event , a @xmath77-driven `` avalanche '' , in which all sites of a compact region become liquid . this is illustrated in fig . [ figek11 ] . considering the two consecutive configurations with average densities @xmath77 and @xmath126 ( before and after the jump ) , sites are considered as turning liquid ( respectively , gas ) when the variation of the local fluid density is larger than @xmath84 ( respectively , lower than @xmath127 ) and are shown in red ( respectively , grey ) in the figure . the size of the condensed region varies for each jump along the isotherm , but the number of particles involved in the local condensation can go much beyond the controlled increment @xmath128 : in fig . [ figek11 ] , the condensation event ( in red ) corresponds to a local increase in the number of particles that is @xmath129 times larger than the initial increment . therefore , others regions of the porous sample must be drained : as shown in the figure , the most important evaporation events ( in grey ) remain smaller and less compact than the condensation event and appear at random throughout the material . ( whereas the definition of `` turning liquid '' is well characterized since the histogram of positive variations of the local density is always peaked at a value close to @xmath130 , this is much less so for its counterpart `` turning gas '' : the histogram of negative variations is continuously decreasing between @xmath131 and @xmath132 , so that the decrease of the average density that balances the condensation event mostly comes from small but widespread local variations . ) this suggests the following interpretation : adding a small amount of fluid in the sample may trigger a condensation `` avalanche '' whose size is larger than the added fluid amount ; a decrease of the chemical potential allows a slight draining of the rest of the sample and provides the required amount of fluid to `` feed '' the avalanche . finally , fig . [ figek10 ] presents our results in the mixed - ensemble for different sizes of the reservoir . as the size increases , the @xmath109-isotherms exhibit flat segments with decreasing length and jumps in the @xmath53-@xmath77 plane that are more and more pronounced this behavior is similar , albeit in a more complex systems and in the presence of a much larger number of metastable states , to that found in the atomistic model and illustrated in fig . [ figisothalpha ] . when the volume of the reservoir becomes @xmath88 times greater than the volume of the porous material , the isotherm displays a unique jump between the grand - canonical low - density branch and the grand - canonical high - density branch , but it still comes with a decrease of the chemical potential ( lower right panel of fig . [ figek10 ] ) . for larger reservoir sizes , the jump gets closer and closer to the grand - canonical isotherm . convergence however is rather slow . on the other hand , for the smallest reservoir sizes ( a reservoir volume @xmath133 times the volume of the porous material or less : see the two upper panels of fig . [ figek10 ] ) , the isotherms tend to superimpose on the canonical isotherm . it can be however noticed that the explored metastable states are not necessarily identical . as explained in section iii , one indeed has to be very careful with the representation of states in the @xmath53-@xmath77 plane which is only a projection from the highly multi - dimensional configurational space : it is possible to cross branches as the mixed - ensemble constraint allows the system to explore other areas of the underlying free - energy landscape . in this paper , we have presented a study of the out - of - equilibrium , hysteretic response of a fluid adsorbed in inhomogeneous porous materials when one couples the porous sample with a finite - size reservoir and controls the total number of particles . varying the relative size of the sample and the reservoir allows one to interpolate between a canonical situation with a controled adsorbed density and a grand - canonical situation with a controled chemical potential . we have considered both an atomistic model of a fluid in simple , yet structured pore and a coarse - grained model for adsorption in a disordered mesoporous material . the adsorption isotherms have been computed by molecular monte carlo simulation in the former case and by a density functional approach in a local mean - field approximation in the latter . the monte carlo simulations give us a clear picture of what occurs at the molecular scale in a small and weakly disordered system whereas the density functional approach provides insights at the mesoscopic scale for a fully disordered system . in both cases , we have found _ metastable states _ that appear as branches of finite extent in the @xmath134 plane . the metastable states correspond to inhomogeneous configurations of the fluid and the number of branches increases rapidly with the complexity of the material . it is worth stressing two points : first , these states are not unstable , _ i.e. _ not purely stabilized by the constraint on the total number of particles@xcite ; second , metastability is induced by the intrinsic inhomogeneity of the solid . accordingly , the picture of metastability in such systems is quite different than that encountered in mean - field - like descriptions of homogeneous ( bulk ) systems undergoing first - order transitions . we have shown that the way the system evolves between these metastable states may depend on the protocol , controlled here by the relative size of the reservoir . in particular , our results suggest that a discontinuity in the grand - canonical adsorption isotherm ( an out - of - equilibrium `` avalanche transition '' ) is associated with the absence of metastable states in a whole region of the @xmath134 plane and that the corresponding canonical adsorption isotherm ( and , more generally , isotherms performed with a small enough reservoir size ) differs from the grand - canonical one and displays a reentrance , even in the thermodynamic limit . what is the relevance of our results for experimental set - ups ? for illustration , we consider two examples . in order to determine the adsorption isotherm with the volumetric method using for instance nitrogen , known amounts of fluid are admitted stepwise in the sample cell . the amount of adsorbed fluid is the difference between the admitted gas and the amount of gas that fills the `` dead volume '' , _ i.e. _ the free space in the sample cell including connections : this is equivalent to the procedure discussed in this paper , with the dead volume playing the role of a finite - size reservoir for the sample . the amount of gas in this reservoir is calculated from the fluid equation of state and from measurements of the pressure , the temperature and the dead volume . in an experiment with a sample size of 20 mm@xmath135 , a cell of a few cm@xmath135 could be used . in such a situation , the ratio of volumes ( porous medium versus total ) is around @xmath137 and the ratio of adsorbed amount on the total amount around @xmath138 . on the other hand , to measure helium adsorption in aerogels , e. wolf et al @xcite use a different experimental set - up that was first proposed by herman et al@xcite , in which the adsorbed amount is controlled through the temperature of a helium gas reservoir connected to the experimental cell . the total amount of helium is fixed , but varying the temperature of the reservoir transfers atoms from the reservoir to the cell , or conversely . in the recent experiments,@xcite with temperatures ranging from 4k to 5k , the adsorbed amount at saturation corresponds to around 20% of the total amount . in both above examples , the ratio between the adsorbed amount and the total amount of fluid is fully in the range of the parameters of our study . however , in most experiments in disordered porous materials the adsorption isotherms display hysteresis but are smooth and continuous : in consequence , as predicted by the present study , no effects of the reservoir size are expected for macroscopic systems . we have shown here that for such effects to be observable , the temperature should be low enough for the grand - canonical adsorption isotherm to exhibit a true discontinuity and not only a very steep variation . it is unlikely that this true discontinuity could be observed in common disordered porous materials such as vycor or xerogels : the disorder and the confinement are too strong and prevent the appearance of the avalanche transition at temperatures higher than the triple point . the case of helium adsorption in aerogels is however more promising since at low temperature ( below 4k ) , jumps have been predicted@xciteand may have already been observed in some experiment@xcite . it would therefore be interesting in this case to perform controlled experiments for different relative sizes of the reservoir to see if the predicted reentrance of the isotherm as one moves away from the grand - canonical situation is indeed encountered , as observed in hysteretic martensitic transformations.@xcite we thank m .- rosinberg for many fruitful discussions . 10 detcheverry , f. ; kierlik , e. ; rosinberg , m. l. ; tarjus , g. _ phys . e _ * 2006 * , _ 72 _ , 051506 . kierlik , e. ; monson , p. a. ; rosinberg , m. l. ; sarkisov , l. ; tarjus , g. _ phys . lett _ * 2001 * , _ 87 _ , 054701/1 . kierlik , e. ; monson , p. a. ; rosinberg , m. l. ; tarjus , g. _ j. phys . : condens . matter _ * 2002 * , _ 12 _ , 9295 . woo , h.j . ; monson , p. a. _ phys . e _ * 2003 * , _ 67 _ , 041207 . see e.g. : bertotti , g. _ the science of hysteresis _ ; academic press : san diego , 2006 and references therein . sethna , j. p. ; dahmen , k. a. ; perkovi , o. in _ the science of hysteresis _ ; bertotti , g. , mayergoyz i. eds . ; elsevier : 2004 . sethna , j. p. ; dahmen , k. ; kartha , s. ; krumhansl , j. a. ; roberts , b. w. ; shore , j. d. _ phys . * 1993 * , _ 70 _ , 3347 . tulimieri , d. j. ; yoon , j. ; chan , m. h. w. _ phys . * 1999 * , _ 82 _ , 121 . herman , t. ; day , j. ; beamish , j. _ phys . b _ * 2005 * , _ 72 _ , 184202 . bonnet , f. ; lambert , t. ; cross , b. ; guyon , l. ; despetis , f. ; puech l. ; wolf , p. e. _ eur . lett . _ * 2008 * , _ 82 _ , 56003 . bonnot , e. ; romero , r. ; illa , x. ; manosa , l. ; planes , a. ; vives , e. _ phys . b _ * 2007 * , _ 77 _ , 064422 . illa , x. ; rosinberg , m. l. ; vives , e. _ phys . b _ * 2006 * , _ 74 _ , 224403 . in general , both the system , _ i.e. _ the sample with the porous medium , and the reservoir are taken in the thermodynamic limit ; yet the reservoir can be of negligible size compared to the system . ball , p. ; evans , r. _ langmuir _ * 1989 * , _ 5 _ , 714 . burgess , c. g. v. ; everett , d. h. ; nuttall , s. _ pure appl . chem . _ * 1989 * , _ 61 _ , 1845 . lilly , m. p. ; finley , p. t. ; hallock , r. b. _ phys . * 1993 * , _ 71 _ , 4186 . neimark , a. v. ; vishniakov , a. _ phys . * 2000 * , _ 62 _ , 4611 . peterson , b. k. ; walton , j. p. r. b. ; gubbins , k. e. _ j. chem . . 2 _ , * 1986 * , _ 82 _ , 1763 . peterson , b. k. ; gubbins , k. e. ; heffelfinger , g. s. _ j. chem . phys . _ * 1988 * , _ 88 _ 6487 . puibasset , j. _ j. phys . chem . _ * 2005 * , _ b109 _ , 4700 . puibasset , j. _ j. chem . phys . _ * 2005 * , _ 122 _ , 134710 . puibasset , j. _ j. chem . phys . _ * 2006 * , _ 125 _ , 074707 . sarkisov , l. ; monson , p. a. _ phys . e _ * 2002 * , _ 65 _ , 011202/1 . rosinberg , m. l. ; kierlik , e. ; tarjus , g. _ europhys . lett . _ * 2003 * , _ 62 _ , 377 . everett , d. h. _ colloids and surfaces a : physicochemical and engineering aspects _ * 1998 * , _ 141 _ , 279 . cross , b. ; puech , l. ; wolf , p. e. _ j. of low temp . _ * 2007 * , _ 148 _ , 903 . detcheverry , f. ; kierlik , e. ; rosinberg , m. l. ; tarjus , g. _ physica b _ * 2004 * , _ 343 _ , 303 . detcheverry , f. ; kierlik , e. ; rosinberg , m. l. ; tarjus , g. _ adsorption _ * 2005 * , _ 11 _ , 115 . as a starting point , it is convenient to rewrite the condition of minimization of the grand potential , @xmath65 , as [ eqa1 ] ( - ) _ i= ( _ i-_i ) and sum over @xmath44 so as to express @xmath63 as a function of the densities @xmath139 \ } } { \sum_i \rho_i}.\end{aligned}\ ] ] our algorithm is then the following : changing the total density @xmath140 by a small step @xmath141 , _ i.e. _ @xmath142 , one first supposes that the supplementary amount is adsorbed in the porous material : @xmath143 . one then computes the new lagrange parameter from @xmath144 \ } } { v_p \rho^{new}}\end{aligned}\ ] ] ( with the old local densities ) and the new local densities from @xmath145}.\end{aligned}\ ] ] the new reservoir density @xmath146 is obtained by inverting [ eqa5 ] ^new = k_bt -w_ff_r^new . one then checks if the constraint is satisfied , _ i.e. _ if @xmath147 within a given precision and , if not , one iterates using @xmath148 $ ] until convergence is reached . in practice , to improve the convergence of the iteration procedure , we use a mixing scheme , retaining a part of the previous iteration for the subsequent iteration . it appears that the configurations visited do not depend on the mixing parameter when the convergence criterions are strong enough . note that other equivalent algorithms can be devised as well and , interestingly , the output of the calculation appears to be quite robust to the choice of the algorithm . for instance , changing eq . [ eqa2 ] by the equivalent equation , [ eqa6 ] ( - ) = , yields the same trajectory for the converged states even if the intermediate stages and the speed of convergence greatly differ . one can also add the increments @xmath141 in the reservoir as in the molecular simulation : @xmath149 ; then , one computes the new lagrange multiplier with eq . [ eqa5 ] and the new local densities with eq . [ eqa4 ] , checks the constraint , and iterates using @xmath150 $ ] . this does not change the isotherm when the evolution is adiabatic . this algorithm seems simpler and possibly closer to the experimental protocol . however , it becomes problematic in the limit of the canonical ensemble ( @xmath151=1 ) . therefore , we have preferred the first algorithm described above . our calculations were performed with samples of linear sizes varying from @xmath41 to @xmath152 and convergence was assumed when for the @xmath153 iteration , @xmath154 and @xmath155 with @xmath156 for @xmath41 and @xmath157 for @xmath158 and @xmath108 . in addition , the steps @xmath141 in @xmath159 were taken as small as @xmath160 so that most of the avalanches could be resolved ( see ref . @xcite for more details about identification of avalanches ) . enforcing further the convergence criterions ( _ e.g. _ with @xmath161 for @xmath158 ) does not change the path followed by the system : adiabatic regime has been reached . however , for the smallest system studied , with linear size @xmath41 , diminishing too much @xmath141 could prevent convergence as the system is not able to find a metastable state with the required density .

we present a study of the relative sizes of planets within the multiple candidate systems discovered with the mission . we have compared the size of each planet to the size of every other planet within a given planetary system after correcting the sample for detection and geometric biases . we find that for planet - pairs for which one or both objects is approximately neptune - sized or larger , the larger planet is most often the planet with the longer period . no such size location correlation is seen for pairs of planets when both planets are smaller than neptune . specifically , if at least one planet in a planet - pair has a radius of , of the planet pairs have the inner planet smaller than the outer planet , while no preferred sequential ordering of the planets is observed if both planets in a pair are smaller than .

1212.1859

approximately 20% of the planetary candidate systems discovered thus far by @xmath0 @xcite have been identified as having multiple transiting candidate planets @xcite . as @xmath0 continues its mission , the number of multiple planet systems is likely to grow - not only because the total number of systems known to host planets will increase , but also because previously identified `` single''-candidate systems may be found to have additional planets not previously detected . for example , 96 ( 11% ) of the `` single '' candidate systems from the @xcite @xmath0 object of interest ( koi ) list are now listed on the @xcite koi list as having multiple planetary candidates . thus , understanding planets in multiple systems is not only important for placing into context our own solar system ( the best studied planetary system ) , but multiple planetary systems may turn out to be the rule rather than the exception . typical detailed confirmation of individual planetary systems takes a concerted ground - based observational and modeling effort to rule out false positives caused by blends ( e.g. , * ? ? ? * ; * ? ? ? however , the multiple candidate systems provide additional information that can be used to confirm planets directly via transit timing variations @xcite or via statistical arguments that multiple transiting systems discovered by @xmath0 are almost all true planetary systems @xcite . indeed , based upon orbital stability arguments , @xmath4 of the pairs within multiple candidate systems are most likely real planets around the same star @xcite . the higher statistical likelihood that candidates in multiple systems are true planetary systems has enabled studies of the properties of planetary systems with a lower level of false positive contamination than would be expected if all transiting systems were studied . the overall false positive rate within the @xmath0 candidate sample has been estimated to be @xmath5% ( e.g. , * ? ? ? * ; * ? ? ? * ) , but in comparison , the overall false positive rates among the multiple transiting systems is expected to be @xmath6% @xcite . the multiple candidate systems also provide another level of certainty to the studies of planetary systems . the quality of knowledge of planetary characteristics for individual planets often is dependent upon the quality of knowledge of the host stellar characteristics . for example , planetary radii uncertainties for transiting planets are dominated by the uncertainties in the stellar radii @xmath7 , and changes in our understanding of the stellar radius can greatly alter our understanding of the radii of individual planets ( e.g. , * ? ? ? by studying the _ relative _ properties of planets within multiple systems , systematic uncertainties associated with the stellar properties are minimized , understanding the relative sizes and , hence , the relative bulk compositions and structures of the planets within a system can yield clues on the formation , migration , and evolution of planets within an individual system and on planetary systems as a whole ( e.g. , * ? ? ? * ) . within our solar system , the distribution of the unique pair - wise radii ratios for each planet compared to the planets in orbits exterior to its orbit ( e.g. , mercury - venus , mercury - earth , @xmath8 mercury - neptune , venus - earth , venus - mars , @xmath8 etc . ) is dominated by ratios less than unity ( i.e. , the inner planets are smaller than the outer planets ) . for the 8 planets in the solar system , 20 of the 28 ( @xmath9 ) unique radii ratios are @xmath10 , but only 3 out of 7 ( @xmath11 ) of neighboring pairs display this sequential size hierarchy . if only the terrestrial planets are considered ( mercury , venus , earth , mars ) , the fraction is @xmath12 with only mars being smaller than its inner companions . how the planets sizes are ordered and distributed is likely a direct result of the how the planets formed and evolved . for example , mars , being the only terrestrial planet in our solar system that does not follow the sequential planet size distribution , may be a direct result of the formation and migration of jupiter , inwards and then back outwards leaving a truncated and depleted inner disk out of which mars was formed @xcite . @xcite noted , for adjacent planets within the @xmath0 multiple - candidate systems , that the distribution of radii ratios for adjacent planets is symmetrically centered around unity , but hints of a planetary size hierarchy can be seen in multi - planet systems discovered by kepler . @xmath0 - 11 , which has 6 transiting planets @xcite with the smallest planets residing preferentially inside the orbits of larger planets . @xmath0 - 11 displays an anti - correlation between the mean density of the planets and the semi - major axis distance from the host star ; i.e. , the larger and lower density planets are located further out than the smaller and denser planets , possibly indicative of the formation and/or evolution of the planetary system @xcite . @xmath0 - 47 , the only known circumbinary multiple planet system , also displays a size hierarchy with the inner planet ( @xmath13 ) being smaller than the outer planet ( @xmath14 ; * ? ? ? yet , in @xmath0 - 20 ( a 5-planet system ) , the relative sizes of the planets do not appear to correlate with the orbital periods of the planets @xcite . but these are only three systems . is there an overall correlation of planetary size with orbital period ? here we explore the _ relative _ sizes of the planetary radii for all planet pairs within the multiple candidate systems discovered thus far by @xmath0 @xcite . in this work , we refer to the kepler candidates as `` planets '' though the majority have not been formally validated or confirmed as planets ; as discussed above , candidates in multiple candidate systems are statistically more likely to be true planetary systems @xcite . we seek to characterize the planetary size hierarchy , as a function of the number of planets detected in the systems and the properties of the stellar hosts , and explore if the size hierarchy seen within the inner solar system also occurs in the @xmath0 multiple - candidate sample . the sample used here is based upon the 2012 koi candidate list published by @xcite ; a full description of the koi list , the vetting that list underwent , and the characteristics of the sample set as a whole are described in the catalog paper . there are 1425 systems with a single candidate , 245 systems with two planet candidates , 84 systems with three planet candidates , 27 systems with four planet candidates , 8 systems with five planet candidates and one system with 6 planet candidates . here we wish to investigate the overall distribution of planet sizes as a function of orbit ; that is , do outer planets , in general , tend to be larger than inner planets ? a smaller planet ( e.g. , shallower transit ) will be more easily detected if the period is short , simply from the fact that the number of observed transiting events increases with shorter orbital period . in a complementary manner , the larger planets are more easily detected at all periods , up to some period threshold where only @xmath15 transits are detected . @xcite tabulate for each candidate the transit period , the transit duration , the impact parameter , and the signal - to - noise of the transit fits . the signal - to - noise ratios , coupled with the transit periods and the impact parameters , enables us to debias the sample for these observational detection efficiencies ( see also * ? ? ? the multiple candidate systems were identified primarily by searching the single candidate systems specifically for additional transiting planets . thus , there could potentially be a detection bias of which we are unaware that is not found in the single candidate systems . however , no substantial difference between single planet systems and the multiple planet systems were found , except for the lack of hot jupiters in the multiple planet systems @xcite . the overall size distribution of the planets and the signal - to - noise detection thresholds appear to be similar between the single candidate and multiple candidate systems . to understand better the detection thresholds for the multiple candidate systems , we have used the distribution of the total transit signal - to - noise to estimate where the the distribution of planets in multiple systems appears to be complete ( figure [ snr - fig ] ) . assuming the signal - to - noise distribution can be characterized with an exponential function where the distribution is complete , we have used the point in the signal - to - noise distribution function where the exponential no longer adequately describes the distribution as the fiducial for completeness . fitting the exponential , we find a turnover in detected samples at a @xmath16 . we use this snr threshold to debias the radii - ratio distribution of planets in multiple - candidate systems . the total signal - to - noise of all the detected transits for each planetary candidate is predicted for all the observed orbital periods within a system . a planet - pair is retained in the analysis only if the predicted total transit signal - to - noise ratios for both planets at the orbital period of the other planet exceeds the signal - to - noise threshold of @xmath17 . the predicted signal - to - noise ratio is determined by scaling the measured total signal - to - noise ratio of the planet by the ratio of the orbital periods . assuming all else is equal , the total signal - to - noise of all the detected transits for a given planet scales with the orbital period as @xmath18 . for a given planet , the total signal - to - noise of all detected transits is proportional to the total number of points ( @xmath19 ) detected in all of the transits ( e.g. , * ? ? ? * ) @xmath20 where @xmath21 is the total number of individual transits detected , and @xmath22 is the number of points detected within a single transit . the number of transits detected is inversely proportional to the orbital period @xmath23 , and the number of points detected per transit is proportional to the transit duration @xmath24 ; thus , the total signal to noise of the detected transits can be parameterized as @xmath25 the transit duration ( @xmath26 ) is proportional to the cube root of the orbital period ( @xmath27 ) ; thus , the total signal - to - noise ratio of the detected transits scales with the orbital period as @xmath28 an additional restriction on the sample was made such that no planet candidate was included that has an impact parameter of @xmath29 . at such high impact parameters , the transit parameters , particularly the transit depth ( i.e. , the planet radius ) , are less certain . using the @xmath17 detection threshold and the impact parameter restrictions , there are 96 multiple - candidate systems with 159 pairs of planets ( 228 individual planets ) in the analysis . all of the planet - pairs used in the analysis of this paper are summarized in table [ sum - tab ] ; planets are labeled with roman numerals ( i , ii , iii , iv , v , vi ) in the order of increasing orbital period . these letters do not necessarily correspond to koi fraction numbers ( e.g. , .01 , .02 , ... ) nor do they correspond to the confirmed planet letters ( e.g. , @xmath0 - 11b , @xmath0 - 11c , ... ) . after the signal - to - noise and impact parameter cuts , the largest planet in the sample is @xmath30 , and the median planet radius is @xmath31 ; the smallest planet retained in the sample has a radius of @xmath32 . the orbital periods of the planets in this sample span 0.45 days to 331 days , with an median period of @xmath33 days . the distributions of the radii and orbit periods for the 228 planets retained in the sample are shown in figure [ planets - fig ] . previous work indicates that the detected planets in the @xmath0 multiple - candidate systems have mutual inclinations of @xmath34 @xcite . due to the usual limitation of the transit technique in only detecting planets that are very nearly edge - on , the kepler sample used here is naturally biased to systems of nearly coplanar planets . the prevalence of kepler multiple - candidate systems shows that there is a large population of such systems with small planets and orbital periods of tens of days ( figure [ planets - fig ] ) . it may be that other system architectures with higher mutual inclinations or different period ranges do not show the same trend in planet sizes that we describe herein . we have calculated the ratios of the inner planet radius to the outer planet radius for each unique pair of planets within a system ( @xmath35 ) , and the cumulative fraction distribution is displayed in figure [ ratiocfrac - fig ] . if there were no preference for the ordering of planet sizes , the chance that a given planet - radii ratio is larger than unity would be equal to the chance that the ratio is below unity , and the cumulative fraction distribution would pass through 50% at @xmath36 . in contrast , @xmath37% of the planet pairs are ordered such that the outer planet is larger than the inner planet ( @xmath38 ) . the resulting fraction deviates from the null - hypothesis expectation value of 50% by @xmath39 . the @xmath40 upper and lower confidence intervals are based upon the clopper - pearson binomial distribution confidence interval @xcite . the clopper - pearson interval is a two - sided confidence interval and is based directly on the binomial distribution rather than an approximation to the binomial distribution . we do caution that in applying the clopper - pearson confidence interval there is an implicit assumption that all of elements of the sample are uncorrelated . given that not all planet - pairs are from independent stellar systems , there may be a correlation between individual planet - pairs within a system ( i.e. , planet i is smaller than planet iii because it is smaller than planet ii ) , and the clopper - pearson confidence intervals may underestimate the true uncertainties in the fractions . in addition to the confidence intervals , the significance of the observed fraction in comparison to the null hypothesis can be evaluated through the @xmath41 statistic . the probability of the observed fraction ( @xmath42 ) when compared to the expected fraction ( @xmath43 ) yields @xmath41 = 6.0 with a probability of only 1.4% that the observed fraction is observed only by chance ( see table [ frac - tab ] ) . we also have performed three additional observational tests to assess if the observed fraction may be the result of an unrecognized bias in the sample . the first test repeated the above analysis , but for each unique planet pair , the measured radii were replaced with radii randomly drawn from the overall radius distribution as measured by kepler ; the same set of periods within a given system were retained . the same impact parameter and signal - to - noise restrictions described above ( [ sample - sec ] ) were applied ( e.g. , if a planet radius was drawn that was too small to be detected at the orbital period ( @xmath44 ) , a new radius was randomly drawn until the @xmath45 threshold was met ) . the random draw was performed 10,000 times for each unique pair of planets , and the cumulative distribution of the planet - radii ratios for all of the random draws is displayed in figure [ ratiocfrac - fig ] . as expected , the random draw distribution displays no size ordering preference ; i.e. , the fraction of planets with @xmath38 is @xmath46 . based upon a kolmogorov - smirnov test , the observed distribution and the random draw distribution are not drawn from different parent distributions with only a probability of @xmath47 indicating that the observed planet radii ratio distribution has a preferential ordering such that smaller planets are in orbits interior to larger planets . we also tested if the observed fraction of planet - radii ratios with @xmath38 is dependent upon the signal - to - noise ratio threshold chosen . in figure [ snrratio - fig ] , the fraction of planet pairs where the inner planet is smaller than the outer planet is plotted as a function of the required signal - to - noise threshold . if no signal - to - noise threshold is required , the fraction is @xmath48 , and as the signal - to - noise threshold is increased , the fraction systematically decreases and levels out near @xmath49 . the higher fractions at a lower signal - to - noise threshold result from incompleteness of the sample ( i.e. , it is easier to detect larger planets at all orbital periods ) . as the signal - to - noise threshold is increased the fraction of planet - radii ratios below unity decreases , but does not systematically approach 50% , as would be expected if there was no preferential size ordering of the planets . rather , the fraction asymptotically approaches 60% for @xmath50 , indicating the @xmath51 threshold ensures that the analysis presented in fig . [ ratiocfrac - fig ] is based upon a sample not significantly biased by completeness . the final test performed was to determine if the observed fraction is dependent upon the maximum period of the outer planet ( see figure [ ratioperiod - fig ] ) . transit surveys are typically more complete at shorter orbit periods , and if there truly was no preference for planet size ordering , the fraction would decline and approach 50% as the maximum orbit period was decreased . within the confidence intervals , the observed fractions are independent of the maximum outer orbital period and are consistent with the observed @xmath52 fraction for the entire sample . the fractions do not approach 50% , at shorter orbital period as would be expected for a random distribution . in fact , a slight ( but not statistically significant ) hint of a higher fraction is observed if the maximum period is @xmath53 days . overall , if the ordering of planets within a given system was random such that there was no sequential ordering of the planets by size , the probabilities of a planet - radii ratio being above or below unity would be equal . but that is not what is observed ; the three above tests ( i.e. , the random radius draw , the @xmath54 , and the maximum outer orbital period ) , in combination with the @xmath41 probability statistic and the confidence intervals , indicate that for @xmath52 ( @xmath55 ) of the unique planet pairs within the kepler multiple planet systems , the inner planet is smaller than the outer planet , with only a 1.4% chance of this being observed by chance . in the following subsections , we explore if and how the number of planets in the system , the orbital separation of the planets , the temperature of the stars , and the size of the planets themselves may affect the observed size hierarchy of the planets in multiple - planet systems . if formation and evolutionary mechanisms within a planetary system affect the size distribution and ordering of planets within a system , one might expect to observe differences in the size ordering as a function of the number of planets within a system . when the planets are divided into groups based upon the number of detected planetary candidates in the system ( e.g. , 2- , 3- , 4- , 5- , and 6-candidates ) , the fractions do not change appreciably from the fractions calculated for the entire sample set . this can be seen in the both @xmath41 statistic ( table [ frac - tab ] ) where the @xmath41 and associated probabilities are all comparable to each other and in figure [ fracnum - fig ] , where the observed fractions ( with confidence intervals ) are plotted and display no dependence on the number of planets in the system . while the number of systems with more than 3 planets have relatively large uncertainties and poor statistics because of the small numbers of the 4- , 5- , and 6-candidate systems , there is no correlation of the fraction of size - ordered planets with the number of planets within a planetary system . it might be expected that planets nearer to the central star would undergo a higher level of photo - evaporation than planets at longer orbital periods and , thus , the relative sizes of the planets would be more extreme if the planets are more widely separated ( larger orbital period ratio ) . to test this , we have plotted the planet - radii ratios vs. the orbital period ratios and the inner and outer planet orbital periods ( fig . [ periodratio - fig ] ) . using the spearman non - parametric rank correlation function , we find that the distributions are likely uncorrelated . the correlation coefficients for the three distributions shown in fig . [ periodratio - fig ] are 0.14 , 0.22 , and 0.10 with probabilities to be exceeded in the null hypothesis of 0.06 , 0.004 and 0.08 , respectively . additionally , to search for a non - zero slope that might indicate the planet - radii ratios are related to or dependent upon the period or period spacings , we have fitted a linear model to each of the distributions . each of the distributions are statistically consistent with a flat distribution as a function of period ratio and orbital period ; the slopes of the fitted linear models were found to be @xmath56 for the radii ratios vs. period ratios ( fig . [ periodratio - fig ] top ) , @xmath57 for the radii ratios vs. the inner orbital period ( fig . [ periodratio - fig ] middle ) , and @xmath58 for the radii ratios vs. the outer orbital period ( fig . [ periodratio - fig ] bottom ) . in general , we find no correlation between the orbital period separation ( or orbital periods themselves ) and the size - ordering of the planets . we have also explored whether there is a correlation between the distribution of planet - radii ratios and the effective temperature of the host stars . we have produced planet - radii ratio distributions , but separated out by the stellar temperature . we chose temperature ranges that roughly correspond to the spectral classification of m and k - stars ( @xmath59 k ) , g - stars ( @xmath60 k ) , and f - stars ( @xmath61 k ; * ? ? ? * ) . the cumulative distributions of the planet - radii ratios for each of the stellar temperature groups are displayed in figure [ radiiteff - fig ] . overall , all three distributions are shifted to lower ratios in comparison to the random draw distribution from fig [ ratiocfrac - fig ] . kolmogorov - smirnov tests between the three distributions indicate that the three distributions do not result from different parent distributions with probabilities of @xmath62 - indicating that , in general , the preferential planet size ordering occurs at approximately the same level ( i.e. , @xmath63 the unique planet pairs have @xmath64 ) across the stellar temperature range . a summary of the planet - radii ratios , the fraction of ratios that exhibit smaller inner planets , and the significance of those fractions based upon the confidence intervals and @xmath41 statistic is given in table [ fracstellar - tab ] for each of the stellar temperature groups . there does appear to be an slight trend such that as the stars become warmer they contain a larger fraction of planet - pairs sequentially ordered by radius . for the cooler stars ( @xmath59 k ) , this fraction is @xmath65 , while for the warmer stars ( @xmath61 k ) , this fraction is @xmath66% . the resulting @xmath41 statistics for these three samples indicates that the observed fractions differ from the null hypothesis fraction of 50% more significantly for the warmer stars than for the cooler stars ( table [ fracstellar - tab ] ) . this trend could be a result of higher planetary evaporation rates associated with the higher luminosity stars or perhaps could result from the warmer ( i.e. , more massive ) stars tending to contain larger ( i.e. , more massive ) planets ( see figure [ radiiteff - fig ] and table [ fracstellar - tab ] ) . the median planet radius does not vary significantly between the three stellar groups ( @xmath67 ; see table [ fracstellar - tab ] ) , but the size of the largest planets in each group does . the g and f stars ( @xmath68 k ) contain saturn - sized and jupiter - sized planets ( in addition to smaller planets ) , while the cooler ( @xmath69 k ) contain neptune - sized planets and smaller ( see figure [ radiiteff - fig ] and table [ fracstellar - tab ] ) . in this sample , 15% of the planets around g - stars , and 25% of the planets around the f - stars , are neptune - sized or larger ; the cooler stars host only 6 ( 4% ) planets larger than neptune , and these four planets orbit three stars with effective temperatures of @xmath70k ( koi 757 , 884 , 941 ) . the absence of large planets around cooler stars is evident in the cumulative distribution of the planet radii ( fig . [ radiiteff - fig ] ) and in the smaller median absolute deviations of the median planet radius for each stellar group ( table [ fracstellar - tab ] ) . this paper is not intended to provide a discussion of the occurrence rates of planets or a detailed analysis of the size distribution of planets ( e.g. , * ? ? ? * ) , but this may imply that the cooler ( i.e. , smaller ) stars only produce smaller planets that do not preferentially follow a size hierarchy of planets . in fact , large planets around small stars may not form at all @xcite or , if they do , they may not survive their youth ( e.g. , * ? ? ? * ) . thus , cool stars may only be left with a population of relatively small planets @xcite . the warmer stars , in contrast , appear to host a larger array of planet sizes with a slightly higher preference for the larger planets ( @xmath71neptune - sized ) to be in longer orbits exterior to the orbits of the smaller planets . given that there appears to be only a weak correlation with the planet - radii ratio and the orbital period ( ratio ) , perhaps the formation and migration of large planets ( perhaps in conjunction with photo - evaporation of the innermost planets ) is necessary for a radius hierarchy to present . if hotter stars tend to show a planet radius hierarchy and the hotter stars also host relatively larger planets , then the planet hierarchy might be expected to be correlated with the planet size . in figure [ largeratio - fig ] , we compare the sizes of the planet radii for the inner and outer planets for each pair of planets within the sample . the overall planet - radii ratio is near unity but there is more scatter below than above the unity line ( i.e. , smaller planets are interior to larger planets ) . there are 36 planet - pairs where the outer planet is @xmath72 larger than the inner planet ( @xmath73 ) ; in contrast , there are only 13 planet - pairs where the inner planet is @xmath72 larger than the outer planet ( @xmath74 . for the planet - pairs that are in these relatively extreme ratio pairs , we have plotted the radii distributions of the candidates , separated out by inner and outer planet and by the sense of the ratio ( see figure [ largeratiohist - fig ] ) . for the pairs where the outer planet is significantly larger , the ratio tends to be dominated by relatively large planets . half ( 18/36 ) of the outer planets are neptune - sized or larger ( @xmath75 ) , while only 25% ( 9/36 ) of the inner planets are very small ( e.g. , earth - sized @xmath76 ) planets . in contrast , for those planet - pairs where the inner planet is much larger than the outer planet , the pairs are evenly split between relatively small outer planets ( 6/13 are @xmath76 ) and relatively large inner planets ( 5/13 @xmath75 ) . it appears that for there to be a planet radius hierarchy , one of the planets most often needs to be @xmath77neptune - sized or larger . to explore this more closely , we have calculated the fraction of planet - pairs with @xmath78 , if at least one planet is larger than some maximum planet radius ( @xmath79 ) , or if both planets are smaller than that same maximum planet radius ( @xmath79 ) , or if both planets are larger than that same maximum planet radius . the fractions were calculated for maximum planet radii of @xmath80r_\oplus$ ] ( see figure [ fracrad - fig ] ) and are tabulated in table [ fracrp - tab ] . if planets - pairs with planets larger than @xmath81 are excluded , the fraction is very near 50% with little preference for sequential planet ordering , but if one planet in the pair is larger than @xmath82 , the observed fraction of planet - pairs with @xmath78 is @xmath83 a @xmath84 separation from the random ordering fraction of 50% ( see table [ fracrp - tab ] ) . based upon the @xmath41 statistic and probability and the confidence intervals , the fraction of planet pairs is significantly above 50% only when neptune - sized planets or larger are allowed in the pairs . the fraction is most significant when neptune - sized or larger planets are compared to planets of all sizes ( top panel fig [ fracrad - fig ] ) ; the @xmath41 statistic are all @xmath85 with a @xmath86 probability that the non@xmath87 fractions are achieved solely by chance . when the sizes of both planets are restricted to radii smaller than neptune , the fractions remain near or below 50% ( middle panel fig [ fracrad - fig ] and table [ fracrp - tab ] ) , with a @xmath88 probability of the observed fractions being generated by chance for the null - hypothesis distribution . when both planets within a planet pair are restricted to planets larger than a certain radius , sequential ordering of the planet sizes is still apparent , but is less significant ( bottom panel fig [ fracrad - fig ] and table [ fracrp - tab ] ) . if both planets in a pair are @xmath89 , then the observed fraction of planet - pairs where the inner planet is smaller than the outer planet are consistent with no planet size hierarchy . these results suggest that the size ordering primarily occurs when a system contains both superearth - sized ( or earth - sized ) planets _ and _ neptune - sized or larger planets , and may be a direct result shepherding of smaller inner planets by larger outer planets @xcite . we have performed a study of the relative sizes of planets within the multiple candidate systems using the @xmath0 planetary candidate list @xcite . for each multiple planet system , we have compared the radius of each planet to the radius of every other planet within a planetary system , after correcting for detection biases . we find that for planet - pairs for which one or both objects is approximately neptune - sized for larger , the larger planet is most often the planet with the longer period , while no such size location correlation is seen for pairs of planets when both planets are smaller than neptune . overall , for all planets - pairs in the sample , @xmath90 of the unique planet pairs are structured in a hierarchical manner such that the inner planet is smaller than the outer planet . if at least one planet in the planet - pair is neptune - sized or larger ( @xmath91 ) , the fraction of inner planets being smaller than outer planets is @xmath92 . however , if both planets are smaller than neptune , then the fraction is consistent with random planet ordering ( @xmath93% ) with no apparent size hierarchy . the planet radius size hierarchy may be a natural consequence of planetary formation and evolution . in particular , the sequential ordering may be a result of a combination of core accretion , migration , and evolution . core accretion models predict that smaller planets are expected to form prior to and interior to the giant planets @xcite . additionally , larger planets formed further out may migrate inwards , shepherding smaller planets inward as the planets move @xcite . this scenario is consistent with the planet size hierarchy being observed for planets pairs involving neptune - sized planets , but being absent for small planet pairs and for the planets around cool stars . for the cooler stars , the forming planets may begin migrating prior to starting rapid gas accretion ( e.g. , * ? ? ? * ) , coupled with a lower extreme ultraviolet luminosity that is not capable of substantially evaporating the planets @xcite . for example , @xcite suggest that the @xmath0 - 11 planets did not form _ in situ _ , but rather , the planets formed beyond the snow line , migrated inwards , and were evaporated by the star to their present sizes . these scenarios ( perhaps , all in concert ) predict a radius hierarchy of planets as a function of orbital distance from the central host star , in particular , predicting that neptune - sized and larger planets are outside super - earth and terrestrial - sized planets . as @xmath0 discovers more planets in longer orbital periods , it will be interesting to learn if our solar system is indeed typical of the planetary architectures found in the galaxy . @xmath0 was competitively selected as the 10th nasa discovery mission . the authors thank the many people who have made @xmath0 such a success . this paper includes data collected by the @xmath0 mission ; funding for the @xmath0 mission is provided by the nasa science mission directorate.this research has made use of the nasa exoplanet archive , which is operated by the california institute of technology , under contract with the national aeronautics and space administration under the exoplanet exploration program . d.c.f . acknowledges nasa support through hubble fellowship grant hf-51272.01-a , awarded by stsci , operated by aura under contract nas 5 - 26555 . drc would like to thank the referee , the @xmath0 team , bill borucki , geoff marcy , stephen kane , peter plavchan , kaspar von braun , teresa ciardi , and jim grubbs for very insightful and inspirational comments and discussions in the formation of this paper . ccc|cccc|cccc|c + k00072 & 5627 & i / ii & 1.38 & 0.837 & 139 & 0.17 & 2.19 & 45.294 & 122 & 0.11 & 0.630 + k00108 & 5975 & i / ii & 2.94 & 15.965 & 84 & 0.79 & 4.45 & 179.600 & 111 & 0.62 & 0.661 + k00119 & 5380 & i / ii & 3.76 & 49.184 & 152 & 0.55 & 3.30 & 190.310 & 63 & 0.69 & 1.139 + k00123 & 5871 & i / ii & 2.64 & 6.482 & 98 & 0.56 & 2.71 & 21.222 & 78 & 0.01 & 0.974 + k00150 & 5538 & i / ii & 2.63 & 8.409 & 106 & 0.73 & 2.73 & 28.574 & 81 & 0.76 & 0.963 + k00209 & 6221 & i / ii & 5.44 & 18.795 & 217 & 0.51 & 8.29 & 50.790 & 287 & 0.40 & 0.656 + k00222 & 4353 & i / ii & 2.03 & 6.312 & 85 & 0.71 & 1.66 & 12.794 & 43 & 0.72 & 1.223 + k00223 & 5128 & i / ii & 2.52 & 3.177 & 90 & 0.78 & 2.27 & 41.008 & 32 & 0.69 & 1.110 + k00271 & 6169 & i / ii & 2.48 & 29.392 & 72 & 0.71 & 2.60 & 48.630 & 54 & 0.79 & 0.954 + k00275 & 5795 & i / ii & 1.95 & 15.791 & 55 & 0.03 & 2.04 & 82.199 & 27 & 0.60 & 0.956 + k00312 & 6158 & i / ii & 1.91 & 11.578 & 41 & 0.70 & 1.84 & 16.399 & 40 & 0.58 & 1.038 + k00313 & 5348 & i / ii & 1.61 & 8.436 & 54 & 0.32 & 2.20 & 18.735 & 55 & 0.73 & 0.732 + k00386 & 5969 & i / ii & 3.25 & 31.158 & 57 & 0.02 & 2.94 & 76.732 & 28 & 0.55 & 1.105 + k00413 & 5236 & i / ii & 2.75 & 15.228 & 52 & 0.67 & 2.10 & 24.674 & 26 & 0.65 & 1.310 + k00431 & 5249 & i / ii & 2.80 & 18.870 & 60 & 0.60 & 2.48 & 46.901 & 34 & 0.68 & 1.129 + k00433 & 5237 & i / ii & 5.60 & 4.030 & 127 & 0.65 & 12.90 & 328.240 & 158 & 0.77 & 0.434 + k00446 & 4492 & i / ii & 1.76 & 16.709 & 33 & 0.62 & 1.72 & 28.551 & 34 & 0.05 & 1.023 + k00448 & 4264 & i / ii & 1.77 & 10.139 & 56 & 0.74 & 2.31 & 43.608 & 66 & 0.66 & 0.766 + k00464 & 5362 & i / ii & 2.63 & 5.350 & 75 & 0.58 & 6.73 & 58.362 & 263 & 0.20 & 0.391 + k00475 & 5056 & i / ii & 2.04 & 8.181 & 37 & 0.58 & 2.26 & 15.313 & 34 & 0.68 & 0.903 + k00509 & 5437 & i / ii & 2.24 & 4.167 & 45 & 0.09 & 2.68 & 11.463 & 32 & 0.74 & 0.836 + k00518 & 4565 & i / ii & 2.11 & 13.981 & 68 & 0.71 & 1.54 & 44.000 & 33 & 0.49 & 1.370 + k00638 & 5722 & i / ii & 3.60 & 23.636 & 53 & 0.02 & 3.78 & 67.093 & 39 & 0.27 & 0.952 + k00657 & 4632 & i / ii & 1.63 & 4.069 & 43 & 0.63 & 2.08 & 16.282 & 46 & 0.74 & 0.784 + k00672 & 5524 & i / ii & 2.60 & 16.087 & 43 & 0.71 & 3.15 & 41.749 & 70 & 0.11 & 0.825 + k00676 & 4367 & i / ii & 2.56 & 2.453 & 255 & 0.70 & 3.30 & 7.972 & 348 & 0.57 & 0.776 + k00693 & 6121 & i / ii & 1.87 & 15.660 & 48 & 0.50 & 1.76 & 28.779 & 23 & 0.64 & 1.062 + k00708 & 6036 & i / ii & 1.82 & 7.693 & 43 & 0.74 & 2.54 & 17.406 & 68 & 0.70 & 0.717 + k00800 & 5938 & i / ii & 3.27 & 2.711 & 48 & 0.76 & 3.39 & 7.212 & 36 & 0.77 & 0.965 + k00841 & 5399 & i / ii & 5.44 & 15.334 & 83 & 0.74 & 7.05 & 31.331 & 117 & 0.74 & 0.772 + k00842 & 4497 & i / ii & 2.04 & 12.718 & 43 & 0.31 & 2.46 & 36.065 & 44 & 0.42 & 0.829 + k00853 & 4842 & i / ii & 2.38 & 8.204 & 48 & 0.12 & 1.78 & 14.496 & 22 & 0.02 & 1.337 + k00870 & 4590 & i / ii & 2.52 & 5.912 & 56 & 0.76 & 2.35 & 8.986 & 47 & 0.73 & 1.072 + k00877 & 4500 & i / ii & 2.42 & 5.955 & 54 & 0.73 & 2.37 & 12.039 & 36 & 0.79 & 1.021 + k00896 & 5190 & i / ii & 3.22 & 6.308 & 68 & 0.52 & 4.52 & 16.239 & 91 & 0.57 & 0.712 + k00936 & 3684 & i / ii & 1.54 & 0.893 & 72 & 0.75 & 2.69 & 9.468 & 96 & 0.76 & 0.572 + k00951 & 4767 & i / ii & 3.74 & 13.197 & 101 & 0.39 & 2.87 & 33.653 & 36 & 0.58 & 1.303 + k00988 & 5218 & i / ii & 2.20 & 10.381 & 67 & 0.02 & 2.17 & 24.570 & 37 & 0.67 & 1.014 + k01236 & 6562 & i / ii & 2.04 & 12.309 & 40 & 0.01 & 3.02 & 35.743 & 64 & 0.37 & 0.675 + k01270 & 5145 & i / ii & 2.19 & 5.729 & 45 & 0.58 & 1.55 & 11.609 & 22 & 0.01 & 1.413 + k01781 & 4977 & i / ii & 1.94 & 3.005 & 81 & 0.19 & 3.29 & 7.834 & 172 & 0.01 & 0.590 + k01824 & 5978 & i / ii & 1.75 & 1.678 & 53 & 0.70 & 1.97 & 3.554 & 50 & 0.69 & 0.888 + + k00085 & 6172 & i / ii & 1.27 & 2.155 & 72 & 0.71 & 2.35 & 5.860 & 154 & 0.77 & 0.540 + k00085 & 6172 & i / iii & 1.27 & 2.155 & 72 & 0.71 & 1.41 & 8.131 & 53 & 0.78 & 0.901 + k00085 & 6172 & ii / iii & 2.35 & 5.860 & 154 & 0.77 & 1.41 & 8.131 & 53 & 0.78 & 1.667 + k00111 & 5711 & i / ii & 2.14 & 11.427 & 96 & 0.53 & 2.05 & 23.668 & 71 & 0.55 & 1.044 + k00111 & 5711 & i / iii & 2.14 & 11.427 & 96 & 0.53 & 2.36 & 51.756 & 75 & 0.57 & 0.907 + k00111 & 5711 & ii / iii & 2.05 & 23.668 & 71 & 0.55 & 2.36 & 51.756 & 75 & 0.57 & 0.869 + k00115 & 6202 & ii / iii & 3.33 & 5.412 & 144 & 0.43 & 1.88 & 7.126 & 41 & 0.55 & 1.771 + k00137 & 5385 & i / ii & 1.82 & 3.505 & 44 & 0.71 & 4.75 & 7.641 & 299 & 0.50 & 0.383 + k00137 & 5385 & i / iii & 1.82 & 3.505 & 44 & 0.71 & 6.00 & 14.858 & 305 & 0.72 & 0.303 + k00137 & 5385 & ii / iii & 4.75 & 7.641 & 299 & 0.50 & 6.00 & 14.858 & 305 & 0.72 & 0.792 + k00152 & 6187 & i / ii & 2.59 & 13.484 & 58 & 0.34 & 2.77 & 27.402 & 57 & 0.07 & 0.935 + k00152 & 6187 & i / iii & 2.59 & 13.484 & 58 & 0.34 & 5.36 & 52.091 & 158 & 0.00 & 0.483 + k00152 & 6187 & ii / iii & 2.77 & 27.402 & 57 & 0.07 & 5.36 & 52.091 & 158 & 0.00 & 0.517 + k00156 & 4619 & i / ii & 1.18 & 5.188 & 38 & 0.53 & 1.60 & 8.041 & 54 & 0.64 & 0.738 + k00156 & 4619 & i / iii & 1.18 & 5.188 & 38 & 0.53 & 2.53 & 11.776 & 131 & 0.69 & 0.466 + k00156 & 4619 & ii / iii & 1.60 & 8.041 & 54 & 0.64 & 2.53 & 11.776 & 131 & 0.69 & 0.632 + k00284 & 5925 & i / ii & 1.09 & 6.178 & 32 & 0.38 & 1.14 & 6.415 & 35 & 0.19 & 0.956 + k00343 & 5744 & i / ii & 1.86 & 2.024 & 72 & 0.63 & 2.68 & 4.762 & 108 & 0.63 & 0.694 + k00343 & 5744 & i / iii & 1.86 & 2.024 & 72 & 0.63 & 1.58 & 41.809 & 20 & 0.53 & 1.177 + k00343 & 5744 & ii / iii & 2.68 & 4.762 & 108 & 0.63 & 1.58 & 41.809 & 20 & 0.53 & 1.696 + k00351 & 6103 & ii / iii & 6.62 & 210.590 & 144 & 0.22 & 9.32 & 331.640 & 211 & 0.19 & 0.710 + k00377 & 5777 & ii / iii & 8.28 & 19.273 & 136 & 0.35 & 8.21 & 38.907 & 81 & 0.62 & 1.009 + k00398 & 5101 & i / ii & 1.76 & 1.729 & 37 & 0.24 & 3.33 & 4.180 & 102 & 0.00 & 0.529 + k00398 & 5101 & ii / iii & 3.33 & 4.180 & 102 & 0.00 & 8.66 & 51.846 & 215 & 0.66 & 0.385 + k00408 & 5631 & i / ii & 3.72 & 7.382 & 96 & 0.79 & 2.91 & 12.560 & 52 & 0.77 & 1.278 + k00408 & 5631 & i / iii & 3.72 & 7.382 & 96 & 0.79 & 2.70 & 30.827 & 34 & 0.79 & 1.378 + k00408 & 5631 & ii / iii & 2.91 & 12.560 & 52 & 0.77 & 2.70 & 30.827 & 34 & 0.79 & 1.078 + k00481 & 5227 & i / ii & 1.54 & 1.554 & 43 & 0.71 & 2.37 & 7.650 & 56 & 0.76 & 0.650 + k00481 & 5227 & ii / iii & 2.37 & 7.650 & 56 & 0.76 & 2.44 & 34.260 & 39 & 0.73 & 0.971 + k00528 & 5448 & i / iii & 2.62 & 9.577 & 55 & 0.48 & 3.27 & 96.671 & 32 & 0.71 & 0.801 + k00620 & 5803 & i / iii & 7.05 & 45.155 & 132 & 0.03 & 9.68 & 130.180 & 279 & 0.06 & 0.728 + k00658 & 5676 & i / ii & 2.03 & 3.163 & 68 & 0.53 & 2.02 & 5.371 & 52 & 0.65 & 1.005 + k00658 & 5676 & i / iii & 2.03 & 3.163 & 68 & 0.53 & 1.14 & 11.329 & 17 & 0.01 & 1.781 + k00665 & 5864 & i / ii & 1.16 & 1.612 & 34 & 0.69 & 1.09 & 3.072 & 27 & 0.64 & 1.064 + k00701 & 4807 & i / ii & 1.27 & 5.715 & 55 & 0.46 & 1.95 & 18.164 & 71 & 0.66 & 0.651 + k00701 & 4807 & ii / iii & 1.95 & 18.164 & 71 & 0.66 & 1.57 & 122.390 & 34 & 0.29 & 1.242 + k00711 & 5502 & ii / iii & 3.18 & 44.699 & 51 & 0.61 & 2.83 & 124.520 & 35 & 0.49 & 1.124 + k00718 & 5788 & i / ii & 2.57 & 4.585 & 67 & 0.34 & 3.06 & 22.714 & 58 & 0.29 & 0.840 + k00718 & 5788 & i / iii & 2.57 & 4.585 & 67 & 0.34 & 2.67 & 47.904 & 27 & 0.68 & 0.963 + k00718 & 5788 & ii / iii & 3.06 & 22.714 & 58 & 0.29 & 2.67 & 47.904 & 27 & 0.68 & 1.146 + k00723 & 5244 & i / iii & 3.26 & 3.937 & 72 & 0.79 & 3.61 & 28.082 & 60 & 0.58 & 0.903 + k00757 & 4956 & i / ii & 2.09 & 6.253 & 34 & 0.00 & 4.73 & 16.068 & 120 & 0.25 & 0.442 + k00757 & 4956 & ii / iii & 4.73 & 16.068 & 120 & 0.25 & 3.21 & 41.192 & 40 & 0.35 & 1.474 + k00806 & 5461 & ii / iii & 13.24 & 60.322 & 366 & 0.37 & 9.52 & 143.210 & 69 & 0.34 & 1.391 + k00864 & 5337 & i / iii & 2.51 & 4.312 & 63 & 0.14 & 2.23 & 20.050 & 30 & 0.00 & 1.126 + k00884 & 4931 & ii / iii & 4.13 & 9.439 & 143 & 0.48 & 4.23 & 20.476 & 73 & 0.70 & 0.976 + k00898 & 4648 & i / ii & 2.18 & 5.170 & 35 & 0.64 & 2.83 & 9.771 & 47 & 0.68 & 0.770 + k00898 & 4648 & ii / iii & 2.83 & 9.771 & 47 & 0.68 & 2.36 & 20.089 & 28 & 0.59 & 1.199 + k00921 & 5046 & ii / iii & 2.66 & 10.281 & 43 & 0.69 & 3.09 & 18.119 & 46 & 0.70 & 0.861 + k00941 & 4998 & i / ii & 2.37 & 2.383 & 38 & 0.53 & 4.14 & 6.582 & 99 & 0.03 & 0.572 + k00961 & 4188 & i / ii & 2.63 & 0.453 & 100 & 0.69 & 2.86 & 1.214 & 73 & 0.77 & 0.920 + k01576 & 5445 & i / ii & 3.20 & 10.415 & 64 & 0.68 & 2.84 & 13.084 & 50 & 0.65 & 1.127 + k01835 & 5004 & i / ii & 2.69 & 2.248 & 43 & 0.44 & 3.11 & 4.580 & 37 & 0.67 & 0.865 + k01860 & 5708 & ii / iii & 2.44 & 6.319 & 46 & 0.46 & 2.36 & 12.209 & 36 & 0.35 & 1.034 + k01867 & 3892 & i / ii & 1.21 & 2.549 & 37 & 0.40 & 1.08 & 5.212 & 25 & 0.31 & 1.120 + + k00094 & 6217 & i / ii & 1.41 & 3.743 & 35 & 0.16 & 3.43 & 10.423 & 78 & 0.01 & 0.411 + k00094 & 6217 & ii / iii & 3.43 & 10.423 & 78 & 0.01 & 9.25 & 22.343 & 455 & 0.30 & 0.371 + k00094 & 6217 & ii / iv & 3.43 & 10.423 & 78 & 0.01 & 5.48 & 54.319 & 206 & 0.38 & 0.626 + k00094 & 6217 & iii / iv & 9.25 & 22.343 & 455 & 0.30 & 5.48 & 54.319 & 206 & 0.38 & 1.688 + k00191 & 5495 & ii / iii & 2.25 & 2.418 & 53 & 0.55 & 10.67 & 15.358 & 642 & 0.59 & 0.211 + k00191 & 5495 & iii / iv & 10.67 & 15.358 & 642 & 0.59 & 2.22 & 38.651 & 21 & 0.52 & 4.806 + k00245 & 5288 & ii / iii & 0.75 & 21.301 & 49 & 0.76 & 2.00 & 39.792 & 282 & 0.79 & 0.375 + k00571 & 3881 & i / ii & 1.31 & 3.887 & 36 & 0.76 & 1.53 & 7.267 & 36 & 0.76 & 0.856 + k00571 & 3881 & ii / iii & 1.53 & 7.267 & 36 & 0.76 & 1.68 & 13.343 & 36 & 0.79 & 0.911 + k00571 & 3881 & ii / iv & 1.53 & 7.267 & 36 & 0.76 & 1.53 & 22.407 & 27 & 0.78 & 1.000 + k00571 & 3881 & iii / iv & 1.68 & 13.343 & 36 & 0.79 & 1.53 & 22.407 & 27 & 0.78 & 1.098 + k00720 & 5123 & i / ii & 1.41 & 2.796 & 58 & 0.08 & 2.66 & 5.690 & 124 & 0.40 & 0.530 + k00720 & 5123 & i / iii & 1.41 & 2.796 & 58 & 0.08 & 2.53 & 10.041 & 87 & 0.64 & 0.557 + k00720 & 5123 & ii / iii & 2.66 & 5.690 & 124 & 0.40 & 2.53 & 10.041 & 87 & 0.64 & 1.051 + k00733 & 5038 & ii / iii & 2.54 & 5.925 & 60 & 0.14 & 2.21 & 11.349 & 35 & 0.31 & 1.149 + k00812 & 4097 & i / iii & 2.19 & 3.340 & 54 & 0.13 & 2.11 & 20.060 & 28 & 0.35 & 1.038 + k00834 & 5614 & iii / iv & 1.98 & 13.233 & 33 & 0.38 & 5.33 & 23.653 & 154 & 0.38 & 0.371 + k00869 & 5085 & ii / iv & 2.73 & 7.490 & 43 & 0.47 & 3.20 & 36.280 & 34 & 0.42 & 0.853 + k00880 & 5512 & iii / iv & 4.00 & 26.442 & 48 & 0.64 & 5.35 & 51.530 & 109 & 0.02 & 0.748 + k00952 & 3911 & ii / iii & 2.25 & 5.901 & 47 & 0.72 & 2.15 & 8.752 & 35 & 0.77 & 1.047 + k00952 & 3911 & ii / iv & 2.25 & 5.901 & 47 & 0.72 & 2.64 & 22.780 & 38 & 0.77 & 0.852 + k00952 & 3911 & iii / iv & 2.15 & 8.752 & 35 & 0.77 & 2.64 & 22.780 & 38 & 0.77 & 0.814 + k01557 & 4783 & ii / iv & 3.60 & 3.296 & 122 & 0.52 & 3.01 & 9.653 & 70 & 0.30 & 1.196 + k01567 & 5027 & ii / iii & 2.46 & 7.240 & 34 & 0.36 & 2.22 & 17.326 & 23 & 0.00 & 1.108 + k01930 & 5897 & ii / iii & 2.21 & 13.726 & 38 & 0.60 & 2.14 & 24.310 & 31 & 0.62 & 1.033 + k01930 & 5897 & ii / iv & 2.21 & 13.726 & 38 & 0.60 & 2.46 & 44.431 & 29 & 0.66 & 0.898 + k01930 & 5897 & iii / iv & 2.14 & 24.310 & 31 & 0.62 & 2.46 & 44.431 & 29 & 0.66 & 0.870 + + k00070 & 5443 & i / ii & 1.92 & 3.696 & 134 & 0.60 & 0.91 & 6.098 & 23 & 0.66 & 2.110 + k00070 & 5443 & i / iii & 1.92 & 3.696 & 134 & 0.60 & 3.09 & 10.854 & 260 & 0.55 & 0.621 + k00070 & 5443 & i / iv & 1.92 & 3.696 & 134 & 0.60 & 1.02 & 19.577 & 18 & 0.74 & 1.882 + k00070 & 5443 & i / v & 1.92 & 3.696 & 134 & 0.60 & 2.78 & 77.611 & 103 & 0.55 & 0.691 + k00070 & 5443 & iii / v & 3.09 & 10.854 & 260 & 0.55 & 2.78 & 77.611 & 103 & 0.55 & 1.112 + k00082 & 4908 & iii / iv & 1.29 & 10.311 & 79 & 0.71 & 2.45 & 16.145 & 172 & 0.73 & 0.527 + k00082 & 4908 & iii / v & 1.29 & 10.311 & 79 & 0.71 & 1.03 & 27.453 & 28 & 0.79 & 1.252 + k00082 & 4908 & iv / v & 2.45 & 16.145 & 172 & 0.73 & 1.03 & 27.453 & 28 & 0.79 & 2.379 + k00232 & 5868 & i / ii & 1.73 & 5.766 & 41 & 0.40 & 4.31 & 12.465 & 207 & 0.43 & 0.401 + k00232 & 5868 & i / iii & 1.73 & 5.766 & 41 & 0.40 & 1.69 & 21.587 & 21 & 0.67 & 1.024 + k00232 & 5868 & ii / iii & 4.31 & 12.465 & 207 & 0.43 & 1.69 & 21.587 & 21 & 0.67 & 2.550 + k00232 & 5868 & ii / iv & 4.31 & 12.465 & 207 & 0.43 & 1.79 & 37.996 & 21 & 0.56 & 2.408 + k00232 & 5868 & ii / v & 4.31 & 12.465 & 207 & 0.43 & 1.69 & 56.255 & 19 & 0.38 & 2.550 + k00500 & 4613 & iii / iv & 1.63 & 4.645 & 30 & 0.72 & 2.64 & 7.053 & 59 & 0.71 & 0.617 + k00500 & 4613 & iv / v & 2.64 & 7.053 & 59 & 0.71 & 2.79 & 9.522 & 55 & 0.79 & 0.946 + k00707 & 5904 & ii / iii & 3.42 & 13.175 & 37 & 0.77 & 5.69 & 21.775 & 80 & 0.77 & 0.601 + k00707 & 5904 & ii / iv & 3.42 & 13.175 & 37 & 0.77 & 4.26 & 31.784 & 42 & 0.77 & 0.803 + k00707 & 5904 & ii / v & 3.42 & 13.175 & 37 & 0.77 & 4.77 & 41.029 & 51 & 0.77 & 0.717 + k00707 & 5904 & iii / iv & 5.69 & 21.775 & 80 & 0.77 & 4.26 & 31.784 & 42 & 0.77 & 1.336 + k00707 & 5904 & iii / v & 5.69 & 21.775 & 80 & 0.77 & 4.77 & 41.029 & 51 & 0.77 & 1.193 + k00707 & 5904 & iv / v & 4.26 & 31.784 & 42 & 0.77 & 4.77 & 41.029 & 51 & 0.77 & 0.893 + k01589 & 5755 & ii / iii & 2.23 & 8.726 & 30 & 0.72 & 2.36 & 12.882 & 27 & 0.79 & 0.945 + + k00157 & 5685 & i / ii & 1.89 & 10.304 & 38 & 0.34 & 2.92 & 13.024 & 67 & 0.35 & 0.647 + k00157 & 5685 & i / iii & 1.89 & 10.304 & 38 & 0.34 & 3.20 & 22.686 & 73 & 0.33 & 0.591 + k00157 & 5685 & i / iv & 1.89 & 10.304 & 38 & 0.34 & 4.37 & 31.995 & 87 & 0.79 & 0.432 + k00157 & 5685 & ii / iii & 2.92 & 13.024 & 67 & 0.35 & 3.20 & 22.686 & 73 & 0.33 & 0.913 + k00157 & 5685 & ii / iv & 2.92 & 13.024 & 67 & 0.35 & 4.37 & 31.995 & 87 & 0.79 & 0.668 + k00157 & 5685 & ii / v & 2.92 & 13.024 & 67 & 0.35 & 2.60 & 46.687 & 40 & 0.49 & 1.123 + k00157 & 5685 & ii / vi & 2.92 & 13.024 & 67 & 0.35 & 3.43 & 118.360 & 54 & 0.36 & 0.851 + k00157 & 5685 & iii / iv & 3.20 & 22.686 & 73 & 0.33 & 4.37 & 31.995 & 87 & 0.79 & 0.732 + k00157 & 5685 & iii / v & 3.20 & 22.686 & 73 & 0.33 & 2.60 & 46.687 & 40 & 0.49 & 1.231 + k00157 & 5685 & iii / vi & 3.20 & 22.686 & 73 & 0.33 & 3.43 & 118.360 & 54 & 0.36 & 0.933 + k00157 & 5685 & iv / v & 4.37 & 31.995 & 87 & 0.79 & 2.60 & 46.687 & 40 & 0.49 & 1.681 + k00157 & 5685 & iv / vi & 4.37 & 31.995 & 87 & 0.79 & 3.43 & 118.360 & 54 & 0.36 & 1.274 + k00157 & 5685 & v / vi & 2.60 & 46.687 & 40 & 0.49 & 3.43 & 118.360 & 54 & 0.36 & 0.758 lcccccc # of stellar systems & 96 & 42 & 33 & 14 & 6 & 1 + # of @xmath35 pairs & 159 & 42 & 55 & 27 & 22 & 13 + number @xmath94 & 95 & 26 & 33 & 16 & 11 & 9 + number @xmath95 & 64 & 16 & 22 & 11 & 11 & 4 + @xmath41 statistic & 6.0 & 2.4 & 2.2 & 0.93 & 0.0 & 1.9 + @xmath41 probability & 0.014 & 0.13 & 0.14 & 0.34 & 1.0 & 0.16 + fraction @xmath94 & 0.597 & 0.619 & 0.600 & 0.592 & 0.500 & 0.692 + lower @xmath40 confidence & 0.042 & 0.089 & 0.076 & 0.115 & 0.126 & 0.179 + upper @xmath40 confidence & 0.041 & 0.082 & 0.072 & 0.107 & 0.126 & 0.140 lccc # of stellar systems & 22 & 47 & 27 + # of @xmath35 pairs & 40 & 79 & 40 + number @xmath94 & 26 & 47 & 22 + number @xmath95 & 14 & 32 & 18 + @xmath41 statistic & 3.6 & 2.9 & 0.4 + @xmath41 probability & 0.06 & 0.09 & 0.50 + fraction @xmath94 & 0.650 & 0.594 & 0.550 + lower @xmath40 confidence & 0.091 & 0.062 & 0.091 + upper @xmath40 confidence & 0.082 & 0.060 & 0.088 + median planet radius @xmath96 $ ] & 2.59 & 2.63 & 2.19 + med . @xmath96 $ ] & 1.49 & 1.18 & 0.61 + min . planet radius @xmath96 $ ] & 1.09 & 0.75 & 1.03 + max . planet radius @xmath96 $ ] & 9.68 & 13.2 & 4.73 lccccccc + # of stellar systems & 93 & 86 & 63 & 44 & 29 & 23 & 18 + # of @xmath35 pairs & 154 & 138 & 106 & 74 & 49 & 42 & 29 + number @xmath94 & 93 & 87 & 71 & 50 & 32 & 30 & 22 + number @xmath95 & 61 & 51 & 35 & 24 & 17 & 12 & 7 + @xmath41 statistic & 6.7 & 9.4 & 12.2 & 9.1 & 4.6 & 7.7 & 7.8 + @xmath41 probability & 0.009 & 0.002 & 0.0005 & 0.003 & 0.03 & 0.005 & 0.005 + fraction @xmath94 & 0.603 & 0.630 & 0.670 & 0.676 & 0.653 & 0.714 & 0.759 + lower @xmath40 confidence & 0.043 & 0.045 & 0.052 & 0.063 & 0.081 & 0.086 & 0.105 + upper @xmath40 confidence & 0.041 & 0.043 & 0.048 & 0.058 & 0.073 & 0.075 & 0.084 + + # of stellar systems & 5 & 15 & 38 & 58 & 73 & 78 & 82 + # of @xmath35 pairs & 5 & 20 & 53 & 85 & 110 & 117 & 130 + number @xmath94 & 2 & 7 & 24 & 45 & 63 & 65 & 73 + number @xmath95 & 3 & 13 & 29 & 40 & 47 & 52 & 57 + @xmath41 statistic & 0.0 & 1.8 & 0.47 & 0.29 & 2.3 & 1.4 & 1.9 + @xmath41 probability & 1.0 & 0.18 & 0.49 & 0.59 & 0.13 & 0.23 & 0.16 + fraction @xmath94 & 0.400 & 0.350 & 0.453 & 0.529 & 0.572 & 0.555 & 0.561 + lower @xmath40 confidence & 0.253 & 0.088 & 0.076 & 0.060 & 0.052 & 0.050 & 0.048 + upper @xmath40 confidence & 0.302 & 0.164 & 0.078 & 0.059 & 0.051 & 0.049 & 0.046 + + # of stellar systems & 89 & 67 & 42 & 21 & 13 & 11 & 10 + # of @xmath35 pairs & 138 & 98 & 64 & 30 & 15 & 13 & 10 + number @xmath94 & 82 & 59 & 42 & 21 & 10 & 8 & 6 + number @xmath95 & 56 & 39 & 22 & 9 & 5 & 5 & 4 + @xmath41 statistic & 4.9 & 4.1 & 6.2 & 4.8 & 1.7 & 0.69 & 0.4 + @xmath41 probability & 0.03 & 0.04 & 0.01 & 0.03 & 0.20 & 0.40 & 0.50 + fraction @xmath94 & 0.594 & 0.602 & 0.656 & 0.700 & 0.667 & 0.615 & 0.600 + lower @xmath40 confidence & 0.046 & 0.055 & 0.069 & 0.106 & 0.162 & 0.141 & 0.204 + upper @xmath40 confidence & 0.044 & 0.053 & 0.064 & 0.091 & 0.135 & 0.172 & 0.180

collapse of the rotating magnetized molecular cloud core is studied with the axisymmetric magnetohydrodynamical ( mhd ) simulations . due to the change of the equation of state of the interstellar gas , the molecular cloud cores experience several different phases as collapse proceeds . in the isothermal run - away collapse ( ) , a pseudo - disk is formed and it continues to contract till the opaque core is formed at the center . in this disk , a number of mhd fast and slow shock pairs appear running parallelly to the disk . after the equation of state becomes hard , an adiabatic core is formed , which is separated from the isothermal contracting pseudo - disk by the accretion shock front facing radially outwards . by the effect of the magnetic tension , the angular momentum is transferred from the disk mid - plane to the surface . the gas with excess angular momentum near the surface is finally ejected , which explains the molecular bipolar outflow . two types of outflows are observed . when the poloidal magnetic field is strong ( magnetic energy is comparable to the thermal one ) , a u - shaped outflow is formed in which fast moving gas is confined to the wall whose shape looks like a capital letter u. the other is the turbulent outflow in which magnetic field lines and velocity fields are randomly oriented . in this case , turbulent gas moves out almost perpendicularly from the disk . the continuous mass accretion leads to the quasistatic contraction of the first core . a second collapse due to dissociation of h in the first core follows . finally another quasistatic core is again formed by atomic hydrogen ( the second core ) . it is found that another outflow is ejected around the second atomic core , which seems to correspond to the optical jets or the fast neutral winds .

astro-ph0105527

star formation has been a long - standing target in astrophysics . the infrared protostar distribution revealed that the molecular cloud cores , which coincide with relatively high - density part ( @xmath2 ) of the molecular clouds , are the sites of star formation . the observed molecular cloud cores are divided into two categories : those observed associated with and without protostars . the molecular cloud cores without protostars are called starless cores or prestellar cores and are considered younger than the cores associated with protostars ( protostellar cores ) . from a theoretical point of view , clouds or cloud cores experience the isothermal run - away collapse first and then accretion on to the stellar core develops @xcite . prestellar cores which indicates inflow motions ( e.g. in l1544 rotation and infall velocities @xmath3 are observed by @xcite ) do indicate that they are in the dynamically contracting phase , in other words , in the run - away collapse @xcite . after the epoch when the dust thermal emissions are trapped in the central part of the cloud ( @xmath4 ) , an adiabatic core is formed and isothermal gas continues to accrete on to the core . the molecular cloud core in this phase is observed as a protostellar core . it is shown that the dynamical evolution of the cloud core is characterized by the sequence from the prestellar cores to protostellar cores . dynamical collapse of the magnetized clouds are studied by many authors @xcite . rotating clouds collapse has been attacked seriously with numerical simulations @xcite . however , a restricted number of articles are published regarding the dynamical contraction of the cloud with both rotation and magnetic fields @xcite ; for quasistatic evolution see @xcite . these researches are confined to the relatively early prestellar stage . is it sufficient to consider the effects of the rotational motion and the magnetic fields separately ? in the dynamical contraction phase , it is shown that the molecular outflow is driven by the cooperative effect of the magnetic fields and rotation motion @xcite . the toroidal magnetic fields are generated from the poloidal ones by the effect of rotation motions . the magnetic torque works only when the poloidal and toroidal magnetic fields coexist . the magnetic torque and thus magnetic angular momentum transfer along the magnetic field line is important to eject the outflow . since the outflow brings the excess angular momentum , the angular momentum that remain in the adiabatic core and thus new - born star is reduced by a factor from @xmath5 to @xmath6 from that of the parent molecular cloud core @xcite . the outflow have not observed either in the magnetized but non - rotating cloud @xcite or the rotating but non - magnetized cloud @xcite . therefore , rotation and magnetic fields are both essential to the evolution of molecular cloud cores . in the present paper , we present the dynamical contraction of the magnetized and rotating cloud . the cooperative effect of magnetic fields and the rotation motions becomes important after the adiabatic core is formed near the center of the cloud core @xcite . therefore , the evolution throughout from the prestellar to protostellar core should be studied . plan of this paper is as follows : in @xmath72 model and numerical method are described . as the initial condition , we choose a slowly rotating cloud with purely poloidal magnetic fields ( no toroidal magnetic fields ) . and we follow the evolution using magnetohydrodynamical simulations . section 3 is devoted to the numerical results . in this section we compare clouds with strong magnetic fields and those with weak magnetic fields . this shows that completely different two types of outflows are ejected in respective clouds . another comparison is made between fast rotators and slow rotators . in @xmath74 , we discuss the evolution till the second core , which becomes actually a new - born star , is formed . it is found that another outflow is found around the second core , which seems to correspond to the optical jets or high speed neutral winds . we also discuss whether the mass inflow / outflow rate and the momentum outflow rate observed in molecular bipolar outflows are explained or not . to study the dynamical contraction , we consider an isothermal cylindrical cloud in hydrostatic balance with infinite length as the initial state . in terms of the gravitational potential @xmath8 and isothermal sound speed @xmath9 , the radial distributions of density @xmath10 , rotation speed @xmath11 , and magnetic field density @xmath12 are calculated using the equation of hydrostatic balance as @xmath13 and the poisson equation for the self - gravity as @xmath14 where @xmath15 represents the gravitational constant . this equation has a solution as follows @xcite @xmath16 @xmath17 and @xmath18 with using the scale - height at the center , @xmath19 where @xmath20 , @xmath21 , and @xmath22 represent respectively the density , the angular rotation speed , and the magnetic flux density at the center of the cylindrical cloud ( @xmath23 ) . this is the same hydrostatic configuration studied by @xcite . however , we assume no initial toroidal magnetic field @xmath24 in contrast to them . the density distribution extends till the radius where the thermal pressure @xmath25 becomes equal to the external pressure @xmath26 . the solution contains three non - dimensional parameters characterizing the distribution after adopting natural normalization such as for distance @xmath27 $ ] , time @xmath28 $ ] , and density @xmath29 . we summarized the conversion factors from non - dimensional to physical quantities in table [ table1 ] . the first parameter characterizing the initial state is related to the magnetic to thermal pressure ratio as @xmath30 and the second one is related to the angular rotation speed as @xmath31 finally the surface to center density ratio @xmath32 the scale - height at the center @xmath33 is expressed using these non - dimensional parameters as @xmath34 from this equation , it is shown that a hydrostatic balance is achieved only when @xmath35 , i.e. , @xmath36 . @xcite s @xmath37 parameter , which represents the ratio of the centrifugal force to the thermal pressure force , is expressed using our parameters as @xmath38 to initiate the cloud collapse , we added a small density perturbation . the wavelength of the perturbation is taken equal to that of the most unstable jeans mode . the linear instability of the rotating , magnetized , isothermal cylindrical cloud is studied by @xcite . they gave a fitting formula for the most unstable wavelength as follows : @xmath39^{-1 } \simeq \frac{2\pi \left(1+\alpha/2+\beta \right)^{1/2 } } { 0.72 \left[\left(1+\alpha/2+\beta\right)^{1/3}-0.6 \right ] f^{1/2 } } , \label{eq12}\ ] ] using equations ( [ eq7 ] ) , ( [ eq9 ] ) , and ( [ eq11 ] ) . therefore , we take this most unstable mode and the initial density is assumed equal to @xmath40 , \label{eq13}\ ] ] for @xmath41 . the amplitude of the perturbation @xmath42 is taken equal to 0.1 . hereafter , we omit the primes which indicate the non - dimensional quantities , unless the quantities are confused with dimensional ones . from calculations of one - dimensional , non - rotating , non - magnetized cloud , the effective equation of state of the gas forming a star is summarized as follows @xcite . gas in the molecular cloud core with @xmath43 obeys the isothermal equation of state with the temperature of @xmath44k . however , the molecular gas becomes optically thick against the thermal emission from the dust , which cools the molecular core , after the contraction proceeds and the central density reaches @xmath45 @xcite . the gas becomes adiabatic beyond the density and the first core forms @xcite . hydrogen molecules ( h@xmath1 ) dissociate into atomic hydrogen ( h ) when its temperature reaches @xmath46k . typical density at which the temperature reaches @xmath47 is equal to @xmath48 . the dissociation ends at the density @xmath49 . finally forms an atomic hydrogen core , which is call the second core by @xcite . to include these changes , we adopt the multiple polytropes @xmath50 with @xmath51 , @xmath52 , and @xmath53 . the basic equations to be solved are the magnetohydrodynamical equations and the poisson equation for the gravitational potential . in cylindrical coordinates ( @xmath54 , @xmath55 , @xmath56 ) with @xmath57 , the equations are expressed as follows : @xmath58 @xmath59 , \label{eq15}\end{aligned}\ ] ] @xmath60 , \label{eq16}\end{aligned}\ ] ] @xmath61 \label{eq17}\ ] ] @xmath62 , \label{eq18}\ ] ] @xmath63 @xmath64 @xmath65 where the variables have their ordinary meanings . equation ( [ eq14 ] ) is the continuity equation ; equations ( [ eq15 ] ) , ( [ eq16 ] ) and ( [ eq17 ] ) are the equations of motion . the induction equations for the poloidal magnetic fields are equations ( [ eq18 ] ) and ( [ eq19 ] ) and for the toroidal magnetic field is equations ( [ eq20 ] ) . the last equation ( [ eq21 ] ) is the poisson equation . the mhd equations were solved using @xcite s monotonic interpolation and the constrained transport method by @xcite with the method of characteristics ( moc ) modified by @xcite . to ensure the specific angular momentum @xmath66 and the toroidal magnetic fields @xmath67 are convected similar to the density , advections of such values are calculated using the consistent advection which is first pointed out by @xcite . we rewrote equations ( [ eq17 ] ) and ( [ eq20 ] ) as @xmath68 -\frac{\partial}{\partial r } \left[\left(\frac{b_\phi}{\rho}\right ) \left(\rho v_r\right)\right],\\ \frac{\partial \rho r v_\phi}{\partial t } & = & -\frac{\partial}{\partial z } \left[\left(r v_\phi \right ) \left(\rho v_z\right)\right ] -\frac{\partial}{\partial r } \left[\left(r v_\phi \right ) \left(\rho v_r\right)\right].\end{aligned}\ ] ] to evaluate the right - hand - side of equations ( [ eq17 ] ) and ( [ eq20 ] ) , we chose the same numerical mass flux @xmath10*v * used in equations ( [ eq14 ] ) and ( [ eq15 ] ) and multiplied @xmath69 and @xmath70 to the mass flux to get numerical flux for the angular momentum density and the toroidal magnetic fields . to solve the poisson equation ( [ eq21 ] ) , we adopted milucgs [ modified incomplete lu decomposition preconditioned conjugate gradient squared method : @xcite ; @xcite ] . the gravitational contraction proceeds in a significantly non - homologous way . thus , to see the late phase evolution , high spatial resolution is needed especially near the center . similarly to the previous papers @xcite , we adopt the nested grid method @xcite . in this method , a number of levels of grids with different spacings are prepared ; finer grids cover the central high - density portion and the coarser ones cover the cloud as a whole . the grids are named as l0 ( the coarsest ) , l1 , l2 , and the grid spacing of l@xmath71 is chosen equal to a half of that of l@xmath72 , that is , the grid spacings of l@xmath71 is equal to @xmath73 and @xmath74 . the l0 grid covers the region @xmath41 and @xmath75 . in the nested grid method , the boundary condition is applied to the outer boundary of the coarsest grid l0 , on which we adopted the fixed boundary condition at @xmath76 and the periodic boundary condition at @xmath77 . the code has been tested comparing the results obtained with and without the nested grid technique [ for detail see @xcite ] . the the most unstable growth rates of perturbations are slightly different ( @xmath78 ) : @xmath79 was attained at @xmath80 for without nested grid ( but using relatively large number of grid points 400@xmath81400 ) , but at @xmath82 in the calculation with the nested grid method . this is mainly due to the fact the eigen - mode of the most unstable mode is as wide as the whole size of the numerical box and this mode is captured better by a simple calculation with a large number of zones ( 400@xmath81400 ) rather than the l0 grid ( 64 @xmath81 64 ) of the present scheme . however , after the perturbation has grown to be non - linear , the growth rate agrees well . in model a , we calculated the evolution with @xmath83 , @xmath84 , @xmath85 , and @xmath86 . we summarized the adopted parameters in table [ table2 ] . similar to the _ non - rotational _ magnetized cloud [ see figs . 2b and 2c of @xcite ] , the cylindrical cloud fragments into prolate spheroidal shape whose wavelength is equal to @xmath87 . this prolate spheroidal shape coincides with the structure expected from the linear stability analysis by @xcite . next , this density enhanced region begins to contract along the major axis of the cylindrical cloud , since the magnetic fields are assumed to run parallelly to the major axis . finally it forms a contracting disk ( pseudo - disk ) perpendicular to the magnetic field lines [ fig . 2d of @xcite ] . the snapshot at this stage is shown in figure [ fig1 ] . using the conversion factor shown in table[table1 ] , the epoch @xmath88 corresponds to @xmath89 from the beginning of calculation . respective panels of figure [ fig1 ] have different spatial coverage . figure [ fig1]a , which shows l1 , captures global structure of contracting disk extending horizontally which is perpendicular to the magnetic field lines . the spatial resolution of l1 is so limited that there seems no internal structures in the contracting disk . however , l5 which has 16 times finer resolution than l1 shows shock fronts facing outer directions extend parallel to the @xmath55-direction . fronts near @xmath90 are the fast - mode mhd shock front , because the magnetic fields bend toward the front passing the shock front . we can see another density jump near @xmath91 ( hereafter we will omit the sign @xmath92 and mention the upper half part of the figure since the structure is symmetric ) . the shock fronts parallel to the disk are known in non - magnetized rotating isothermal clouds @xcite . this is not due to the rotation ; multiple shock fronts are also found in the contracting magnetized cloud without rotation @xcite . however , situation becomes a bit complicated in this cloud . in the outer region @xmath93 , the magnetic field lines run almost vertically ( @xmath94 and @xmath67 ) . passing the mhd fast shock , in the intermediate region ( @xmath95 ) , the radial and toroidal components are amplified and the density increases compared to the outer region . finally after passing another front near @xmath96 , the toroidal component @xmath67 decreases . that is , since the magnetic field lines deflect departing from the front at the second front , this is the slow mhd shock . the density range captured in this panel is from @xmath97 to @xmath98 . it should be noticed that these shocks occurs in the isothermal gas . this phase is called `` run - away collapse , '' in which the central density ( @xmath99 ) increases greatly in a finite time - scale . figure [ fig1]c shows the structure captured by l10 . almost all the gas in this figure is isothermal . however , a central small part of the contracting disk @xmath100 and @xmath101 enters the adiabatic regime . at this stage , we can see another density jump is forming just outside the adiabatic part of the disk . it seems to grow into an accretion shock front , since it is known that an accretion shock forms outside the core when the adiabatic core develops @xcite . this is easily understood as follows : the adiabatic gas with specific heat ratio @xmath102 has a hydrostatic equilibrium irrespective of its mass . the scale - height in the @xmath54-direction of the core becomes larger than that of the disk . the adiabatic part of the contracting disk forms a spherical static core . figure [ fig2 ] shows the cross - cut view along two axes ( panel a : along the disk mid - plane @xmath103 and panel b : along the @xmath54-axis @xmath23 ) . the lines with 0.6066 represent the stage shown in figure [ fig1 ] . at this stage ( @xmath88 ) , inflowing gas is almost isothermal ( @xmath104 ) . in figure [ fig2]a , the radial distributions of the density , the magnetic flux density , and the radial and toroidal components of velocity are shown . we can see that density and magnetic flux density distributions are approximately expressed by power - laws as @xmath105 and @xmath106 except for the central part . at @xmath107 , the density in the core exceeds @xmath108 . and at @xmath109 , it reaches @xmath110 . at this stage , a radially outward - facing shock front is seen even inside the disk ; infall motion is abruptly decelerated and the density and magnetic flux density are amplified . this shows that a compact core is formed inside the accretion shock front . the central density increases with time and inside @xmath111 adiabatic gas ( @xmath112 ) distributes . the size of the core is equal to @xmath113 . this reduces with time since the mass of the core increases by the effect of continuous accretion . before the shock front is formed @xmath114 , the radial inflow velocity takes the maximum about @xmath115 near @xmath116 . for the larson - penston self - similar solution for the spherically symmetric dynamical collapse @xcite , this maximum inflow speed is expected equal to @xmath117 . on the other hand , it equals to @xmath118 for non - rotating isothermal _ disk _ @xcite . therefore , it is shown that the actual inflow speed ranges between the spherically symmetric self - similar solution and the axially symmetric thin disk solution . after the shock front is formed around the core , the inflow velocity takes the maximum just outside the shock front and the maximum speed increases with time . inflow motion is accelerated toward the shock front . similar acceleration is also seen in the toroidal velocity , @xmath11 . before the core formation the toroidal speed @xmath11 takes the maximum near @xmath119 . however , @xmath11 increases toward the accretion shock after core formation . at @xmath109 it reaches @xmath120 ( see also fig.1 of @xcite ) . structure seen in the cross - cut along the @xmath54-axis is more complicated ( fig.[fig2]b ) . two shock fronts mentioned earlier ( fig.[fig1]b ) correspond to the jumps near @xmath121 and @xmath122-axis it is found near @xmath123 , while departing from the @xmath54-axis ( @xmath124 ) it is found near @xmath125 . ] at @xmath88 , the density and the inflowing velocity distributions have no discontinuities besides these two shock fronts . however , at @xmath107 , the inflowing velocity distribution begins to indicate a clear discontinuity near @xmath126 . this is a newly formed shock front and propagate spatially . comparing two curves of @xmath107 and @xmath109 , it is shown that this shock front breaks into two fronts and the inner one ( @xmath126 ) is standing still , while another outer one ( @xmath127 ) is propagating outwardly . these two shock fronts are outwardly facing . thus the inwardly propagation of the inner fronts is due to the infalling gas motion . figure [ fig3 ] illustrates the structure at @xmath128 . although the gas is inflowing both inside and outside of the disk at @xmath107 ( fig.[fig2 ] ) , at this stage @xmath128 ( @xmath129 ) represents the time from the beginning of calculation but @xmath130 represents the time after the core formation . we assumed that the core consists of the gas with density @xmath112 . ] prominent outflow is formed . it is shown that the flow pattern is completely changed in @xmath131yr . outflow sweeps a sphere with radius of @xmath132 ( fig.[fig3]a ) . figure [ fig3]b indicates that the gas near the disk surface flows inwardly for @xmath133 . however , the direction of the flow is changed upwardly near @xmath134 . finally this gas is ejected . while the gas near the mid - plane of the disk ( @xmath135 ) continues to contract . this is reasonable because the total amount of angular momentum in one magnetic flux tube must be conserved in the axisymmetric ideal mhd simulation ; for the outflow gas to get angular momentum , a part of the gas in the same magnetic flux tube has to lose its angular momentum and falls further . in the acceleration process of the gas , the angular momentum is transferred from the gas near the mid - plane to the gas near the surface of the disk . considering the angular rotation speed , the angular momentum is transferred from the fast - rotating mid - plane to the slowly rotating surface gas . from figures [ fig1]c and [ fig3]a ( both show the structure captured by l10 ) , we can see the magnetic field lines run completely differently comparing before ( @xmath136 : fig.[fig1]c ) and after ( @xmath137 : fig.[fig3]a ) the adiabatic core formation . that is , in the isothermal runaway collapse phase ( fig.[fig1]c ) the magnetic fields lines run vertically , in other words , perpendicularly to the disk . in contrast , after the adiabatic core is formed , the disk continues to contract and drags the magnetic field lines inwardly . thus the angle between the magnetic field lines and the disk decreases . figure [ fig3]b is a close - up view whose spatial resolution is 4-times finer than that of figure [ fig3]a . this panel shows us that the angle between the flow and the disk is about @xmath138 . the reason why the outflow begins after the core formation is related to the angle between the magnetic field lines and the disk , @xmath139 . @xcite have pointed out that for a cold gas rotating with the keplerian speed to get angular momentum from the keplerian disk via infinitely strong magnetic fields , @xmath139 must be smaller than a critical value @xmath140 . this is understood as follows : consider the gas on one magnetic flux tube . when the magnetic flux tube rising steeply from the disk as @xmath141 , the gas has to climb the effective potential well even if it rotates with the same angular speed of the keplerian disk . on the other hand , when @xmath142 , gas can escape from the gravitational well by getting angular momentum from the disk , if the gas has the same angular speed of the keplerian disk . although the exact value of @xmath143 depends on the disk rotation speed and the disk - to - central star mass ratio , small angle is preferable to acceleration . it should be noted that this configuration is achieved only after the core formation @xcite . figure [ fig3]c shows the ratio of the toroidal magnetic component to poloidal one by contour lines , which is overlaid on the magnetic field lines . since the toroidal - to - poloidal ratio is as small as @xmath144 in the disk , the disk is poloidal - dominated . however , in the region where the gas flows outwardly the toroidal component grows and the toroidal - to - poloidal ratio reaches @xmath145 . the outflow gas is toroidal - field - dominated . the coincidence of acceleration region with the toroidal - dominant region seems to indicate that the toroidal fields play an important role to accelerate the gas . the toroidal component of the lorentz force , @xmath146 works below this toroidal - dominant region , that is , @xmath147 . this toroidal component @xmath148 accelerates the toroidal velocity @xmath11 and resultant toroidal motion amplifies the toroidal component of the magnetic fields . outflow speed exceeds the sound speed and the fastest speed reaches @xmath149 at this time . it increases with time . although the outflow seems to continue , the further evolution is hard to study , because the time - scale ( the free - fall time - scale at the central core ) becomes shorter and shorter . therefore , we study model ah with a constant polytropic index larger than that of model a. in models ah1 ( fig.[fig4]a ) and ah2 ( fig.[fig4]b ) , the polytropic indices are chosen @xmath150 and @xmath151 , respectively , for @xmath112 . ( models whose name have `` h '' have simple polytropic relation with @xmath150 or @xmath151 for @xmath112 . ) due to the hard polytropic index , the size of the adiabatic core , whose surface is determined by the jump in @xmath152 , becomes large ; for example at @xmath128 the size is equal to @xmath153 for model ah1 and @xmath154 for model ah2 while it is @xmath155 for model a. similar to model a , just outside the core , outflow begins to be accelerated . the region which the outflow sweeps expands and the surface which separates inflow and outflow forms another mhd shock front . and the expansion of the front is very similar to that of model a ( the front reaches @xmath156 at this time which is similar to model a ) . to see the similarity in more detail , we calculated the mass of the core @xmath157 for models a , ah1 , and ah2 . these are equal to @xmath158 , @xmath159 , and @xmath160 at the time @xmath128 ( @xmath129 ) for models a , ah1 and ah2 , respectively . at later epoch @xmath161 ( @xmath162yr ) , @xmath163 ( model ah1 ) and @xmath164 ( model ah2 ) . from these results , it is shown that the core mass increases with time due to the continuous accretion and the mass does not depend on the exact equation of state in the core . this is understood as follows : the core mass is determined by the accretion rate of the _ isothermal gas _ which is independent from the polytropic @xmath165 in the core . the gravity by the core has an effect on the outer inflow and outflow . since the effect depends only on its mass , the difference of the polytropic index of the core does not play an important role for the inflow and outflow . therefore , we will study this model ah to see the long time evolution of the outflow . in figure [ fig4]c , the snapshot at @xmath166 ( @xmath167 ) is plotted for model ah1 . comparing this with figure [ fig1]b for model a ( both have the same resolution but for different epochs ) , it is shown that the shock front which separates the inflow and the outflow passed the slow - mode mhd shock near @xmath168 and has just reached the outer fast - mode shock front near @xmath169 . the evolution of model ah2 is essentially the same . the maximum speed of the outflow reaches @xmath170 . this maximum speed seems smaller than that observed in the molecular outflow . since the mass accumulated in the core is only equal to @xmath171 , the outflow speed seems to be much faster than this value , when the mass has grown to typical t tauri stars . figure [ fig3]b indicates the outflow is accelerated near the core and the opening angle of the outflow in this region is wide . however , departing from the acceleration region the flow changes its direction toward the @xmath54-axis . figure [ fig4]c show that the opening angle decreases as the outflow proceeds . this indicates the flow is collimated . to see the effect of the initial rotation speed , we compare models ah1 ( @xmath172 ) , bh ( @xmath173 ) , and ch ( @xmath174 ) . these models have the same magnetic field strength , @xmath175 . in figure [ fig5 ] panels ( a - c ) , the structures at the final epoch of the isothermal run - away collapse phase are plotted for respective models . models bh and ch indicate no prominent discontinuity in l6 ( figs . [ fig5]b and c ) , while model ah1 has several shock fronts as described in @xmath7[sec3.1 ] . however , in l10 ( not shown ) , there are discontinuities near @xmath176 ( model bh ) and @xmath177 ( model ch ) as well as in model ah1 ( @xmath178 ) . comparing panels ( b ) and ( c ) , distributions of the density and magnetic field lines are similar each other . this indicates that the evolution in the isothermal phase is slightly dependent on the initial angular momentum if @xmath179 . figures [ fig5]d - f show the structure at the age @xmath180 after the core formation . in all models the outflows are formed . however , the size of the region swept by the outflow is different for each model . with increasing @xmath181 , more energetic outflow is driven . from the flow vectors , it is shown that model ah1 ( fig.[fig6]d ) forms a bit more collimated outflow than models bh and ch ( fig.[fig6]e and f ) . this seems to correspond to the differences in density distribution and magnetic field configuration . that is , in model ah1 ( also a and ah2 ) there is a relatively thick disk seen in l6 which is bounded by the shock fronts . this thick disk seems to confine the outflowing gas in model ah1 . while , in models bh and ch the disk is relatively thin , which seems to make the flow isotropic . further , the opening angle of the magnetic field lines in models bh and ch is larger than that of model ah1 . this causes the flow also open . difference between models bh and ch comes from the fact that the epochs when the outflow begins are different . since the initial angular momentum in model bh is five - times larger than that of model ch , in model bh the outflow begins earlier than model ch . at @xmath182 , however , even in model ch the top of the outflow reaches @xmath169 and the structure looks very similar to model bh at @xmath183 ( fig.[fig5]e ) . at the epoch when figure [ fig5]d - f is taken ( @xmath184^{-1/2}$ ] ) , the mass in the adiabatic core reaches @xmath185 ( model ah1 ) , @xmath186 ( model bh ) , and @xmath187 ( model ch ) . the instantaneous rate of mass accretion to the adiabatic core for each model attains @xmath188 ( model ah1 ) , @xmath189 ( model bh ) , and @xmath190 ( model ch ) , respectively . therefore , the core mass is approximately proportional to the mass accretion rate and the mass accretion rate is larger for models with smaller angular rotation speed @xmath181 . accretion rate expected from the inside - out collapse model @xcite is equal to @xmath191 . therefore , the accretion rates calculated here are @xmath192 times larger than that expected by the inside - out collapse model . consider the reason why the mass of the core decreases with increasing @xmath181 . since the gas is supplied to the core mainly through the disk , we will consider the mass inflow / outflow transported in the disk . at that time , the gas disk can be divided into three regions . outermost region is occupied with isothermal gas and the gas is contracting or inflowing ( pseudo - disk ) . therefore the inflow mass rate @xmath193 and the outflow mass rate @xmath194 in this outermost region . inside of this region , outflow is generated , although a large part of the gas is still inflowing . therefore in this middle region , the inflow rate is smaller than that of the outermost region , @xmath195 , and the excess mass is transported to the outflow , @xmath196 . innermost is the adiabatic core . since the mass accretion rate to the core is equal to the net mass inflow rate from the middle region , @xmath197 . mass inflow driven by the self - gravity is more important in a model with small @xmath181 in which the self - gravity is ineffectively counterbalanced with the centrifugal force . therefore @xmath198 becomes larger for slow rotator . this is the first effect of the rotation . furthermore , the outflow brings away appreciable amount of gas . as mentioned previously , the outflow is strongly generated in the fast rotator . thus , the mass outflow rate increases with increasing @xmath181 as @xmath199 ( model ah1 ) , @xmath200 ( model bh ) , and @xmath201 ( model ch ) . as a result , increasing @xmath181 , the portion of outflow gas to the inflow gas @xmath202 becomes large as @xmath203 40% for model ah1 , @xmath203 10% for model bh , and @xmath204 5% for model ch . these two effects works cooperatively to reduce the mass accretion rate @xmath205 to the core for the cloud with a large @xmath181 . the maximum outflow speed realized in respective figures are equal to @xmath206 ( model ah1 ) , @xmath207 ( model bh ) , and @xmath208 ( model ch ) . since the outflow is accelerated by the toroidal magnetic fields which are generated by the rotation motion , this @xmath209 increases with increasing @xmath181 . as shown in @xcite , since the excess angular momentum of the inflowing gas is effectively removed by the outflow , the total angular momentum of the first core which is defined as a gas with @xmath112 is equal to @xmath210 contained in the mass of @xmath211 ( model ah1 ) , @xmath212 in @xmath213 ( model bh ) , and @xmath214 in @xmath215 ( model ch ) . the total angular momenta contained in the first cores are only 1.1% , 0.07% , and 0.9% of the initial ones contained in respective mass @xmath216 . to see the effect of the magnetic field strength , we calculated models nh , dh and eh in which we took @xmath217 , @xmath218 , and @xmath219 . model nh has no magnetic fields . in figure [ fig6]a , a snapshot at @xmath220 captured by l8 is shown for model nh . at this stage , whole the cloud is in isothermal regime and the disk experiences the run - away collapse even if the centrifugal force may work to support the cloud . this confirms the earlier studies by @xcite and @xcite . the physical reason why the centrifugal force does not stop the contraction in the isothermal run - away collapse phase is explained in @xcite as follows : due to the the centrifugal force , the mass contained in the jeans scale ( @xmath221 ) from the center is _ decreasing _ throughout the collapse . in this sense the centrifugal force does work ! only small part of the mass that resides near the center becomes high - density . but contraction itself continues and the central density rises greatly in a finite time - scale , if the isothermal equation of state is valid . similar to the previous magnetized models , a small adiabatic core is formed first . since there is no magnetic fields , magnetic braking does not work , however , in this model . therefore , gas that accretes on the core has relatively large angular momentum in contrast to the magnetized model . as a result , a ring forms by the gas which accreted on the adiabatic core . since the specific angular momentum , @xmath222 , increases further with time , radius of the ring grows . another snapshot in panel ( b ) at @xmath223 ( @xmath224 ) shows the ring clearly . the ring seems unstable for non - axisymmetric perturbation . this may form a spiral structure similar to that found by @xcite . however this is beyond the scope of this paper . therefore , it is concluded that a rotating but non - magnetic cloud leads to a rotating ring in the adiabatic accretion stage . to see the effect of the magnetic field strength , in figure [ fig7 ] we compare models bh ( @xmath83 ) , dh(@xmath225 ) , and eh ( @xmath226 ) . all models have the same initial rotation speed @xmath227 , and polytropic gamma @xmath150 . in panels ( a)-(c ) , the structures at the epoch when an adiabatic core begins to form are plotted . comparing these panels , it is shown that in model b ( @xmath83 ) a flare - up disk is formed whose isodensity lines are departing from the disk mid - plane leaving from the center . decreasing the initial magnetic field strength , the shape of dense part of the disk becomes rounder . similar effect is already reported for non - rotating magnetized cloud collapse @xcite , that is , deceasing @xmath175 the shape of the isothermal contracting disk becomes rounder and finally reaches a sphere for @xmath217 . in panel ( d ) we plotted a snapshot for model bh at @xmath228 ( @xmath229 ) captured in l7 . this is the same as illustrated in figure [ fig5]e but captured in l7 which has twice as fine as figure [ fig5]e . figure [ fig7]d which shows the structure near the root of outflow indicates that it is very similar to that of model ah1 . for example , the outflow leaves from the disk in the direction almost parallel to the disk but it changes its direction to the @xmath54-direction . time @xmath228 is equivalent to @xmath229 . it is concluded that in a time - scale of @xmath230 , the flow pattern is completely changed from the run - away collapse to the outflow plus continuous inflow in the disk . the outflow gas flows along the surface whose shape resembles a capital letter u. the flow departs from the disk with a wide opening angle but it change its direction parallel to the @xmath54-axis . in panel ( e ) , we plotted the structure expected from a model with weaker magnetic field ( model dh : @xmath225 and @xmath227 ) . the snapshot corresponds to the epoch of @xmath231 . this corresponds to @xmath232 which is similar to the time - scale between panels ( a ) and ( d ) . in contrast to the previous model bh , the outflow gas forms a sphere and the magnetic field lines are folded inside this sphere . the magnetic field lines are folded by the pinch or hoop stress of the toroidal magnetic field . the toroidal magnetic field component is confined in the region where the adjacent poloidal magnetic field lines are running in the opposite directions . for example , the regions around @xmath233 and @xmath234 in figure [ fig7]e . in this model , the initial poloidal magnetic fields are weak compared to model bh . therefore , rotational motion amplifies the toroidal fields and their strength surpasses easily that of the poloidal ones . thus , the hoop stress by the toroidal field pinches efficiently the magnetic field lines . in the outflow acceleration region , the toroidal component is predominant over the poloidal one . magnetic field lines are pinched locally and folded . as a result , a spherical magnetic bubble is formed in this process , in which the toroidal magnetic field is predominant . toroidal component of the magnetic fields is continuously generated by the twist motion driven by the disk rotation . the disk angular momentum is transferred by this process . as a result , we do not see the ring of model nh which is supported by the centrifugal force . for the model with extremely weak field , we calculated model eh ( @xmath226 and @xmath235 ) . in panel ( f ) , we plotted the snapshot at @xmath236 ( @xmath237 ) . density distribution and magnetic field configuration show that the flow in the magnetic bubble is more complicated or turbulent in this model . the shape of the bubble is more elongated than that formed in model dh . distribution of toroidal field lines does not show any systematic pattern inside the bubble . size of the bubble both in the @xmath54- and @xmath55-directions are smaller than that of models bh and dh . it is concluded that the size of the outflow region increases with increasing the magnetic field strength @xmath175 . comparing these three models , it is shown that there are at least two types of outflows . that is , a laminar u - type flow in which fast moving gas flows along the surface whose shape resembles a capital letter u and a turbulent outflow in which the magnetic fields and the velocity are randomly oriented . the masses accumulated in @xmath238 are equal to @xmath239 for model bh , @xmath240 for model dh , and @xmath241 for model eh . this shows that the mass accretion rate , @xmath242 , is an increasing function of the initial magnetic field strength , @xmath175 . this seems strange if we remember @xmath242 is an decreasing function of the initial rotation speed , @xmath181 , since both @xmath175 and @xmath181 have an effect to counterbalance against the self - gravity . this means that the mass inflow rate in the isothermal run - away collapse region , @xmath198 , increases with increasing @xmath175 . this seems to come from a number of reasons , that is , the initial cylindrical cloud becomes massive with increasing @xmath175 . another reason is related to the characteristic wave speed in the magnetized medium . that is , the characteristic speed of the fast mode mhd wave is equal to @xmath243 in the case that the wave is propagating perpendicular to the magnetic field lines . this implies that the mass inflow rates are proportional to @xmath244 ( a : b : c = 2.83 : 1.15 : 1.02 ) . this is not inconsistent with the actual values . in the previous section , we introduced a simple polytrope above the critical density @xmath112 except for model a. however , this is an approximation to see a long evolution . in reality , the composite polytrope ( eq . [ eqn : toh82 ] ) should be used . here , further evolution later than that shown in figures [ fig1 ] and [ fig2 ] is shown . in model r , we used the composite polytrope to include the combined effects such as the dynamical compressional heating , the radiative cooling through the dust thermal emissions , and the energy loss associated with the h@xmath1 dissociation . this corresponds to the continuation of model a but we consider here a more compact cloud as @xmath245 . it should be addressed since @xmath246 is assumed hundred - times larger than other models , the size scale @xmath247 is 10-times smaller than other models with @xmath248 as @xmath249 . the evolution up to the first core formation is similar to model a , that is , after the equation of state becomes hard , @xmath112 , collapse of the adiabatic core slows down and isothermal gas begins to accrete on to the core . in this model since @xmath246 is assumed equal to @xmath250 , the size scale @xmath247 is 10-times smaller than other models with @xmath86 . in figure [ fig8]a , we plotted the structure captured by l6 just before the first core formation @xmath251 . since the accretion continues , the mass of the adiabatic core increases with time , which leads to collapse the core quasi - statically . the structure in this stage ( at @xmath252 or @xmath253 ) is shown in panel ( b ) . the radius of the first core shows in this panel is equal to @xmath254 au . this quasistatic contraction phase ends when the central density reaches @xmath255 . the structure at this stage ( @xmath256 [ @xmath257 ) is plotted in panel ( c ) . comparing this with panel ( b ) , both of which illustrate l12 , it is clear that the adiabatic core is contracting . after that , the second collapse begins . since the thermal energy is lost by the process of the dissociation of h@xmath1 , the equation of state is assumed soft ( @xmath258 ) again . in this phase , the flow is very similar to that realized in the isothermal run - away collapse phase ( @xmath104 ) . the similarity comes from the fact that the equation of states of the both phases are soft . in panel ( d ) , we plotted the structure captured in l16 at @xmath259 ( @xmath260 ) , which represents the typical structure in the second collapse phase . this continues until the central density exceeds @xmath261 , beyond which another adiabatic core ( the second core ) is formed shown in panel ( e ) . the size of the second core is equal to @xmath262 . this meets a similar situation when the first adiabatic core is formed , that is , a central part of the gas obeys a harder equation of state and forms a quasistatic core , while the outer part obeys a softer equation of state and continues to collapse . analogy between the first and the second core leads a bipolar outflow . in panel ( f ) , it is shown a second outflow is formed around the second core . this is a snapshot of l16 at @xmath263 ( @xmath264 ) , which resembles the structure seen in figure [ fig3]b . this model strongly indicates us that there is another kind of outflow accelerated around the second core ( the second outflow ) besides that formed around the first core ( the first outflow ) . the outflow speed , @xmath265 , is much faster than that of the first outflow , due to the fast thermal speed in the adiabatic core and the deep local gravitational potential . since the time span of the simulation shown here is restricted , we can not trace the evolution further . however , the simulation predicts that at least two different outflows are formed each of which is related to different types of adiabatic cores . since the flow speed of the second outflow is much faster than that of the first outflow , the respective outflows correspond to the molecular bipolar outflow ( the first outflow ) and the fast neutral wind or optical jets ( the second outflow ) . the radial size of the outflow is approximately equal to @xmath266 ( = 0.14 au or @xmath267 ) . this indicates that optical jets are found inside molecular bipolar outflows . in model r , we assumed the ideal magnetohydrodynamics . however , the ionization fraction of high - density gas is quite low and the electric conductivity decreases as collapse proceeds . calculation of ionization equilibrium @xcite shows us that after @xmath268 charged grains are more abundant than ions , as a carrier of electric charges . since the mass - to - charge ratio of grains is much larger than that of ions , the electric conductivity decrease greatly and the magnetic field decays mainly through joule dissipation . therefore , in the late phase of the first core ( quasistatic contraction ) , the magnetic field in the core decreases its strength till the field configuration becomes force - free . after the dissociation of h@xmath1 begins , the temperature of the central region is enough high to achieve the thermal ionization of metals . thus , in the second collapse phase , coupling of magnetic fields is recovered . from these , the ideal mhd is consistent in the first and the second collapse phases . however , the magnetic flux density at the beginning of the second collapse phases seems to be much weaker than that obtained here . as shown in @xmath7[sec : weak - b ] , when the poloidal magnetic field is weak , the flow pattern is different from the case that the magnetic energy is comparable to the thermal one . therefore , if the magnetic flux is partly lost from the central part of the first core , a turbulent outflow around the second core seems to be formed similar to figure [ fig7]f . here in this subsection , mass inflow / outflow rate and linear momentum outflow rate are seen more closely . the outflow mass loss rate through the boundary of each grid level is measured as @xmath269 ds , \label{eqn : mdotout}\ ] ] where * n * represents the unit vector outwardly normal to the surface . the integrand means that only the outwardly running mass flux is summed up . similarly , the mass inflow rate is calculated as @xmath270 ds . \label{eqn : mdotin}\ ] ] these rates are calculated for respective levels of the nested grid system . the time variations of mass inflow and outflow rates are illustrated for respective models in figure [ fig9 ] . the time ( horizontal axis ) is measured from the epoch of the core formation . time variations , especially longer time variations , in inflow rates are common for all panels . just after the core formation , the inflowing mass flux decreases with the distance from the center . this means that the inner part of the isothermal collapse region is well expressed by the @xcite-@xcite self - similar solution @xcite which leads to the mass inflow rate of @xmath271 ; departing from the center , the approximation of the larson - penston solution becomes worse and the inflow becomes lower . at @xmath272 the inflow rates for various surfaces converge on @xmath273 , although intense outflow reduces the inflow rate below this value . we will see each model more closely . outflow rate rises in the deeper levels ( l@xmath71 with larger @xmath71 ) first and propagates to lower levels ( l@xmath71 with smaller @xmath71 ) . this indicates the outflow region expands both in @xmath55- and @xmath54-directions . comparing models ah1 ( panel a ) , bh ( panel b ) , and ch ( panel c ) , the effects of the initial rotation speed , @xmath274 , are apparent . + ( 1 ) with increasing @xmath181 , the outflow begins earlier . + ( 2 ) with increasing @xmath181 , the mass outflow rate increases ; although the ratio @xmath275 is equal to only @xmath276 in model ch ( @xmath277 ) , in model ah1 ( @xmath84 ) it attains @xmath278 . we compare models bh ( panel b ) , dh ( panel d ) , and eh ( panel e ) to see the effects of the initial magnetic field strength , @xmath175 . this shows that with decreasing @xmath175 ( from models bh to dh ) the mass outflow rate increases . however , in model e , in which extremely weak poloidal magnetic fields is assumed , the time variation in the inflow and outflow rates is rather chaotic which is lead by the turbulent flow pattern realized in model eh . averaging the rates as @xmath279 , the rate of model e is @xmath280 _ smaller _ than that of model dh . considering the disk after rotating at a fixed angle , the disk of model eh can generate weaker toroidal magnetic fields than model dh , since model eh has only weak poloidal ( source ) magnetic fields . therefore , it is reasonable that the mass outflow rate in model eh , @xmath281 , is smaller than that of model dh . how about the increase from model bh to dh ? this is inconsistent with the above discussion . this increase seems related to the fact that the flow patterns of models bh and dh are different . that is , in model dh a magnetic bubble , in which magnetic field lines are folded and amplified , is formed and the bubble expands . in contrast , in model bh , the gas is flowing outward along specific magnetic field lines and outflowing gas forms a wall of a chimney whose shape looks like a capital letter u. mass outflow rate seems larger in the magnetic bubble type outflow rather than the chimney wall type . the linear momentum outflow rates are calculated by @xmath282 ds . \label{eqn : mvdotout}\ ] ] the time variation of momentum outflow rate is shown in figure [ fig10 ] . panels ( a ) , ( b ) , and ( c ) have a prominent common feature that the momentum outflow rates ( @xmath283 ) calculated at the boundaries of l@xmath71 ( @xmath284 ) are much larger than that of the boundaries of l@xmath71 ( @xmath285 ) . this feature is common in models with @xmath83 ( models ah1 , ah2 , bh , and ch ) . however , models with weaker magnetic field does not show this feature . recall the fact that models ah1 , ah2 , bh , and ch form a u - shaped outflow and in contrast models dh and eh form turbulent magnetic bubble type outflow . difference in the momentum outflow rate seems to be related to the outflow pattern . in the laminar u - shaped outflow , the outflow is ejected with a rather wide opening angle and collimated to the symmetric axis . since equation ( [ eqn : mvdotout ] ) counts the momentum passing the upper and lower boundaries , the outflow rate increases after the outflow is collimated and the gas flows parallel to the @xmath54-axis . this occurs in the larger scale than l7 . typical momentum outflow rate for molecular bipolar outflows observed by @xmath286co(@xmath287 ) @xcite are distributed between @xmath288 for objects associated with active class 0 ir sources and @xmath289 for objects associated with late class 1 ir sources . this range corresponds to @xmath290 , respectively , if we assume @xmath291 . figure 10 shows us that the momentum outflow rate in the range from @xmath292 to @xmath293 is expected . thus , the momentum outflow rate for almost all the co bipolar outflow sources associated with low - mass young stellar objects except for the active early class 0 sources are explained by this model . comparing models ah1 , bh , and ch , it is shown that the momentum outflow rate increases with @xmath181 . comparing mass outflow and momentum outflow rates , models ah1 , bh , and ch indicate that the momentum outflow rate is approximately proportional to the mass outflow rate , which means that the outflow speed is approximately equal irrespective of @xmath181 . however , changing @xmath175 is more complicated . compare panels ( b ) , ( d ) , and ( e ) . as shown in figure [ fig9 ] , model dh ( @xmath225 ) shows larger mass outflow rate than model bh ( @xmath83 ) . in contrast , as for the momentum outflow rate , the maximum rate of model bh is larger than that of model dh ( fig.[fig10 ] ) . this represents the outflow speed of model bh ( @xmath83 ) is faster than model dh ( @xmath225 ) . panel ( e ) shows that the momentum outflow rate has a chaotic time variation as well as the mass outflow rate ( fig.[fig9]e ) , for model eh , which seems to correspond the turbulent outflow shown in figure [ fig7]f . summarizing the effect of changing @xmath175 , + ( 1 ) the mass outflow rate increases with increasing @xmath294 but decreases after @xmath295 . + ( 2 ) the outflow speed is an increasing function of @xmath175 . + ( 3 ) there are distinctly different two types of flow patterns : models with strong magnetic fields lead the u - shape outflow , while the flow becomes turbulent for models with weak magnetic fields . here , we intend to clear where the outflow is ejected . in the run - away collapse phase , the rotational motion is relatively unimportant . after the core formation , rotation motion increases especially around the first core . this occurs near the centrifugal radius , which is defined for a gas with specific angular momentum @xmath296 as @xmath297 where @xmath216 denotes the mass inside the radius @xmath298 and @xmath299 represents a numerical factor of the order of unity . considering the self - similar solution for the run - away collapse phase @xcite , the specific angular momentum @xmath296 is proportional to @xmath216 . further a uniform - density cylinder which is rotating with a uniform rotation speed @xmath300 has a @xmath296 distribution proportional to @xmath216 . assuming @xmath301 ( @xmath302 is a numerical factor ) , the centrifugal radius is written @xmath303 as shown in section [ sec3 ] , the angular momentum is redistributed in one magnetic flux tube . that is , gas near the disk surface obtains large amount of the specific angular momentum but that near the disk mid - plane loses the angular momentum . if the gas near the disk surface has the angular momentum of @xmath304 , the condition that the effective potential at the centrifugal radius @xmath305 is larger than or equal to zero leads to a minimum angular momentum of @xmath306 we have explored the evolution of a magnetized interstellar cloud rotating around the symmetric axis . following the change in the equation of state of the interstellar gas @xcite , the cloud experiences several phases before going to a star , that is , the isothermal run - away collapse , the slowly contracting core composed of the molecular hydrogen ( the first core ) , the second run - away collapse in the high - density gas where the dissociation of hydrogen molecules proceeds , and finally the second core which is made up of the atomic hydrogen . the magnetized cloud forms a pseudo - disk in these first and second run - away collapses , in which a supersonically contracting disk is formed and magnetic field lines are running perpendicularly to the disk . in the pseudo - disk , a number of fast- and slow - mode mhd shock pairs are formed which is extending parallelly to the disk . just after the core is formed at the center , an accretion shock front appears which surrounds the core , through which the supersonic inflow motion is decelerated . while the first and second cores are slowly contacting , the outer pseudo disk continues to contract . just outside the accretion shock front , the infall motion is accelerated and thus rotational motion becomes important from the conservation of angular momentum . by the effect of rotational motion , the toroidal magnetic fields and the poloidal currents are amplified , which bring a strong magnetic torque . the magnetic torque leads the angular momentum transfer from mid - plane to surface of the disk . this is the origin of the outflow found in star forming regions . large scale bipolar molecular outflows are made outside of the first core , while optical jets and fast neutral winds are expected to be accelerated outside of the second core . matter lost its excess angular momentum continues to contract further to form a star . this work was supported partially by the grants - in - aid ( 11640231 , 10147105 ) from mext ( the ministry of education , culture , sports , science and technology ) . numerical calculations were carried out by fujitsu vpp5000 at the astronomical data analysis center , the national astronomical observatory , an inter - university research institute of astronomy operated by mext , japan . [ [ section ] ] basu , s. , & mouschovias , t. ch . 1994 , , 432 , 720 basu , s. , & mouschovias , t. ch . 1995 , , 452 , 386 bentz , w. 1984 , , 139 , 378 berger , m.j . & colella , p. 1989 , j. comput . , 82 , 64 berger , m.j . & oliger , j. 1984 , j. comput . phys . , 53 , 484 bodenheimer , p. 1995 , , 33 , 199 bodenheimer , p. , tohline , j. e. , & black , d. c. 1980 , apj , 242 , 209 bontemps , s. , andre , p. , terebey , s. , cabrit , s. , 311 , 858 blandford , r. d. , & peyne , d. g. 1982 , , 199 , 883 ciolek , g.e . , & basu , s. 2000 , , 529 , 925 dorfi , e. 1982 , , 114 , 151 dorfi , e. 1989 , , 225 , 507 evans , c. r. , hawley , j. f. 1988 , , 332 , 659 fiedler , r. a. , & mouschovias , t. ch . 1992 , , 391 , 199 fiedler , r. a. , & mouschovias , t. ch . 1993 , , 415 , 680 gustafsson , i. 1978 , bit , 18 , 142 hayashi , c. 1980 , in `` star forming regions , '' ed . m. peimbert , & j. jugaku ( dordrecht : d.reidel ) , p.403 klein , r.i . , fisher , r.t . , mckee , c.f . , truelove , j.k . 1999 , in `` numerical astrophysics , '' ed . miyama , k. tomisaka , & t. hanawa . ( boston : kluwer academic ) , p.131 kompaneets , a.s . , soviet phys . dokl . , 5 , 46 larson , r.b . 1969 , , 145 , 271 matsumoto , t. , nakamura , f. , & hanawa , t. 1994 , , 46 , 243 matsumoto , t. , nakamura , f. , & hanawa , t. 1997 , , 478 , 569 masunaga , h. , & inutsuka , s. , 510 , 822 meijerink . j. a. , & van der vorst , h. a. 1977 , math . , 31 , 148 mouschovias , t. ch . , & morton s. a. 1991 , , 371 , 296 mouschovias , t. ch . , & morton s. a. 1992 , , 390 , 144 nakamura , f. , matsumoto , t. , hanawa , t. & tomisaka , k. 1999 , , 510 , 274 nakamura , f. , hanawa , t. & nakano , t. 1995 , , 444 , 770 nakano , t. , umebayshi , t. 1986 , , 218 , 663 narita , s. , hayashi , c. , & miyama , s. m. 1984 , prog . theor . phys . 72 , 1118 norman , m.l . , wilson , j.r . , & barton , r.t . 1980 , , 239 , 968 ogino , s. , tomisaka , k. , & nakamura , f. 1999 , , 51 , 637 ohashi , n. , lee , s.w . , wilner , d.j . , & hayashi , m. 2000 , , 518 , l41 penston , m.v . 1969 , , 144 , 425 phillips , g. l. 1986a , , 221 , 571 phillips , g. l. 1986b , , 222 , 111 saigo , k. & hanawa , t. , 493 , 342 scott , e. h. & black , d. c. 1980 , , 239 , 166 shu , f.h . 1977 , , 214 , 488 stodkiewicz , j.s . 1963 , acta astron . , 13 , 30 stone , j. m. & norman , m. l. , 80 , 753 tohline , j. e. 1982 , fundamentals of cosmic physics , 8 , 1 tomisaka , k. , ikeuchi , s. , & nakamura t. 1990 , , 362 , 202 tomisaka , k. 1995 , , 438 , 226 tomisaka , k. 1996 , , 48 , 701 tomisaka , k. 1998 , , 502 , l163 tomisaka , k. 2000 , , 528 l41 truelove , j. k. , klein , r. i. , mckee , c. f. , holliman , j. h. , ii , howell , l. h. , & greenough , j. a. 1997 , , 489 , l179 truelove , j. k. , klein , r. i. , mckee , c. f. , holliman , j. h. , ii , howell , l. h. , greenough , j. a. , & woods , d. t. 1998 , , 495 , 821 tsuribe , t. , & inutsuka , s. 1999a , , 523 , l155 tsuribe , t. , & inutsuka , s. 1999b , , 526 , 307 van leer , b. 1977 , 23 , 276 wood , d. 1982 , mnras , 199 , 331 .conversion from the non - dimensional to the physical quantities .[table1 ] [ cols="<,<,<",options="header " , ] lccccc model & @xmath175 & @xmath181 & @xmath307 & @xmath246 & polytrope + & & & & @xmath308 + a & 1 & 5 & @xmath309 & @xmath310 & realistic + ah1 & 1 & 5 & @xmath309 & @xmath310 & @xmath150 + ah2 & 1 & 5 & @xmath309 & @xmath310 & @xmath151 + b & 1 & 1 & @xmath309 & @xmath310 & realistic + bh & 1 & 1 & @xmath309 & @xmath310 & @xmath150 + ch & 1 & 0.2 & @xmath309 & @xmath310 & @xmath150 + dh & 0.1 & 1 & @xmath309 & @xmath310 & @xmath150 + eh & 0.01 & 1 & @xmath309 & @xmath310 & @xmath150 + n & 0 & 5 & @xmath309 & @xmath310 & realistic + nh & 0 & 5 & @xmath309 & @xmath310 & @xmath150 + r & 1 & 1 & @xmath311 & @xmath312 & realistic +

a partial interference cancellation ( pic ) group decoding based space - time block code ( stbc ) design criterion was recently proposed by guo and xia , where the decoding complexity and the code rate trade - off is dealt when the full diversity is achieved . in this paper , two designs of stbc are proposed for any number of transmit antennas that can obtain full diversity when a pic group decoding ( with a particular grouping scheme ) is applied at receiver . with the pic group decoding and an appropriate grouping scheme for the decoding , the proposed stbc are shown to obtain the same diversity gain as the ml decoding , but have a low decoding complexity . the first proposed stbc is designed with multiple diagonal layers and it can obtain the full diversity for two - layer design with the pic group decoding and the rate is up to symbols per channel use . but with pic - sic group decoding , the first proposed stbc can obtain full diversity for any number of layers and the rate can be full . the second proposed stbc can obtain full diversity and a rate up to with the pic group decoding . some code design examples are given and simulation results show that the newly proposed stbc can well address the rate - performance - complexity tradeoff of the mimo systems . diversity techniques , space - time block codes , linear receiver , partial interference cancellation .

0904.1812

space - time ( st ) coding is a bandwidth - efficient transmission technique that can improve the reliability of data transmission in mimo wireless systems @xcite . orthogonal space - time block coding ( ostbc ) is one of the most attractive st coding approaches because the special structure of orthogonality guarantees a full diversity and a simple ( linear ) maximum - likelihood ( ml ) decoding . the first ostbc design was proposed by alamouti in @xcite for two transmit antennas and was then extended by tarokh _ et . _ in @xcite for any number of transmit antennas . a class of ostbc from complex design with the code rate of @xmath2 was also given by tarokh _ et _ in @xcite . later , systematic constructions of complex ostbc of rates @xmath3 for @xmath4 or @xmath5 transmit antennas for any positive integer @xmath6 were proposed in @xcite . however , the ostbc has a low code rate not more than @xmath7 for more than two transmit antennas @xcite . to enhance the transmission rate of the stbc , various stbc design approaches were proposed such as quasi - ostbc @xcite and algebraic number theory based stbc @xcite . the quasi - ostbc increases the code rate by relaxing the orthogonality condition on the code matrix , which was originally proposed in @xcite , @xcite , and @xcite , independently . due to the group orthogonality , the ml decoding is performed pair - wise or group - wise with an increased complexity compared to the single - symbol decoding . in @xcite , quasi - ostbc was studied in the sense of minimum decoding complexity , i.e. , a real pair - wise symbols decoding . in @xcite , the pair - wise decoding was generalized to a general group - wise decoding . the decoding for these codes is the ml decoding and their rates are basically limited by that of ostbc . the algebraic number theory based stbc are designed mainly based on the ml decoding that may have high complexity and even though some near - ml decoder , such as sphere decoder @xcite can be used , the expected decoding complexity is still dominated by polynomial terms of a number of symbols which are jointly detected @xcite . to reduce the large decoding complexity of the high rate stbc aforementioned , several fast - decodable stbc were recently proposed @xcite@xcite . the stbc proposed in @xcite achieves a high rate and a reduced decoding complexity at the cost of loss of full diversity . the fast - decodable stbc in @xcite can obtain full rate , full diversity and the reduced ml decoding complexity , but the code design is limited to @xmath8 and @xmath9 mimo transmissions only . another new perspective of reducing the decoding complexity was recently considered in @xcite and @xcite to resort to conventional linear receivers such as zero - forcing ( zf ) receiver or minimum mean square error ( mmse ) receiver instead of the ml receiver to collect the full diversity . the outage and diversity of linear receivers in flat - fading mimo channels were studied in @xcite , but no explicit code design was given to achieve the full diversity when the linear receivers are used . based on the new stbc design criterion for mimo systems with linear receivers , toeplitz stbc @xcite and overlapped - alamouti codes @xcite were proposed and shown to achieve the full diversity with the linear receivers . recently , some other new designs of stbc with linear receivers were proposed @xcite . however , the code rate of stbc achieving full diversity with linear receivers is upper bounded by one . later , guo and xia proposed a partial interference cancellation ( pic ) group decoding scheme @xcite which can be viewed as an intermediate decoding approach between the ml receiver and the zf receiver by trading a simple single - symbol decoding complexity for a high code rate larger than one symbol per channel use . moreover , in @xcite an stbc design criterion was given to achieve full diversity when the pic group decoding is applied at the receiver . the proposed pic group decoding in @xcite was also connected with the successive interference cancellation ( sic ) strategy to aid the decoding process , referred to as pic - sic group decoding . a few code design examples were presented in @xcite , but a general design of stbc achieving full diversity with the pic group decoding remains an open problem . in this paper , we propose two designs of stbc which can achieve full diversity with the pic group decoding for any number of transmit antennas . the first proposed stbc have a structure of multiple diagonal layers and for each diagonal layer there are exactly @xmath10 coded symbols embedded , being equal to the number of transmit antennas , which are obtained from a cyclotomic lattice design . indeed , each diagonal layer of the coded symbols can be viewed as the conventional rate - one diagonal stbc @xcite . the code rate of the proposed stbc can be from one to @xmath10 symbols per channel use by adjusting the codeword length , i.e. , embedding different number of layers in the code matrix . with the pic group decoding the code rate of the first proposed full - diversity stbc can be only up to @xmath0 symbols per channel use , i.e. , for two layers . for more than two layers embedded in the codeword , the code rate is increased at the cost of losing full diversity with the pic group decoding . however , with the pic - sic group decoding , the proposed stbc with arbitrary number of layers can obtain full diversity and the code rate can be up to @xmath10 . the second proposed stbc is designed with three layers of information symbols embedded in the codeword and the pic group decoding can be performed in three separate groups accordingly . without loss of decoding complexity compared to the first proposed stbc , the second proposed stbc can achieve full diversity and a code rate larger than @xmath0 . note that the code rate for the first proposed full - diversity stbc with pic group decoding can not be above @xmath0 . in the pic group decoding of the proposed stbc , every @xmath10 neighboring columns of the equivalent channel matrix are clustered into one group . this paper is organized as follows . a system model of st transmission over mimo channels with the pic group decoding is introduced in section [ sec : system ] . in section [ sec : new ] , a design of high rate stbc with the pic group decoding is proposed , which contains multiple diagonal layers of coded symbols . for a particular code design with two diagonal layers , the full diversity with the pic group decoding is proved . for the code with pic - sic group decoding , the full diversity is shown for any number of diagonal layers . several full - diversity code design examples are given in section [ sec : example ] . in section [ sec : xu ] , another design of high rate stbc with the pic group decoding is proposed , which can achieve full diversity with three layers . simulation results are presented in section [ sec : sim ] . finally , in section [ sec : conclusion ] , we draw our conclusions . _ notations _ : column vectors ( matrices ) are denoted by boldface lower ( upper ) case letters . superscripts @xmath11 and @xmath12 stand for transpose and conjugate transpose , respectively . @xmath13 denotes the field of complex numbers . @xmath14 denotes the @xmath15 identity matrix , and @xmath16 denotes the @xmath17 matrix whose elements are all @xmath18 . @xmath19 is the vectorization of matrix @xmath20 by stacking the columns of @xmath20 on top each other . in this section , we first briefly describe the system model and then describe the pic group decoding proposed in @xcite . we consider a mimo transmission with @xmath10 transmit antennas and @xmath21 receive antennas over block fading channels . the received signal matrix @xmath22 is @xmath23 where @xmath24 is the codword matrix , transmitted over @xmath25 time slots , @xmath26 is a noise matrix with independent and identically distributed ( i.i.d . ) entries being circularly symmetric complex gaussian distributed @xmath27 , @xmath28 is the channel matrix whose entries are also i.i.d . with the distribution @xmath27 , @xmath29 denotes the average signal - to - noise ratio ( snr ) per receive antenna and @xmath30 is the normalization factor to ensure that the average energy of the coded symbols transmitting from all antennas during one symbol period is @xmath31 . the realization of @xmath32 is assumed to be known at the receiver , but not known at the transmitter . therefore , the signal power is allocated uniformly across the transmit antennas . let @xmath33 be the number of independent information symbols @xmath34 per codeword @xmath20 , selected from a complex constellation @xmath35 . the code rate of the stbc is defined as @xmath36 symbols per channel use . if @xmath37 , the stbc is said to have full rate , i.e. , @xmath38 symbols per channel use . in this paper , we consider that information symbols @xmath34 are coded by linear dispersion stbc as @xmath39 where @xmath40 is the linear stbc matrix . to decode the transmitted sequence @xmath41 at the receiver , we need to extract @xmath41 from @xmath20 . this can be done by as follows . by substituting ( [ eqn : lstbc ] ) into ( [ eqn : y ] ) , we get @xmath42 then , by taking vectorization of the matrix @xmath43 we have @xmath44 where @xmath45 , @xmath46 , @xmath47^t$ ] , and @xmath48 is an equivalent channel matrix , @xmath49\end{aligned}\ ] ] with the @xmath50-th column @xmath51 , @xmath52 . for a zf receiver , the estimate @xmath53 of the transmitted symbol sequence @xmath41 is , @xmath54 where @xmath55 . equivalently , it can be written as the single - symbol decoding as follows , @xmath56_{l , : } \mathbf{y}-s_l\right\|^2,\,\,\,\,l=1,2\cdots , l , \end{aligned}\ ] ] where @xmath57_{l,:}$ ] denotes the @xmath50-th row of @xmath58 . for an ml receiver , the estimate of @xmath59 that achieves the minimum of the squared frobenius norm is given by @xmath60 in the ml decoding , computations of squared frobenius norms for all possible codewords are needed and therefore result in prohibitively huge computational complexity when the length of the information symbols vector to be decoded is large . in the following , we give a metric to evaluate the computational complexity of the ml decoding , which is the same as the one shown in ( * ? ? ? * definition 2 ) . the decoding complexity @xmath61 is defined as the number of squared frobenius norms @xmath62 that should be computed in the decoding process . with the above definition , we have the following two remarks . the decoding complexity of the zf detection is @xmath63 , i.e. , @xmath33 times of the cardinality of the signal constellation . it is equivalent to the single - symbol decoding complexity . the decoding complexity of the ml detection is @xmath64 , i.e. , the complexity of the _ full _ exhaustive search of all @xmath33 information symbols drawn from the constellation @xmath35 . we next describe the pic group decoding studied in @xcite . define index set @xmath65 as @xmath66 where @xmath33 is the number of information symbols in @xmath41 . we then partition @xmath65 into @xmath67 groups : @xmath68 with @xmath69 where @xmath70 is the cardinality of the subset @xmath71 . we call @xmath72 a grouping scheme . for such a grouping scheme , we have @xmath73 define @xmath74^t,\,\,\,p=1 , \cdots , p.\\ \mathbf{g}_{p}&=&\left[\begin{array}{cccc } \mathbf{g}_{i_{p,1 } } & \mathbf{g}_{i_{p,2 } } & \cdots & \mathbf{g}_{i_{p , l_p } } \end{array } \right],\,\,\,p=1 , \cdots , p.\end{aligned}\ ] ] with these notations , ( [ eqn : y2 ] ) can be written as @xmath75 suppose we want to decode the symbols embedded in the group @xmath76 . the pic group decoding first implements linear interference cancellation with a suitable choice of matrix @xmath77 in order to completely eliminate the interferences from other groups @xcite , i.e. , @xmath78 , @xmath79 and @xmath80 . then , we have @xmath81 where the interference cancellation matrix @xmath77 can be chosen as follows @xcite , @xmath82 in case @xmath83 , \end{aligned}\ ] ] has full column rank . if @xmath84 does not have full column rank , then we need to pick a maximal linear independent vector group from @xmath84 and in this case a projection matrix @xmath77 can be found too @xcite . afterwards , the symbols in the group @xmath76 are decoded with the ml decoding algorithm as follows , @xmath85 the above pic group decoding is connected to some of the known decodings as in the following remarks . for one special case of @xmath86 , the grouping scheme is @xmath87 with @xmath88 . from ( [ eqn : qp ] ) , we have @xmath89 . then , the pic group decoding is equivalent to the ml decoding where all information symbols are jointly decoded . for the special case of @xmath90 , the grouping scheme is @xmath91 i.e. , every single symbol is regarded as one group . then , the pic group decoding is equivalent to the zf decoding where every single symbol is separated from all the other symbols and then decoded . the pic group decoding with @xmath92 can be viewed as an intermediate decoding approach between the ml decoding and the zf decoding . alternatively , the ml decoding and the zf decoding can both be regarded as the special cases of the pic group decoding corresponding to @xmath86 and @xmath90 , respectively . for the pic group decoding , the following two steps are needed : the group zero - forcing to cancel the interferences coming from all the other groups as shown in ( [ eqn : gzf ] ) and the group ml decoding to jointly decode the symbols in one group as shown in ( [ eqn : gml ] ) . therefore , the decoding complexity of the pic group decoding should reside in the above two steps . note that the interference cancellation process shown in ( [ eqn : gzf ] ) mainly involves with linear matrix computations , whose computational complexity is small compared to the ml decoding for an exhaustive search of all candidate symbols . therefore , to evaluate the decoding complexity of the pic group decoding , we mainly focus on the computational complexity of the ml decoding within the pic group decoding algorithm . according to _ definition 2 _ , the ml decoding complexity in the pic group decoding algorithm is @xmath93 . it can be seen that the pic group decoding provides a flexible decoding complexity which can be from the zf decoding complexity @xmath94 to the ml decoding complexity @xmath95 . in @xcite , an sic - aided pic group decoding algorithm , namely pic - sic group decoding was proposed . similar to the blast detection algorithm @xcite , the pic - sic group decoding is performed after removing the already - decoded symbol set from the received signals to reduce the interference . if each group has only one symbol , then the pic - sic group will be equivalent to the blast detection . the performance of a decoding algorithm for a wireless communication system is related to the diversity order . if the average probability of a detection error for communication over a fading channel usually behaves as : @xmath96 where @xmath97 is a constant and @xmath98 is called the _ diversity order _ of the system . for an mimo communication system , the maximum diversity order is @xmath99 , i.e. , the product of the number of transmit antennas and the number of receiver antennas . in order to optimize the reception performance of the mimo system , a full diversity is usually pursued which can be achieved by a proper signal transmission scheme or data format ( e.g. , stbc ) . in @xcite , the `` rank - and - determinant criterion '' of stbc design was proposed to maximize both the diversity gain @xmath98 and the coding gain @xmath100 of the mimo system with an ml decoding . recently , in @xcite an stbc design criterion was derived to achieve full diversity when the pic group decoding is used at the receiver . in the following , we cite the main result of the stbc design criterion proposed in @xcite . [ prop1 ] ( * ? ? ? * theorem 1 ) [ _ full - diversity criterion under pic group decoding _ ] for an stbc @xmath20 with the pic group decoding , the full diversity is achieved when 1 . the code @xmath20 satisfies the full rank criterion , i.e. , it achieves full diversity when the ml receiver is used ; _ and _ 2 . @xmath101 are linearly independent vector groups for any @xmath102 . in @xcite , the stbc achieving full diversity with pic group decoding were proposed for @xmath0 and @xmath103 transmit antennas . however , a systematic code design of the full - diversity stbc with pic group decoding remains an open problem . [ prop2 ] @xcite [ _ full - diversity criterion under pic - sic group decoding _ ] for an stbc @xmath20 with the pic - sic group decoding , the full diversity is achieved when 1 . the code @xmath20 satisfies the full rank criterion , i.e. , it achieves full diversity when the ml receiver is used ; _ and _ 2 . at each decoding stage , @xmath104 , which corresponds to the current to - be decoded symbol group @xmath105 , the remaining groups @xmath106 corresponding to yet uncoded symbol groups are linearly independent vector groups for any @xmath102 . in this section , we first propose a systematic design of high - rate stbc which has a rate up to @xmath10 symbols per channel use and achieves full diversity with the ml decoding . the systematic design of the stbc is structured with multiple diagonal layers . then , we prove that the proposed stbc with two diagonal layers can obtain full diversity with the pic group decoding and the code rate can be up to @xmath0 symbols per channel use . finally , we prove that the proposed stbc with any number of diagonal layers can obtain full diversity with pic - sic group decoding and the code rate can be up to @xmath10 symbols per channel use . our proposed space - time code @xmath107 , i.e. , @xmath20 in ( [ eqn : y ] ) , is of size @xmath108 ( for any given @xmath25 , @xmath10 and @xmath109 ) and will be transmitted from @xmath10 antennas over @xmath25 time slots . let @xmath110 . the symbol stream @xmath34 ( composed of @xmath111 complex symbols chosen from qam constellation and then scaled by @xmath112}$ ] ) is first parsed into @xmath113 symbol vectors @xmath76 ( @xmath114 ) . each symbol vector is linearly precoded by an @xmath115 matrix @xmath116 , which is a chosen constellation rotation matrix . next , the @xmath113 vector @xmath117 is used to form the space - time code matrix @xmath107 , in which the @xmath118-th descending diagonal from left to right is the diagonal form of @xmath117 . the resulting transmitted code matrix @xmath107 is given by @xmath119 , \end{aligned}\ ] ] where the @xmath118-th descending diagonal from left to right , denoted by @xmath120^t$ ] is given by @xmath121 and the @xmath113 information symbol vector @xmath76 is given by @xmath122^t , \end{aligned}\ ] ] for @xmath123 . the proposed stbc in ( [ eqn : code ] ) has asymptotically full rate when the block length is sufficiently large . in the codeword in ( [ eqn : code ] ) , a total number of @xmath124 independent information symbols are encoded into the codeword @xmath107 , which is then transmitted from @xmath10 antennas over @xmath25 time slots . the code rate of transmission is therefore @xmath125 for a very large block length @xmath25 , it can be seen that the rate @xmath126 of the proposed st coding scheme approaches @xmath10 symbols per channel use , i.e. the full rate . in @xcite , the rotation matrix @xmath116 was designed for diagonal stbc to achieve the full diversity gain and the optimal diversity product . with the optimal cyclotomic lattices design for @xmath10 transmit antennas , from ( * ? ? ? * table i ) we can get a set of integers @xmath127 and let @xmath128 . then , the optimal lattice @xmath116 is given by ( * ? ? ? * eq . ( 16 ) ) @xmath129 . \end{aligned}\ ] ] where @xmath130 with @xmath131 and @xmath132 are distinct integers such that @xmath133 and @xmath134 are co - prime for any @xmath135 . _ example 1 _ : for @xmath103 transmit antennas we can choose @xmath136 and @xmath137 according to ( * ? ? ? * table i ) . then , in order to ensure that @xmath133 and @xmath134 are co - prime for any @xmath135 we can obtain @xmath138 , @xmath139 , @xmath140 . when @xmath141 , the signal constellation is located on the equal literal triangular lattice . when @xmath142 , @xmath143 can be @xmath103 and @xmath144 can be @xmath145 , and in this case the signal constellation is located on the square lattice . _ example 2 _ : for @xmath146 transmit antennas we can select @xmath147 and @xmath148 . then , @xmath138 , @xmath139 , @xmath149 , @xmath150 . the cyclotomic design of the matrix @xmath116 is vital for the design of the algebraic stbc . in the following , we show some properties of the matrix @xmath116 that will be used later for our design . @xcite the diagonal cyclotomic st code @xmath151 defined by @xmath152\right\}$ ] achieves full diversity under ml decoding , where @xmath153^t=\mathbf{\theta } \left[\begin{array}{cccc } s_1 & s_2 & \cdots & s_m \end{array } \right]^t$ ] and @xmath116 is given by ( [ eqn : cyclo ] ) . every entry of the matrix @xmath116 in ( [ eqn : cyclo ] ) is non - zero . this property is obvious from ( [ eqn : cyclo ] ) . we show the main result of the proposed stbc when an ml decoding is used at the receiver , as follows . consider a mimo transmission with @xmath10 transmit antennas and @xmath21 receive antennas over block fading channels . the stbc @xmath107 as described in ( [ eqn : code ] ) achieves full diversity under the ml decoding . in order to prove that the st code @xmath107 in ( [ eqn : code ] ) can obtain full diversity under ml decoding , it is sufficient to prove that @xmath154 achieves full rank for any distinct pair of st codewords @xmath107 and @xmath155 . for any pair of distinct codewords @xmath107 and @xmath155 , there exists at least one index @xmath118 ( @xmath156 ) such that @xmath157 , where @xmath158 and @xmath159 are related to @xmath76 and @xmath160 from ( [ eqn : rm ] ) , respectively . let @xmath118 denote the minimum index of vectors satisfying @xmath157 . then , for any index @xmath161 with @xmath162 , it must have @xmath163 . define @xmath164 as the difference between symbols @xmath165 and @xmath166 . then , from ( [ eqn : code ] ) @xmath167 can be expressed as @xmath168 , \end{aligned}\ ] ] where @xmath169 for @xmath170 . this is because for @xmath157 , it exists @xmath171 . due to the suitably chosen constellation rotation matrix @xmath116 in ( [ eqn : cyclo ] ) , @xmath172 must have nonzero entries for any @xmath173 . then , the matrix @xmath167 has full rank . the full rankness of @xmath167 can be examined similar to that for the toeplitz code ( or delay diversity code ) @xcite by checking if the columns of @xmath167 are linearly independent . specifically , we establish @xmath174 with @xmath175^t$ ] . first , we examine the @xmath118-th equation in @xmath174 and get @xmath176 . because @xmath169 for @xmath170 , @xmath177 . then , we examine the @xmath178-th equation and get @xmath179 . immediately , @xmath180 . likewise , we examine the @xmath181-th equation until @xmath182-th equation , and we can get @xmath183 . therefore , all columns of @xmath167 are linearly independent and @xmath167 has full rank . this property will be used in next section in the proof of the full diversity property under the pic group decoding . we show the main result of the proposed stbc when a pic group decoding with a particular grouping scheme is used at the receiver , as follows . consider a mimo transmission with @xmath10 transmit antennas and @xmath21 receive antennas over block fading channels . the stbc @xmath107 as described in ( [ eqn : code ] ) with two diagonal layers ( i.e. , @xmath184 ) is used at the transmitter . the equivalent channel matrix is @xmath185 . if the received signal is decoded using the pic group decoding with the grouping scheme @xmath186 where @xmath187 for @xmath188 , i.e. , the size of each group is equal to the number of transmit antennas @xmath10 , then the code @xmath107 achieves the full diversity . the code rate of the full - diversity stbc can be up to @xmath0 symbols per channel use . in order to prove _ theorem 2 _ , let us first introduce the following lemmas . consider the system as described in _ theorem 2 _ with @xmath189 and the stbc @xmath107 as given by ( [ eqn : code ] ) , 1 . the equivalent channel matrix @xmath190 can be expressed as @xmath191 \end{aligned}\ ] ] where @xmath192,\,\,\,\ , p=1,2 , \cdots , p , \end{aligned}\ ] ] 2 . when @xmath184 , @xmath193 and @xmath194 are linearly independent vector groups as long as @xmath195 , where @xmath196 . a proof of _ lemma 1 _ is given in appendix i. consider the system as described in _ theorem 2 _ and the stbc @xmath107 as given by ( [ eqn : code ] ) . for the equivalent channel matrix @xmath197 , @xmath101 are linearly independent vector groups for @xmath195 when @xmath198 if and only if @xmath101 are linearly independent vector groups for @xmath195 when @xmath189 . the proof of _ lemma 2 _ is straightforward and is also the same as what is mentioned in @xcite . as shown in _ proposition [ prop1 ] _ , a codeword @xmath107 with pic group decoding can obtain the full diversity if 1 ) : : @xmath107 achieves the full diversity with the ml receiver , and 2 ) : : @xmath101 are linearly independent vector groups as long as @xmath195 . for the proposed code @xmath107 in ( [ eqn : code ] ) with @xmath184 , the first condition is satisfied as shown in _ theorem 1_. the second condition is satisfied as shown in _ lemma 1 _ for @xmath189 and _ lemma 2 _ for @xmath198 , respectively . therefore , the code @xmath107 in ( [ eqn : code ] ) with @xmath184 can obtain full diversity with the pic group decoding provided that the grouping scheme is @xmath186 where @xmath187 for @xmath188 . the code rate of the full - diversity stbc with the pic group decoding can be derived by substituting @xmath184 and @xmath199 into ( [ eqn : rate ] ) as @xmath200 for a large number of transmit antennas , the rate approaches to ( but not larger than ) @xmath0 symbols per channel use . the decoding complexity of the pic group decoding of the proposed stbc with the grouping scheme as described in _ theorem 2 _ is @xmath201 . the decoding complexity of the proposed stbc with the pic group decoding is equivalent to the ml decoding of @xmath10 independent information symbols jointly . as shown in ( [ eqn : rate ] ) , the code rate of the proposed stbc in ( [ eqn : code ] ) for a given @xmath10 can be increased by embedding larger number of groups in the codeword , i.e. , increasing the value of @xmath67 . it is noteworthy to mention that the increase of the code rate does not result in the increase of the decoding complexity . for the proposed stbc with any number of layers and the pic - sic group decoding we have the following results . consider a mimo transmission with @xmath10 transmit antennas and @xmath21 receive antennas over block fading channels . the stbc @xmath107 as described in ( [ eqn : code ] ) with @xmath67 diagonal layers is used at the transmitter . the equivalent channel matrix is @xmath185 . if the received signal is decoded using the pic - sic group decoding with the sequential order and with the grouping scheme being @xmath202 where @xmath203 for @xmath204 , i.e. , the size of each group is equal to the number of transmit antennas @xmath10 , then the code @xmath107 achieves the full diversity . the code rate of the full - diversity stbc can be up to @xmath10 symbols per channel use . the proof of this theorem is simple . observing that @xmath205 $ ] and the group @xmath206 is linearly independent from the groups @xmath207 for any @xmath118 and @xmath208 where @xmath209 is given by ( [ eqn : g ] ) , according to _ proposition 2 _ the full diversity can be easily proved . the detailed proof is omitted . in this section , we show a few code design examples . we denote @xmath210 the code constructed by ( [ eqn : code ] ) for given parameters : @xmath10 the number of transmit antennas , @xmath25 the block length of the code , and @xmath67 the number of groups to be decoded in the pic group decoding . for notational brevity , we only show the equivalent channel of the proposed codes for miso systems . consider a code for 2 transmit antennas with 3 time slots . according to the code structure ( [ eqn : code ] ) , we have @xmath211,\end{aligned}\ ] ] where @xmath212^t=\mathbf{\theta}[\begin{array}{cc } s_1 & s_2 \end{array } ] ^t$ ] and @xmath213^t=\mathbf{\theta}[\begin{array}{cc } s_3 & s_4 \end{array } ] ^t$ ] . the constellation rotation matrix @xmath116 can be chosen as @xmath214,\ ] ] where @xmath215 and @xmath216 with @xmath217 @xcite . the code rate of the code is @xmath218 . in fact , this code is equivalent to the one proposed in ( * ? ? ? * section vi - _ example 1 _ ) . the equivalent channel of the code @xmath219 is given by @xmath220 . \end{aligned}\ ] ] the grouping scheme for the pic group decoding is @xmath221 and @xmath222 . it can be seen that @xmath193 and @xmath194 are linearly independent . then , the code can obtain full diversity with the pic group decoding . for given @xmath223 , the code achieving full diversity with the pic group decoding can be designed as follows , @xmath224 . \end{aligned}\ ] ] this code has a code rate of @xmath225 and two groups to be decoded . the equivalent channel of the code @xmath226 is @xmath227 & \left[\begin{array}{c } \mathbf{0}_{1\times 4 } \\ \mathbf{b } \end{array } \right ] \end{array } \right ] , \end{aligned}\ ] ] where @xmath228 is given by @xmath229 \end{aligned}\ ] ] with @xmath230 being the ( @xmath231)-th entry of the matrix @xmath116 for @xmath232 . the grouping scheme for the pic group decoding is @xmath233 and @xmath234 . it can be seen that the groups @xmath193 and @xmath194 are linearly independent to each other . then , the code can obtain full diversity with the pic group decoding . consider @xmath235 time slots . we can get @xmath236 . \end{aligned}\ ] ] the code rate of the code @xmath237 is @xmath218 which has the same rate as the one proposed in ( * ? ? ? * section vi - _ example 2 _ ) . the equivalent channel of the code @xmath237 is @xmath238 & \left[\begin{array}{c } \mathbf{0}_{2\times 4 } \\ \mathbf{b } \end{array } \right ] \end{array } \right ] , \end{aligned}\ ] ] where @xmath228 is given by ( [ eqn : b ] ) . because the groups @xmath193 and @xmath194 are linearly independent to each other . then , the code can obtain full diversity with the pic group decoding . moreover , we can also design the code for @xmath235 with @xmath239 layers ( i.e. , @xmath240 ) as follows , @xmath241 . \end{aligned}\ ] ] the code rate of the code @xmath242 is @xmath0 and the equivalent channel is given by @xmath243 , \end{aligned}\ ] ] where @xmath244 , & \mathbf{g}_2=\left[\begin{array}{c } \mathbf{0}_{1\times 4}\\ \mathbf{b } \\ \mathbf{0}_{1\times 4 } \end{array } \right ] , & \mathbf{g}_3=\left[\begin{array}{c } \mathbf{0}_{1\times 4 } \\ \mathbf{0}_{1\times 4}\\ \mathbf{b } \\ \end{array } \right ] . \end{array}\nonumber \end{aligned}\ ] ] it can be proved that the groups @xmath193 , @xmath194 , and @xmath245 are not linearly independent groups . therefore , according to _ proposition 1 _ , the code @xmath242 can not achieve the full diversity with the pic group decoding . however , the code @xmath242 can obtain full diversity with pic - sic group decoding . this is because @xmath193 is linearly independent from @xmath194 and @xmath245 , and @xmath194 is linearly independent from @xmath245 . according to _ proposition 2 _ , with pic - sic group decoding and a proper decoding order @xmath246 or @xmath247 , the code @xmath242 can achieve the full diversity . for given @xmath235 and @xmath184 , the code is designed as follows , @xmath248 . \end{aligned}\ ] ] the code rate of the code @xmath249 is @xmath250 . the equivalent channel is @xmath251 & \left[\begin{array}{c } \mathbf{0}_{1\times 5 } \\ \mathrm{diag}(\mathbf{h})\mathbf{\theta}_5 \end{array } \right ] \end{array } \right ] , \end{aligned}\ ] ] where @xmath252 is the rotation matrix of size @xmath253 . the grouping scheme for the pic group decoding of @xmath249 is @xmath254 and @xmath255 . it can be seen that the groups @xmath193 and @xmath194 are linearly independent to each other . then , the code @xmath249 can obtain full diversity with the pic group decoding . notice that the code design in ( [ eqn : code ] ) can only achieve the full diversity with pic group decoding for two diagonal layers ( r.f . _ theorem 2 _ ) and the code rate is not larger than @xmath0 symbols per channel use . with @xmath67 ( @xmath256 ) diagonal layers in the code ( [ eqn : code ] ) , the rate can be increased but the independence among @xmath67 channel groups @xmath257 in ( [ eqn : h0 ] ) is not satisfied , thereby may lose the full diversity gain . in this section , we propose a new code design which can achieve full diversity with pic group decoding and a rate above @xmath0 . for @xmath258 ( @xmath118 is an integer ) , our proposed stbc @xmath259 for @xmath10 transmit antennas is given by @xmath260,\ ] ] where the symbol vector @xmath261^t$ ] is given by @xmath262 @xmath116 is an @xmath115 constellation rotation matrix given by ( [ eqn : cyclo ] ) and @xmath263^t$ ] is the @xmath264 information symbol vector . for @xmath265 , our proposed stbc @xmath259 is given by @xmath266.\ ] ] for @xmath267 , our proposed stbc @xmath259 is given by @xmath268.\end{aligned}\ ] ] the proposed stbc has the rate as follows : @xmath269 for @xmath258 , in the codeword ( [ eqn : codeword ] ) @xmath270 information symbols are sent over @xmath271 time slots . thus , the code rate is @xmath272 . similarly , it is easy to prove the code rate for cases of @xmath265 and @xmath267 . [ asymptotic rate ] it is obvious that the code rate ( [ eqn : rate2 ] ) approaches to @xmath1 when a large number of transmit antennas are used . its full diversity property will be proved in the next subsection . however , the full diversity code proposed by the first design in ( [ eqn : code ] ) can not achieve a rate more than 2 , which was shown in _ theorem 2_. [ rate comparison ] note that the code design in ( [ eqn : code ] ) can achieve full diversity with the pic group decoding for @xmath184 only and the rate is @xmath273 . in table i , the comparison of the code rate between the first code design in ( [ eqn : code ] ) and the second design in ( [ eqn : codeword])-([eqn : codeword3 ] ) is given . c + + @xmath274 + @xmath275 + @xmath276 + @xmath277 + @xmath278 + @xmath279 + @xmath280 + & c code ( [ eqn : code ] ) + .comparison in code rate @xmath126 ( symbols per channel use ) [ cols="^",options="header " , ] + + [ decoding complexity ] the decoding complexity of the proposed stbc with the pic group decoding is equivalent to the ml decoding of @xmath10 independent information symbols jointly . according to _ definition 2 _ , the ml decoding complexity in the pic group decoding algorithm is @xmath281 . next , we show that the proposed stbc in ( [ eqn : codeword ] ) achieves full diversity when a pic group decoding is used at the receiver . let the stbc @xmath107 as described in ( [ eqn : codeword ] ) be used at the transmitter . there are @xmath10 transmit antennas and @xmath31 receive antenna . if the received signal is decoded using the pic group decoding with the grouping scheme @xmath282 , where @xmath283 for @xmath284 , then the code @xmath107 achieves the full diversity . in order to prove the _ theorem 4 _ , let us first introduce the following lemma . consider the system as described in _ theorem 4 _ with @xmath189 , and the channel matrix @xmath285^t$ ] . let @xmath286 , where @xmath287 denotes the @xmath288th row of the rotation matrix @xmath116 for @xmath289 . then , the equivalent channel matrix is given by @xmath290=\left [ \begin{array}{ccc } \mathbf{g}_1&0&0\\ 0&\mathbf{g}_2&0\\ 0&0&\mathbf{g}_3\\ \mathbf{g}_4&\mathbf{g}_1&0\\ 0&\mathbf{g}_5&\mathbf{g}_2\\ \mathbf{g}_3&0&\mathbf{g}_6\\ \vdots&\vdots&\vdots\\ \mathbf{g}_{3p-2}&\mathbf{g}_{3p-5}&0\\ 0&\mathbf{g}_{3p-1}&\mathbf{g}_{3p-4}\\ \mathbf{g}_{3p-3}&0&\mathbf{g}_{3p}\\ \mathbf{g}_2&\mathbf{g}_{3p-2}&0\\ 0&\mathbf{g}_3&\mathbf{g}_{3p-1}\\ \mathbf{g}_{3p}&0&\mathbf{g}_1\\ \mathbf{g}_5&\mathbf{g}_6&\mathbf{g}_4\\ \vdots&\vdots&\vdots\\ \mathbf{g}_{3p-1}&\mathbf{g}_{3p}&\mathbf{g}_{3p-2 } \end{array } \right].\ ] ] the proof of _ lemma 3 _ is straightforward and it is easy to verify that @xmath291^t$ ] . firstly , we show that the proposed stbc @xmath107 in ( [ eqn : codeword ] ) achieves the full diversity with ml decoding , i.e. , @xmath292 achieves full rank for any distinct pair of codewords @xmath20 and @xmath293 . since @xmath20 and @xmath293 are distinct , at least one pair of the information symbol vectors @xmath294 and @xmath295 are different . suppose only one pair of the information symbol vectors are distinct , say @xmath296 and @xmath297 . by the _ property 1 _ , @xmath298 for all @xmath299 . in this case , the matrix @xmath300 is exactly a diagonal matrix with the rows rearranged . thus , it achieves full rank . if all three pairs of the information symbol vectors are distinct , then @xmath301 for all @xmath284 and @xmath299 . it is easy to see from the codeword structure that @xmath300 is exactly a lower triangular matrix with nonzero diagonal entries . thus , it achieves full rank . if two pairs of the information symbol vectors are distinct and the other one pair of the information symbol vectors are the same , say @xmath302 . after replacing the ( @xmath303)th row by the ( @xmath304)th row for @xmath305 , the matrix @xmath300 becomes a lower triangular matrix with nonzero diagonal entries . thus , it achieves full rank . as shown above , @xmath300 has full rank for all @xmath306 \neq[\begin{array}{ccc } \mathbf{s}'_1 & \mathbf{s}'_2 & \mathbf{s}'_3 \end{array } ] $ ] , i.e. , @xmath307 . next , we show that @xmath193 , @xmath194 and @xmath245 are linearly independent vector groups as long as @xmath308 . if @xmath309 , it is obvious that @xmath193 is linearly independent of @xmath194 and @xmath245 . otherwise , @xmath310 , if @xmath311 , @xmath193 is also independent of @xmath194 and @xmath245 . therefore , we can reorder @xmath32 as @xmath312 suppose @xmath313 is the first one in the sequence such that @xmath314 . according to the _ property 2 _ , the entries of @xmath315 in @xmath193 are not zeros , but the entries in the same row of @xmath194 and @xmath245 are either 0 or @xmath286 with @xmath316 , which implies that @xmath193 can not be expressed as the linear combination of @xmath194 and @xmath245 . similarly , in order to prove @xmath194 and @xmath245 are independent of the remaining vector groups , we only need to rearrange the channel matrix as @xmath317 and @xmath318 respectively . hence , @xmath193 , @xmath194 and @xmath245 are linearly independent vector groups as long as @xmath308 . according to _ proposition 1 _ , the code @xmath107 in ( [ eqn : codeword ] ) achieves full diversity using pic grouping decoding with the group scheme @xmath282 . similarly , we can prove that the codes @xmath319 and @xmath320 achieve full diversity with pic grouping decoding . for @xmath276 . it has @xmath321 . @xmath322 , \end{aligned}\ ] ] the rate of this code is @xmath323 . the equivalent channel of the code @xmath324 is @xmath325 , \end{aligned}\ ] ] where the @xmath326 row vector @xmath327 , @xmath328 , with @xmath329 being the @xmath288th row of the matrix @xmath116 . the grouping scheme for the pic group decoding is @xmath233 , @xmath330 and @xmath331 . obviously , the code can obtain the full diversity with the pic group decoding . for @xmath278 . it has @xmath332 . @xmath333 , \end{aligned}\ ] ] the rate of this code is @xmath334 . the equivalent channel of the code @xmath335 is @xmath336 , \end{aligned}\ ] ] where the @xmath337 row vector @xmath327 , @xmath338 , with @xmath329 being the @xmath288th row of the matrix @xmath116 . the grouping scheme for the pic group decoding is @xmath339 , @xmath340 and @xmath341 . obviously , the code can obtain the full diversity with the pic group decoding . a code example of @xmath342 is given by @xmath343.\ ] ] in this section , simulation results of the proposed stbc with the pic group decoding over rayleigh fading channels are presented . we first show bit error rate ( ber ) performance of the codes proposed in this paper for four transmit antennas and compare them to the one proposed in @xcite . specifically , we consider three stbc for four transmit antennas proposed in this paper , i.e. , @xmath242 in ( [ eqn : c463 ] ) , @xmath237 in ( [ eqn : c462 ] ) and @xmath226 in ( [ eqn : c452 ] ) , and then compare them with guo - xia s code given in ( * ? ? ? * section vi - _ example 2 _ ) . in order to make a fair performance comparison , we keep the same bandwidth efficiency of @xmath344 bps / hz . thus , we use 16qam for the code @xmath242 ( the code rate is @xmath345 ) and 64qam for both codes @xmath237 ( the code rate is @xmath346 ) and guo - xia s code ( the code rate is @xmath346 ) . for the code @xmath226 , because it has a code rate @xmath347 we use 64qam and thus its bandwidth efficiency is @xmath348 bps / hz higher than the other three codes . since we use square qam , in the rotation matrix @xmath116 we use @xmath142 in _ example 1 _ of section [ sec : new ] . [ fig : c452 ] shows the performance of the proposed code @xmath226 with various detection approaches for a @xmath349 mimo system over rayleigh fading channels . it can be seen that the code with pic and pic - sic group decoding algorithms has both less than @xmath31 db snr performance loss compared to the one with the ml decoding . the performance loss of the pic and pic - sic group decoding is trade for the largely reduced decoding complexity . specifically , for 16qam signaling the ml decoding complexity is @xmath350 and the complexity of the pic group decoding is @xmath351 . moreover , it is obvious that the code @xmath226 with the pic group decoding and the pic - sic group decoding can both obtain full diversity at high snr . blast and zf detection of the code @xmath226 can not obtain full diversity . [ fig : c463 ] shows the performance of the proposed code @xmath242 with various detection approaches for a @xmath349 mimo system over rayleigh fading channels . it is shown that the code @xmath242 with pic group decoding does not guarantee the full diversity at high snr . however , with pic - sic group decoding the code @xmath242 can achieve the full diversity . the performance gap between the ml decoding and the pic group decoding is around 2 db snr loss , but the decoding complexity is significantly reduced by the pic group decoding from @xmath352 to @xmath353 . again , zf and blast detection can not obtain the full diversity . [ fig : tast ] presents the performance comparison between tast code @xcite and the proposed code @xmath242 with ml , pic and pic - sic group decoding algorithms , respectively , for @xmath349 systems over rayleigh fading channels . note the rate of tast code is full , i.e. , @xmath103 symbols per channel use for @xmath103 transmit antennas . to keep the bandwidth efficiency of @xmath344 bps / hz , 4-qam is used for tast code and 16-qam is used for the code @xmath242 . it is shown in fig . [ fig : tast ] that tast code with the ml decoding gives the best performance . however , with pic and pic - sic group decoding algorithms tast code will lose full diversity . the proposed code @xmath242 can obtain much better performance than tast code when both use the pic and pic - sic group decoding algorithms . [ fig : perfect ] gives the performance comparison between perfect stbc @xcite and the proposed code @xmath242 with ml , pic and pic - sic group decoding algorithms , respectively , for @xmath349 systems over rayleigh fading channels . similar to tast code , perfect stbc can not obtain full diversity with the pic and pic - sic group decoding . the proposed code @xmath242 has a much better performance than perfect stbc while admitting low complexity in the decoding . fig . [ fig : codes ] shows the ber performance of various codes with the pic group decoding for @xmath103 transmit and @xmath103 receive antennas . it can be seen that the code @xmath237 has very similar performance to guo - xia s code . moreover , the code @xmath226 has @xmath31 db loss compared to guo - xia s code . this is because it has a higher bandwidth efficiency than guo - xia s code . in particular , the code @xmath242 in ( [ eqn : c463 ] ) achieves the best ber performance among all the simulated codes . this is attributed to more information symbols embedded in the code @xmath242 than the other codes and its code rate is the highest . we also consider the simulation of the proposed code @xmath249 in ( [ eqn : c5 ] ) with the pic group decoding for five transmit antennas . [ fig : code5tx ] shows the ber performance of the code @xmath249 for three , four and five receive antennas , respectively . it demonstrates that an increase of the number of receive antennas results in a larger diversity gain as illustrated by the slope of the ber curves . in this paper , two designs of stbc that can achieve full diversity with pic group decoding was proposed . the first proposed stbc are constructed with multiple diagonal layers and each layer is composed of a fixed number of coded symbols equal to @xmath10 , i.e. , the number of the transmit antennas . the code rate of the first proposed stbc is varied in accordance with the number of layers embedded in the codeword . for the pic group decoding , a grouping scheme was proposed to cluster every @xmath10 neighboring columns of the equivalent channel matrix into one group . with the proposed stbc and the pic group decoding in mimo systems , it was proved that full diversity can be achieved when two diagonal layers are embedded in the code matrix . moreover , for the full - diversity stbc with the pic group decoding , the code rate is up to @xmath0 symbols per channel use . when pic - sic group decoding is used , the proposed full - diversity stbc can have a rate up to @xmath10 . a few examples of code design achieving full diversity pic group decoding were given . the second proposed stbc are based on three - layers and can achieve full diversity with the pic group decoding and their code rates can be up to @xmath1 . simulation results confirmed the theoretical analysis and show the full diversity performance of the proposed codes when the pic group decoding is applied at receiver . it was also demonstrated that the proposed stbc outperform perfect stbc and tast code when pic group decoding is applied . define the @xmath108 matrix @xmath354 ( @xmath355 ) as follows , @xmath356,\,\,\,p=1,2,\cdots , p , \end{aligned}\ ] ] where @xmath110 and @xmath120^t$ ] is given by ( [ eqn : rm ] ) . then , ( [ eqn : code ] ) can be written as @xmath357 for miso systems , we have @xmath358^t$ ] with @xmath359 ( @xmath360 ) being the channel gain from the @xmath361-th transmit antenna to the receiver . using ( [ eqn : code2 ] ) , we can express ( [ eqn : y ] ) as @xmath362 where @xmath363 , @xmath364 and @xmath365 is given by @xmath366,\,\,\,p=1,2,\cdots , p . \end{aligned}\ ] ] using ( [ eqn : rm ] ) , we can further write ( [ eqn : ya ] ) as @xmath367 where the equivalent channel matrix @xmath368 is given by @xmath369 \end{aligned}\ ] ] and @xmath370^t$ ] . [ def3]@xcite . let @xmath379 be @xmath380 groups of vectors . the vector groups @xmath379 are said to be linearly independent if for @xmath384 , @xmath385 is independent of the remaining vector groups @xmath386 . using ( [ eqn : g ] ) , we can express @xmath193 and @xmath194 as , respectively , @xmath388,\,\,\,\,\,\,\ , \mathbf{g}_2= \left[\begin{array}{c } \mathbf{0}_{1\times m } \\ \mathrm{diag}(\mathbf{h})\mathbf{\theta } \end{array } \right ] . \end{aligned}\ ] ] let @xmath393 be the @xmath361-th ( @xmath394 ) column of the matrix @xmath193 . in order to prove that @xmath193 is independent of @xmath194 , from _ definition [ def2 ] _ we see it is equivalent to prove that the vector @xmath393 is independent of @xmath194 for all @xmath170 . further , using _ definition [ def1 ] _ it is equivalent to prove that @xmath395 where @xmath396 denotes an @xmath113 vector . equivalently , ( [ eqn : g1 ] ) can be expressed as @xmath397 where @xmath398 is a constant and @xmath399 . in order to prove ( [ eqn : g1b ] ) , we can use proof by contradiction . that is , we assume that @xmath400 and @xmath399 . then , we examine the @xmath389-th equation ( from top to bottom ) of ( [ eqn : g1b ] ) and get @xmath401 for @xmath394 , where @xmath402 denotes the @xmath403-th entry of the matrix @xmath193 . this is because all top @xmath389 rows of @xmath194 are all zeros when @xmath392 as seen from ( [ eqn : g ] ) . again from ( [ eqn : g ] ) , we have @xmath404 where @xmath405 denotes the @xmath403-th entry of the matrix @xmath116 . for given @xmath406 and @xmath407 for all @xmath408 , we get @xmath409 . this contradicts with the assumption @xmath399 . therefore , ( [ eqn : g1b ] ) holds and _ step 1 _ is proved . let @xmath414 denote the @xmath361-th ( @xmath394 ) column of @xmath194 . in order to prove that @xmath194 is independent of @xmath387 , from _ definition [ def2 ] _ it is equivalent to prove that the vector @xmath414 is independent of @xmath193 for all @xmath170 . further , using _ definition [ def1 ] _ it is equivalent to prove that @xmath415 where @xmath416 denotes an @xmath113 vector . equivalently , ( [ eqn : gq ] ) can be expressed as @xmath417 where @xmath398 is a constant and @xmath399 . in order to prove ( [ eqn : gqb ] ) , we can use proof by contradiction . we assume that @xmath418 and @xmath399 . note the first row of @xmath194 is all - zero . then , we examine the @xmath419-th equation(from top to bottom ) of ( [ eqn : gqb ] ) and get @xmath420 for @xmath394 , where @xmath421 denotes the @xmath422-th entry of the matrix @xmath194 . this is because the @xmath419-th row of @xmath387 is all - zero when @xmath413 as seen from ( [ eqn : g ] ) . again from ( [ eqn : g ] ) , we have @xmath423 where @xmath424 denotes the @xmath425-th entry of the matrix @xmath116 . for given @xmath426 and @xmath427 for all @xmath408 , we then get @xmath409 . this contradicts with the assumption @xmath399 . therefore , ( [ eqn : gqb ] ) holds and _ step 2 _ is proved . s. m. alamouti , `` a simple transmit diversity technique for wireless communication , '' _ ieee j. sel . areas commun . 14511458 , oct . v. tarokh , h. jafarkhani , and a. r. calderbank , `` space - time block codes from orthogonal designs , '' _ ieee trans . inf . theory _ 14561467 , july 1999 . x. liang , `` orthogonal designs with maximal rates , '' _ ieee trans . inf . theory _ 49 , pp . 24682503 , oct . 2003 . k. lu , s. fu , and x .- xia , `` closed form designs of complex orthogonal space - time block codes of rates @xmath3 for @xmath428 or @xmath429 transmit antennas , '' _ ieee trans . inf . theory _ , 43404347 , dec . 2005 . h. wang and x .- g xia , `` upper bounds of rates of complex orthogonal space - time block codes , '' _ ieee trans . inf . theory _ , vol 49 , pp . 27882796 , oct . 2003 . o. tirkkonen , a. boariu , and a. hottinen , `` minimal non - orthogonality rate @xmath31 space - time block code for @xmath430 tx antennas , '' in _ proc . ieee 6th int . symp . on spread - spectrum tech . and appl . ( isssta 2000 ) _ , sep . 2000 , . 429432 . n. sharma and c. b. papadias , `` full - rate full - diversity linear quasi - orthogonal space - time codes for any number of transmit antennas , '' _ eurasip j. applied signal processing _ , vol . 9 , pp . 12461256 , aug . 2004 . p. elia , k. r. kumar , s. a. pawar , p. v. kumar , and h .- f . lu , `` explicit space - time codes achieving the diversity - multiplexing gain tradeoff , '' _ ieee trans . 38693884 , sep . 2006 . g. wang , h. liao , h. wang , and x .- xia , `` systematic and optimal cyclotomic lattices and diagonal space - time block code designs , '' _ ieee trans . inf . theory _ , vol . 3348 - 3360 , dec . 2004 . g. j. foschini , `` layered space - time architecture for wireless communication in a fading environment when using multi - element antennas , '' _ bell labs tech . vol . 1 , no . 2 , pp . 4159 , 1996 .

in this paper , we study the distribution and behaviour of internal equilibria in a-player-strategy random evolutionary game where the game payoff matrix is generated from normal distributions . the study of this paper reveals and exploits interesting connections between evolutionary game theory and random polynomial theory . the main contributions of the paper are some qualitative and quantitative results on the expected density , , and the expected number , , of ( stable ) internal equilibria . firstly , we show that in multi - player two - strategy games , they behave asymptotically as as is sufficiently large . secondly , we prove that they are monotone functions of . we also make a conjecture for games with more than two strategies . thirdly , we provide numerical simulations for our analytical results and to support the conjecture . as consequences of our analysis , some qualitative and quantitative results on the distribution of zeros of a random bernstein polynomial are also obtained . example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore

1505.04676

evolutionary game theory ( egt ) has been proven to be a suitable mathematical framework to model biological and social evolution whenever the success of an individual depends on the presence or absence of other strategies @xcite . egt was introduced in 1973 by smith and price @xcite as an application of classical game theory to biological contexts , and has since then been widely and successfully applied to various fields , not only biology itself , but also ecology , population genetics , and computational and social sciences @xcite . in these contexts , the payoff obtained from game interactions is translated into reproductive fitness or social success @xcite . those strategies that achieve higher fitness or are more successful , on average , are favored by natural selection , thereby increase in their frequency . equilibrium points of such a dynamical system are the compositions of strategy frequencies where all the strategies have the same average fitness . biologically , they predict the co - existence of different types in a population and the maintenance of polymorphism . as in classical game theory with the dominant concept of nash equilibrium @xcite , the analysis of equilibrium points in random evolutionary games is of great importance because it allows one to describe various generic properties , such as the overall complexity of interactions and the average behaviours , in a dynamical system . understanding properties of equilibrium points in a concrete system is important , but what if the system itself is not fixed or undefined ? analysis of random games is insightful for such scenarios . to this end , it is ambitious and desirable to answer the following general questions : * how are the equilibrium points distributed ? how do they behave when the number of players and strategies change ? * mathematical analysis of equilibrium points and their stability in a general ( multi - player multi - strategy ) evolutionary game is challenging because one would need to cope with systems of multivariate polynomial equations of high degrees ( see section [ sec : pre ] for more details ) . nevertheless , some recent attempts , both through numerical and analytical approaches , have been made . one approach is to study the probabilities of having a concrete number of equilibria , whether all equilibrium points or only the stable ones are counted , if the payoff entries follow a certain probability distribution @xcite . this approach has the advantage that these probabilities provide elaborate information on the distribution of the equilibria . however , it consists of sampling and solving of a system of multivariate polynomial equations ; hence is restricted , even when using numerical simulations , to games of a small number of players and/or small number of strategies : it is known that it is impossible to ( analytically ) solve an algebraic equation of a degree greater than @xmath5 @xcite . another possibility is to analyze the attainability of the patterns and the maximal number of evolutionarily stable strategies ( ess ) @xcite , revealing to some extent the complexity of the interactions . this line of research has been paid much attention in evolutionary game theory and other biological fields such as population genetics @xcite . more recently , in @xcite , the authors investigate the expected number of internal equilibria in a multi - player multi - strategy random evolutionary game where the game payoff matrix is generated from normal distributions . by connecting egt and random polynomial theory , they describe a computationally implementable formula of the mean number of internal equilibrium points for the general case , lower and upper bounds the multi - player two - strategy random games , and a close - form formula for the two - player multi - strategy games . in this paper , we address the aforementioned questions , i.e. , of analysing distributions and behaviours of the internal equilibria of a random evolutionary game , in an _ average _ manner . more specifically , we first analyse the expected density of internal equilibrium points , @xmath6 , i.e. the expected number of such equilibrium points per unit length at point @xmath7 , in a @xmath0-player @xmath1-strategy random evolutionary game where the game payoff matrix is generated from a normal distribution ( for short , normal evolutionary games ) . here the parameter @xmath8 , with @xmath9 , denotes the ratio of frequency of strategy @xmath10 to that of strategy @xmath1 , respectively ( more details in section [ sec : pre ] ) . in such a random game , we then analyse the expected number of internal equilibria , @xmath3 , and , as a result , characterize the expected number of internal _ stable _ equilibria , @xmath11 . we obtain both quantitative ( asymptotic formula ) and qualitative ( monotone properties ) results of @xmath2 and @xmath3 , as functions of the ratios , @xmath7 , the number of players , @xmath0 , and that of strategies , @xmath1 . to obtain these results , we develop further the connection between egt and random polynomial theory explored in @xcite , and more importantly , establish appealing ( previously unexplored ) connections to the well - known classes of polynomials , the bernstein polynomials and legendre polynomials . in contrast to the direct approach used in @xcite , our approach avoids sampling and solving a system of multivariate polynomial equations , thereby enabling us to study games with large numbers of players and/or strategies . we now summarise the main results of the present paper . the main analytical results of the present paper can be summarized into three categories : asymptotic behaviour of the density function and the expected number of ( stable ) equilibria , a connection between the density function with the legendre polynomials , and monotonic behaviour of the density function . in addition , we provide numerical results and illustration for the general games when both the numbers of players and strategies are large . to precisely describe our main results , we introduce the following notation regarding asymptotic behaviour of two given functions @xmath12 and @xmath13 @xmath14 note that throughout the paper we sometimes put arguments of a function as subscripts . for instance , the expected density of internal equilibrium points , @xmath6 , besides @xmath7 , is also analyzed as a function of @xmath1 and @xmath0 . we will explicitly state which parameter(s ) is being varied whenever necessary to avoid the confusion . the main results of the present paper are the following . as described above , @xmath6 denotes the expected number of internal equilibrium points per unit length at point @xmath7 , in a @xmath0-player @xmath1-strategy random evolutionary game where the game payoff matrix is generated from a normal distribution ; @xmath3 the expected number of internal equilibria ; and @xmath11 the expected number of internal stable equilibria . the formal definitions of these three functions are given in section [ sec : pre ] . in theorem [ theo : concentration ] , we prove the following asymptotic behaviour of @xmath15 for all @xmath16 : @xmath17 . we also prove that @xmath15 is always bounded from above and @xmath18 . 2 . in theorem [ theo : behavior of e2 ] , we prove a novel upper bound for the expected number of multi - player two - strategy random games , @xmath19 and obtain its limiting behaviour : @xmath20 . this upper bound is sharper than the one obtained in ( * ? ? ? * theorem 2 ) , which is , @xmath21 . these results lead to two important corollaries . first , we obtain a sharper bound for the expected number of stable equilibria , @xmath22 , and the corresponding limit , @xmath23 , see corollary [ cor : expected stable equi ] . the second corollary , corollary [ cor : expected zeros bernstein ] , is mathematically significant , in which we obtain lower and upper bounds and a limiting behaviour of the expected number of real zeros of a random bernstein polynomial . 3 . in theorem [ theo : fd interm of pd ] , we establish an expression of @xmath15 in terms of the legendre polynomial and its derivative . 4 . in theorem [ theo : fd interms of pd and pd-1 ] , we express @xmath15 in terms of the legendre polynomials of two consecutive order . 5 . in theorem [ theo : fd / d decreases ] , we prove that @xmath24 is a decreasing function of @xmath0 for any given @xmath25 . consequently , @xmath26 and @xmath27 are decreasing functions of @xmath0 . 6 . in proposition [ prop : condition for f increase ] , we provide a condition for @xmath15 being an increasing function of @xmath0 for any given @xmath25 . we conjecture that this condition holds true and support it by numerical simulation . . in theorem [ theo : en2 ] , we provide an upper bound for @xmath28 . we also make a conjecture for @xmath6 and @xmath3 in the general case ( @xmath29 ) . we offer numerical illustration for our main results in section [ sec : simulation ] . the density function @xmath6 provides insightful information on the distribution of the internal equilibria : integrating @xmath6 over any interval produces the expected number of real equilibria on that interval . in particular , the expected number of internal equilibria @xmath3 is obtained by integrating @xmath6 over the positive half of the space . theorem [ theo : fd / d decreases ] and proposition [ prop : condition for f increase ] , which are deduced from theorems [ theo : fd interm of pd ] and [ theo : fd interms of pd and pd-1 ] , are qualitative statements , which tell us _ how _ the expected number of internal equilibria per unit length @xmath30 in a @xmath0-player two - strategy game changes when the number of players @xmath0 increases . theorem [ theo : concentration ] quantifies its behaviour showing that @xmath30 is approximately ( up to a constant factor ) equal to @xmath4 . the function @xmath31 , as seen in theorem [ theo : concentration ] , certainly satisfies the properties that @xmath32 increases but @xmath33 decreases . thus , it strengthens theorem [ theo : fd / d decreases ] and further supports conjecture [ cojecture : fd increases ] . theorem [ theo : behavior of e2 ] is also a quantitative statement which provides an asymptotic estimate for the expected number of internal ( stable ) equilibria . furthermore , it is important to note that the expected number of real zeros of a random polynomial has been extensively studied , dating back to 1932 with block and plya s seminal paper @xcite ( see , for instance , @xcite for a nice exposition and @xcite for the most recent progress ) . therefore , our results , in theorems [ theo : behavior of e2 ] , [ theo : fd interm of pd ] and [ theo : fd interms of pd and pd-1 ] , provide important , novel insights within the theory of random polynomials , but also reveal its intriguing connections and applications to egt . the rest of the paper is structured as follows . in section [ sec : pre ] , we recall relevant details on egt and random polynomial theory . section [ sec : 2d games ] presents full analysis of the expected density function and the expect number of internal equilibria of a multi - player two - strategy game . the results on asymptotic behaviour and on the connection to legendre polynomials and its applications are given in sections [ subsec : asymptotic ] and [ subsec : connection ] , respectively . in section [ sec : general games ] , we provide analytical results for the two - player multi - strategy game and numerical simulations for the general case . therein we also make a conjecture about an asymptotic formula for the density and the expected number of internal equilibria in the general case . in section [ sec : conclusion ] , we sum up and provide future perspectives . finally , some detailed proofs are presented in the appendix . this section describes some relevant details of the egt and random polynomial theory , to the extent we need here . both are classical but the idea of using the latter to study the former has only been pointed out in our recent paper @xcite . [ sec : pre ] the classical approach to evolutionary games is replicator dynamics @xcite , describing that whenever a strategy has a fitness larger than the average fitness of the population , it is expected to spread . formally , let us consider an infinitely large population with @xmath1 strategies , numerated from 1 to @xmath1 . they have frequencies @xmath34 , @xmath35 , respectively , satisfying that @xmath36 and @xmath37 . the interaction of the individuals in the population is in randomly selected groups of @xmath0 participants , that is , they play and obtain their fitness from @xmath0-player games . we consider here symmetrical games ( e.g. the public goods games and their generalizations @xcite ) in which the order of the participants is irrelevant . let @xmath38 be the payoff of the focal player , where @xmath39 ( @xmath40 ) is the strategy of the focal player , and @xmath41 ( with @xmath42 and @xmath43 ) be the strategy of the player in position @xmath44 . these payoffs form a @xmath45-dimensional payoff matrix @xcite , which satisfies ( because of the game symmetry ) @xmath46 whenever @xmath47 is a permutation of @xmath48 . this means that only the fraction of each strategy in the game matters . the equilibrium points of the system are given by the points @xmath49 satisfying the condition that the fitnesses of all strategies are the same , which can be simplified to the following system of @xmath50 polynomials of degree @xmath51 @xcite @xmath52 where @xmath53 , and @xmath54 are the multinomial coefficients . assuming that all the payoff entries have the same probability distribution , then all @xmath55 , @xmath56 , have symmetric distributions , i.e. with mean 0 @xcite . we focus on internal equilibrium points @xcite , i.e. @xmath57 for all @xmath58 . hence , by using the transformation @xmath59 , with @xmath60 and @xmath58 , dividing the left hand side of the above equation by @xmath61 we obtain the following equation in terms of @xmath62 that is equivalent to @xmath63 hence , finding an internal equilibria in a general @xmath1-strategy @xmath0-player random evolutionary game is equivalent to find a solution @xmath64 of the system of @xmath65 polynomials of degree @xmath45 in ( [ eq : eqn for fitnessy ] ) . this observation links the study of generic properties of equilibrium points in egt to the theory of random polynomials . suppose that all @xmath55 are gaussian distributions with mean @xmath66 and variance @xmath67 , then for each @xmath68 ( @xmath58 ) , @xmath69 is a multivariate normal random vector with mean zero and covariance matrix @xmath70 given by @xmath71 the density function @xmath2 and the expected number @xmath3 of equilibria can be computed explicitly . the lemma below is a direct consequence of ( * ? ? ? * theorem 7.1 ) ( see also ( * ? ? ? * lemma 1 ) ) . for a clarity of notation , we use bold font to denote an element in high - dimensional euclidean space such as @xmath72 . [ lemma : e(n , d ) ] assume that @xmath73 are independent normal random vectors with mean zero and covariance matrix @xmath70 as in . the expected density of real zeros of eq . at a point @xmath74 is given by @xmath75 where @xmath76 denotes the gamma function and @xmath77 the matrix with entries @xmath78 with @xmath79 as a consequence , the expected number of internal equilibria in a _ d_-player _ _ n-__strategy random game is determined by @xmath80 note that the assumption in lemma [ lemma : e(n , d ) ] is quite limited when applying to games with more than two strategies as in that case the independence of the @xmath81 terms does not carry over into the independence of @xmath82 terms , see remark [ rem : assumption ] for a detailed discussion . we provide mathematical analysis of the expected density function @xmath15 and the expected number of equilibria @xmath83 for a multi - player two - strategy game . section [ subsec : asymptotic ] presents asymptotic behaviour . a connection to legendre polynomials and its applications are given in [ subsec : connection ] . in section [ subsec : connection ] , further applications of this connection to study monotonicity of the density function are explored . in the case of multi - player two - strategy games ( @xmath84 ) , the system of polynomial equations in becomes a univariate polynomial equation @xmath85 where @xmath86 is the fraction of strategy 1 ( i.e. , @xmath87 is that of strategy 2 ) and @xmath88 is the payoff to strategy 1 minus that to strategy 2 obtained in a @xmath0-player interaction with @xmath44 other participants using strategy 1 . it is worth noticing that @xmath89 is the bernstein basis polynomials of degree @xmath51 @xcite . therefore , the left - hand side of is a random bernstein polynomial of degree @xmath51 . as a by - product of our analysis , see theorem [ theo : behavior of e2 ] , we will later obtain an asymptotic formula of the expected real zeros of a random bernstein polynomial . letting @xmath90 ( @xmath91 ) , eq . is simplified to @xmath92 the expected density of real zeros of this equation at a point @xmath93 is @xmath15 . for simplicity of notation , from now on we write @xmath94 instead of @xmath15 . we recall some properties of the density function @xmath94 from ( * ? ? ? * proposition 1 ) that will be used in the sequel . [ pro : properties of f ] the following properties hold 1 . @xmath95 . 2 . @xmath96^\frac{1}{2}$ ] , where @xmath97 3 . @xmath94 has an alternative representation @xmath98 where @xmath99 satisfies that @xmath100 . 4 . @xmath101 is a decreasing function . @xmath103 . below are some examples of @xmath104 , with @xmath105 and 4 @xmath106 we recall that @xmath90 , where @xmath86 is the fraction of strategy 1 . we can write the density in terms of @xmath86 using the change of variable formula as follows . @xmath107 define @xmath108 to be the density on the right - hand side of the above expression , i.e. , @xmath109 the following lemma is an interesting property of the density function @xmath110 , which explains the symmetry of the strategies ( swapping the index labels converts an equilibrium at @xmath86 to one at @xmath87 ) . numerical simulations in section [ sec : simulation ] further illustrate this property ( see figure [ fig : f2d_depend_frequency_x ] ) . [ lemma : symmetry of g ] the function @xmath111 is symmetric about the line @xmath112 , i.e. , @xmath113 we have @xmath114 where we have used the fifth property in proposition [ pro : properties of f ] to obtain the second equality above . the following theorem provides an upper bound and asymptotic behaviour for @xmath94 and its scaling with respect to @xmath0 . the following statement holds [ theo : concentration ] @xmath115 as a consequence , for any given @xmath16 @xmath116 and furthermore @xmath117 it follows from the first and the fourth properties in proposition [ pro : properties of f ] that @xmath118 on the other hand , according to the third property in proposition [ pro : properties of f ] , we have @xmath119 since @xmath120 , it follows that for any @xmath121 @xmath122 which is equivalent to @xmath123 . hence , the upper bound of @xmath94 in holds . the upper bound and limit in are obvious consequences of . it remains to prove the asymptotic behaviour of @xmath94 as a function of @xmath0 . using and the fourth property in proposition [ pro : properties of f ] , it follows that , for @xmath124 @xmath125 and for @xmath126 @xmath127 from these estimates , we deduce that @xmath128 for any @xmath16 . we numerically illustrate theorem [ theo : concentration ] in section [ sec : simulation ] , see figures [ fig : fd ] . [ theo : behavior of e2 ] it holds that @xmath129 furthermore , we obtain the following asymptotic formula for @xmath83 @xmath130 we first provide a proof of this theorem . then we discuss its implications for the expected number of stable equilibrium points in a random game and that of real zeros of a random bernstein polynomial in corollaries [ cor : expected stable equi ] and [ cor : expected zeros bernstein ] . a comparision with well - known results in the theory of random polynomials is given in remark [ re : elliptic polynomial ] . the lower bound was derived previously in ( * ? ? ? * theorem 2 ) . for the sake of completeness , we provide it again here . using the sixth and the fourth properties in proposition [ pro : properties of f ] , we have @xmath131 . therefore , @xmath132 . the underlying idea of the proof for the upper bound is to split the integral range in the formula of @xmath83 into two intervals . the first one is from @xmath66 to @xmath81 , for some @xmath133 ; we then estimate @xmath94 in this interval by @xmath134 . the second one is from @xmath81 to @xmath67 , which is estimated using the upper bound of @xmath94 given in . the value of @xmath81 will then be optimized . @xmath135\nonumber \\&\leq 2\left[\alpha f_{d}(0)+\frac{\sqrt{d-1}}{2\pi}\int_\alpha^1\frac{1}{t } \,dt\right]\nonumber \\&=\frac{1}{\pi}\left[2(d-1)\,\alpha-\sqrt{d-1}\,\ln \alpha\right].\label{e2 : 1}\end{aligned}\ ] ] let @xmath136 be the expression inside the square brackets in the right - hand side of . to obtain the optimal ( i.e. smallest ) upper bound , we minimize @xmath136 with respect to @xmath81 the optimal value of @xmath81 satisfies the following equation @xmath137 which leads to @xmath138 . substituting this value into , we obtain @xmath139 it follows that @xmath19 , which is . we now prove . by taking logarithm in , we obtain @xmath140 therefore @xmath141 since @xmath142 , we achieve . theorem [ theo : behavior of e2 ] has two interesting implications about the expected number of stable equilibrium points in a random game and that of real zeros of a random bernstein polynomial . [ cor : expected stable equi ] the expected number of stable equilibrium points in a random game with @xmath0 players and two strategies , @xmath143 , satisfies @xmath144 and furthermore , satisfies the following limiting behaviour @xmath145 from ( * ? ? ? * theorem 3 ) , it is known that an equilibrium in a random game with two strategies and an arbitrary number of players , is stable with probability @xmath146 . thus , @xmath147 . hence , the corollary is clearly followed from theorem [ theo : behavior of e2 ] . [ cor : expected zeros bernstein ] the expected number of real zeros , @xmath148 , of a random bernstein polynomial @xmath149 where @xmath88 are independent standard normal distributions , satisfies @xmath150 as mentioned beneath , solving @xmath151 is equivalent to solving eq . . it follows that @xmath104 is the expected density of real zeros of @xmath152 . therefore , @xmath148 given by @xmath153 note that the second equality in holds because @xmath104 is even in @xmath93 due to . the asymptotic behaviour of @xmath154 is then followed directly from theorem [ theo : behavior of e2 ] . [ re : elliptic polynomial ] the study of the distribution and expected number of real zeros of a random polynomial is an active research field with a long history dating back to 1932 with block and plya @xcite , see for instance @xcite for a nice exposition and @xcite for the most recent results and discussions . consider a random polynomial of the type @xmath155 the most well - known classes of polynomials studied extensively in the literature are : flat polynomials or weyl polynomials for @xmath156 , elliptic polynomials ( ep ) or binomial polynomials for @xmath157 and kac polynomials for @xmath158 . we emphasize the difference between the polynomial studied in this paper , i.e. the right - hand side of eq . , with the elliptic polynomial : in our case @xmath159 are binomial coefficients , not their square root as in the elliptic polynomial . in the former case , @xmath160 , and as a result the density function and the expected number of real zeros have closed formula , see @xcite @xmath161 our case is more challenging , because of the square in the coefficients , + @xmath162 is no longer a generating function . nevertheless , theorem [ theo : behavior of e2 ] shows that @xmath83 still has interesting asymptotic behaviour as in and . in this section , we first establish a connection between the expected density function @xmath163 and the well - known legendre polynomials . then using the connection and known properties of legendre polynomials , we prove some qualitative properties of @xmath163 and the expected number of equilibria . the main results of this section can be summarised as follows . a. theorem [ theo : fd interm of pd ] and theorem [ theo : fd interms of pd and pd-1 ] derive an expression for the expected density @xmath163 in terms of the legendre polynomials . b. theorem [ theo : fd / d decreases ] shows that @xmath164 is an increasing function of @xmath0 for any given @xmath16 . c. corollary [ cor : monotonicity of e ] proves that @xmath165 and @xmath166 are decreasing functions of @xmath0 . technically , keys to these theorems are lemma [ lem : connection btw md and pd ] that connects the legendre polynomials @xmath167 to @xmath168 in , and lemma [ lem : turan inequality ] showing a reverse turan s inequality . these lemmas are of interest in their own right . we recall some relevant details on legendre polynomials . legendre polynomials , denoted by @xmath169 , are solutions to legendre s differential equation @xmath170+d(d+1)p_d(x)=0,\ ] ] with initial data @xmath171 . the following are some important properties of the legendre polynomials that will be used in the sequel . 1 . explicit representation @xmath172 2 . recursive relation @xmath173 3 . first derivative relation @xmath174 the first few legendre polynomials are @xmath175 the legendre polynomials were first introduced in 1782 by a. m. legendre as the coefficients in the expansion of the newtonian potential @xcite . this expansion gives the gravitational potential associated to a point mass or the coulomb potential associated to a point charge . in the course of time , legendre polynomials have been widely used in physics and engineering . for instance , they occur when one solves laplace s equation ( and related partial differential equations ) in spherical coordinates @xcite . our approach reveals an appealing , and previously unexplored relationship between legendre polynomials and evolutionary game theory . we explore this relationship and its applications in the next sections . the starting point of our analysis is the following relation between the polynomial @xmath176 in and the legendre polynomial @xmath167 . [ lem : connection btw md and pd ] it holds that @xmath177 see appendix [ app : lem : connection btw md and pd ] . note that traditionally in the legendre polynomials , the arguments are in @xmath178 $ ] . in this paper , however , the arguments are not in this interval since @xmath179 . legendre polynomials with arguments greater than unity have been used in the literature , for instance in ( * ? ? ? * chapter 2 ) . we now establish a connection between the density @xmath104 and the legendre polynomials . according to the second property in proposition [ pro : properties of f ] , we have @xmath180^\frac{1}{2 } , % = \frac{1}{2\pi}\left[\frac{1}{t}\frac{m'_{d+1}(t)}{m_{d+1}(t)}+\frac{m''_{d+1}(t)m_{d+1}(t)-(m'_{d+1}(t))^2}{m^2_{d+1}(t)}\right]^\frac{1}{2},\ ] ] where @xmath181 denotes the derivative with respect to @xmath93 . using this formula and lemma [ lem : connection btw md and pd ] , we obtain the following expression of @xmath182 in terms of @xmath183 and its derivative . [ theo : fd interm of pd ] the following formula holds @xmath184 see appendix [ app : proof of theorem fd interm of pd ] . as a direct consequence of , we obtain the following bound for @xmath15 . @xmath185 in comparison with the estimate obtained in theorem [ theo : concentration ] , this inequality is weaker for @xmath16 since it is of order @xmath186 . however , it does not blow up as @xmath93 approaches @xmath66 . we provide another expression of @xmath182 in terms of two consecutive legendre polynomials @xmath187 and @xmath188 . in comparison with , this formula avoids the computations of the derivative of the legendre polynomial @xmath167 . [ theo : fd interms of pd and pd-1 ] it holds that @xmath189 ^ 2.\ ] ] see appendix [ app : proof of theorem fd interms of pd and pd-1 ] theorem [ theo : fd interms of pd and pd-1 ] is crucial for the subsequent qualitative study of the density @xmath15 for varying @xmath0 . [ lem : turan inequality ] the following inequality holds for all @xmath190 @xmath191 see appendix [ app : proof of lemma turan inequality ] . note that this inequality is the reverse of the turn inequality @xcite where the author considered the case @xmath192 $ ] . [ theo : fd / d decreases ] for any given @xmath193 , @xmath194 is a decreasing function of @xmath0 . we need to prove that @xmath195 from , we have @xmath196 ^ 2.\ ] ] assume that @xmath197 . then @xmath198 for all @xmath0 and from , we have @xmath199 which implies that @xmath200 therefore @xmath201 for all @xmath197 . we first consider the case @xmath202 . from lemma [ lem : turan inequality ] for @xmath203 , we have @xmath204 or equivalently @xmath205 it follows that @xmath206 ^ 2\leq \left[\frac{1+t^2}{1-t^2}-\frac{p_{d}}{p_{d+1}}\left(\frac{1+t^2}{1-t^2}\right)\right]^2.\ ] ] from this inequality and as well as a similar formula for @xmath207 , we obtain @xmath208 which is the claimed property for the case @xmath202 . the case @xmath126 is then followed due to the relation @xmath209 . since @xmath210 and @xmath211 , the following statement is an obvious consequence of theorem [ theo : fd / d decreases ] . [ cor : monotonicity of e ] @xmath165 and @xmath212 are decreasing functions of @xmath0 . the following proposition is a necessary and sufficient condition , which is stated in terms of the legendre polynomials , for @xmath182 being increasing as a function of @xmath0 . [ prop : condition for f increase ] @xmath182 is an increasing function of @xmath0 if and only if , for @xmath213 @xmath214\cdot [ p^2_{d+1}(x)+p^2_{d}(x)-2x p_{d+1}(x)p_d(x)]\nonumber \\&\qquad+(2d+1)(x^2 - 1)p^2_d(x)p^2_{d+1}(x)\geq 0.\end{aligned}\ ] ] furthermore , if @xmath215\ ] ] then holds . see appendix [ app : proof of lemma condition for f increase ] . we numerically verify the inequality in figure [ fig : fd increase ] ; however , it is unclear to us how to prove it rigorously . we also recall that it is shown in theorem [ theo : concentration ] that @xmath104 behaves like @xmath4 , which is an increasing function of @xmath0 , as @xmath0 is sufficiently large . motivated by these observations , we make the following prediction . [ cojecture : fd increases ] for any given @xmath25 , @xmath104 is an increasing function of @xmath0 . we provide further numerical simulation to support this conjecture by directly plotting @xmath104 in figure [ fig : fd]c . [ remark : from_t_to_x ] we recall that in the case @xmath84 , the variable @xmath93 is defined by @xmath90 , where @xmath86 is the fraction of strategy @xmath67 and @xmath87 is that of strategy @xmath216 . equivalently , @xmath86 can be expressed in terms of @xmath93 as @xmath217 . hence one can also transform the statements of the theorem above ( and later ) in terms of @xmath86 . as has been shown in the beginning of section [ sec : pre ] , the transformation from @xmath86 to @xmath93 has the advantage that it transforms a complex equation ( eq . ) to a univariate polynomial equation ( eq . ) . this enables us to exploit many available results and techniques from the literature of random polynomial theory . moreover , from the relationship between @xmath218 and @xmath219 , it follows that all the monotonicity properties with respect to @xmath0 are reserved for @xmath220 ( see a numerical illustration in figure [ fig : f2d_depend_frequency_x ] ) . in this section , first we prove an estimate for the density @xmath28 similarly as in theorem [ theo : concentration ] for a two - player multi - strategy game . the expected number of internal equilibria in this case has been computed explicitly in ( * ? ? ? * theorem 3 ) . we then conjecture an asymptotic formula for the general case . finally , we provide numerical simulations to support our conjectures , as well as the main results in the previous sections . + [ theo : en2 ] assume that @xmath221 are independent random vectors , then it holds that @xmath222 as a consequence , for any @xmath7 such that @xmath223 @xmath224 according to ( * ? ? ? * theorem 3 ) @xmath225 the first assertion is then followed . now suppose that @xmath223 . by the cauchy - schwartz inequality , which states @xmath226 for @xmath1 all @xmath1 positive numbers @xmath227 , we have @xmath228 therefore @xmath229 on the other hand , since @xmath230 , it follows that @xmath231 therefore , we obtain . note that the expected number of internal equilibria for a two - player multi - strategy game is given by ( * ? ? ? * theorem 3 ) @xmath232 l c l*7cr d & 2 & 3 & 4 & 5 + @xmath233 from theory & 0.25 & 0.57 & 0.92 & 1.29 + @xmath233 from random sampling & 0.249496 & 0.569169 & 0.910236 & 1.28898 + [ table : ed ] [ rem : assumption ] in theorem [ theo : en2 ] , we need an assumption that the random vectors @xmath221 are independent . recalling from section [ sec : pre ] that @xmath234 , where @xmath235 is the payoff of the focal player , with @xmath68 ( @xmath35 ) being the strategy of the focal player , and @xmath41 ( with @xmath42 and @xmath43 ) is the strategy of the player in position @xmath44 . the assumption clearly holds for @xmath236 . for @xmath237 , the assumption holds only under quite restrictive conditions such as @xmath238 is deterministic or @xmath239 are essentially identical . note that the assumption is necessary to apply ( * ? ? ? * theorem 7.1 ) . hence , to remove this assumption , one would need to generalize ( * ? ? ? * theorem 7.1 ) . this is difficult and is still an open problem @xcite . nevertheless , since the system has not been studied in the mathematical literature , it is interesting on its own to investigate the number of real zeros of this system under the assumption of independence of @xmath240 . as such , the investigation not only provides new insights into the theory of zeros of systems of random polynomials but also gives important hint on the complexity of the game theoretical question , i.e. the number of expected number of equilibria . in figure [ fig : theory vs samplings ] and table [ table : ed ] , we numerically compare the density function @xmath241 and the expected number of equilibria @xmath242 computed from the theory with the above assumption and from samplings without the assumption . we observe that @xmath241 have the same shape ( behaviour ) in both cases . in addition , the values of @xmath242 computed from samplings are slightly smaller than those computed from the theory . we now make a conjecture for the expected density and expected number of equilibria in a general multi - player multi - strategy game . [ con : conjecture ] in a multi - player multi - strategy game , it holds that @xmath243 we envisage that even a stronger statement holds @xmath244 this conjecture is motivated from theorem [ theo : behavior of e2 ] and a similar result for the system of multivariate elliptic polynomial as follows ( see also remark [ re : elliptic polynomial ] ) . consider a system of @xmath50 random polynomials , each of the form @xmath245 where the coefficients @xmath246 are independent standard multivariate normal distribution . then according to ( * ? ? ? * section 7.2 ) , the expected density and the expected number of real zeros are given by @xmath247 these formula are direct consequences of lemma [ lemma : e(n , d ) ] and the fact that @xmath248 is a generating function , which generalises the univariate case @xmath249 as mentioned in remark [ re : elliptic polynomial ] , in our case @xmath250 is no longer a generating function due to the square of the multinomial coefficients . motivated by this and theorem [ theo : behavior of e2 ] for the univariate case , we expect that the conjecture holds for the general case ( i.e. @xmath0-player @xmath1-strategy normal evolutionary games ) . in this section , we provide numerical simulations for all the ( main ) results obtained in the previous sections . the following figures are plotted . in the following list , ( 1 ) to ( 4 ) are to illustrate and confirm the analytical results obtained in the previous sections . the others , from ( 5 ) to ( 10 ) , are numerical simulations for the games with large @xmath0 and @xmath1 . they also provide numerical support for conjecture [ con : conjecture ] . 1 . @xmath104 and @xmath251 as functions of @xmath93 for different values of @xmath0 , see figures [ fig : fd]a - b . figure [ fig : fd]a illustrates the fourth property in proposition [ pro : properties of f ] that @xmath104 is increasing as a function of @xmath93 . figure [ fig : fd]b explains the scaling by showing that @xmath251 is a bounded and decreasing function . 2 . @xmath104 and @xmath251 as functions of @xmath0 for different values of @xmath93 , see figures [ fig : fd]c - d . these figures show that @xmath104 increases with @xmath0 while @xmath251 decreases , which are in agreement with conjecture [ cojecture : fd increases ] and theorem [ theo : fd / d decreases ] . 3 . @xmath252 and @xmath253 as functions of the frequency @xmath254 . 4 . @xmath255 as a function of @xmath0 , see figure [ fig : lnedivlnd]a . this figure demonstrates theorem [ theo : behavior of e2 ] on the convergence of @xmath255 to @xmath146 as @xmath0 tends to infinity . @xmath256 as a function of @xmath257 for different values of @xmath0 , see figures [ fig : f_n3n4]a - c . 6 . @xmath256 as a function of @xmath0 for different values of @xmath257 , see figure [ fig : lnfd34_depend_d]a . @xmath258 as a function of @xmath259 for different values of @xmath0 , see figures [ fig : f_n3n4]d - f . 8 . @xmath258 as a function of @xmath0 for different values of @xmath259 , see figure [ fig : lnfd34_depend_d]b . 9 . @xmath260 as a function of @xmath0 , see figure [ fig : lnedivlnd]b . 10 . @xmath261 as a function of @xmath0 , see figure [ fig : lnedivlnd]c . in figures [ fig : f_n3n4 ] , we provide numerical results of @xmath6 for @xmath262 and @xmath263 . we observe that the density function decreases with @xmath264 ( namely , @xmath265 and @xmath266 for @xmath262 , and @xmath265 , @xmath266 and @xmath267 for @xmath263 ) and increases with @xmath0 . we conjecture that for the general @xmath0-player @xmath1-strategy normal evolutionary game , the density function decreases with @xmath264 and increases with @xmath0 . figures [ fig : lnedivlnd]b and [ fig : lnedivlnd]c support conjecture [ con : conjecture ] . from these two figures , one also can see the complexity of the problem when @xmath0 increases . we are able to run simulations for @xmath0 up to @xmath268 for @xmath84 , up to @xmath269 for @xmath270 and only up to @xmath271 for @xmath272 . . [ fig : fd increase ] how do equilibrium points in a general evolutionary game distribute if the payoff matrix entries are randomly drawn , and furthermore , how do they behave when the numbers of players and strategies change ? to address these important questions regarding generic properties of general evolutionary games , we have analyzed here the density function , @xmath2 , and the expected number of ( stable ) equilibrium points , @xmath3 ( respectively , @xmath11 ) , in a normal @xmath0-player @xmath1-strategy evolutionary game . we have shown , analytically and using numerical simulations , that @xmath15 monotonically decreases with @xmath93 while it increases with @xmath0 . the latter implies that , as the number of players in the game increases , it is more likely to see an equilibrium at a given point @xmath93 . we also proved that its scaling with respect to the number of players in a game @xmath0 , @xmath24 , decreases with @xmath0 . more interestingly , we proved that this density function asymptotically behaves in the same manner as @xmath4 at any given @xmath93 ( i.e. regardless of the equilibrium point ) . similar monotonicity behaviors of the density function were observed numerically for games with larger numbers of strategies @xmath1 . additionally , we proved an upper bound for two - player game with an arbitrary number of strategies , @xmath1 . briefly , our analysis of the expected density of equilibrium points has offered clear and fresh understanding about equilibrium distribution . related to this distribution analysis , there have been some works analyzing patterns of esss and their attainability in concrete games , i.e. with a fixed payoff matrix @xcite . differently , our analysis addresses distribution of the general equilibrium points , and for generic games . note also that those works dealt with two - player games while we address here the general case ( i.e. with arbitrary @xmath0 ) . regarding the expected numbers of internal ( stable ) equilibrium points , first of all , as a result of the described monotonicity properties of the density function , we established analytically that @xmath83 and @xmath143 increase with @xmath0 while their scaled forms , @xmath26 and @xmath27 , decrease with @xmath0 . next , we proved a new upper bounds for @xmath236 , with arbitrary @xmath0 : @xmath19 . this upper bound is sharper than the one described in @xcite ( which is also the only known one , to the best of our knowledge ) . as a consequence , a sharper upper bound for the number of expected number of stable equilibria can be established : @xmath22 . more importantly , that allowed us to derive close - form limiting behaviors for such numbers : @xmath20 and @xmath23 . as such , apart from the mathematical elegance of our results , the analysis here has significantly improved existing results on the studies of random evolutionary game theory @xcite . moreover , generalizing these formulas for the general case , i.e. when the number of strategies @xmath1 is also arbitrary , our conjecture that , @xmath273 , is nicely corroborated by numerical simulations . related to this analysis , there have been some works analyzing ess in random evolutionary games @xcite . in both @xcite and @xcite , the authors focused on the asymptotic behaviour of ess with large support sizes , i.e. considering also equilibrium points which are not internal , in random pairwise games . in a similar context , in @xcite , the authors studied ess but with support size of two , showing the asymptotic behaviors of such ess when the number of strategies varies . differently from all these works ( which dealt exclusively with two - player interactions ) , our analysis copes with multi - player games . hence , our results have led to further understanding with respect to the asymptotic behaviour of the expected number of stable equilibria for multi - player random games . last but not least , as the density functions we analyzed here are closely related to legendre polynomials , and actually , they are of the same form as the bernstein polynomials , we have made a clear contribution to the longstanding theory of random polynomials @xcite(see again discussion in introduction ) . we have derived asymptotic behaviors for the expected real zeros @xmath148 of a random bernstein polynomial , which , to our knowledge , had not been provided before . note that the asymptotic behaviors and close forms of the expected real zeros of some other well - known polynomials have been derived by other authors . for instance , the weyl polynomials for @xmath274 ; the elliptic polynomials or binomial polynomials with @xmath275 ; and kac polynomials with @xmath276 . the difference between the polynomial studied here ( i.e. bernstein ) with the elliptic case : @xmath277 are binomial coefficients , not their square root . in the elliptic case , @xmath160 , and as a result , @xmath278 ; while in our case , because of the square in the coefficients , @xmath162 , is no longer a generating function . indeed , due to this difficulty , the analysis of the bernstein polynomials is still rather limited , see for instance @xcite for a detailed discussion . in short , our analysis has provided new understanding about the generic behaviors of equilibrium points in a general evolutionary game , namely , how they distribute and change in number when the number of players and that of the strategies in the game , are magnified . detailed proofs of some lemmas and theorems in the previous sections are presented in this appendix . by taking the derivative of both sides in , we obtain @xmath283 it follows that @xmath284 now we compute the expression inside the square - root of the right - hand side of . we have @xmath285 and @xmath286 substituting this expression into , we get @xmath287 according to , the legendre polynomial @xmath167 satisfies the following equation for all @xmath288 @xmath289 as a consequence , we obtain @xmath290 substituting this expression into with @xmath213 , we get @xmath291-\left(\frac{p_d'}{p_d}\right)^2\left(\frac{1+t^2}{1-t^2}\right)\right\ } \\&=\frac{4d^2}{(1-t^2)^2}-\frac{16t^2}{(1-t^2)^4}\left(\frac{p_d'}{p_d}\right)^2\left(\frac{1+t^2}{1-t^2}\right),\end{aligned}\ ] ] which is the claimed relation . using the following relation of the legendre polynomials for all @xmath292 @xmath293 we get @xmath294 in particular , taking @xmath213 , we obtain @xmath295 \\&=\frac{d(1-t^2)^2}{4t^2}\left[\frac{1+t^2}{1-t^2}-\frac{p_{d-1}}{p_d}\left(\frac{1+t^2}{1-t^2}\right)\right].\end{aligned}\ ] ] substituting this expression into , we achieve @xmath296 ^ 2,\ ] ] which is . we will prove that @xmath299 from , we have @xmath300 ^ 2 \\&\qquad-\frac{4d^2}{(1-t^2)^2}+\frac{d^2}{t^2}\left[\frac{1+t^2}{1-t^2}-\frac{p_{d-1}}{p_d}\left(\frac{1+t^2}{1-t^2}\right)\right]^2 \\&=\frac{4(2d+1)}{(1-t^2)^2}-\frac{1}{t^2}\left[\frac{1+t^2}{1-t^2}-\frac{(d+1)p_d^2-dp_{d-1}p_{d+1}}{p_dp_{d+1}}\left(\frac{1+t^2}{1-t^2}\right)\right]\times \\&\qquad\left[(2d+1)\frac{1+t^2}{1-t^2}-\frac{(d+1)p_d^2+dp_{d-1}p_{d+1}}{p_dp_{d+1}}\left(\frac{1+t^2}{1-t^2}\right)\right].\end{aligned}\ ] ] therefore @xmath104 is increasing as a function of @xmath0 if and only if @xmath301\times\left[(2d+1)\frac{1+t^2}{1-t^2}-\frac{(d+1)p_d^2+dp_{d-1}p_{d+1}}{p_dp_{d+1}}\left(\frac{1+t^2}{1-t^2}\right)\right ] \\&\qquad\leq \frac{4(2d+1)t^2}{(1-t^2)^2}.\end{aligned}\ ] ] we re - write the expression above using the variable @xmath254 , using the relation @xmath302 , as follows @xmath303\times\left[(2d+1)x-\frac{(d+1)p_d^2+dp_{d-1}p_{d+1}}{p_dp_{d+1}}(x)\right]\leq ( 2d+1)(x^2 - 1).\ ] ] we now simplify this expression using the recursion relation of the legendre polynomials , i.e. @xmath304 . namely , we have @xmath305p_{d+1 } \\&=(d+1)[-p_d^2-p_{d+1}^2 + 2xp_dp_{d+1}]\end{aligned}\ ] ] and @xmath306p_{d+1 } \\&\qquad=(d+1)(p^2_{d+1}-p_d^2).\end{aligned}\ ] ] substituting these calculations into we obtain . to prove the second assertion of proposition [ prop : condition for f increase ] , we proceed as follows . let @xmath307 \\ & \qquad= \left[(2d+1)x p_{d}(x)- d p_{d-1}(x)\right]^2 + ( d+1)^2 p^2_{d}(x ) \\&\qquad\qquad-2x ( d+1 ) \left[(2d+1)x p_{d}(x)- d p_{d-1}(x)\right]p_d(x ) \\ & \qquad=\left[(2d+1)^2 x^2 + ( d+1)^2 - 2(d+1)(2d+1 ) x^2 \right ] p^2_{d}(x ) \\&\qquad\qquad-\left[2d(2d+1)x - 2d(d+1 ) x \right ] p_{d}(x ) p_{d-1}(x ) + d^2 p^2_{d-1}(x ) \\ & \qquad = \left[d^2 + ( 2d+1)(1-x^2 ) \right ] p^2_{d}(x)+d^2 p^2_{d-1}(x)-2x d^2 p_{d}(x)p_{d-1}(x ) \\ % & \qquad\geq \left[(2d+1)x - d-1 \right]^2 p^2_{d}(x)+d^2 p^2_{d-1}(x)-2x d^2 p_{d}(x)p_{d-1}(x ) \ \ \text{(since $ x \geq 1$)}\\ & \qquad= d^2 \left[p^2_{d}(x)+p^2_{d-1}(x)-2x p_{d}(x)p_{d-1}(x)\right ] + ( 2d+1)(1-x^2 ) p^2_{d}(x ) \\ & \qquad= h_d + ( 2d+1)(1-x^2 ) p^2_{d}(x).\end{aligned}\ ] ] hence , the expression in can be simplified as follows @xmath308 + ( 2d+1)(x^2 - 1)p^2_d(x)p^2_{d+1}(x ) \\ & = [ h_d + ( 2d+1)(1-x^2 ) p^2_{d}(x)][p^2_{d+1}(x)-p^2_{d}(x ) ] + ( 2d+1)(x^2 - 1)p^2_d(x)p^2_{d+1}(x ) \\ & = h_d [ p^2_{d+1}(x)-p^2_{d}(x ) ] + ( 2d+1)(x^2 - 1)p^4_d(x ) \\ & = \frac{p^2_{d+1}(x)-p^2_{d}(x)}{p^2_{d}(x)-p^2_{d-1}(x)}\left[h_d \left(p^2_{d}(x)-p^2_{d-1}(x)\right ) + ( 2d-1)(x^2 - 1)p^2_d(x)p^2_{d-1}(x ) \cdot q \right ] , % % % \\ % & \geq \frac{p^2_d(x)}{p^2_{d-1}(x)}\left[h_d \cdot \frac { p^2_{d-1}(x ) p^2_{d+1}(x)- p^2_{d-1}(x)p^2_{d}(x)}{p^2_d(x ) } + ( 2d-1)(x^2 - 1)p^2_d(x)p^2_{d-1}(x)\right ] \text { ( since } \ x \geq 1)\\ % & \geq \frac{p^2_d(x)}{p^2_{d-1}(x)}\left[h_d \cdot \frac { p^4_{d}(x)- p^2_{d-1}(x)p^2_{d}(x)}{p^2_d(x ) } + ( 2d-1)(x^2 - 1)p^2_d(x)p^2_{d-1}(x)\right ] \text{(see lemma \ref{lem : turan inequality } ) } \\&\text { wrong again $ h_d < 0$}\\ % & = \frac{p^2_d(x)}{p^2_{d-1}(x)}\left[h_d \cdot \left ( p^2_{d}(x)- p^2_{d-1}(x)\right ) + ( 2d-1)(x^2 - 1)p^2_d(x)p^2_{d-1}(x)\right ] , \\ % & \cdots \\ % & \geq\frac{p^2_d(x)}{p^2_1(x)}\left[h_1 \cdot \left(p^2_{1}(x)- p^2_{0}(x)\right ) + ( x^2 - 1)p^2_1(x)p^2_0(x)\right ] \\ % & = p^2_d(x ) ( x^2 - 1 ) \\ % & \geq 0,\end{aligned}\ ] ] where @xmath309 . suppose that is true , i.e. , @xmath310.\ ] ] this implies that @xmath311 for all @xmath254 and @xmath0 . then it follows that @xmath312 + ( 2d+1)(x^2 - 1)p^2_d(x)p^2_{d+1}(x)\nonumber \\&\qquad\geq \frac{p^2_{d+1}(x)-p^2_{d}(x)}{p^2_{d}(x)-p^2_{d-1}(x)}\left[h_d \left(p^2_{d}(x)-p^2_{d-1}(x)\right ) + ( 2d-1)(x^2 - 1)p^2_d(x)p^2_{d-1}(x)\right]\nonumber \\&\qquad\geq \frac{p^2_{d+1}(x)-p^2_{d}(x)}{p^2_{d}(x)-p^2_{d-1}(x)}\times \frac{p^2_{d}(x)-p^2_{d-1}(x)}{p^2_{d-1}(x)-p^2_{d-2}(x)}\nonumber \\&\qquad\qquad\times \left[h_{d-1 } \left(p^2_{d-1}(x)-p^2_{d-2}(x)\right ) + ( 2d-3)(x^2 - 1)p^2_{d-1}(x)p^2_{d-2}(x)\right]\nonumber \\&\qquad\geq\cdots\nonumber \\&\qquad\geq \prod\limits_{i=1}^{d}\frac{p^2_{i+1}(x)-p^2_{i}(x)}{p^2_{i}(x)-p^2_{i-1}(x)}\times\left[h_1 \left(p^2_{1}(x)-p^2_{0}(x)\right ) + ( x^2 - 1)p^2_{1}(x)p^2_{0}(x)\right].\end{aligned}\ ] ] by definition of @xmath313 , we have @xmath314 substituting this into , we obtain @xmath308 + ( 2d+1)(x^2 - 1)p^2_d(x)p^2_{d+1}(x ) \\&\geq ( x^2 - 1)\prod\limits_{i=1}^{d}\frac{p^2_{i+1}(x)-p^2_{i}(x)}{p^2_{i}(x)-p^2_{i-1}(x)}\\&=p^2_{d+1}(x)-p_d^2(x)\geq 0,\end{aligned}\ ] ] i.e. , the condition is satisfied . e. kostlan . on the expected number of real roots of a system of random polynomial equations . in _ foundations of computational mathematics ( hong kong , 2000 ) _ , pages 149188 . world sci . publ . , river edge , nj , 2002 .

the excitation of a non - linear ion - wake by a train of ultra - relativistic electron bubble wake- is modeled . the ion - wake is shown to be a driven non - linear ion - acoustic wave in the form of a cylindrical ion - soliton- . the phases of the oscillating radial electric fields of the slowly - propagating electron bubble is asymmetric in time and excites time - averaged inertial ion motion radially . the electron compression in the back of the bubble sucks - in the ions and the space - charge within the bubble cavity expels them , driving a cylindrical ion - soliton structure with on - axis and bubble - edge density - spikes . once formed , the channel - edge density - spike is driven radially outwards by the thermal pressure of the wake energy . its channel - like structure due to the flat - residue left behind by the propagating ion - soliton , is independent of the energy - source driving the bubble electron wake . we explore the use of the partially - filled channel formed by the cylindrical ion - soliton for a novel regime of positron acceleration . osiris pic simulations are used to study the ion - wake soliton structure , its driven propagation and its use for positron acceleration .

1504.03735

plasma ions are generally assumed to be stationary in the theory of ultra - relativistic non - linear plasma electron waves excited as wakefields of a laser or particle beam @xcite-@xcite . important exceptions to this occur when the drive beam ( or a witness beam being accelerated ) in a plasma wakefield accelerator becomes denser than the background plasma density by a factor larger than the ion to electron mass ratio @xcite@xcite . such is the case for the parameters of ultra - low emittance future collider designs at the tev energy scale . ion motion also becomes important in the wake train behind the initial wake oscillation or bubble . understanding the long - term ion behavior @xcite@xcite is important to determine the state of the plasma for succeeding pulses in a future plasma collider of high repetition rate @xcite@xcite . we also show that the resulting ion - wake channel formed is of interest for positron wakefield acceleration . driven by a matched laser pulse ( vector potential @xmath0 and frequency @xmath1 ) with @xmath2 . ( b ) non - linear ion - wake in the form of a cylindrical ion - soliton of radius @xmath3 excited behind the bubble electron wake in a proton plasma . ( c ) transverse ion - density profile at @xmath4 . notice that the ion density perturbation in this excitation phase is still building up and is a fraction of the background ion density , @xmath5 . ] in this paper , we model the excitation of a nonlinear ion - wake in the trail of an electron plasma wake in the bubble regime , driven by either relativistically intense laser or particle beam energy sources . the ion - wake is shown to be a non - linear ion - acoustic wave in the form of a cylindrical ion - acoustic soliton . its characteristics are similar to the solutions of the cylindrical korteweg - de vries equation ( ckdv ) @xcite-@xcite . however , the ion - wake shown here is different from a cylindrical - kdv soliton due to several reasons : the bubble wake electron oscillations do not thermalize into an isothermal plasma ; the ion - wake soliton breaks up into n - solitons as it evolves and there is an ion - density spike on the axis . the soliton propagates radially outwards leaving behind a flat residue resulting in a partially filled ion - wake channel . the dynamics of the ion wake is shown to have two distinct phases . the excitation phase where the time asymmetry of the radial electric fields of the bubble excite the ions into a soliton - like structure . at later times , the propagation phase where the soliton is driven outwards by the pressure of thermalized electrons after the electron oscillations undergo phase mixing @xcite . representative pic simulation results in fig.[fig1:ion - wake - laser ] and fig.[fig2:ion - wake - beam ] illustrate the salient features of the non - linear ion - wake . figure[fig1:ion - wake - laser ] shows the excitation phase at an early time when the bubble wake train is executing ordered oscillations and its fields have begun to excite the ion - wake structure seen in fig.[fig1:ion - wake - laser](b),(c ) . the radial oscillations later phase mix and the electrons thermalize , their thermal pressure then drives the ion - soliton outwards as shown in fig.[fig2:ion - wake - beam ] . as we show below , it is the longitudinal or time asymmetry of the radial electron wakefields that excites the ion soliton which propagates leaving behind a partially - filled channel shown in the pic simulations in fig.[fig1:ion - wake - laser ] and fig.[fig2:ion - wake - beam ] . , @xmath6 , @xmath7 , @xmath8 , these beam - plasma parameters are quite similar to @xcite . ( b ) corresponding ion density in cylindrical coordinates ( fixed - box ) . note the n - soliton formation in the ion - density . this is seen as the ion - wake evolves further , for instance at @xmath9 . ( c ) radial electron and ion density profile at @xmath10 . ] in section [ ion - wake - model - considerations ] we present considerations and assumptions made to derive the ion - wake model . in section [ radial - ion - wave ] using the fluid equations for ion dynamics we model the non - linear ion - acoustic waves as a driven cylindrical ion soliton . the driven wave models are shown to have two distinct phases : the excitation phase and the propagation phase . we present a model for the excitation of the ion - soliton by the electron bubble radial fields in section [ ion - soliton - excitation - phase ] and verify it using simulation results . in section [ ion - soliton - propagation - phase ] we present simulation results showing the dynamics of propagation of the ion - soliton driven by a radial temperature gradient . finally , in section [ ion - soliton - positron - wake ] we analyze the properties of a positron beam driven wakefield in an ion - wake channel . we recognize that to study the time evolution of a wake - excited plasma in order to establish the duration over which it relaxes to thermal equilibrium both collisional and collision - less dynamics have to be considered along with the physics of recombination modes such as electron - ion recombination . however , in this work the dynamics of plasma is modeled under the collision - less approximation . thus , diffusion is not important during the timescales over which the ion - wake is studied . we do not discuss recombination except mentioning that the afterglow " is dominated by volume recombination while localized effects can not be ruled out . in order to formally establish the difference between the density wave processes that occur over collision - less timescale in contrast to the ones that start dominating under collisions , we show the assumptions made to arrive at the dynamics of diffusion . the process of diffusion is modeled with a parabolic partial differential equation which is deduced from the ion - fluid equations under the assumption that the inertial response of the ions is much faster than the collisional timescales . the effect of collisions is introduced as a drag force , @xmath11 . the collisional drag force modifies the ion - fluid equation of motion as , @xmath12 where @xmath13 is the average electron - ion collision frequency and is obtained from the mean free path . diffusion of plasma is thus driven by the charge - separation field , @xmath14 and the thermal pressure , @xmath15 while being impeded by the collisional drag . upon ignoring the inertia of the ions , @xmath16 , the equation for the ion velocity by diffusion in an isothermal plasma is , @xmath17 . the characteristic parameter of diffusion is the diffusion coefficient or diffusivity @xmath18 and mobility @xmath19 which depend upon the collision frequency . using the gradient of the velocity in the continuity equation and ignoring the mobility @xmath20 , leads to a fick s law diffusion equation , @xmath21 ; characteristic of a parabolic equation . when the mobility is retained , the fluid equation is a moment of the fokker - planck equation which is the kinetic model of the collision - driven drift and diffusion . the diffusion equation thus can not support wave - like solution because such solutions are characteristic of a hyperbolic partial differential equation . the solutions of linear and non - linear diffusion equations show the evolution of density profile by diffusion and can be obtained using the self - similar formulation . the self - similar solutions show the spatial and temporal evolution of the density to be exponentially decaying . in the non - linear case , the density can have a sharp - front as it decays . however , a soliton - like propagating solution can not be described with diffusion equation . hence , the cylindrical ion - soliton presented here is not diffusion but a wave phenomenon . the electron bubble wake is excited by a sub - wavelength impulse of an ultra - short driver . in contrast , the ion - wake is excited as the ions undergo sustained interactions with the bubble fields within the spatial extent of the wake over several plasma electron oscillations . this happens because the electron wake - plasmon oscillations @xcite have a near speed - of - light phase - velocity ( @xmath22 ) but negligible group - velocity @xcite @xmath23 ( in the 1-d limit ) , where , @xmath24 is the mean electron thermal velocity of the background plasma . therefore a slowly - propagating train of coupled electron plasmons is excited in a cold collision - less plasma @xcite . a large difference between phase - velocity and the group velocity of the electron oscillations allows sustained field - ion interactions . it should be noted that high phase - velocity plasma electron waves are possible only in a cold plasma with appropriate density , @xmath25 that allows near speed - of - light propagation of the energy sources , @xmath26 . ion - soliton modeled here is assuming a significant difference between the phase - velocity and the group velocity of the plasma - electron waves . a time symmetric electron wakefield would excite time symmetric ion oscillations where the ion velocities average to zero . however , the bubble wake is asymmetric in time as the back of the bubble electron compression is a small fraction of the length of electron cavitation . the electron oscillations become non - linear at high driver intensities as all the interacting electrons are displaced radially , @xmath27 , forming a non - linear bubble - shaped electron spatial structure enclosing ions in its cavity . the wakefields excited in the bubble are useful for accelerating electrons @xcite@xcite@xcite . high intensities also lead to fields that can directly drive the plasma electrons to velocities near the speed - of - light . this occurs when for a laser pulse @xmath28 and an electron beam @xmath29 where @xmath30 is the peak normalized laser vector potential , @xmath31 the peak beam density and radius . the radially expelled electrons oscillate radially under the force of the plasma ions . these oscillations are excited over plasma electron oscillation timescales , @xmath32 ( @xmath33 ) where @xmath34 is the temporally anharmonic relativistic electron quiver momentum . the normalized quiver momentum of the electrons in the bubble - oscillations is relativistic @xmath35 and the quiver frequency is @xmath36 @xcite . we show that non - linear ultra - relativistic electron wakefields interacting with the plasma ions lead to the excitation of a non - linear ion - wake . the non - linear ion - wake @xmath37 in fig.[fig1:ion - wake - laser ] and fig.[fig2:ion - wake - beam ] is excited over timescales @xmath38 in the trail of a bubble - wake train . by shaping the energy source it can be matched or guided to excite a long train of nearly identical plasmons , fig.[fig3:bubble - train ] . since it is the electric field @xmath39 of a nearly stationary bubble plasmon that excite collective ion - motion we model the ion dynamics in a single bubble . using the single bubble ion dynamics , fig.[fig5:bubble - ion - dynamics ] we model the ion - wake over the whole bubble - train spanning several hundred plasma skin - depths ( @xmath40 ) . the wake - plasmon energy density ( @xmath41 , where @xmath42 is the wakefield amplitude ) is continually partitioned between the field energy and the coherent electron quiver kinetic energy . in our model we do not include heavy beam - loading of the bubble electron wake . under heavy beam - loading the bubble field energy is efficiently coupled to the kinetic energy of the accelerated beam . in this scenario the bubble collapses and the magnitude of the ion - wake is smaller . the decoherence of the ordered electron quiver to random thermal energy , @xmath43 due to the phase - mixing @xcite of individual electron trajectories caused by the non - linearities and inhomogeneities is further stimulated by the ion motion . the details of the thermalization of the wake electrons under ion motion is beyond the scope of this paper . it is over these timescales upon thermalization that the steepened ion - density expands outwards radially as a non - linear ion - acoustic wave driven by the electron thermal pressure . the energy transfer process observed here is a coupling from the non - linear plasma electron - mode to a non - linear ion - acoustic mode @xcite . we also observe energy coupling to the bow - shock which is formed behind the bubble , fig.[fig5:bubble - ion - dynamics ] . the ion - wake is a partially - filled channel with sub - skin - depth density - spikes on - axis and at the bubble - edge located at the bubble - radius , @xmath44 @xcite@xcite of several @xmath40 . the ion accumulation in both the density - spikes is many times the background density as shown in fig.[fig1:ion - wake - laser](c ) and fig.[fig2:ion - wake - beam](c ) . the wake energy is coupled to the ions and to the thermal energy @xcite of the de - cohering wake electrons . the interplay between the electron thermal energy and ion kinetic energy then form a cylindrically symmetric non - linear ion - acoustic wave which further empties the near - void region of the ion - wake channel . the time - scale of dissipation of the ion - wake and relaxation of the plasma distribution to @xmath45 sets an upper limit on the repetition - rate @xcite of the future plasma colliders . it is well known that the ion acoustic wave is damped by collisions and ion - wave landau damping . the ion dynamics also opens important questions on the design the plasma container walls in a future high - repetition rate accelerator because if a significant flux of ions strikes the wall it can cause significant damage . it was suggested that the wakefield energy in the plasma wave could be replenished and sustained @xcite by a train of energy source in order to achieve high repetition rate . however , as shown in this paper due to ion motion this is not possible . we explore the use of the ion - wake channel for positron - beam driven positron acceleration in a novel and relevant `` suck - in '' regime @xcite where the positron beam radius @xmath46 . such channels are also promising @xcite@xcite@xcite for exciting electron - wakefields with independent transverse and longitudinal field spatial structures . the formation of plasma channels which are vaguely similar to the cylindrical soliton shown here but are excited by significantly different processes have been shown previously . these processes are : using a collimated laser with annular profile @xcite@xcite , by an electrical discharge @xcite , by linear wakes in the self - modulated regime @xcite@xcite and by the ponderomotive force of a short - pulse laser - driven linear wake @xcite . in the linear electron wake regime the wakefields are symmetric so they average out over the electron oscillation period and only the second - order ponderomotive force ( @xmath47 ) drives the ion - motion ( @xmath48 ) @xcite . spatially , in the linear - wake regimes the radial ion excitation is limited to around @xmath40 and the ion - density perturbation is linear in the background plasma density , @xmath25 @xcite@xcite . additionally , in the earlier studies no ion acoustic - waves have been examined . to develop insight into the ion wake physics , we consider the 1-d simplified dispersion relation of the ion - acoustic plane waves , @xmath49 where , @xmath50 and @xmath51 under the collision - less condition , @xmath52 and @xmath53 is the adiabatic index with @xmath54 being the degrees of freedom of the ions . during the initial phase of the interaction between the bubble fields and the ions , the ion inertia dominates their motion . at times when the fields of the bubble interact with the ions , the inertia of ions leads to very small spatial displacement scales and so @xmath55 so the term @xmath56 where @xmath57 is the debye wavelength . thus in this phase @xmath58 and the ion - soliton spike at @xmath59 grows over the plasma frequency timescale , @xmath60 . at initial times the plasma is also cold so @xmath61 . when the ions gain significant momentum and start oscillating over larger spatial scales in response to the electron dynamics then @xmath62 . in this phase the acoustic wave propagation becomes dispersion - less with @xmath63 . however , when the ion density perturbation undergoes steepening in a large amplitude acoustic wave forming a spike over much smaller spatial scales , @xmath64 becomes large and dispersion becomes important . the ion - acoustic dispersion relation as @xmath64 increases has higher - order corrections in @xmath64 , @xmath65 . also , as the ions are heated at later time @xmath66 increases and modifies the sound speed to @xmath67 . thus when the ions are heated the sound speed lies between @xmath68 and @xmath69 . during the time when the oscillations in the electron plasma - wave wake are undergoing phase - mixing the wakefields still persist and the ion motion is driven by the fields as well as the thermal pressure . we show under the linear approximation how an ion - wake differs from a freely propagating ion - acoustic wave . it should be noted that as the wake electrons phase - mix , the plasma is not isothermal in this case . the linear ion - acoustic wave can be obtained by perturbative expansion of ion density , @xmath70 and ion fluid velocity @xmath71 in the zeroth - order ion fluid continuity equation , @xmath72 . taking a partial derivative with time , @xmath73 . the ion - fluid equation of motion where the electron temperature has a spatial gradient is , @xmath74 . upon substituting the equation of motion in the time - derivative of the linearized continuity equation , @xmath75 . thus , a driven ion - acoustic wave linearized to the first - order in density perturbation has the form , @xmath76 ( easily inferred from the radial electric field ) . ( a ) electron density in 2d cylindrical real - space , ( b ) corresponding longitudinal electric field profile and ( c ) corresponding radial - field profile . the beam - plasma parameters are the same as in fig.[fig2:ion - wake - beam ] but the electron - wake is shown at an earlier time @xmath77 . ] in this first - order approximation the right - hand side of the eq.[ion - wave - equation - first - order ] shows two separate timescales of the ion - wake . at earlier times , the first term on the right - hand side dominates . this is the _ formation _ phase of the ion - wake where the bubble electron oscillations undergo ordered radial motion and the bubble radial electric field excites the inertial response of the ions . this leads to the formation of the on - axis and @xmath59 ion density spikes . at later times after the phase - mixing between radial oscillators the electrons thermalize and @xmath78 . this is the _ propagation _ phase where the electron thermal pressure gradient drives the cylindrical soliton at @xmath59 radially outwards . the pic simulations show the two phases of the electron dynamics . the coherent and phase - mixing phases of the electron oscillations are evident in fig.[fig3:bubble - train](a ) . the electron motion in the first oscillation bucket is coherent whereas several oscillations behind the driver , the electron begin to de - cohere . this is also evident in the time - evolution of the ion soliton in fig.[fig6:ion - motion - lineouts](b ) where the radial electric field goes to zero around 200@xmath79 and the ion - soliton is seen moving outwards radially in [ fig6:ion - motion - lineouts](a ) . the linearized ion - acoustic wave equation is however inadequate to describe the propagating solitary density spike at the ion - wake edge , with ion density accumulation many times the background density . in the linear regime the homogenous ion - acoustic wave equation predicts sinusoidal radial ion oscillations that support the wave . when the density in the ion perturbation begins to rise to the order of the background density , the electrostatic potential due to charge - separation between the ions and the thermal electrons correspondingly rises . this leads to _ wave - steepening _ due to the preferential acceleration of ions in the direction of the ion - acoustic wave velocity . when the potential of the wave is large enough the ions get trapped and co - propagate with the ion - acoustic wave phase velocity , this non - linearity is the basis of the soliton . it should be noted that the linearized kinetic theory does not formally incorporate the trapping of particles at the wave phase - velocity . in this limit the density perturbation shape is therefore not sinusoidal as the co - propagating background ions accumulate and their density perturbation takes the form of an ion - soliton . the co - propagating ion velocity in the soliton can therefore exceed the ion - acoustic phase velocity , @xmath80 and @xmath81 where @xmath82 is the mach number . therefore , non - linear acoustic waves are in the form of a soliton and propagate faster than the ion - acoustic velocity . in the second - order , the non - linear ion - density spike @xmath83 propagation is governed by the korteweg - de vries ( kdv ) equation @xcite which has propagating solutions of the form @xmath84 @xcite where @xmath85 is the ion - acoustic waveform , a soliton solution and @xmath86 is the mach number of the propagating solution . higher - order contributions to the kdv equation have also been considered . however , the standard form of the kdv equation assumes an isothermal plasma whereas the bubble - wake phase - mixes into a plasma with radial temperature gradient . in a non - isothermal plasma the effect of trapped electrons in the ion - soliton have been considered using the bernstein - greene - kruskal model at the ion - acoustic velocity @xcite . it is also known that a single ion - soliton under the appropriate conditions can break - up into multiple solitons leading to a n - soliton solution . this is also a phenomenon we observe in the simulations shown in the ion density of the beam - driven case at @xmath87 in fig.[fig2:ion - wake - beam ] . we consider a description of the non - linear cylindrical ion - acoustic waves with a radial temperature gradient by retaining the second - order of perturbation in ion - density and fluid velocity . we also assume that the background electron trapping does not significantly modify the distribution function . we assume that the temperature does not change by several orders of magnitude in vicinity of the ion soliton . this assumption is also validated by the pic simulations in fig.[fig4:elec - temperature - profile ] . / \sigma_{k } n_e(k , r ) } $ ] , profile of the wake electrons corresponding to the time in fig.[fig2:ion - wake - beam ] at 460 @xmath79 . this represents the square - root of the electron temperature , @xmath88 . the radial gradient of the temperature , @xmath89 is thus computed at the peak of the soliton ( red ) and in its vicinity ( green ) . it is interesting to note that @xmath90 . ] to obtain the kdv equation in cylindrical coordinates with radial temperature gradient we normalize with respect to the local electron temperature , the radius : @xmath91 , time : @xmath92 , electric field : @xmath93 , potential : @xmath94 , ion - density perturbation : @xmath95 , electron - density perturbation : @xmath96 , ion - fluid velocity : @xmath97 . under this normalization the cylindrical coordinate equations transform as : electron boltzmann distribution equation @xmath98 , ion - fluid continuity equation @xmath99 , ion - fluid equation of motion @xmath100 and the poisson equation @xmath101 . the electric field @xmath102 is both due to the thermal pressure and the radial fields of the wake , @xmath103 . but , in the following analysis the propagation of a non - linear ion - acoustic wave is considered , so we assume that the electron oscillations are thermalized and thus the effect of the fields of the wake is negligible , @xmath104 . we look for a propagating disturbance of @xmath105 , @xmath106 , @xmath107 and @xmath102 in a stationary background plasma with uniform background density @xmath25 . we consider weakly non - linear ion - acoustic wave and expand all the wave quantities in the powers of @xmath108 . we expand the ion - density : @xmath109 , electron density : @xmath110 , electric - field : @xmath111 , electrostatic potential : @xmath112 , electron temperature : @xmath113 and ion - fluid velocity : @xmath114 . note that we have assumed that before the electron wake excitation the plasma is cold , @xmath115 . we transform to a moving frame of the steepened ion density perturbation using the coordinate transform @xmath116 and @xmath117 . therefore , the radial @xmath118-derivative is , @xmath119 and the @xmath120-derivative is @xmath121 . using this , @xmath122 and @xmath123 . we renormalize the electric field as , @xmath124 . note that in the moving frame the potential gradient is , @xmath125 , so @xmath126 is a more appropriate quantity . under the assumption that in the moving - frame the quantities of the disturbance change with small @xmath108 , the terms in equations are perturbatively expanded and the terms with same powers of @xmath127 are collected . from the @xmath128 order terms of all the equations above , we infer @xmath129 and @xmath130 . collecting the @xmath131 terms from the boltzmann s equation we obtain @xmath132 . from the ion - fluid equation of motion we obtain @xmath133 and from the poisson equation we obtain @xmath134 . taking the @xmath127-order terms of the continuity equation and substituting @xmath85 we obtain , @xmath135 \right ) - \delta \left(u^2 + v^{(2 ) } \right ) = 0 $ ] . neglecting quantities with @xmath127 times the second - order terms and using the @xmath136 result above , @xmath137 = - \mathcal{u}\frac{\partial } { \partial \xi}t_e^{(1)}$ ] . using the @xmath138 terms of the poisson equation we obtain the driven korteweg - de vries equation in cylindrical coordinates @xcite , @xmath139 it differs from the cartesian - kdv equation by the term @xmath140 and the driver term @xmath141 . the cartesian kdv equation can be analytically solved to obtain two classes of solutions : ( a ) self - similar solutions which are shown in @xcite to be airy functions and ( b ) soliton solutions . a `` soliton '' is a single isolated pulse which retains its shape as it propagates at some velocity , @xmath142 . this means that for a soliton - like solution the @xmath85 only depends upon the soliton - frame variable , @xmath143 and not on space - like @xmath144 and time - like @xmath145 variables separately . the solution of the cartesian kdv equation in this co - moving frame is @xmath146 @xcite . the ckdv equation and the driven ckdv equation obtained here can not be solved analytically . however , the numerical analysis and experimental verification @xcite of the cylindrical - kdv ( ckdv ) equation show that it supports functions of the form @xmath147 in the form of a cylindrical ion - soliton . but , the amplitude of the cylindrical soliton changes as it propagates . the velocity of the soliton in the cylindrical case is higher than in cartesian case @xcite . since the ion - wake is excited in a non isothermal plasma its velocity changes as it is driven . the mean electron temperature reduces because the volume of the ion - wake expands as the soliton moves radially outwards . it is also known to support n - soliton solution . we computationally seek the dependence of the non - linear ion - density spike on @xmath148-coordinate . it should be noted that such soliton solutions are supported under certain limiting condition on the mach number , @xmath86 . the strict condition on the existence and stability of an ion - soliton arises from the requirements on the magnitude of the potential required for trapping the background ions for a given velocity of the soliton . here we find that the speed of ion soliton is approximately equal to the ion - acoustic speed using the mean temperature . as this is not an isothermal plasma there is no well - defined ion - acoustic speed . so , the ion - acoustic wave is phase - mixed as its phase - velocity is spatially varying . however , we confirm from the simulations that this condition is strongly satisfied as @xmath149 . in this scenario the mean electron temperature provides an estimate of the sound - speed . the local electron temperature of the ion soliton as shown in fig.[fig4:elec - temperature - profile ] , is used to calculate mach number , @xmath86 and thus a stability criteria can be derived . this problem is represented using the condition on the sagdeev psuedo - potential , @xmath150 that it has to be a real number . this condition is satisfied when @xmath151 therefore , @xmath152 and @xmath153 . using this we find the well - known condition , @xmath154 as will be shown later , we find from simulations that the mach number calculated using the mean temperature is well within these bounds and thus the soliton is stable . since the characteristic time of ion - motion is much longer than the electron oscillations , the longitudinal field @xmath155 averages out over the full bubble electron oscillation . so , the ions do not gain any net longitudinal momentum . however , atypical radial ion - dynamics arise because the radial fields , @xmath156 are asymmetric in time as shown in fig.[fig5:bubble - ion - dynamics ] and do not average to zero , driving an average radial ion - momentum . the first stage of the ion - wake formation is controlled by the different time - asymmetric phases in fig.[fig5:bubble - ion - dynamics ] of the bubble radial impulses namely , suck - in " due to the electron compression in the back of the bubble @xmath157 during @xmath158 shown in fig.[fig5:bubble - ion - dynamics](c ) , and the push - out " due to the mutual - ion space - charge coulomb repulsion force @xmath159 during @xmath160 shown in fig.[fig5:bubble - ion - dynamics](d ) . the suck - in force is spatially - periodic at non - linear plasma wavelength , @xmath161 with a duty - cycle @xmath162 . in addition to the plasma wake , the propagating energy sources themselves impart impulses such as the laser ponderomotive force @xmath163 where @xmath164 and the radial force of the drive beam @xmath165 where @xmath166 . we neglect the driver impulses ( below threshold intensity for direct non - linear ion excitation @xcite@xcite ) ) because they act on the ions over their sub - wavelength short duration . this is unlike the slowly - propagating wake - plasmon bubbles that undergo continual interaction over many plasma periods . the validity of this assumption is evident from the laser ion - wake in fig.[fig1:ion - wake - laser ] . since the ponderomotive force of a laser driver is an outward force for both the electrons and ions , the on - axis density - spike can not be from this force . similarly the ion - density - spike at the radial wake - edge in an electron beam driven ion - motion can not be excited directly by the force of the beam , and is caused by the electron wake s radial - edge density compression . the lagrangian fluid model of the ions in a bubble consists of ion - rings under cylindrical symmetry with @xmath167 . the bared - ion region inside the bubble is assumed to be a positively charged cylinder under steady - state approximation ( @xmath168 , back of the bubble electron compression radius ) . the force on the ions from the non - linear electron compression @xmath169 in the back of the bubble and radius @xmath170 , pulls the ion rings inward ; and within the bubble , the space - charge force of the ions opposes it and prevents full collapse . the `` suck - in '' force on ions is @xmath171 . the space - charge force on the ions in the cavity is @xmath172 . the equation of motion is @xmath173 using , @xmath174 , we have , @xmath175 where we have assumed that @xmath176 . therefore the equilibrium radius where the impulses balance out is @xmath177 . the ion - rings at @xmath178 collapse inwards towards the axis resulting in an on - axis density spike . whereas the ion - rings at @xmath179 move out away from the axis . for @xmath180 the ion - response is slower but similar . when the radially outward moving ion - rings reach beyond @xmath59 , there is excess net negative charge of the wake electrons within the bubble - sheath . as a result the radially propagating ion rings get trapped and start accumulating just inside the bubble and can not freely move beyond , forming a density compression at @xmath59 . so , the cylindrical ion soliton is formed around @xmath59 . this is seen in fig.[fig1:ion - wake - laser ] and fig.[fig2:ion - wake - beam ] where the ion and electron density peak at @xmath59 . in the non - linear regime @xmath181 , the spatial - scale of the ion - wake is over several @xmath40 . this is due to the balance between the opposite radial forces on the electrons at @xmath59 , from the driver and the ion cavity @xcite@xcite . in the laser - driven bubble @xmath182 gives @xmath183 when simplified using @xmath184 and @xmath185 , @xmath186 ( computationally , @xmath187 @xcite ) . in the beam - driven bubble @xmath188 which gives , @xmath189 , where @xmath190 is the line charge density of the beam . we verify the above model using @xmath191 osiris pic simulations @xcite of the ion - wake in the bubble regime by simulating various laser - pulses in cartesian coordinates and electron - beams in cylindrical coordinates . we initialize eulerian specification of the plasma ( non - moving window ) of pre - ionized @xmath192 homogeneous density with a density - ramp over 20 @xmath40 at each of the vacuum - plasma interfaces . we resolve the smallest spatial scale , @xmath40 in the beam case and @xmath193 in the laser case ( laser frequency @xmath1 ) , with 20 cells . the time - step satisfies causality and minimizes numerical dispersion . it is normalized to @xmath79 for beam and to @xmath194 for lasers . we use absorbing boundary conditions for fields and particles . the laser pulse is circularly polarized with radially gaussian and longitudinally polynomial profile @xcite with @xmath0 ( not shown @xmath195 to @xmath196 ) , pulse length of @xmath197 , matched focal spot - size radius of @xmath198 , and laser frequency @xmath199 . the electron beam is initialized with @xmath200 , @xmath201 ( not shown @xmath202 to @xmath203 ) and spatial gaussian - distribution with @xmath204 and @xmath205 . we compare the electron - beam driven ion - wake soliton structure in theory to the simulations in fig.[fig5:bubble - ion - dynamics](a ) and [ fig3:bubble - train](a ) . the observed @xmath206 ( just behind the beam ) whereas the estimated bubble radius is @xmath207 ( @xmath208 , the assumption @xmath209 is not well satisfied ) . in fig.[fig2:ion - wake - beam ] which is in the propagation - phase , the observed ion - wake soliton is located at @xmath210 at 460 @xmath79 which is about @xmath211 . the ion - soliton is excited at an early time around @xmath59 and in the snapshot in fig.[fig2:ion - wake - beam ] it has propagated outwards . the on - axis density spike drops to a minimum at @xmath212 in fig.[fig2:ion - wake - beam ] whereas the estimated @xmath213 ( @xmath214 , @xmath215 , @xmath216 ) . the radial ion momentum @xmath217 phase - space in fig.[fig7:radial - phase - spaces](b ) shows the ions accumulate at the axis and the channel edge at a time corresponding to fig.[fig2:ion - wake - beam](b ) . the ions at the channel edge are seen to have a drift velocity and a thermal spread . the radial electron momentum @xmath217 phase - space in fig.[fig7:radial - phase - spaces](a ) shows that a large density of thermalized electrons are trapped within the ion soliton which is confirmed from the density plots in fig.[fig2:ion - wake - beam](a ) . in the laser - driven bubble simulations the expected and observed @xmath218 as shown in fig.[fig1:ion - wake - laser](a ) . in fig.[fig1:ion - wake - laser](c ) the ion - wake soliton is created at @xmath219 . the expected and observed on - axis density - spike radius is @xmath220 ( @xmath221 , @xmath215 , @xmath216 ) . the model for the excitation of this structure of the non - linear wake has been verified for a range of laser and beam parameters from quasi - linear to strongly non - linear electron wake regime . as described in section [ ion - soliton - excitation - phase ] the electron bubble - wake train fields excite a cylindrical ion soliton . eventually , the electron oscillations phase - mix and thermalize as electron thermal energy on the time - scale of about an ion plasma period . the cylindrical soliton then propagates radially outwards driven by the temperature gradient as shown in eq.[cylindrical - kdv - equation ] . this is the modified ckdv equation in a non - equilibrium condition when an electron temperature gradient also contributes to the force on the ions . ) with time and a third - order fit ( green ) for the position of the ion density - spike of the soliton . ( b ) radial wakefields of the electron bubble oscillations ( in terms of @xmath222 ) at the electron density spike ( magenta ) and at the ion density spike ( blue ) . ( c ) radial velocity of the ion density spike of the soliton calculated from the third - order fit curve ( red ) . an estimate of the sound speed ( green ) using the mean temperature , between the axis & the soliton location ( green ) and in the vicinity of the soliton peak ( blue ) , in the expression @xmath223 . since the plasma is not isothermal the mean temperature is calculated by averaging the temperature of electrons over the indicated spatial region . ( d ) gradient of the electron temperature at the soliton ion density peak ( blue ) and in the vicinity of the peak ( red ) . the vicinity of the ion density peak of the soliton is defined as shown in fig.[fig4:elec - temperature - profile ] . , width=460 ] the channel - edge density spike , with a form similar to the ckdv - solution in the @xmath224 frame as shown in fig.[fig1:ion - wake - laser ] and fig.[fig2:ion - wake - beam ] to be propagating radially outwards . the propagation phase starts around @xmath225 as the radial electric fields @xmath226 as shown in fig.[fig6:ion - motion - lineouts](b ) . the propagation phase is evident in fig.[fig6:ion - motion - lineouts](a ) where the red curve is the position of the peak of the ion - soliton in time . the cylindrical ion - soliton has propagated from @xmath227 ( also seen in fig.[fig2:ion - wake - beam ] ) to @xmath228 which corresponds to an average speed of @xmath229 . we compare the time - averaged soliton speed @xmath230 to the average speed of sound , @xmath231 where the average @xmath232 from the electron phase - space ( not shown ) . this gives @xmath233 ( @xmath234 for 2d ) in agreement with the average soliton velocity . using this time - averaged analysis we see that @xmath235 and so the stability criteria in eq.[kdv - solution - existence ] is satisfied . however , as the soliton moves out the volume between the axis and the soliton edge increases thus the electron thermal energy spreads over a larger volume . the re - distribution of the thermal energy over a larger volume leads to the reduction in the temperature with time . the soliton is not freely propagating but is driven by the radial gradient of the electron temperature as shown in eq.[cylindrical - kdv - equation ] . the soliton speed thus changes in time as shown in the red curve of fig.[fig6:ion - motion - lineouts](c ) . the sound speed also varies with time and it is estimated using the temperature at that instant using , @xmath236 in terms of the root mean squared electron radial momentum . as the plasma is not in thermal equilibrium and its temperature varies radially as shown in fig.[fig4:elec - temperature - profile ] . the root - mean - square radial momentum represents the temperature at an instant of time and is calculated over radial dimension - from the axis to the soliton location : channel-@xmath237 / \sigma^{r_{sol}}_{r=0 } n_e(r ) } ~ \propto ~ \sqrt{t_e}$ ] or in the vicinity of the soliton : soliton-@xmath238 / \sigma^{r_{sol}+\epsilon}_{r_{sol}-\epsilon } n_e(r ) } ~ \propto ~ \sqrt{t_e}$ ] from the @xmath217 phase - space . the instantaneous sound speed , @xmath239 computed with channel-@xmath240 is in the green curve in fig.[fig6:ion - motion - lineouts](c ) and @xmath239 computed with soliton-@xmath240 is in the blue curve in fig.[fig6:ion - motion - lineouts](c ) . the extent of @xmath241 around the soliton peak is shown in fig.[fig4:elec - temperature - profile ] . we compare the curves in ( i ) red : @xmath242 ( from the 3rd - order curve - fitting to the position of the ion - density peak ) , ( ii ) green : @xmath239 from channel-@xmath243 and ( iii ) blue : @xmath239 from soliton-@xmath243 in fig.[fig6:ion - motion - lineouts](c ) . from the comparison it is observed that they are in good agreement . it can be seen that the velocity of the soliton estimated using the location of the ion - density peak ( red ) lies between @xmath239 calculated using the average temperature over the channel ( green ) which is the upper limit and @xmath239 calculated using the average temperature over the soliton ( blue ) which is the lower limit . note , that since @xmath244 the blue curve which is lower than the @xmath142 curve in red , is a better estimate of the sound speed in the vicinity of the ion - soliton . the sound - speed with radial momentum averaged over the channel is over - estimated because the thermal electron spread estimation also includes the velocity of the electrons drifting with the ion - soliton . we also present the radial gradient of the electron temperature , @xmath245 in fig.[fig6:ion - motion - lineouts](d ) . it is interesting to note from the blue curve in fig.[fig6:ion - motion - lineouts](d ) that the temperature gradient at the peak of the ion - soliton is zero , @xmath246 . in the vicinity of the soliton peak we see that the gradient of the temperature follows the variation in the ion soliton velocity , this follows from eq.[cylindrical - kdv - equation ] . the vicinity of the soliton peak is shown as the green curve overlaid on the thermal momentum curve in fig.[fig4:elec - temperature - profile ] . in fig.[fig2:ion - wake - beam](b ) n - soliton formation is observed in the ion - density at around @xmath247 . the single - ion soliton is seen splitting into several solitons . the n - soliton solution is formed because different ion - rings that are driven in the push - out " phase have a radial position dependent defocussing force acting on them , @xmath248 . this is shown in fig.[fig5:bubble - ion - dynamics](d ) . thus the ion - rings originating at a larger radii from the axis are pushed outwards with a force of a higher magnitude and the rings originating at a smaller radii just outside @xmath249 are pushed outwards by a smaller force . so , over a longer time the set of ion - rings with a higher initial momentum propagate radially outwards at a larger radial velocity . this break - up of a single ion - soliton into n - solitons occurs over a longer time - scale because the difference in momentum is small compared to the average momentum . the thermal momentum , @xmath250 at this time is less than one - tenth of the peak wake quiver momentum . there are several reasons for the cooling , such as , transfer of the wake energy to the ions and the trapped electrons @xcite , escape of the highest energy electrons and un - trapped ions from the channel edge , energy loss to the bow - shock and the re - distribution of the energy over an expanding volume . the peak radial ion - momentum is @xmath251 which shows that not all the radially propagating ions are trapped . the un - trapped free - streaming ions at @xmath252 can be distinguished from the ions at the channel - edge in @xmath217 phase - space . radial momentum phase - space showing the accumulation of thermalized electrons within the ion - soliton . ( b ) ion @xmath217 radial momentum phase - space showing the on - axis and ion - wake edge ion accumulations . , width=460 ] positron acceleration using the ion - wake channel is explored in the non - linear suck - in regime of positron - beam radii @xmath253 and peak density @xmath254 . in this regime @xmath255 , so the on - axis ion density has a radially limited defocussing force . it is well known @xcite that in a homogeneous plasma positron beam driven wakes have two major problems @xcite - ( i ) the plasma electrons collapsing to the axis under the force of a positron beam start at different radii . the positron beam radial force driving the suck - in " decreases with the radii . as a result , the electron compression is not optimal . ( ii ) the plasma ions located in the path of the positron beam result in a de - focussing force on it . the transport of the positron in a positron - beam driven wake in homogenous plasma is not ideal . as the ion - wake channel is a practical realization of the hypothesized ideal hollow - channel plasma @xcite we examine its excitation by a positron - beam @xcite . in this section we analyze whether the positron - beam driven wake - fields excited in the ion - wake can be used for the acceleration and transport of a positron beam . we describe the analytical model of the radial electron suck - in " based excitation of a positron beam wake in the plasma . the equation of motion of the plasma electron rings at @xmath256 from the axis , under the positron beam suck - in force is @xmath257 where @xmath258 is the space just behind the positron beam with velocity @xmath259 driving the collapse . this is a non - linear second - order differential equation of the form , @xmath260 where @xmath54 is not linear in @xmath256 . under the assumption about the positron - beam properties , @xmath261 and @xmath262 being constant during the entire interaction of the positron - beam with the hollow - channel over its full length so , upon dropping the dependence on @xmath144 the equation simplifies to its _ special case _ which has analytical solutions , @xmath263 . the solution to this equation is @xcite , @xmath264 where @xmath265 . therefore the collapse time - duration is @xmath266 . we note that there is an anomaly that exists in our problem formulation and the solution because we have not taken into account the space - charge force of the compressing electrons as they collapse to the axis and this force balances the suck - in force of the positron beam . under these approximations the collapse time is @xmath267 also , note that we have neglected the initial expansion velocity of the channel , @xmath268)@xcite . for optimal compression avoiding phase - mixing , the electron rings should collapse over , @xmath269 where @xmath270 is the non - linear wavelength of the positron - driven wake and @xmath271 is the duty - cycle of compression phase . so , the optimal channel radius is @xmath272 . the scaling of the @xmath273 with positron beam parameters is show in @xcite . , @xmath274 , @xmath275 ) . the radial profile of the accelerating - wakefield and normalized focussing - wakefield potential ( radial field integrated from the edge of the box to a radius).,width=460 ] using 2-@xmath276d pic simulations in a moving window we study the positron beam driven wakefields in cylindrical geometry . we compare positron acceleration in an ideal ( heaviside density function , @xmath277 ) and an ion - wake channels ( on - axis and channel - edge density - spike , channel minimum density of @xmath278 ) with @xmath279 . for non - linear wake parameters @xmath280 , @xmath281 and @xmath282 ( @xmath283 ) . fig.[fig8:positron - acc - field - radial - profiles ] shows that the peak on - axis accelerating field is @xmath284 for an ideal channel and @xmath285 for the ion - wake channel . fig.[fig8:positron - acc - field - radial - profiles ] also shows that the focussing potential ( normalized to 27.6 @xmath286 ) is similar and overall focussing in both cases . however , in the ion - channel the radial field is defocussing around the on - axis ion - spike . thus the ion - wake channel is useful for accelerating and transporting positrons despite the lower accelerating fields in comparison with the ideal channel . we also note that the on - axis density spike has detrimental effect on the focussing fields near the axis . ideal channels of a few @xmath40 are technologically challenging whereas the ion - wake channel of radius @xmath287 is formed behind every bubble - wake . in conclusion , using theory and pic simulations we have shown the dynamics of the formation and evolution of a non - linear ion - wake excited by the well - characterized electron bubble - wakefields @xcite-@xcite . we have shown that the non - linear ion - wake has the characteristic cylindrical ion - soliton solution and evolves to an n - soliton solution over longer time as described by the ckdv equation . thus over the period of persistence of the ion - soliton a second electron bunch can not be accelerated in the plasma . this establishes an upper limit on the repetition rate of the plasma collider . we have also shown the feasibility of using the ion - wake channel for positron acceleration in an experimentally relevant parameter regime . work supported by the us department of energy under de - sc0010012 and the national science foundation under nsf - phy-0936278 . we acknowledge the osiris collaboration @xcite for providing us with the code and support for pic simulations . we thank dr . v. yakimenko for useful ideas and suggestions . we acknowledge encouragement for experiments from dr . m. hogan and the facet group at stanford linear accelerator laboratory and prof . m. downer s group at university of texas at austin . we acknowledge the 256-node _ chanakya _ server at duke university supported and maintained by mr . mahmood sayed . tajima , t. , dawson , j. m. , _ laser electron accelerator _ , http://link.aps.org/doi/10.1103/physrevlett.43.267[phys . lett . * 43 * , pp.267 - 270 ( 1979 ) ] , doi : 10.1103/physrevlett.43.267 ; + chen , p. , dawson , j. m. , huff , r. w. , katsouleas , t. , _ acceleration of electrons by the interaction of a bunched electron beam with a plasma _ , http://link.aps.org/doi/10.1103/physrevlett.54.693[phys . * 54 * ( 7 ) , 693 - 696 ( 1985 ) ] , doi : 10.1103/physrevlett.54.693 rosenzweig , j. b. , breizman , b. , katsouleas , t. , su , j. j. , _ acceleration and focusing of electrons in two - dimensional nonlinear plasma wake fields _ , http://dx.doi.org/10.1103/physreva.44.r6189[phys . a , * 44 * , ( 1991 ) ] , doi:10.1103/physreva.44.r6189 pukhov , a. , meyer - ter - vehn , j. , _ laser wake field acceleration : the highly non - linear broken - wave regime _ , http://dx.doi.org/10.1007/s003400200795[appl . b * 74 * , pp.355 - 361 ( 2002 ) ] . doi:10.1007/s003400200795 lu , w. , huang , c. , zhou , m. , tzoufras , m. , tsung , f. s. , mori , w. b. , and katsouleas , t. , _ a nonlinear theory for multidimensional relativistic plasma wave wakefields _ , http://dx.doi.org/10.1063/1.2203364[physics of plasmas * 13 * , 056709 ( 2006 ) ] , doi : 10.1063/1.2203364 mangles , s.p.d . , http://dx.doi.org/10.1038/nature02939[nature * 431 * , 535 ( 2004 ) ] ; geddes , c. g. r. , et . al . , http://dx.doi.org/10.1038/nature02900[nature * 431 * , 538 ( 2004 ) ] ; faure , j. , et . al . , http://dx.doi.org/10.1038/nature02963[nature * 431 * , 541 ( 2004 ) ] . rosenzweig , j. b. , cook , a. m. , scott , a. , thompson , m. c. , yoder , r. b. , _ effects of ion motion in intense beam - driven plasma wakefield accelerators _ , http://dx.doi.org/10.1103/physrevlett.95.195002[phys . lett . * 95 * , 195002 ( 2005 ) ] . doi:10.1103/physrevlett.95.195002 gholizadeh , r. , katsouleas , t. , muggli , p. , huang , c. k. , mori , w. b. , _ preservation of beam emittance in the presence of ion motion in future high - energy plasma - wakefield - based colliders _ , http://dx.doi.org/10.1103/physrevlett.104.155001[phys . lett . , * 104 * , 155001 ( 2010 ) ] doi:10.1103/physrevlett.104.155001 stephen maxon and james viecelli , cylindrical solitons , http://dx.doi.org/10.1063/1.1694941[physics of fluids ( 1958 - 1988 ) 17 , 1614 ( 1974 ) ] ; doi : 10.1063/1.1694941 ; stephen maxon and james viecelli , spherical solitons , http://dx.doi.org/10.1103/physrevlett.32.4[physical review letters , * 32 * , p.4 ( 1974 ) ] ; doi : 10.1103/physrevlett.32.4 noah hershkowitz and thomas romesser , observations of ion - acoustic cylindrical solitons , http://dx.doi.org/10.1103/physrevlett.32.581[physical review letters , * 32 * , iss . 11 , pp.581 - 583 ( 1974 ) ] ; doi : 10.1103/physrevlett.32.581 berezin , yu . a. , karpman , v. i. , theory of non - stationary finite - amplitude waves in a low - density plasma , http://www.jetp.ac.ru/cgi-bin/e/index/e/19/5/p1265?a=list[jetp , vol . 19 , no . 5 , p. 1265 ( november 1964 ) ] [ zhetf , vol . 46 , no . 5 , p. 1880 , november ( 1964 ) ] korteweg , d. j. and de vries , g. , _ on the change of form of long waves advancing in a rectangular canal , and on a new type of long stationary waves _ , http://dx.doi.org/10.1080/14786449508620739[philosophical magazine * 39 * , 240 , pp.422 - 443 , ( 1895 ) ] . doi:10.1080/14786449508620739 ; sahai , a. a. , katsouleas , t. c. , tsung , f. s. , and mori , w. b. , _ long term evolution of plasma wakefields _ , http://accelconf.web . cern.ch/accelconf/pac2013/papers/mopac10.pdf[proceedings of north american particle accelerator conference , sep 2013 , pasadena , ca , usa ] ( mopac10 , isbn 978 - 3 - 95450 - 138 - 0 ) , 03-alternative acceleration schemes , a23 ? laser - driven plasma acceleration . sahai , a. a. , katsouleas , t. c. , gessner , s. , hogan , m. , joshi , c. , mori , w. b. , _ excitation of wakefields in a relativistically hot plasma created by dying non - linear plasma wakefields _ , http://dx.doi.org/10.1063/1.4773768[aip conf . 1507 , 618 ( 2012 ) ] . doi:10.1063/1.4773768 ( also slac - pub-15367 ) lee , s. , katsouleas , t. c. , hemker , r. g. , dodd , e. s. , mori , w . b. , _ plasma - wakefield acceleration of a positron beam _ , http://dx.doi.org/10.1103/physreve.64.045501[phys . e , * 64 * , 045501 ( 2001 ) ] , doi:10.1103/physreve.64.045501 sahai , a. a. and katsouleas , t. c. , _ optimal positron - beam excited wakefields in hollow and ion - wake channels _ , proceedings of international particle accelerator conference 2015 , richmond , virginia , may ( 2015 ) sen - gupta , s. , kaw , p. k. , _ phase mixing of nonlinear plasma oscillations in an arbitrary mass ratio cold plasma _ , http://dx.doi.org/10.1103/physrevlett.82.1867[phys . lett . * 82 * , 1867 ( 1999 ) ] . doi:10.1103/physrevlett.82.1867 vlasov , a. a. , _ the vibrational properties of an electron gas _ , http://ufn.ru/en/articles/1968/6/a/[zh . , * 8 * , pp.291 ( 1938 ) ] . physics uspekhi , * 93 * , no.3 and 4 , may - june 1968 ] . tonks , l. and langmuir , i. , _ oscillations in ionized gases _ , http://dx.doi.org/10.1103/physrev.33.195[physical review , * 33 * , pp.195 - 210 , feb . ( 1929 ) ] ; chen , f. f. , _ introduction to plasma physics and controlled fusion , plenum press , new york ( 1984 ) , isbn-13 : 978 - 0306413322 _ sahai , a. a. , katsouleas , t. c. , tsung , f. s. , mori , w. b. , _ long term evolution of plasma wakefields _ , http://accelconf.web.cern.ch/accelconf/pac2013/papers/mopac10.pdf[proc . of north am . part . conf . , * mopac10 * , sep 2013 , pasadena , ca , usa ] ruth , r. d. , chao , a. d. , morton , p. l. , wilson , p. b. , _ a plasma wake field accelerator _ , http://www.slac.stanford.edu/cgi-wrap/getdoc/slac-pub-3374.pdf[particle accelerator , * 17 * , pp.171 - 189 ( 1985 ) ] slac - pub-3374 , july 1984 fonseca , r. a. , et al . _ osiris , a three - dimensional fully relativistic particle in cell code for modeling plasma based accelerators _ , http://dx.doi.org/10.1007/3-540-47789-6_36[lect . note comput . sci . * 2331 * , pp.342 - 351 ( 2002 ) ] . doi:10.1007/3 - 540 - 47789 - 6_36 gorbunov , l. m. , mora , p. , solodov , a. a. , _ plasma ions dynamics in the wake of a short laser pulse _ , http://dx.doi.org/10.1103/physrevlett.86.3332[phys . * 86 * , ( 2001 ) . ] doi:10.1103/physrevlett.86.3332 chiou , t. c. , katsouleas , t. c. , _ high beam quality and efficiency in plasma - based accelerators _ , http://dx.doi.org/10.1103/physrevlett.81.3411[phys . rev . lett . * 81 * , 3411 , ( 1998 ) ] . doi:10.1103/physrevlett.81.3411 vieira , j. , fonseca , r. a. , mori , w. b. , silva , l. o. , _ ion motion in self - modulated plasma wakefield accelerators _ , http://dx.doi.org/10.1103/physrevlett.109.145005[phys . 109 * , 145005 ( 2012 ) ] . doi:10.1103/physrevlett.109.145005 ting , a. , moore , c. i. , krushelnick , k. , manka , c. , esarey , e. , sprangle , p. , et . al . , _ plasma wakefield generation and electron acceleration in a self - modulated laser wakefield accelerator experiment _ , http://dx.doi.org/10.1063/1.872332[phys . plasmas * 4 * , iss.5 , ( 1997 ) ] . doi : 10.1063/1.872332 ; durfee , c. g. , and milchberg , h. m. , _ light pipe for high intensity laser pulses _ , http://dx.doi.org/10.1103/physrevlett.71.2409[phys . lett . * 71 * , 2409 , ( 1993 ) ] . doi:10.1103/physrevlett.71.2409 kimura , w. d. , milchberg , h. m. , muggli , p. , li , x. , and mori , w. b. , _ hollow plasma channel for positron plasma wakefield acceleration _ , http://dx.doi.org/10.1103/physrevstab.14.041301[phys . beams * 14 * , 041301 ( 2011 ) ] . doi:10.1103/physrevstab.14.041301 hubbard , r. f. , ehrlich , y. , kaganovich , d. , cohen , c. , moore , c. i. , sprangle , p. , ting , a. , zigler , a. , _ intense laser pulse propagation in capillary discharge plasma channels _ , http://dx.doi.org/10.1063/1.58909[aip conf . 472 , 394 ( 1999 ) ] . doi:10.1063/1.58909 tzoufras , m. , tsung , f. s. , mori , w. b. , sahai , a. a. , _ improving the self - guiding of an ultraintense laser by tailoring its longitudinal profile _ , http://dx.doi.org/10.1103/physrevlett.113.245001[phys . rev . lett . * 113 * , 245001 ( 2014 ) ] . doi:10.1103/physrevlett.113.245001 sahai , a. a. , katsouleas , t. c. , muggli , p. , _ self - injection by trapping of plasma @xmath288 oscillating in rising density gradient at the vacuum - plasma interface _ , http://accelconf.web.cern.ch/accelconf/ipac2014/papers/tupme051.pdf[*tupme51 * , proceedings of ipac2014 , dresden , germany , pp . 1479 - 1482 , 2014 ] schroeder , c. b. , whittum , d. h. , wurtele , j. s. , _ multimode analysis of the hollow plasma channel wakefield accelerator _ , http://dx.doi.org/10.1103/physrevlett.82.1177[phys . lett . , * 82 * , 1177 ( 1999 ) ] . doi:10.1103/physrevlett.82.1177 schroeder , c. b. , benedetti , c. , esarey , e. , leemans , w. p. , _ beam loading in a laser - plasma accelerator using a near - hollow plasma channel _ , http://dx.doi.org/10.1063/1.4849456[physics of plasmas , * 20 * , iss.12 , 123115 , ( 2013 ) ] doi:10.1063/1.4849456 ; schroeder , c. b. , esarey , e. , benedetti , c. , _ control of focusing forces and emittances in plasma - based accelerators using near - hollow plasma channels _ , http://dx.doi.org/10.1063/1.4817799[physics of plasmas , * 20 * , iss.8 , 080701 , ( 2013 ) ] .

an isophotal atlas of 75 star cluster pairs and multiplets in the small magellanic cloud is presented , comprising 176 objects . they are concentrated in the smc main body . the isophotal contours were made from digitized sky survey images and showed relevant structural features possibly related to interactions in about 25% of the sample . previous n - body simulations indicate that such shapes could be due to tidal tails , bridges or common envelopes . the diameter ratio between the members of a pair is preferentially in the range , with a peak at . the projected separation is in the range pc with a pronounced peak at pc . for 91 objects it was possible to derive ages from colour - magnitude diagrams using the ogle - ii photometric survey . the cluster multiplets in general occur in ob stellar associations and/or hii region complexes . this indicates a common origin and suggests that multiplets coalesce into pairs or single clusters in a short time scale . pairs in the smc appear to be mostly coeval and consequently captures are a rare phenomenon . we find evidence that star cluster pairs and multiplets may have had an important role in the dynamical history of clusters presently seen as large single objects . the images in this study are based on photographic data obtained using the uk schmidt telescope , which was operated by the royal observatory edinburgh , with funding from the uk science and engineering research council , until 1988 june , and thereafter by the anglo - australian observatory . original plate material is copyright by the royal observatory edinburgh and the anglo - australian observatory . the plates were processed into the present compressed digital form with their permission . the digitized sky survey was produced at the space telescope science institute under us government grant nag w-2166 . ]

astro-ph0006369

star cluster pairs are common objects in the magellanic clouds and it is important to understand their formation and evolution processes . a list of 30 cluster pairs in the smc was first presented by hatzidimitriou & bhatia ( 1990 ) . bica & schmitt ( 1995 , hereafter bs95 ) revised previous data on smc extended objects ( star clusters , associations and emission nebulae ) and identified new ones using sky survey eso / serc r and j films . they presented a list of 40 pairs and 2 triple star clusters . pietrzyski & udalski ( 1999a ) reported 23 pairs and 4 triplets derived from pietrzyski et al.s ( 1998 ) star cluster catalogue in the ogle survey area . bica & dutra ( 2000 ) updated bs95 s catalogue considering the entries in pietrzyski et al s catalogue . bica & dutra ( 2000 ) indicated 75 pairs and multiplets comprising 176 individual objects . in recent years there has been growing evidence of interacting star clusters in the magellanic clouds , especially in the lmc . bhatia & hatzidimitriou ( 1988 ) concluded that more than 50% of lmc pairs must be physical systems . bhatia & mcgillivray ( 1988 ) found evidence that ngc2214 is a merging binary star cluster , based on the presence of a flattened core and an extended halo . indeed sagar et al . ( 1991 ) detected two turnoff points revealing the presence of two interacting clusters . bica et al . ( 1992 ) studied cluster pairs and multiplets in the lmc bar by means of integrated colours and found systems which resulted coeval and some with age differences . vallenari et al ( 1998 ) confirmed such scenarios by means of colour - magnitude diagrams . they also presented isophotal contours indicating physical interaction . several other studies have found binarity evidence in lmc cluster pairs ( e.g. kontizas et al . 1993 , dieball & grebel 2000a , 2000b ) . bhatia et al . ( 1991 ) presented a photographic atlas of binary star cluster candidates in the lmc . for the smc no morphological atlas is available and isophotal maps are required to test possible physical interactions . indeed comparisons of isophotes with isopleth maps of n - body simulations proved to be a useful tool ( rodrigues et al . 1994 , de oliveira et al . 1998 , hereafter odb98 ) , since the simulations produce features such as bridges , common envelopes and extensions . the observational importance of tidal tails as interaction signatures was also indicated by leon et al . ( 1999 ) . in this work we provide isophotal maps for smc pairs and multiplets to study the following properties of these candidate physical systems : ( i ) angular distribution ; ( ii ) projected centre - to - centre separation of members ; ( iii ) isophotal structures using the digitized sky survey@xmath6 ( hereafter dss ) ; ( iv ) ages derived by means of isochrone fitting , when possible . we discuss candidate physical systems and infer a scenario for their formation and evolution . in sect . 2 we gather the objects providing coordinates , sizes , centre - to - centre separations and other details for the smc pairs and multiplets . in sect . 3 we provide the isophotal contour atlas together with classifications of interaction features whenever present . a preliminary version of the present isophotal smc atlas together with one for the lmc was given in de oliveira ( 1996 ) . in sect . 4 we derive cluster ages by means of colour - magnitude diagrams extracted from the ogle - ii bvi photometric survey ( udalski et al . 1998 ) . in sect . 5 we discuss the properties of the systems and the possible scenarios for their origin and evolution . finally , concluding remarks are given in sect . table [ tab : catalog ] shows data for the 176 smc objects which form 75 star cluster pairs and multiplets ( bica & dutra 2000 ) , considering that the pair bs63/b67 in the latter study is possibly a triplet with ngc294 ( odb98 ) . by columns : ( 1 ) object cross - identification in the different catalogues : n ( henize 1956 ) , k ( kron 1956 ) , l ( lindsay 1958 ) , h ( hodge 1960 ) , sl ( shapley & lindsay 1963 ) , b ( brck 1976 ) , dem ( davies et al . 1976 ) , eso ( lauberts 1982 ) , h86- ( hodge 1986 ) , ma ( meyssonier & azzopardi 1993 ) , bs ( bica & schmitt 1995 ) , ogle ( pietrzyski et al . 1998 ) . note that some are embedded objects , named after the corresponding hii region ; ( 2 ) and ( 3 ) right ascension and declination for the epoch j2000 , respectively ; ( 4 ) object type following bs95 : c for star cluster , nc for small hii regions with embedded star clusters , cn for clusters which show some traces of emission , a for associations , ca and/or ac for star clusters of low density and objects with intermediate properties , an for associations which show some traces of emission and na for hii regions with embedded associations ; ( 5 ) and ( 6 ) major and minor sizes , respectively ; ( 7 ) position angle of major axis ( @xmath7=n , @xmath8=e ) ; ( 8) centre - to - centre angular separation ( 1 arcsec = 0.28 pc assuming an absolute distance modulus ( m - m ) = 18.9 for the smc , westerlund 1990 ) . for triplets and multiplets we measured the separation between the two main members , as indicated in the corresponding object line in the table ; ( 9 ) remarks : @xmath9 and @xmath10 indicate groups with six , five and four members , triplets and pairs , respectively . a running number identifies each candidate system . abbreviations ` br * ' indicates that a bright star is present ; ` att ' means attached to . a hierarchical indication is given for objects embedded in or superimposed on larger ones : ` in ' suggests a possible physical connection while ` sup ' suggests a projection . there are 56 pairs , 15 triplets , 2 quadruplets , 1 quintuplet and 1 sextuplet . the quintuplet is located in the star forming region ngc395 with dimensions 25 pc x 18 pc which in turn is embedded in the hii complex smc - dem126 with 66 pc x 35 pc according to angular sizes in bs95 . the sextuplet is mostly contained in the ob association h - a35 ( hodge 1985 ) with dimensions 57 pc x 47 pc . one quadruplet is located in h - a60 with 44 pc x 24 pc . note that dimensions ( columns 5 and 6 ) and position angle ( column 7 ) are from bica & dutra ( 2000 ) and were measured on plates . the dimensions are in general larger than the isophotal sizes in the subsequent analyses . l r r r r r r r l name & ra(2000 ) & dec(2000 ) & t & dmax & dmin & pa & separ . & comments & h : m : s & @xmath11 : : & & & & @xmath11 & & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) bs3 & 0:30:01 & -73:20:01 & ac & 33 & 33 & - & 31 & mt-1 h86 - 2 & 0:30:04 & -73:20:35 & ac & 39 & 36 & 80 & 31 & mt-1 bs4 & 0:30:07 & -73:20:58 & c & 33 & 27 & 70 & & mt-1 bs7 & 0:32:43 & -73:37:59 & c & 33 & 33 & - & 59 & mp-1 bs8 & 0:32:56 & -73:38:33 & ac & 36 & 36 & - & & mp-1 h86 - 22 & 0:34:53 & -73:02:08 & c & 18 & 18 & - & 42 & mp-2 bs9 & 0:35:02 & -73:02:14 & ac & 45 & 39 & 100 & & mp-2 hw9 & 0:36:25 & -73:00:05 & c & 45 & 45 & - & 52 & mp-3 hw10 & 0:36:31 & -72:59:13 & c & 57 & 57 & - & & mp-3 b9 & 0:37:13 & -72:57:53 & c & 27 & 27 & - & 68 & mp-4 in h - a1 h86 - 41 & 0:37:29 & -72:57:48 & c & 33 & 24 & 50 & & mp-4 h86 - 38 & 0:37:24 & -73:01:50 & a & 72 & 72 & - & 46 & mp-5 bs10 & 0:37:34 & -73:01:30 & c & 39 & 39 & - & & mp-5 bs14,smc_ogle165 & 0:39:12 & -73:14:46 & c & 33 & 33 & - & 44 & mp-6 smc_ogle5 & 0:39:22 & -73:15:28 & ca & 51 & 45 & 90 & & mp-6 hw12a & 0:39:26 & -73:22:59 & c & 33 & 27 & 10 & 36 & mp-7 h86 - 54 & 0:39:35 & -73:22:58 & c & 33 & 33 & - & & mp-7 b15 & 0:39:42 & -72:58:39 & ca & 36 & 27 & 100 & 30 & mp-8 h86 - 57 & 0:39:47 & -72:58:56 & ca & 30 & 30 & - & & mp-8 bs13 & 0:40:08 & -72:45:30 & ca & 54 & 45 & 40 & 48 & mp-9 bs248 & 0:40:15 & -72:46:02 & ac & 39 & 30 & 70 & & mp-9 ngc220,k18,l22,eso29sc3 & 0:40:31 & -73:24:10 & c & 72 & 72 & - & 88 & mt-2,in h - a3 & smc_ogle8 ngc222,k19,l24,eso29sc4 & 0:40:44 & -73:23:00 & c & 72 & 72 & - & 88 & mt-2,in h - a3 & smc_ogle9 b23,smc_ogle170 & 0:40:55 & -73:24:07 & c & 36 & 30 & 40 & & mt-2,in h - a3 b19 & 0:40:44 & -73:03:42 & c & 27 & 21 & 170 & 32 & mt-3 b20 & 0:40:49 & -73:04:09 & c & 18 & 18 & - & 32 & mt-3 h86 - 62,smc_ogle10 & 0:40:48 & -73:05:17 & c & 36 & 36 & - & & mt-3 ngc231,k20,l25,eso29sc5 & 0:41:06 & -73:21:07 & c & 108 & 108 & - & 36 & mp-10,in h - a3 & smc_ogle11 bs15 & 0:41:21 & -73:20:31 & a & 138 & 138 & - & & mp-10,in h - a3 b21,smc_ogle171 & 0:41:15 & -72:49:55 & c & 21 & 21 & - & 34 & mp-11 b22 & 0:41:21 & -72:49:32 & c & 21 & 21 & - & & mp-11 b29 & 0:42:11 & -73:43:51 & ca & 30 & 21 & 100 & 46 & mp-12 hw16,smc_ogle13 & 0:42:22 & -73:44:03 & cn & 36 & 36 & - & & mp-12,in smc - dem7 ngc241,k22w , l29w , & 0:43:33 & -73:26:25 & c & 57 & 57 & - & 23 & mp-13 & eso29sc6w , smc_ogle17 ngc242,k22e , l29e , & 0:43:38 & -73:26:37 & c & 45 & 45 & - & & mp-13 & eso29sc6e , bh1,smc_ogle18 b31,smc_ogle19,smc_ogle175 & 0:43:38 & -72:57:31 & c & 30 & 24 & 150 & & mt-4 bs20,smc_ogle20 & 0:43:38 & -72:58:48 & c & 27 & 27 & - & 30 & mt-4 h86 - 70,smc_ogle21 & 0:43:44 & -72:58:36 & c & 39 & 27 & 50 & 30 & mt-4 bs27,smc_ogle177 & 0:44:55 & -73:10:27 & c & 24 & 21 & 80 & 19 & mp-14,in h86 - 72 smc - n10,l61 - 60,smc - dem11 , & 0:44:56 & -73:10:11 & nc & 21 & 21 & - & & mp-14,in h86 - 72 & ma85 ngc248n , smc - n13b , l61 - 67n & 0:45:24 & -73:22:34 & na & 42 & 36 & 110 & 21 & mp-15 & smc - dem16n , eso29en8n , ma101,smc_ogle26n ngc248s , smc - n13a , l61 - 67s & 0:45:26 & -73:23:04 & nc & 42 & 33 & 150 & & mp-15 & smc - dem16s , eso29en8s , ma103,smc_ogle26s b39,smc_ogle27 & 0:45:26 & -73:28:53 & c & 33 & 33 & - & 15 & mp-16 bs30 & 0:45:30 & -73:29:06 & c & 24 & 24 & - & & mp-16 b36 & 0:45:44 & -72:50:35 & c & 42 & 36 & 140 & 10 & mp-17 smc_ogle31 & 0:45:51 & -72:50:25 & ca & 33 & 27 & 90 & & mp-17 , not b36 b38 & 0:45:54 & -72:36:08 & c & 18 & 18 & - & 38 & mp-18 h86 - 79 & 0:45:58 & -72:35:36 & c & 27 & 27 & - & & mp-18 h86 - 76,smc_ogle182 & 0:46:02 & -73:23:44 & c & 27 & 27 & - & & mt-5,in smc - dem21 h86 - 78n , smc_ogle33n & 0:46:12 & -73:23:27 & cn & 27 & 27 & - & 16 & mt-5,in smc - n16 h86 - 78s , smc_ogle33s & 0:46:12 & -73:23:39 & cn & 27 & 24 & 60 & 16 & mt-5,in smc - n16 l31,smc_ogle36 & 0:46:35 & -72:44:32 & c & 66 & 51 & - & & mt-6 h86 - 83,smc_ogle35 & 0:46:34 & -72:46:26 & c & 42 & 42 & - & 30 & mt-6 h86 - 84,smc_ogle185 & 0:46:34 & -72:45:56 & c & 24 & 24 & - & 30 & mt-6 h86 - 86,smc_ogle40 & 0:47:01 & -73:23:35 & c & 48 & 39 & 110 & 71 & mp-19,in h - a9 h86 - 87,smc_ogle187 & 0:47:06 & -73:22:17 & c & 48 & 42 & 90 & & mp-19,in h - a9 h86 - 95 & 0:47:37 & -73:00:51 & ca & 21 & 12 & 100 & 30 & mp-20 h86 - 96 & 0:47:44 & -73:00:46 & c & 18 & 18 & - & & mp-20 bs35,smc_ogle42 & 0:47:50 & -73:28:42 & c & 42 & 42 & - & 55 & mp-21 k25,l35,smc_ogle45 & 0:48:01 & -73:29:10 & c & 72 & 72 & - & & mp-21 ma205 & 0:48:07 & -73:14:49 & nc & 15 & 15 & - & 37 & mt-7 smc - n25,l61 - 106,smc - dem38 , & 0:48:09 & -73:14:19 & na & 51 & 51 & - & & mt-7 & ma208,smc_ogle189 smc - n26,l61 - 107,ma206 & 0:48:08 & -73:14:53 & nc & 27 & 27 & - & 37 & mt-7 h86 - 99,smc_ogle190 & 0:48:13 & -72:47:35 & ca & 39 & 39 & - & 27 & mp-22 h86 - 100,smc_ogle191 & 0:48:20 & -72:47:42 & ca & 45 & 45 & - & & mp-22 b50 & 0:49:02 & -73:21:44 & c & 33 & 33 & - & 66 & mt-8 bs41,smc_ogle194 & 0:49:06 & -73:21:10 & c & 33 & 33 & - & & mt-8 l39,smc_ogle54 & 0:49:18 & -73:22:20 & c & 42 & 33 & 170 & 66 & mt-8,in bs43 smc - n33,l61 - 138,ma297 & 0:49:29 & -73:26:33 & nc & 18 & 15 & 80 & 12 & mp-23,in smc - dem44 ma301 & 0:49:30 & -73:26:23 & nc & 21 & 18 & 80 & & mp-23,in smc - dem44 smc_ogle56 & 0:49:36 & -72:50:13 & ca & 48 & 36 & 100 & 70 & mt-9,in smc - dem46e h86 - 110 & 0:49:44 & -72:51:14 & ca & 45 & 33 & 160 & 70 & mt-9 h86 - 109,smc_ogle58 & 0:49:45 & -72:51:58 & c & 27 & 27 & - & & mt-9 ma317 & 0:49:42 & -73:10:37 & nc & 18 & 15 & 20 & 20 & mp-24 smc - n34,l61 - 142,smc - dem50 , & 0:49:46 & -73:10:25 & nc & 39 & 27 & 120 & & mp-24 & ma322 b53,smc_ogle197 & 0:50:04 & -73:23:04 & c & 57 & 57 & - & 80 & mp-25 b55,smc_ogle60 & 0:50:22 & -73:23:16 & c & 42 & 36 & 110 & & mp-25 smc_ogle199 & 0:50:15 & -73:03:15 & ca & 15 & 15 & - & 35 & mp-26 , in smc - dem51 bs45,smc_ogle59 & 0:50:16 & -73:02:00 & ca & 60 & 54 & 70 & & mp-26,in smc - dem51 h86 - 106w & 0:50:31 & -73:20:11 & c & 30 & 24 & 90 & 23 & mp-27,in smc - dem52 h86 - 106e & 0:50:37 & -73:20:11 & c & 33 & 27 & 80 & & mp-27,in smc - dem52 h86 - 115,smc_ogle63 & 0:50:37 & -73:03:28 & ac & 96 & 72 & 40 & 77 & mp-28,in smc - dem51 smc_ogle65 & 0:50:55 & -73:03:27 & c & 39 & 39 & - & & mp-28 bs46,smc_ogle200 & 0:50:39 & -72:58:44 & c & 30 & 27 & 60 & 47 & mp-29,in l40 h86 - 116,smc_ogle64 & 0:50:40 & -72:57:55 & c & 30 & 30 & - & & mp-29,in l40 bs48,smc_ogle201 & 0:50:42 & -73:23:49 & ac & 51 & 33 & 80 & 46 & mp-30,in smc - dem53 h86 - 108,ma401 & 0:50:53 & -73:24:22 & na & 60 & 60 & - & & mp-30,in smc - dem53 b59,l61 - 183,ma488 , & 0:51:44 & -72:50:25 & cn & 48 & 36 & 80 & 26 & mp-31 & smc_ogle73 [ tab : catalog ] l r r r r r r r l name & ra(2000 ) & dec(2000 ) & t & dmax & dmin & pa & separ . & comments & h : m : s & @xmath11 : : & & & & @xmath11 & & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) & ( 9 ) smc - n46,l61 - 184,smc - dem62 , & 0:51:47 & -72:50:47 & nc & 39 & 39 & - & & mp-31 & ma498 h86 - 120 & 0:51:46 & -73:28:01 & c & 21 & 21 & - & 43 & mt-10,in smc - dem70s bs53 & 0:51:49 & -73:28:38 & a & 39 & 33 & 150 & & mt-10,in smc - dem70s h86 - 122 & 0:51:58 & -73:27:41 & c & 27 & 27 & - & 43 & mt-10,in smc - dem70s bs56,smc_ogle77 & 0:52:13 & -73:00:12 & c & 42 & 33 & 90 & 56 & mp-32 h86 - 130,smc_ogle78 & 0:52:17 & -73:01:04 & c & 45 & 36 & 0 & & mp-32 b64,smc_ogle210 & 0:52:30 & -73:02:59 & c & 42 & 42 & - & 56 & mp-33,in h - a29 bs57,smc_ogle211 & 0:52:32 & -73:02:10 & c & 39 & 27 & 60 & & mp-33,in h - a29 h86 - 134w , smc_ogle212 & 0:52:45 & -72:59:24 & c & 30 & 30 & - & 19 & mt-11,in h - a30 b65,smc_ogle83 & 0:52:44 & -72:58:48 & c & 45 & 45 & - & & mt-11 h86 - 134e , smc_ogle213 & 0:52:48 & -72:59:22 & c & 30 & 27 & 0 & 19 & mt-11,in h - a30 bs61 & 0:52:43 & -73:01:45 & ca & 36 & 27 & 80 & 42 & mp-34,in h - a29 bs255 & 0:52:52 & -73:01:45 & c & 24 & 18 & 70 & & mp-34 bs63,smc_ogle84 & 0:52:47 & -73:24:25 & c & 30 & 24 & 150 & 19 & mt-12,in smc - dem73 b67,smc_ogle87 & 0:52:49 & -73:24:43 & c & 39 & 30 & 110 & & mt-12,in smc - dem73 ngc294,l47,eso29sc22 , & 0:53:06 & -73:22:49 & c & 102 & 102 & - & & mt-12 & smc_ogle90 h86 - 140,smc_ogle214 & 0:53:09 & -72:49:58 & c & 27 & 24 & 50 & 15&mp-35 h86 - 139 & 0:53:11 & -72:50:05 & c & 21 & 18 & 40 & & mp-35 bs67 & 0:53:32 & -73:21:03 & ac & 39 & 39 & - & 51 & mp-36 bs68,smc_ogle95 & 0:53:42 & -73:21:32 & ca & 54 & 45 & 130 & & mp-36 b72 & 0:53:26 & -72:40:57 & c & 72 & 72 & - & & m6,in h - a35 h86 - 143,smc_ogle93 & 0:53:31 & -72:40:04 & c & 48 & 48 & - & & m6,in h - a35 bs257 & 0:53:36 & -72:38:30 & ac & 48 & 39 & 0 & & m6 smc - n52a , l61 - 243 , & 0:53:40 & -72:39:35 & nc & 30 & 30 & - & 20&m6,in h - a35 & smc - dem77sw , ma696,smc_ogle94 smc - n52b , l61 - 244,b73 , & 0:53:42 & -72:39:15 & nc & 30 & 30 & - & 20 & m6,in h - a35 & smc - dem77ne , ma699,smc_ogle96 h86 - 148 & 0:53:55 & -72:40:08 & c & 30 & 30 & - & & m6,in h - a35 bs69,smc_ogle217 & 0:53:56 & -72:51:24 & ca & 36 & 24 & 45 & 75 & mp-37 bs72,smc_ogle97 & 0:54:11 & -72:51:54 & ca & 45 & 36 & 20 & & mp-37 b78 & 0:54:45 & -72:07:46 & c & 66 & 48 & 110 & 56&mp-38,in h - a37 l51,eso51sc7 & 0:54:54 & -72:06:46 & c & 60 & 45 & 170 & & mp-38,in h - a37 h86 - 159,smc_ogle102 & 0:55:12 & -72:41:00 & c & 30 & 24 & 130 & 45 & mp-39,in bs260 h86 - 160 & 0:55:21 & -72:40:10 & c & 24 & 24 & - & & mp-39,in bs260 bs81,smc_ogle223 & 0:56:26 & -72:29:45 & c & 36 & 33 & 0 & 43 & mp-40,in h - a40 h86 - 172,smc_ogle108 & 0:56:34 & -72:30:08 & c & 33 & 33 & - & & mp-40,in h - a40 h86 - 175,smc_ogle227 & 0:57:50 & -72:26:24 & c & 24 & 24 & - & 38 & mp-41 h86 - 179,smc_ogle112 & 0:57:57 & -72:26:42 & c & 24 & 24 & - & & mp-41 h86 - 177,smc_ogle226 & 0:57:50 & -72:30:29 & c & 45 & 45 & - & 46 & mp-42,in b - ob13 h86 - 176 & 0:57:53 & -72:29:48 & c & 36 & 30 & 90 & & mp-42 smc - n62,smc - dem93 & 0:57:56 & -72:39:26 & na & 72 & 72 & - & 75 & mt-13,in h - a42 smc - n63,l61 - 331,smc - dem94 , & 0:58:16 & -72:38:47 & na & 36 & 36 & - & & mt-13,in h - a42 & ma1065,smc_ogle113 smc - n64a , l61 - 335,smc - dem95 & 0:58:26 & -72:39:57 & nc & 48 & 39 & 70 & 75 & mt-13,in smc - n64 & h86 - 182,ma1071,smc_ogle114 bs269 & 0:58:19 & -72:13:10 & ca & 24 & 18 & 60 & 35 & mp-43 bs270 & 0:58:23 & -72:12:43 & ca & 33 & 30 & 130 & & mp-43 bs271 & 0:58:37 & -72:13:27 & nc & 39 & 30 & 30 & 35 & mp-44,in smc - dem98 bs272,smc_ogle229 & 0:58:38 & -72:14:04 & nc & 39 & 39 & - & & mp-44,in smc - dem98 bs93 & 0:59:36 & -71:44:13 & c & 24 & 21 & 10 & 17 & mt-14,in smc - dem105 b97 & 0:59:37 & -71:44:40 & c & 30 & 30 & - & & mt-14,in smc - dem105 bs273 & 0:59:40 & -71:44:34 & ac & 30 & 27 & 120 & 17 & mt-14,in smc - dem105 ic1611,k40,l61,eso29sc27 , & 0:59:48 & -72:20:02 & c & 90 & 90 & - & & m4 - 1 & smc_ogle118 h86 - 186,smc_ogle119 & 0:59:57 & -72:22:24 & c & 36 & 36 & - & 29 & m4 - 1,att smc - dem114 ic1612,k41,l62,eso29sc28 , & 1:00:01 & -72:22:08 & c & 72 & 48 & 20 & 29 & m4 - 1,att smc - dem114 & smc_ogle120 k42,l63,smc_ogle124 & 1:00:34 & -72:21:56 & c & 51 & 51 & - & & m4 - 1,att smc - dem114 b98sw & 1:00:21 & -73:52:51 & c & 36 & 27 & 40 & 33 & mp-45 b98ne & 1:00:28 & -73:52:32 & c & 33 & 27 & 130 & & mp-45 h86 - 189,smc_ogle123 & 1:00:33 & -72:14:23 & c & 24 & 24 & - & 68 & mp-46 h86 - 190,smc_ogle230 & 1:00:33 & -72:15:31 & c & 24 & 24 & - & & mp-46 h86 - 191,smc_ogle231 & 1:00:58 & -72:32:25 & c & 48 & 48 & - & 81 & mp-47,in ? smc - dem114 h86 - 194,smc_ogle232 & 1:01:14 & -72:33:03 & c & 51 & 51 & - & & mp-47,in ? smc - dem114 bs102 & 1:01:14 & -73:47:45 & c & 30 & 21 & 100 & 42 & mp-48 hw44 & 1:01:22 & -73:47:15 & c & 45 & 45 & - & & mp-48 b110 & 1:02:11 & -72:00:11 & c & 33 & 27 & 70 & 65 & mp-49 , br*in , in h - a49 b112 & 1:02:23 & -72:00:11 & c & 51 & 45 & 60 & & mp-49 , br*in , in h - a49 k45w , l69w & 1:02:45 & -73:44:19 & c & 33 & 27 & 40 & 47 & mp-50 k45e , l69w & 1:02:49 & -73:44:25 & c & 24 & 24 & - & & mp-50 ngc376,k49,l72,eso29sc29 , & 1:03:53 & -72:49:34 & c & 108 & 108 & - & 75 & mp-51 & smc_ogle139 bs114,smc_ogle235 & 1:03:59 & -72:48:18 & ac & 42 & 30 & 110 & & mp-51 smc_ogle138 & 1:03:53 & -72:06:11 & ca & 36 & 36 & - & 82 & mp-52,in h - a53 smc_ogle144,smc_ogle236 & 1:04:05 & -72:07:15 & ca & 36 & 36 & - & & mp-52 bs122 & 1:04:18 & -73:10:21 & c & 21 & 21 & - & 26 & mp-53 b119 & 1:04:19 & -73:09:54 & c & 36 & 36 & - & & mp-53 smc - n78a , l61 - 438,ma1512 & 1:05:04 & -71:59:01 & nc & 24 & 21 & 140 & 23 & m5,in ngc395 smc - n78b , l61 - 439 , & 1:05:04 & -71:59:25 & nc & 24 & 18 & 100 & & m5,in ngc395 & ma1508/1514,smc_ogle145 ma1520,smc_ogle147 & 1:05:08 & -71:59:45 & nc & 30 & 27 & 130 & 23 & m5,in ngc395 smc - n78d , smc - dem127 & 1:05:11 & -71:58:28 & na & 54 & 54 & - & & m5,att smc - n78 smc_ogle146 & 1:05:13 & -71 59 42 & na & 27 & 27 & - & & m5,in ngc395 bs130 & 1:05:56 & -72:04:11 & a & 57 & 36 & 140 & 30&mp-54,att smc - n78 bs132 & 1:06:01 & -72:03:37 & ca & 42 & 33 & 150 & & mp-54,att smc - n78 bs133 & 1:06:23 & -71:55:12 & a & 39 & 30 & 30 & & mt-15,in bs134 bs135 & 1:06:40 & -71:55:13 & ca & 39 & 39 & - & 47 & mt-15,in bs134 bs136 & 1:06:48 & -71:54:55 & ca & 42 & 42 & - & 47 & mt-15,in smc - dem132 b134 & 1:09:01 & -73:12:24 & ca & 48 & 33 & 80 & & m4 - 2,in h - a60 bs142 & 1:09:07 & -73:12:01 & c & 24 & 21 & 70 & 31 & m4 - 2,in h - a60 ic1644,smc - n81,l61 - 481 , & 1:09:12 & -73:11:42 & nc & 48 & 39 & 40 & 31 & m4 - 2,in h - a60 & eso29en35,ma1688/1687 b135 & 1:09:19 & -73:11:15 & c & 33 & 24 & 60 & & m4 - 2,in h - a60 ngc422,k62,l87,eso51sc22 & 1:09:25 & -71:46:00 & c & 60 & 60 & - & 64 & mp-55 ic1641,hw62,eso51sc21 & 1:09:39 & -71:46:07 & c & 45 & 39 & 40 & & mp-55 hw71nw & 1:15:30 & -72:22:36 & c & 18 & 18 & - & 17 & mp-56 hw71se & 1:15:33 & -72:22:50 & c & 39 & 33 & 170 & & mp-56 for isophotal analysis purposes we extracted digitized images of pairs and multiplets from the dss . the plates are from the serc southern sky survey and include iiia - j long ( 3600s ) , v band medium ( 1200s ) and v band short ( 300s ) exposures . the pds pixel values correspond to photographic density measures from the original plates , and are not calibrated . the images were processed with the iraf package at the instituto de fsica - ufrgs , applying a 2-d gaussian filter to smooth out individual stars , producing isodensity maps . we show the isophotal atlas of smc star cluster pairs and multiplets in figs . [ fig : atlas1 ] throughout [ fig : atlas7 ] , where morphological evidence of interactions can be searched for . in each panel we point members and provide designations . cluster pairs and multiplets with evidence of physical interaction are marked in column 6 of table [ tab : interact ] . their isophotal maps show features such as isophotal distortions , common envelope , isophotal twistings , etc . we classified these isophotal maps according to the following criteria : ( i ) isophotes of the members are detached ( * d * ) , but showing isophotal distortions ; ( ii ) isophotes of the members are connected by a `` bridge '' ( * b * ) ; ( iii ) the members are embedded in the same isophotal envelope ( * e * ) . these classification criteria are an important tool for a selection of interacting cluster pair candidates , since such isophotal morphologies are predominant in n - body model encounters ( rodrigues et al . 1994 , odb98 ) . the classifications above were made for objects not disturbed by gas emission . it is worth noting that both bridges and envelopes imply a common outer isophotal , however the former includes a dimension which is considerably smaller than the smaller cluster diameter . the pair b59/smc - n46 would be an envelope limiting case ( fig . [ fig : atlas4 ] ) . as examples , the pair h86 - 186/ic1612 ( fig . [ fig : atlas5 ] ) is embedded in the same isophotal envelope and it is classified as * e*. the pair h86 - 140/h86 - 139 ( fig . [ fig : atlas4 ] ) shows a bridge linking its members and is classificated as * b*. in fig . [ fig : atlas6 ] the pair ngc376/bs114 shows isophotal distortions , however the outer isophotes of the components are not connected and it is thus classified as * d*. we detected relevant isophotal features for about 25% of the sample distributed as follows : 6 candidate cluster systems with common envelope , 7 with bridge and 5 detached cases . high isophotal densities such as those observed for b78/l51 ( fig . [ fig : atlas5 ] ) are related to the high surface brightness which often occurs in blue clusters . it is important to note that non relevant isophotal features can appear in the maps caused by bright stars or background fluctuations . by means of previous inspections in the original images we verified the non relevant objects . for example , the compact object in the lower right corner of the b39/bs30 isophotal map ( fig . [ fig : atlas2 ] ) is a bright star , however the compact one to the right of the main cluster can be a star or a compact cluster , as seen in the original form . higher resolution images would be necessary to check the presence of a third cluster . l r r r r r l name & age & e(b - v ) & age@xmath12 & e(b - v)@xmath12 & morphologies & commments & myr & & myr & & & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) bs3 & & & & & b & mt-1 outside ogleii h86 - 2 & & & & & b & mt-1 outside ogleii bs4 & & & & & & mt-1 outside ogleii h86 - 38 & 251@xmath1325 & 0.10&&&&mp-5 bs10 & @xmath14560 & 0.10&&&&mp-5 b15 & @xmath141000 & 0.07&&&&mp-8 h86 - 57 & @xmath141000 & 0.07&&&&mp-8 ngc220,smc_ogle8 & 65@xmath1313 & 0.07&100@xmath1323&0.07&d & mt-2 ngc222,smc_ogle9 & 70@xmath137 & 0.07&100@xmath1323&0.07&d & mt-2 b23,smc_ogle170 & @xmath15100 & 0.07&&&d & mt-2 ngc231,smc_ogle11 & 65@xmath138 & 0.12&79@xmath1318&0.08&&mp-10 bs15 & 70@xmath137 & 0.10&&&&mp-10 b29 & 150@xmath1370 & 0.10 & & & & mp-12 hw16,smc_ogle13 & @xmath1520 & 0.07&20@xmath1315&0.05&&mp-12 ngc241,smc_ogle17 & 55@xmath135 & 0.12&79@xmath1318&0.10&b&mp-13 ngc242,smc_ogle18 & 65@xmath1310 & 0.10&79@xmath1318&0.10&b&mp-13 b31,smc_ogle19,smc_ogle175 & 280@xmath1330 & 0.10&400@xmath1392&0.08&&mt-4 bs20,smc_ogle20 & 450@xmath1350 & 0.08&400@xmath1392&0.08&&mt-4 h86 - 70,smc_ogle21 & 450@xmath1350 & 0.07&&&&mt-4 bs27,smc_ogle177 & @xmath1525 & 0.18&79@xmath1337&0.08&&mp-14 smc - n10 & hii region & & & & & mp-14 ngc248n , smc_ogle26n & hii region & & & & & mp-15 ngc248s , smc_ogle26s & hii region & & & & & mp-15 b39,smc_ogle27 & 450@xmath1350 & 0.07&&&b&mp-16 bs30 & 450@xmath1350 & 0.07&&&b&mp-16 b36 & 315@xmath1350 & 0.03&&&&mp-17 smc_ogle31 & 450@xmath1350 & 0.03&&&&mp-17 h86 - 76,smc_ogle182 & 200@xmath1320 & 0.14&&&&mt-5 h86 - 78n , smc_ogle33n & @xmath1525 & 0.14&16@xmath139&0.15&&mt-5 h86 - 78s , smc_ogle33s & @xmath1530 & 0.14&16@xmath139&0.15&&mt-5 l31,smc_ogle36 & 250@xmath1350 & 0.10&&&&mt-6 h86 - 83,smc_ogle35 & 180@xmath1350 & 0.10&&&&mt-6 h86 - 84,smc_ogle185 & 250@xmath1350 & 0.10&&&&mt-6 h86 - 86,smc_ogle40 & 350@xmath1350 & 0.03&&&&mp-19 h86 - 87,smc_ogle187 & 140@xmath1320 & 0.07&158@xmath1336&0.04&&mp-19 bs35,smc_ogle42 & 400@xmath13100 & 0.03 & & & & mp-21 k25,smc_ogle45 & 250@xmath1350 & 0.07&250@xmath1358&0.07&&mp-21 ma205 & hii region & & & & b&mt-7 smc - n25,smc_ogle189 & hii region & & & & b&mt-7 smc - n26 & hii region & & & & & mt-7 h86 - 99,smc_ogle190 & 225@xmath1350 & 0.14&&&d&mp-22 h86 - 100,smc_ogle191 & 200@xmath1350 & 0.14 & & & d&mp-22 b50 & @xmath1530 & 0.10&&&&mt-8 bs41,smc_ogle194 & 70@xmath1330 & 0.10&79@xmath1318 & 0.07 & & mt-8 l39,smc_ogle54 & 80@xmath1320 & 0.10 & 100@xmath1323 & 0.10 & & mt-8 smc - n33 & hii region & & & & e&mp-23 ma301 & hii region & & & & e&mp-23 smc_ogle56 & 115@xmath1315 & 0.18&&&&mt-9 h86 - 110 & @xmath1520 & 0.18&&&&mt-9 h86 - 109,smc_ogle58 & 180@xmath1320 & 0.18&200@xmath1345&0.17&&mt-9 ma317 & hii region & & & & b&mp-24 smc - n34 & hii region & & & & b&mp-24 b53,smc_ogle197 & 200@xmath1350 & 0.07&250@xmath1355&0.08&&mp-25 b55,smc_ogle60 & 160@xmath1330 & 0.07&250@xmath1355&0.08&&mp-25 smc_ogle199 & & & & & & mp-26 bs45,smc_ogle59 & 65@xmath1330 & 0.10 & 63@xmath1314 & 0.10 & & mp-26 h86 - 106w & 200@xmath1350 & 0.10&&&e&mp-27 h86 - 106e & @xmath1530/200@xmath1350@xmath16 & 0.12/0.10&&&e&mp-27 h86 - 115,smc_ogle63 & & & & & & mp-28 smc_ogle65 & 200@xmath1320 & 0.10&&&&mp-28 bs46,smc_ogle200 & 80@xmath1310 & 0.07&100@xmath1323&0.06&b&mp-29 h86 - 116,smc_ogle64 & 125@xmath1315 & 0.10&126@xmath1329&0.10&b&mp-29 bs48,smc_ogle201 & 160@xmath1320 & 0.10&&&&mp-30 h86 - 108,ma401 & @xmath1530 & 0.12&&&&mp-30 b59,smc_ogle73 & 100@xmath1315 & 0.14&158@xmath1375&0.11&e&mp-31 smc - n46 & hii region & & & & e&mp-31 [ tab : interact ] l r r r r r l name & age & e(b - v ) & age@xmath12 & e(b - v)@xmath12 & morphologies & comments & myr & & myr & & & ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) bs56,smc_ogle77 & 140@xmath1320 & 0.07&79@xmath1338&0.08&d & mp-32 h86 - 130,smc_ogle78 & 65@xmath138 & 0.12&79@xmath1318&0.08&d & mp-32 b64,smc_ogle210 & @xmath1510 & 0.12&158@xmath1375&0.07&&mp-33 bs57,smc_ogle211 & @xmath1510 & 0.12&&&&mp-33 h86 - 134w , smc_ogle212 & & & & & & mt-11 b65,smc_ogle83 & 65@xmath1310 & 0.10&63@xmath1330&0.09&&mt-11 h86 - 134e , smc_ogle213 & & & & & & mt-11 bs61 & 250@xmath1350 & 0.03&&&&mp-34 bs255 & & & & & & mp-34 bs63,smc_ogle84 & 450@xmath1350 & 0.10&&&d&mt-12 b67,smc_ogle87 & 450@xmath1350 & 0.10&500@xmath13115&0.10&d&mt-12 ngc294,smc_ogle90 & 300@xmath1350 & 0.10&316@xmath1373&0.11&d & mt-12 h86 - 140,smc_ogle214 & 55@xmath1325 & 0.10&&&b&mp-35 h86 - 139 & @xmath1530 & 0.10&&&b&mp-35 bs67 & @xmath14300 & 0.07&&&&mp-36 bs68,smc_ogle95 & 500@xmath13100 & 0.07&&&&mp-36 b72 & 80@xmath1310 & 0.10 & & & & m6 h86 - 143,smc_ogle93 & 200@xmath1325 & 0.10&&&&m6 bs257 & 50@xmath1330@xmath16 & 0.10&&&&m6 smc - n52a , smc_ogle94 & hii region & & & & & m6 smc - n52b , smc_ogle96 & hii region & & & & & m6 h86 - 148 & 400@xmath13100 & 0.10&&&&m6 h86 - 159,smc_ogle102 & 500@xmath1350 & 0.10&&&&mp-39 h86 - 160 & & & & & & mp-39 bs81,smc_ogle223 & 80@xmath1320/250@xmath1350@xmath16 & 0.10&&&&mp-40 h86 - 172,smc_ogle108 & 280@xmath1330 & 0.07&&&&mp-40 h86 - 175,smc_ogle227 & 30@xmath1310 & 0.10&&&&mp-41 h86 - 179,smc_ogle112 & @xmath1530 & 0.10 & 32@xmath1323 & 0.11 & & mp-41 h86 - 177,smc_ogle226 & @xmath1530 & 0.10&&&d&mp-42 h86 - 176 & @xmath1530 & 0.10&&&d&mp-42 smc - n62 & hii region & & & & & mt-13 smc - n63,smc_ogle113 & hii region & & & & & mt-13 smc - n64a , smc_ogle114 & hii region & & & & & mt-13 bs269 & & & & & & mp-43 bs270 & @xmath1530 & 0.10&&&&mp-43 bs271 & @xmath1530 & 0.12&&&&mp-44 bs272,smc_ogle229 & @xmath1530 & 0.08&79@xmath1318&0.05&&mp-44 ic1611,smc_ogle118 & 100@xmath1320 & 0.07 & 160@xmath1337 & 0.08 & & m4 - 1 h86 - 186,smc_ogle119 & 180@xmath1320 & 0.07 & & & e&m4 - 1 ic1612,smc_ogle120 & 100@xmath1350 & 0.10 & 50@xmath1324 & 0.07 & e&m4 - 1 k42,smc_ogle124 & 20@xmath1310 & 0.10 & 40@xmath139 & 0.06 & & m4 - 1 h86 - 189,smc_ogle123 & 570@xmath1370 & 0.07 & & & & mp-46 h86 - 190,smc_ogle230 & 125@xmath1313 & 0.07 & 32@xmath1323&0.08&&mp-46 h86 - 191,smc_ogle231 & 200@xmath1325 & 0.07&&&&mp-47 h86 - 194,smc_ogle232 & 200@xmath1325 & 0.07&&&&mp-47 ngc376,smc_ogle139 & 20@xmath132 & 0.10&32@xmath137&0.07&e&mp-51 bs114,smc_ogle235 & 250@xmath1325 & 0.04&&&e&mp-51 smc_ogle138 & @xmath1530 & 0.04 & 25@xmath1320 & 0.04 & & mp-52 smc_ogle144,smc_ogle236 & @xmath1530 & 0.07 & 40@xmath1319 & 0.05 & & mp-52 smc - n78a & hii region & & & & & m5 smc - n78b , smc_ogle145 & hii region & & 79@xmath1337 & 0.07 & & m5 ma1520,smc_ogle147 & hii region & & 12@xmath139 & 0.06 & & m5 smc - n78d & hii region & & & & & m5 smc_ogle146 & hii region & & 20@xmath1315 & 0.06 & & m5 bs130 & 80@xmath1310 & 0.10&&&&mp-54 bs132 & @xmath1530/200@xmath1325@xmath17 & 0.10 & & & & mp-54 hw71nw & & & & & e&mp-56 outside ogleii hw71se & & & & & e&mp-56 outside ogleii notes to table [ tab : interact ] : 1- age depends on membership of bright stars ; 2- upper main sequence is possibly underpopulated . udalski et al . ( 1998 ) provided a bvi photometric survey of the smc central @xmath3 2.4 square degrees with results for about 2.2 million stars . the data were collected during the ogle - ii microlensing search project ( udalski et al . 1997 ) . pietrzyski & udalski ( 1999b ) studied colour - magnitude diagrams ( cmds ) extracted from the bvi database for 93 smc star clusters , taking into account neighbouring field extractions for comparisons . they estimated reddening values for each cluster using : ( i ) the mean i band magnitude of red clump stars in the cluster neighbourhood ; ( ii ) the assumed extinction - free magnitude of the smc s red clump stars , i = 18.34 mag ( udalski 1998 ) . they derived ages using bertelli et al.s ( 1994 ) isochrones , adopting as a rule for the smc a metallicity z = 0.004 , and an absolute distance modulus ( m - m ) = 18.65 . considering the ogle - ii angular coverage we conclude that out of the present 176 objects 133 are therein included . we obtained from the database v and i band cmds for these objects , using their coordinates and diameters to define a box extraction . guided by dss images of each system we selected representative field regions to extract stars and construct cmds for comparison purposes . ages and reddenings for the clusters were estimated by fitting the padova isochrones . we adopted bertelli et al.s ( 1994 ) isochrones rather than the new isochrones by girardi et al . ( 2000 ) , since the previous set includes younger ages . the isochrone age grid suitable for 4 , 5 , 6.3 , 7.9 , 10 , 12 , 16 , 20 , 25 , 32 , 40 , 50 , 63 , 79 , 100 , 126 , 160 , 199 , 251 , 316 , 400 , 500 , 630 , 790 and 1000 myr . the isochrone age range and resolution allow one to estimate errors taking into account the stellar statistics in the cluster and field cmds . we also assumed a smc metallicity of z = 0.004 , a foreground galactic reddening of e(b - v)@xmath18 = 0.03 in the smc direction and e(v - i)/e(b - v ) = 1.31 . table [ tab : interact ] shows the ages , in column 2 and reddening values ( foreground plus smc internal ) in column 3 determined in the present work via isochrone fitting . ages and reddening values derived by pietrzyski & udalski ( 1999b ) are in column 4 and 5 respectively , when available for comparisons . in our sample we could estimate cmd ages for 91 objects , 40 of them in common with pietrzyski & udalski ( 1999b ) . we conclude that despite the reddening method and distance modulus differences between the two studies , there is good overall agreement for the age determinations . we also included ages for 21 objects embedded in hii regions , classified as na and nc ( see column 4 of table [ tab : catalog ] , and also bs95 ) . we assumed for them an age of 3 myr . we overlap two age histograms in fig . [ fig : age ] . the dotted line histogram shows objects from table [ tab : interact ] with ages via cmds . there are three peaks , at about 80 , 220 and 450 myr . the first two peaks appear to be present in grebel et al.s ( 2000 ) sample for 200 smc clusters based on ogle - ii data , reported at t @xmath3 100 and 200 myr respectively . they suggested them as enhanced star formation epochs . considering pietrzyski & udalski s ( 1999b ) sample with cmd ages for 93 smc clusters , no peak is seen at t = 80 - 100 myr , having their histogram a maximum at the youngest bin . in their histogram there occurs a local maximum at t @xmath3 250 - 300 myr , but the statistics is low . the third peak in the present study has no counterpart in both previous analyses , and it is probably an artifact from the low isochrone age step resolution in that range . the solid line histogram in fig . [ fig : age ] shows additionally the embedded objects in hii regions for which we assumed an age of 3 myr , and the objects from table [ tab : interact ] with lower or upper limits by assuming them as the ages themselves . we can observe again a three peak distribution but the first peak is now shifted to the youngest bin similarly to pietrzyski & udalski s ( 1999b ) histogram . this suggests that only the 200 myr peak is relevant , being related to the smc / lmc last encounter ( gardiner et al . 1994 , grebel et al.s 2000 ) . the maximum seen in the youngest bin is possibly related to the cluster formation / destruction rates . since the present sample deals with pairs and multiplets a fast destruction rate might be caused by the internal dynamical evolution in each cluster complex , caused by merger and/or other effects . approximately 55% of cluster pairs and multiplets in table [ tab : interact ] present similar ages between their members indicating that they are coeval . this suggests that most of the pairs and multiplets had a common origin , possibly from the same molecular complex . note that about 60% of the pairs and multiplets embedded in ob associations ( h - a , hodge 1985 ) , as indicated in column 9 of table [ tab : catalog ] , have comparable ages between their members . this could be an explanation for the origin of cluster systems . considering triplet and multiplet members , we found that about 70% of them are younger than 100 myr . these results suggest a possible binary cluster formation scenario : clusters can be born in multiplet systems and coalesce by mergers and tidal disruptions forming binary clusters in a timescale of @xmath3 100 myr . this time is in agreement with dynamical times required for an interacting pair to merge into a single cluster ( odb98 , de oliveira et al . 2000 ) . the pairs with a bridge in the isophotal maps have comparable ages for their components ( table [ tab : interact ] ) . as examples , the pairs ngc241/ngc242 and b39/bs30 in fig . [ fig : atlas2 ] show a bridge linking their members which could be interpreted as a sign of interaction ( see the similarity with the n - body simulation model in fig . 11 of odb98 ) . a typical timescale for the bridge phenomenon is @xmath3 30 myr , as deduced from a series of n - body simulations related to bridge formation and evolution ( odb98 ) . another interesting isophotal feature is related to the cluster triplet ngc220/ngc222/b23 which shows distortions for the small cluster in a direction almost perpendicular to the line connecting itself to the large components ngc220 and ngc222 . this configuration and morphologies are compatible with a fast hyperbolic encounter with small impact parameter ( e.g. fig . 12 of odb98 ) . [ fig : distr_smc ] shows the angular distribution of pairs and multiplets together with smc hi column density isophotes from mathewson & ford ( 1984 ) . it can be seen that most of the objects are concentrated in the smc main body , close to the higher concentration of hi , so it is not unexpected that in general they result young ( sect . 4 ) . the objects appear to be distributed along an axis . such distribution is present in the overall smc cluster sample and there is growing evidence that it is related to a nearly edge - on disk containing the bulk of the young stellar population in the smc ( bica _ et al . _ 1999 , westerlund 1990 ) . a nearly edge - on disk in a low internal reddening galaxy like the smc would imply an increase of projection effects as compared to a simulation such as that carried out by bhatia & hatzidimitriou ( 1988 ) for the nearly face - on lmc disk where the physical pairs would be about 50% . consequently the fraction of physical pairs in the smc would be lower . the present approach including morphological evidence of interaction is an attempt to constrain this aspect . indeed the fraction with isophotal distortions is only 25% ( sect . projection effects can be responsible for the age spread in some multiplets . for example the sextuplet ( table [ tab : interact ] ) has component ages 80 myr ( b72 ) , 200 myr ( h86 - 143 ) , 50 myr ( bs257 ) , 3 myr ( smc - n52a and smc - n52b ) and 400 myr ( h86 - 148 ) . possibly only the 3 or 4 younger components could be related to ob - association h - a35 , the remaining objects would be captures or projection effects . this age spread is also present in ic1611 s quadruplet and in some triplets . on the other hand the quintuplet in the star forming complex ngc395 has all its members with the same age ( 3 myr ) thus forming a physical system . in fig . [ fig : diam_smc ] we show the distribution of the diameter ratio between members for all pairs in the sample . the diameter ratio is mostly in the range @xmath19 , with a peak at @xmath2 indicating that the majority of pair members have a comparable size . this effect was also observed in the lmc ( bhatia et al . 1991 ) . [ fig : separ_smc ] shows the distribution of the centre - to - centre angular separation between pair members . the separation range is @xmath3 @xmath20 arcsec ( @xmath33 - 22 pc ) with a pronounced peak at @xmath3 @xmath21 arcsec ( 13 pc ) . a similar peak was also observed by bhatia et al . ( 1991 ) and de oliveira ( 1996 ) who found a bimodal distribution for the lmc pairs with peaks at @xmath3 5 and 13 pc . the observed upper limit of the projected centre - to - centre linear separation @xmath3 23 pc ( 80 arcsec ) is comparable to bhatia & hatzidimitriou s ( 1988 ) separation criterion for pairs in the lmc ( 18.7 pc ) . the frequent separation value around 13 pc may reflect a preferred survival distance for the systems , combined to projection effects . we presented an isophotal atlas for 75 star cluster pairs and multiplets in the small magellanic cloud , comprising 176 objects . it was possible to derive ages from colour - magnitude diagrams using the ogle - ii photometric survey for 91 objects . in addition we included in the analysis ages for 21 embedded objects in hii regions . the age distribution has a maximum in the youngest bin with a profile related to the cluster formation / destruction rates , in particular cluster multiple systems can have a fast destruction rate caused by their internal dynamical evolution . there is a second peak around 220 myr which is probably related to the smc / lmc last encounter . we find that 55% of the pairs and multiplets in the sample are in general coeval , indicating that captures are a rare phenomenon . most of the cluster multiplets occur in ob stellar associations and/or hii region complexes which indicates a common origin and suggests that multiplets coalesce into pairs or single clusters in a short time scale ( @xmath3 100 myr ) . the majority of the cluster members have comparable sizes , with a diameter ratio ranging mostly between @xmath19 . the projected separation distribution between the members of a pair has a pronounced peak at @xmath3 13 pc . these observational results are important constraints to theoretical models of star cluster pair encounters and could be related with the formation process and subsequent dynamical evolution of cluster systems . the angular distribution of cluster pairs and multiplets shows that most of the objects are located in the smc main body . the overall smc cluster sample presents a similar distribution and there is evidence that it is related to a nearly edge - on disk in the smc . considering this , it is expected an increase of projection effects as compared to estimates for the lmc disk where physical pairs would be about 50% ( bhatia & hatzidimitriou 1988 ) . the atlas shows that around 25% of the isophote maps present relevant structures like bridges , common envelopes or detached distorted isophotes . n - body simulations have indicated that these structures arise from interactions between the members of the cluster systems . indeed cluster pairs as ngc241/ngc242 and b39/bs30 show in their isophotal maps bridges linking their members and have comparable ages . we conclude that multiplicity may have an important role in the early dynamical evolution of star clusters in general , and signatures of that may survive in the long term structure of large single clusters ( de oliveira et al . 2000 ) . grebel e.k . , zaritsky d. , harris j. , 2000 . in _ news views of the magellanic clouds . in : proceedings of the iau symp . 190 _ , eds . you - hua chu , nicholas b. suntzeef , james e. hesser , david a.bohlender ( eds ) . , pasp conf . series , p. 405 .

we report the discovery of one extremely metal - poor ( emp ;$] ) and one ultra metal - poor ( ump ;$] ) star selected from the sdss / segue survey . these stars were identified as emp candidates based on their medium - resolution ( ) spectra , and were followed - up with high - resolution ( ) spectroscopy with the magellan - clay telescope . their derived chemical abundances exhibit good agreement with those of stars with similar metallicities . we also provide new insights on the formation of the ump stars , based on comparison with a new set of theoretical models of supernovae nucleosynthesis . the models were matched with ump stars found in the literature , together with one of the program stars ( ) , with}}}=-4.34 $ ] . from fitting their abundances , we find that the supernovae progenitors , for stars where carbon and nitrogen are measured , had masses ranging from to , and explosion energies from to . these results are highly sensitive to the carbon and nitrogen abundance determinations , which is one of the main drivers for future high - resolution follow - up of ump candidates . in addition , we are able to reproduce the different cno abundance patterns found in ump stars with a single progenitor type , by varying its mass and explosion energy .

1507.03656

the most metal - poor stars in the galactic halo carry important information about the formation and early evolution of the conditions in the early universe , as well as in the assembly of the milky way . in particular , ultra metal - poor ( ump ; @xmath0$]$ ] = @xmath11 , where @xmath12 is the number density of atoms of a given element in the star ( @xmath13 ) and the sun ( @xmath14 ) , respectively . ] @xmath15 , e.g. , * ? ? ? * ; * ? ? ? * ) stars are believed to be formed by gas clouds polluted by the chemical yields of the very first ( population iii ) stars formed in the universe @xcite . even though this scenario for the origin of ump stars is qualitatively widely accepted , there are still many open questions , such as the range of masses and specific characteristics of the population of first stars , and how many progenitors each ump star might have had @xcite . there are different scenarios for the progenitor populations that have been proposed to provide the necessary ingredients to trigger the formation of ump stars : ( i ) pair - instability supernovae from very massive stars ( e.g. , * ? ? ? * ) , however , the predicted abundance pattern from such supernova has not been uniquely observed ; ( ii ) relatively normal massive stars , however , with reduced mixing relative to their modern counterparts ( e.g. , * ? ? ? * ) ; ( iii ) fast - rotating massive stars @xcite and ; ( iv ) `` faint '' supernovae @xcite . analyses of the abundances of ump stars , however , are required to ultimately establish the nature and shape of the initial mass function ( imf ) of these population iii stars . the goal is to constraint the progenitor properties including their masses ( see discussion in * ? ? ? presently , however , there are about @xmath5 ump stars with high - resolution spectra available in the literature , and only five stars known to be hyper metal - poor ( hmp ; @xmath6}}}\leq-5.0 $ ] , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? since the completion of the sloan extension for galactic understanding and exploration ( segue-1 ; * ? ? ? * ) and segue-2 @xcite , the numbers of identified very / extremely metal - poor ( vmp / emp , with @xmath16}<-2.0 $ ] and @xmath17 , respectively ) stars have increased by over an order of magnitude , when compared to previous efforts , such as the hk @xcite and hamburg / eso @xcite surveys . the segue-1 and segue-2 campaigns , sub - surveys of the sloan digital sky survey ( sdss ; * ? ? ? * ) , accomplished this with medium - resolution ( @xmath3 ) spectroscopy . currently , there are several tens of thousands of vmp ( and on the order of @xmath18 emp ) candidates identified by these surveys , but just a small fraction have been studied with high - resolution spectroscopy . recent efforts to increase these numbers using sdss / segue candidates include studies by @xcite , @xcite , @xcite , and @xcite , and have already resulted in the discovery of sdss j1029@xmath191729 , with @xmath6}}}=-4.99 $ ] @xcite , sdss j1742@xmath192531 , with @xmath6}}}=-4.80 $ ] and sdss j1035@xmath190641 , with @xmath6}}}<-5.07 $ ] @xcite , and sdss j1313@xmath190019 , with @xmath6}}}=-5.00 $ ] @xcite . in this paper , we present results from a new selection effort for emp stars from the data base of medium - resolution sdss / segue spectra . we report the discovery of another ump star , sdss [email protected] ( hereafter , @xmath6}}}=-4.34\pm 0.05 $ ] ) and also an emp star , sdss [email protected] ( hereafter , @xmath6}}}=-3.64\pm 0.05 $ ] ) , selected from medium - resolution sdss / segue spectra . we have carried out a detailed chemical - abundance analysis , as well as a comparison with data from the literature and theoretical models for population iii stars , in order to gain insights on the progenitor population(s ) of ump stars . this paper is outlined as follows : section [ secobs ] describes the medium - resolution spectroscopy target selection and high - resolution follow - up observations , followed by the determinations of the stellar parameters and chemical abundances in section [ secatm ] . section [ seccomp ] shows a comparison between ( i ) the abundances of the program stars and other literature data , and ( ii ) the ump stars in the literature and a new set of theoretical models for pop iii stars . our conclusions are provided in section [ final ] . our targets were selected from the segue-2 database , to avoid possible overlap with other high - resolution surveys of metal - poor stars based on sdss / segue data . we applied restrictions to the magnitude ( @xmath20 ) and metallicity ( @xmath6}}}<-3.0 $ ] ) . estimates of the stellar atmospheric parameters from medium - resolution sdss / segue spectra were obtained using the segue stellar parameter pipeline ( sspp - see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * for a detailed description ) . these restrictions cut down the entire sdss / segue sample to ten potential ump stars . although there are many more possible emp star candidates selected from the sdss / segue-2 , we imposed the magnitude constraint in order to obtain candidates which can be observed with high signal - to - noise , sufficient for detailed high - resolution spectral analysis , at reasonable exposure times . after visual inspection of the ten candidate spectra ( and checking their atmospheric parameters from the sspp ) , two objects were followed - up : and , which were the lowest metallicity candidates . table [ candlist ] lists basic information on the program stars . table [ obstable ] lists the sspp - derived @xmath21 , @xmath22 , and @xmath0 $ ] , used as first - pass estimates when determining the parameters for the high - resolution analysis , presented in section [ secatm ] . figure [ spec_med ] shows the sdss / segue-2 spectra for the observed stars , in comparison with the star ( @xmath6}}}=-4.99 $ ] - * ? ? ? * ) , the first confirmed ump star from sdss / segue . the red dashed lines show the sspp spectral templates , used for matching the observed spectra and to determine the atmospheric parameters . the adopted sspp metallicities are also shown . it is possible to note a slight mismatch between the observed and synthetic spectra , in particular for h@xmath23 . this results in an overestimated temperature , which translates into a higher @xmath6}}}$ ] value for a given strength of the ca k line . by comparing the sspp parameters with the high - resolution determinations ( see section [ secatm ] for details ) , the sspp temperatures are overestimated by @xmath24 , leading to higher estimated @xmath6}}}$ ] . high - resolution data were obtained during the 2013a semester , using the magellan inamori kyocera echelle ( mike * ? ? ? * ) spectrograph on the magellan - clay telescope at las campanas observatory . the observing setup included a @xmath25 slit with @xmath26 on - chip binning , which yielded a resolving power of @xmath27 in the blue spectral range and @xmath28 in the red spectral range . the s / n at @xmath29 is @xmath30 per pixel ( using integration times of @xmath31 and @xmath32 minutes for and , respectively ) . mike spectra have nearly full optical wavelength coverage over @xmath33 . table [ candlist ] lists the details of the high - resolution observations for the program stars . the data were reduced using a data reduction pipeline developed for mike spectra , initially described by @xcite . figure [ highres ] shows regions of the mike spectra for the program stars in comparison with the emp star he 1300@xmath190157 , which has similar stellar parameters as our program stars ( @xmath34 , @xmath35 , @xmath6}}}=-3.88 $ ] ; * ? ? ? the lower left panel shows the region around the line at @xmath36 ; the lower right panel shows the mg triplet around @xmath37 ; and the upper panel shows the lines . atomic absorption lines were measured using the same line list as in @xcite , based on lines from @xcite , @xcite , and the vald database @xcite . equivalent - width measurements were obtained by fitting gaussian profiles to the observed absorption lines . table [ eqw ] lists the lines used in this work , their measured equivalent widths , and the derived abundance from each line . for the abundance analysis , we employ one - dimensional plane - parallel model atmospheres with no overshooting @xcite , computed under the assumption of local thermodynamic equilibrium ( lte ) . the 2011 version of the moog synthesis code @xcite was used for the spectral synthesis . in this version , scattering is treated with a source function that sums both absorption and scattering components , rather than treating continuous scattering as true absorption @xcite . elemental abundance ratios , @xmath38 $ ] , are calculated taking solar abundances from @xcite . the average measurements ( or upper limits ) for @xmath39 elements , derived from the mike spectra , are listed in table [ abund ] . the @xmath40 values are the standard error of the mean . the abundance uncertainties , as well as the systematic uncertainties in the abundance estimates due to the atmospheric parameters , were treated in the same way as described in @xcite . for the equivalent - width analysis , any uncertainties calculated to be less than the @xmath40 for was replaced by 0.05 dex . for the spectral synthesis , a best value for the abundance of a given line is assumed , then lower and upper abundance values are set so they enclose the entire spectral feature . that is taken as the uncertainty . for the systematic uncertainties , table [ sys ] shows how changes in each atmospheric parameter affect the determined abundances . also given is the total uncertainty for each element . from the high - resolution mike spectra , effective temperatures of the stars were determined by minimizing trends between the abundances derived from lines and their excitation potentials . the temperatures derived by this procedure are known to be underestimated when compared with @xmath41 based on photometry . as a consequence , such differences also lead to small changes in surface gravities and chemical abundances . @xcite provide a simple linear relation to correct the spectroscopy - derived `` excitation temperatures '' to photometric - based temperatures . we apply this procedure and use the corrected @xmath41 to obtain our final stellar parameters . for warmer stars on the subgiant branch and near the main - sequence turnoff , these corrections are fortunately small . in our cases , the temperature corrections were @xmath42 for , and @xmath43 for . microturbulent velocities were determined by minimizing the trend between the abundances of lines and their reduced equivalent widths . for , the surface gravity was determined from the balance of two ionization stages for iron lines ( and ) . we allowed the difference between the abundances of the and lines to be @xmath44 . for , since no lines could be measured , the surface gravity was estimated from a yale - yonsei isochrone @xcite with @xmath45 , @xmath6}}}=-3.5 $ ] , and @xmath46}}}=+0.4 $ ] . the final atmospheric parameters for the program stars are listed in table [ obstable ] . we were able to determine the lte lithium abundance for . from the doublet at @xmath47 , the spectral synthesis resulted in @xmath48 . @xcite provide nlte corrections for li abundances , based on evolutionary status and @xmath49 . from their figures 1 - 3 , it is possible to see that the corrections ( for @xmath50 and @xmath51 ) are less than @xmath52 , which is within the uncertainty of the observed abundances for . figure [ li6707_32 ] shows the spectral synthesis for the li line in . the dots represent the observed spectrum , and the solid line is the best abundance fit . the shaded area encompasses a @xmath53 difference in @xmath49 . the synthesized spectrum without li is represented by the light gray line . it was possible to measure the carbon abundance for ( @xmath54}}}=+0.49 $ ] ) , using the ch @xmath55-band region around @xmath56 . for , an upper limit was determined ( @xmath54}}}<+1.45 $ ] ) , using the procedure described in @xcite . figure [ ch4313_33 ] shows the ch @xmath55-band spectral synthesis for . the large black dots represent the observed spectrum , and the solid line is the best abundance fit . the dotted and dashed line represent a @xmath57 variation in @xmath58 $ ] , which we conservatively use as the uncertainty . we also calculated the carbon abundance corrections for , based on the procedure described by @xcite . for @xmath59 , @xmath6}}}=-3.64 $ ] , and @xmath54}}}=+0.49 $ ] , the correction is @xmath60 for the @xmath61}}}=0.0 $ ] case . for the elements na , mg , al , si , ca , sc , ti , cr , mn , co , and ni , abundances were determined from equivalent - width analysis only . for , no particular discrepancies were found for elements with abundances determined for more than one line . the standard errors of the average abundances are typically smaller than @xmath62 . in the case of , only na and mg ( besides fe ) have more than one measured line , both with @xmath63 . for , neutron - capture element abundances were determined from spectral synthesis . for , the lines used were 4077 ( @xmath64$]=@xmath65 ) and @xmath66 ( @xmath67}}=-1.20 $ ] ) ; the abundance was measured from the @xmath36 ( @xmath68}}=-1.30 $ ] ) line . figure [ basr_syn33 ] shows the comparison between the observed and synthetic spectra for these three lines in . there is good agreement between the abundances of the two sr lines ( values within @xmath69 ) . only upper limits were determined for ( @xmath67}}<-0.08 $ ] and @xmath68}}<+0.62 $ ] ) . there are many reasons why observations of new stars in the metallicity range @xmath6}}}<-3.5 $ ] are important . these include a proper description of the low - metallicity tail of the galactic halo mdf , as well as the nature of the progenitor populations of ump stars . below we show a comparison between the abundances of the program stars and other stellar abundances from the literature , as well as an attempt to describe the main characteristics ( e.g. , mass and explosion energy ) of ump star progenitors . is a subgiant ( @xmath70 , assuming @xmath71 ) , and has @xmath72 , well below the li plateau for metal - poor dwarf stars described by @xcite . figure [ licomp ] shows the behavior of li abundances , as a function of temperature , for a sample of metal - poor stars from the literature . individual references are listed in the figure caption . the li @xmath72 abundance for is consistent with data from the literature at the same temperature and metallicity , and suggests that li is already being depleting on the lower giant branch , due , at least in part , to internal processing . we also compared the abundances of our sample stars with the emp star samples from @xcite ( giants only ) and @xcite . results are shown in figure [ abfull ] . the carbon abundances from the literature were also corrected using the procedure described in @xcite . no significant differences are found for the light elements , except for al , which falls below the trend for the literature data . the upper limits of c , sr , and ba for are also consistent with typical values for halo stars in this metallicity range . is moderately enhanced in carbon ( @xmath54}}}=+0.49 $ ] ) . given its high gravity , the carbon correction is only @xmath60 , leading to @xmath54}}}_c=+0.50 $ ] . even though in the @xmath6}}}<-3.5 $ ] metallicity range , @xmath73 of the stars exhibit @xmath54}}}>+0.50 $ ] @xcite , the carbon abundance of appears to be within expectations . for , the @xmath54}}}<+1.45 $ ] is also within expectations , given that @xmath74 of the stars with @xmath6}}}<-4.3 $ ] show @xmath54}}}>+1.0 $ ] . however , a higher s / n spectrum is needed for a proper carbon abundance determination for . the abundances of ti , cr , and ni for appear to be higher than the trends presented by the literature data . we caution the reader , however , that the abundances for these species were derived by equivalent - width analysis of only one spectral feature . further measurements are needed to properly address this issue . in the @xmath6}}}<-3.0 $ ] regime , the neutron - capture elements sr and ba are thought to be formed by the @xmath75-process , rather than the @xmath76-process , which dominates at @xmath6}}}>-2.5 $ ] @xcite . the low values determined for are consistent with the literature data . the @xmath77}}=+0.06 $ ] value is also within the range for stars with @xmath6}}}<-3.5 $ ] @xcite . in this section we attempt to assess the properties ( mass distribution and explosion energies ) of the progenitor population of ump stars . for this exercise , we used and @xmath78 ump stars with parameters and abundances determined from high - resolution spectroscopy , gathered by @xcite . individual references and parameters are listed in table [ umptab ] . the abundance data were compared to the theoretical model predictions for non - rotating single massive population iii stars , in the range @xmath79 of @xcite . the model database comprises @xmath80 initial masses , and explores a range of explosion energies from @xmath81 to @xmath82 kinetic energy of the ejecta . the models further vary the amount of mixing in the supernova ejecta due to rayleigh taylor instabilities ( e.g. , * ? ? ? here we also use their @xmath83 matching algorithm @xcite . an online tool including the model database can be found at ` starfit ` . the same fitting procedure was also used in @xcite , @xcite , and @xcite . in this work we adopt a similar procedure as the one described in @xcite , using all available element measurements and upper limits up to atomic number @xmath84 . for cr and sc , the authors assumed that there are additional production sites not included in the model data , e.g. , contributions from neutron star winds , and hence the model yield was taken as lower limit . chemical abundances ( and/or upper limits ) for the ump sample are available for the following species : c , n , o , na , mg , al , si , ca , sc , ti , v , cr , mn , fe , co , ni , and zn . for the solar abundances , the ` starfit ` code uses the photospheric values from @xcite . figures [ starfit1 ] and [ starfit2 ] show the best model fit for the @xmath85 ump stars in table [ umptab ] . the red filled squares represent all the abundances gathered from the individual studies ( including upper limits as arrows ) . the solid black line is the best model fit to the data . among the @xmath85 stars , six do not have nitrogen abundance measurements or upper limits . for those , we estimated the nitrogen abundance using @xmath86}}=0.0 $ ] ( see discussion below and top panel of figure 11 in * ? ? ? * ) . for the carbon abundances , seven stars had their abundances corrected for their evolutionary status using the procedures described in @xcite . a summary of the abundances and corrections applied to the data is shown in table [ umptab ] . from the figures , one can see no relation between the progenitor mass and the iron abundance of the ump stars . this may be expected , since the @xmath6}}}<-4.0 $ ] range is believed to probe second - generation stars , for which the local enrichment from the pop iii star explosion is expected to dominate over global chemical evolution effects . in addition , it is possible to see that the models always predict a very low nitrogen abundance ( @xmath87}}<-6.0 $ ] ) where only upper limits on nitrogen are determined . this has a direct impact on the best model chosen by the ` starfit ` code , and introduces an artifact on the progenitor mass distribution , as explained below . also , for stars where @xmath88 $ ] is higher than @xmath89 $ ] by more than @xmath90 ( e.g. he 0057@xmath915959 , cd@xmath9138@xmath92245 , and cs 22949@xmath91037 ) , the model is not capable of reproducing the cn pattern , and the overall fit yields a higher residual value . figure [ histmass ] shows the progenitor mass ( left panels ) and explosion energy ( right panels ) distributions for the best model fits . the upper panels show the full sample ( @xmath85 stars ) , the middle panels show stars with measured nitrogen abundances ( @xmath93 stars including the estimates from table [ umptab ] ) , and the bottom panels show stars with nitrogen upper limits ( seven stars ) . for the progenitor mass , a clear separation in the distributions is seen in the middle and bottom panels . progenitor masses for stars with upper limits on nitrogen concentrate in the @xmath94 range . in contrast , progenitor masses for stars with measured nitrogen abundance show two preferred ranges : @xmath95 and @xmath96 . concerning the explosion energy , more than @xmath97 of the stars with measured nitrogen show @xmath98 , regardless of the progenitor mass ranges listed above . this can either be an indication of the nature of these progenitors , or an numerical artifact , since this is the lowest available energy within the model grid . in one particular case ( he 2239@xmath915019 ) , the explosion energy is the maximum value allowed by the models ( @xmath82 ) . this spurious result could be explained by the lack of carbon and nitrogen abundance measurements ( see explanation below in sections [ robust ] and [ error ] . ) , or by the behavior of the abundance pattern for the light elements . for the stars with measured nitrogen , interesting differences in the progenitor population arise for the lowest metallicity stars in the sample ; one example is the difference in the cno pattern of he 0107@xmath915240 and he 1327@xmath912326 . even though these two stars have @xmath6}}}$ ] values within @xmath99 of one another , and estimated progenitor masses within @xmath100 , he 1327@xmath912326 has a @xmath101 two orders of magnitude higher than he 0107@xmath915240 . in the context of this work , this difference could be explained by changes in mass and explosion energy ( @xmath102 for he 0107@xmath915240 and @xmath81 for he 1327@xmath912326 ) , and does not require additional models to describe the progenitor population . another less extreme comparison is for cs 30336@xmath91049 and he 1424@xmath910241 : their metallicity , carbon , and nitrogen abundances are within @xmath44 , @xmath103 , and @xmath104 , respectively , and both share the same best model fit ( @xmath105 and @xmath81 ) . for sdss j1742@xmath1925 ( @xmath6}}}<-5.07 $ ] ) , even though the progenitor mass and explosion energy ( @xmath106 and @xmath102 ) are similar to other stars in the same @xmath6}}}$ ] range , the small number of determined abundances ( c and ca ) clearly affects the fitting procedure ( see discussion below ) . additional abundance measurements for these stars , as well as the discovery of more stars in this @xmath0 $ ] range are needed for further investigation . for the most iron - poor star known , smss j0313@xmath916708 , even though our results are consistent with the models described in @xcite , the lack of nitrogen abundance measurement is possibly affecting the progenitor mass determination . the ` starfit ` procedure gives the ten best model fits for a given set of input abundances , ranked by their @xmath83 value . to test the robustness of the best solution , we analyzed the @xmath83 variation between the ten best models for each star , and how they affect the progenitor mass . for instance , a flat @xmath83 distribution with a wide range of progenitor masses is an indication that the solutions are not very robust . ideally , the @xmath83 value should rapidly increase between the first and second - best solutions . figure [ cases4 ] shows the evolution of the @xmath83 values as a function of the model rank . each line represents a star on table [ umptab ] , labeled by its `` star i d '' . the numbers above each point are the variations ( in @xmath107 ) of the @xmath83 value for a given model , compared to the @xmath83 value of the best model fit . the point size is proportional to the progenitor mass , with approximate values labeled on the top left . inspection of figure [ cases4 ] reveals that , for @xmath108 of the sample stars , the variation in @xmath83 between the first and second - best models is above @xmath109 . in addition , for six out of eight stars where the @xmath83 variation is below @xmath109 , the progenitor masses only changes in two cases . for these stars @xmath110 ( he 0057@xmath915959 ) and @xmath111 ( cd@xmath9138 245 ) the variation is between @xmath105 and @xmath112 , which are precisely the two preferred mass values according to figure [ histmass ] ( see discussion above ) . for stars @xmath113 ( cs 30336@xmath91049 ) and @xmath114 ( he 1424@xmath910241 ) , there is a large change in mass between model - rank 2 and 3 , with a small variation in @xmath83 . even though both stars have 14 determined abundances , additional work must be performed to evaluate whether the measurements or the models could be changed to generate a better agreement . for stars @xmath115 ( sdss j1742@xmath1925 ) and @xmath116 ( sdss j1035@xmath1906 ) , the low number of determined abundances compromises the robustness of the best model fits . chemical abundance measurements carry uncertainties related to the choice of model atmosphere , continuum placement when measuring absorption line strengths , and atomic data line lists . in addition , non - lte and 3d effects also have an impact on the determinations . all of these changes in the abundances can generate changes in the progenitor - star properties . keeping the observed uncertainty in mind , we are using c and n as representative examples , as the fitting results appear to most strongly depend on them . as shown above , the existence of a nitrogen measurement is quite important for the model - fit procedure . the stellar evolution of primordial stars , which are the basis of this study , is very peculiar . these primorial stars have no initial cno , so in order to burn hydrogen for which they need the cno cycle they need to produce a trace of cno material themselves . typical mass fractions are @xmath117 or less at the end of core hydrogen burning and in the h - burning shell . because of that , any significant amounts of nitrogen found in the ump stars ( and not made by them _ in situ _ ) must come from another primordial production mechanism in the pop iii stars , during or after the onset of central helium burning , and by mixing of the he burning products with the h envelope . without rotation or other mixing processes , in the models of @xcite this only occurred consistently for initial masses of @xmath118 and above ( their figure 11 ) , whereas most of the lower - mass models did not have large nitrogen yields . hence observational constraints from nitrogen measurements or upper limits can significantly constrain the mass range of models that provide a good fit . in order to evaluate how changes in the nitrogen ( and/or carbon ) abundance reflect on the progenitor mass , we ran the ` starfit ` procedure for the @xmath85 ump stars for ten different scenarios , listed in table [ casestab ] . then , we evaluated how the progenitor mass of the best model changes . figure [ cases2 ] shows how the progenitor mass changes for each star , for the ten cases listed in table [ casestab ] , compared with the result of the best - fit model presented in figures [ starfit1 ] and [ starfit2 ] , and table [ umptab ] . blue filled dots are stars with available carbon and nitrogen measurements , and red filled dots represent stars with available upper limits for carbon and/or nitrogen . when changing the carbon abundances cases ( i ) and ( ii ) the stars affected with available carbon and/or nitrogen measurements are @xmath119 ( cs 22949@xmath91037 ) , and @xmath116 ( sdss j1035@xmath190641 ) . star @xmath119 has its progenitor mass changed from @xmath112 to @xmath105 ( within the mass range shown in the midle panels of figure [ histmass ] ) , and star @xmath116 from @xmath106 to @xmath120 . the latter case has only three determined abundances , which makes the model - fitting more susceptible to changes . for nitrogen , large changes in progenitor mass are seen ( for stars @xmath121 , @xmath122 , @xmath123 , @xmath113 , @xmath114 , @xmath124 , and @xmath115 ) when reducing the @xmath125 $ ] abundances cases ( iii ) and ( iv ) . for positive changes in nitrogen cases ( v ) , ( vi ) , and ( vii ) there are smaller progenitor masses variations for a handful of stars . these variations , however , are always between @xmath126 and @xmath127 , agreeing with the two distributions shown in the middle panel of figure [ histmass ] . when removing the carbon abundances from the fit case ( ix ) the only considerable change ( apart from the stars with upper limits ) is for star @xmath121 ( sdss j2209@xmath910028 ) . this can be explained by the low number of abundances five used for the fit . the most noticeable changes happen for cases ( viii ) and ( x ) , when the nitrogen abundance is removed . then , every progenitor mass is changed by at least @xmath128 . with these tests , we show that the models are sufficiently robust to not be affected by typical uncertainties in @xmath58 $ ] and @xmath125 $ ] ( @xmath129 ) . moreover , regardless of the number of abundances available for the fit , we stress the strong effect the presence ( or absence ) of nitrogen abundances has on the final result . below we compare our results with a similar study , in which abundance patterns of stars with @xmath6}}}<-3.5 $ ] are compared with a set of sn models with fixed masses . @xcite performed abundance profiling for @xmath130 stars with @xmath6}}}<-3.5 $ ] from the literature , to gain insight into the properties of the pop iii progenitors . the authors use pop iii sn models with @xmath131 and @xmath132 . with the fixed mass , free parameters are explosion energy , remnant mass , and ejected mass . they also include models with @xmath131 and enhanced mixing due to rapid rotation . there are ten stars in common between @xcite and this work . for all these stars , the explosion energy of their @xmath131 sn progenitor is higher ( by at least an order of magnitude ) then the values found by this work . regardless , it is interesting to see that the progenitor mass range found by the ` starfit ` ( @xmath133 - when using both c and n for the matching ) is consistent with a @xmath131 progenitor . for the two most iron - poor stars analyzed by @xcite he 0107@xmath912326 and he 1327@xmath912326 the authors find similar explosion energies and remnant masses for both stars , regardless of the different nitrogen abundances . in their figure 7 , however , the models can not properly reproduce the nitrogen abundance for he 1327@xmath912326 . @xcite also performs abundance profiling for stars with @xmath6 } } } < -4.5 $ ] . their best - fit models for he 0107@xmath912326 , he 1327@xmath912326 , and he 0557@xmath914840 ( high carbon abundances ) have @xmath131 and explosion energy of @xmath134 . similar to what was found by @xcite , the high n abundance for he 1327@xmath912326 is also not reproduced , with the models consistently underproducing nitrogen when compared to the observed values , by at least @xmath135 . for sdss j1029@xmath191729 , the authors find a suitable model with @xmath132 and explosion energy of @xmath136 . an alternative formation scenario for this star , which includes dust - induced cooling , is given by @xcite , where the authors find good agreement between observations and the yields of core - collapse sne with metal - free progenitors of @xmath137 and @xmath138 . in the case of smss j0313@xmath916708 , models with @xmath131 and @xmath132 well - reproduce the abundances of carbon , magnesium , and calcium , as well as the upper limit for nitrogen . in contrast , the model presented in this work for smss j0313@xmath916708 also has a good overall abundance fit , even though is has a higher mass ( @xmath139 ) and lower explosion energy @xmath134 as opposed to @xmath140 . in this work we presented the high - resolution abundance analysis of two stars selected from the sdss / segue survey , with @xmath6}}}=-3.64 $ ] and @xmath6}}}=-4.34 $ ] . a detailed chemical abundance analysis reveals that these stars show the expected behavior of stars in the same metallicity range . the addition of these two stars to the @xmath6}}}<-3.5 $ ] range corroborates with the hypothesis presented by @xcite that the galactic halo mdf smoothly decreases down to @xmath6}}}=-4.1 $ ] , instead of the sharp cutoff at @xmath6}}}=-3.6 $ ] suggested by @xcite and @xcite . we also provide new insights on the progenitor population of ump stars , by comparing their abundance patterns to a set of theoretical sn models . we find that all ump stars with carbon and nitrogen abundances available have progenitors with masses in either @xmath141 or @xmath142 ranges , with explosion energies between @xmath9 and @xmath10 . we stress that , even though there could be additional suitable candidates for the ump progenitors , such as the fast rotating massive stars from @xcite and the faint supernovae from @xcite , the models presented in this work are capable of describing the differences in the abundance patterns by adjusting the progenitor mass and explosion energy . comparison with similar studies from the literature shows that , even though higher progenitor masses ( @xmath143 ) also show good agreement for the most extreme cases , such as smss j0313@xmath916708 , models have their best fits for @xmath131 , which is consistent with the mass range found in this work . besides having additional targets at @xmath6}}}<-4.0 $ ] for comparison , it is important to have accurate nitrogen and light element abundances for all ump stars in the literature . there are still a number of similar candidates in need of high - resolution follow - up , and we expect to conduct similar studies in the near future . v.m.p . and t.c.b . acknowledge partial support for this work from grants phy 08 - 22648 ; physics frontier center / joint institute or nuclear astrophysics ( jina ) , and phy 14 - 30152 ; physics frontier center / jina center for the evolution of the elements ( jina - cee ) , awarded by the us national science foundation . a.f . is supported by nsf career grant ast-1255160 . y.s.l . acknowledges partial support by the 2014 research fund of chungnam national university and support from the national research foundation of korea to the center for galaxy evolution research . was supported by a future fellowship of the australian research council ( ft120100363 ) . , r. , shectman , s. a. , gunnels , s. m. , mochnacki , s. , & athey , a. e. 2003 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4841 , society of photo - optical instrumentation engineers ( spie ) conference series , ed . m. iye & a. f. m. moorwood , 1694 lrr @xmath144 ( j2000 ) & 13:22:50.6 & 12:04:41.4 + @xmath23 ( j2000 ) & @xmath1901:23:43.0 & @xmath1912:01:11.5 + @xmath145 ( mag ) & 16.3 & 16.4 + @xmath146 & 0.50 & 0.34 + + date & 2013 05 30 & 2013 05 31 + ut & 02:35:55 & 00:09:39 + exptime ( s ) & 3600 & 5400 + v@xmath147(km / s ) & 87.84 & 83.00 ccccccccc & 5466 ( 150 ) & 3.12 ( 0.35 ) & @xmath913.32 ( 0.20 ) & & 5008 ( 100 ) & 1.95 ( 0.20 ) & 1.95 ( 0.20 ) & @xmath913.64 ( 0.05 ) + & 5894 ( 150 ) & 2.66 ( 0.35 ) & @xmath913.41 ( 0.20 ) & & 5467 ( 100 ) & 3.20 ( 0.20 ) & 1.50 ( 0.20 ) & @xmath914.34 ( 0.05 ) + lrrrrrrrr na i & 5889.950 & 0.00 & 0.108 & 62.8 & 2.30 & & 39.0 & 2.28 + na i & 5895.924 & 0.00 & @xmath910.194 & 61.9 & 2.58 & & 30.3 & 2.41 + mg i & 3829.355 & 2.71 & @xmath910.208 & 98.5 & 4.01 & & & + mg i & 3832.304 & 2.71 & 0.270 & 132.3 & 4.22 & & 89.6 & 3.62 + mg i & 4167.271 & 4.35 & @xmath910.710 & 9.1 & 4.35 & & & + mg i & 4702.990 & 4.33 & @xmath910.380 & 13.6 & 4.18 & & & + mg i & 5172.684 & 2.71 & @xmath910.450 & 105.0 & 4.19 & & 63.5 & 3.78 + mg i & 5183.604 & 2.72 & @xmath910.239 & 117.2 & 4.23 & & 64.1 & 3.59 + mg i & 5528.405 & 4.34 & @xmath910.498 & 12.9 & 4.27 & & & + al i & 3961.520 & 0.01 & @xmath910.340 & 37.6 & 1.49 & & & + si i & 3905.523 & 1.91 & @xmath911.092 & 106.5 & 4.14 & & 71.8 & 3.88 + ca i & 4226.730 & 0.00 & 0.244 & 103.8 & 2.52 & & 77.1 & 2.42 + ca i & 4283.010 & 1.89 & @xmath910.224 & 16.2 & 3.17 & & & + ca i & 4434.960 & 1.89 & @xmath910.010 & 16.8 & 2.97 & & & + ca i & 4454.780 & 1.90 & 0.260 & 24.6 & 2.92 & & & + ca i & 6162.170 & 1.90 & @xmath910.089 & 18.8 & 3.06 & & & + ca i & 6439.070 & 2.52 & 0.470 & 12.0 & 2.96 & & & + ca ii & 3933.663 & 0.00 & 0.105 & & & & 728.3 & 2.35 + sc ii & 4246.820 & 0.32 & 0.240 & 47.7 & @xmath910.86 & & 13.6 & @xmath910.78 + sc ii & 4314.083 & 0.62 & @xmath910.100 & 20.4 & @xmath910.73 & & & + ti ii & 3759.291 & 0.61 & 0.280 & 124.6 & 1.80 & & & + ti ii & 3761.320 & 0.57 & 0.180 & 116.8 & 1.67 & & & + ti ii & 3813.394 & 0.61 & @xmath912.020 & 20.9 & 1.62 & & & + ti ii & 3913.461 & 1.12 & @xmath910.420 & 48.6 & 1.17 & & & + ti ii & 4012.396 & 0.57 & @xmath911.750 & 44.6 & 1.78 & & & + ti ii & 4025.120 & 0.61 & @xmath911.980 & 21.8 & 1.58 & & & + ti ii & 4290.219 & 1.16 & @xmath910.930 & 31.4 & 1.36 & & & + ti ii & 4337.914 & 1.08 & @xmath910.960 & 37.0 & 1.41 & & & + ti ii & 4395.031 & 1.08 & @xmath910.540 & 52.4 & 1.26 & & & + ti ii & 4399.765 & 1.24 & @xmath911.190 & 22.9 & 1.51 & & & + ti ii & 4443.801 & 1.08 & @xmath910.720 & 45.0 & 1.30 & & & + ti ii & 4450.482 & 1.08 & @xmath911.520 & 14.8 & 1.42 & & & + ti ii & 4468.517 & 1.13 & @xmath910.600 & 41.0 & 1.17 & & 19.0 & 1.48 + ti ii & 4501.270 & 1.12 & @xmath910.770 & 37.7 & 1.26 & & & + ti ii & 4533.960 & 1.24 & @xmath910.530 & 38.3 & 1.17 & & & + ti ii & 4563.770 & 1.22 & @xmath910.960 & 30.1 & 1.41 & & & + ti ii & 4571.971 & 1.57 & @xmath910.320 & 32.8 & 1.23 & & & + cr i & 4254.332 & 0.00 & @xmath910.114 & 45.4 & 1.33 & & & + cr i & 4274.800 & 0.00 & @xmath910.220 & 49.8 & 1.52 & & & + cr i & 4289.720 & 0.00 & @xmath910.370 & 26.5 & 1.21 & & & + cr i & 5206.040 & 0.94 & 0.020 & 23.3 & 1.76 & & & + cr i & 5208.419 & 0.94 & 0.160 & 15.0 & 1.39 & & 16.8 & 1.89 + mn i & 4030.753 & 0.00 & @xmath910.480 & 58.8 & 1.28 & & & + mn i & 4033.062 & 0.00 & @xmath910.618 & 40.7 & 1.07 & & & + mn i & 4034.483 & 0.00 & @xmath910.811 & 45.2 & 1.34 & & & + fe i & 3727.619 & 0.96 & @xmath910.609 & 97.6 & 3.81 & & & + fe i & 3743.362 & 0.99 & @xmath910.790 & 96.8 & 4.00 & & & + fe i & 3753.611 & 2.18 & @xmath910.890 & 23.6 & 3.70 & & & + fe i & 3763.789 & 0.99 & @xmath910.221 & 109.3 & 3.76 & & & + fe i & 3765.539 & 3.24 & 0.482 & 41.0 & 3.89 & & & + fe i & 3767.192 & 1.01 & @xmath910.390 & 109.9 & 3.96 & & 58.4 & 3.23 + fe i & 3786.677 & 1.01 & @xmath912.185 & 41.0 & 4.04 & & & + fe i & 3787.880 & 1.01 & @xmath910.838 & 87.0 & 3.78 & & & + fe i & 3815.840 & 1.48 & 0.237 & 107.1 & 3.76 & & 61.6 & 3.17 + fe i & 3820.425 & 0.86 & 0.157 & & & & 77.1 & 3.05 + fe i & 3824.444 & 0.00 & @xmath911.360 & 118.1 & 3.97 & & & + fe i & 3825.881 & 0.91 & @xmath910.024 & & & & 76.7 & 3.27 + fe i & 3827.823 & 1.56 & 0.094 & 98.2 & 3.77 & & & + fe i & 3840.438 & 0.99 & @xmath910.497 & 103.1 & 3.85 & & & + fe i & 3841.048 & 1.61 & @xmath910.044 & 88.9 & 3.69 & & & + fe i & 3846.800 & 3.25 & @xmath910.020 & 19.4 & 3.90 & & & + fe i & 3849.967 & 1.01 & @xmath910.863 & 96.4 & 4.04 & & & + fe i & 3850.818 & 0.99 & @xmath911.745 & 59.4 & 3.94 & & & + fe i & 3856.372 & 0.05 & @xmath911.280 & 120.0 & 3.98 & & & + fe i & 3859.911 & 0.00 & @xmath910.710 & & & & 85.8 & 3.27 + fe i & 3865.523 & 1.01 & @xmath910.950 & 80.2 & 3.67 & & & + fe i & 3878.018 & 0.96 & @xmath910.896 & 85.8 & 3.72 & & & + fe i & 3878.573 & 0.09 & @xmath911.380 & 110.8 & 3.90 & & 57.5 & 3.21 + fe i & 3886.282 & 0.05 & @xmath911.080 & & & & 65.3 & 3.08 + fe i & 3887.048 & 0.91 & @xmath911.140 & & & & 32.5 & 3.24 + fe i & 3895.656 & 0.11 & @xmath911.668 & 95.2 & 3.78 & & 42.8 & 3.15 + fe i & 3899.707 & 0.09 & @xmath911.515 & 106.2 & 3.90 & & & + fe i & 3902.946 & 1.56 & @xmath910.442 & 86.8 & 3.96 & & & + fe i & 3917.181 & 0.99 & @xmath912.155 & 35.3 & 3.85 & & & + fe i & 3920.258 & 0.12 & @xmath911.734 & 98.1 & 3.92 & & 41.4 & 3.19 + fe i & 3922.912 & 0.05 & @xmath911.626 & 107.8 & 4.01 & & 45.4 & 3.10 + fe i & 3940.878 & 0.96 & @xmath912.600 & 26.4 & 4.07 & & & + fe i & 3949.953 & 2.18 & @xmath911.251 & 13.8 & 3.75 & & & + fe i & 3977.741 & 2.20 & @xmath911.120 & 22.3 & 3.90 & & & + fe i & 4005.242 & 1.56 & @xmath910.583 & 79.1 & 3.87 & & & + fe i & 4045.812 & 1.49 & 0.284 & 106.8 & 3.66 & & 59.9 & 3.07 + fe i & 4063.594 & 1.56 & 0.062 & 99.0 & 3.75 & & 46.1 & 3.02 + fe i & 4067.978 & 3.21 & @xmath910.470 & 11.8 & 4.04 & & & + fe i & 4071.738 & 1.61 & @xmath910.008 & 101.7 & 3.94 & & 423.0 & 3.07 + fe i & 4132.058 & 1.61 & @xmath910.675 & 74.5 & 3.88 & & & + fe i & 4134.678 & 2.83 & @xmath910.649 & 11.3 & 3.77 & & & + fe i & 4143.868 & 1.56 & @xmath910.511 & 75.7 & 3.68 & & 28.3 & 3.18 + fe i & 4147.669 & 1.48 & @xmath912.071 & 15.4 & 3.82 & & & + fe i & 4181.755 & 2.83 & @xmath910.371 & 23.6 & 3.88 & & & + fe i & 4187.039 & 2.45 & @xmath910.514 & 31.9 & 3.77 & & & + fe i & 4187.795 & 2.42 & @xmath910.510 & 36.2 & 3.83 & & & + fe i & 4191.430 & 2.47 & @xmath910.666 & 26.7 & 3.84 & & & + fe i & 4199.095 & 3.05 & 0.156 & 26.9 & 3.67 & & & + fe i & 4202.029 & 1.49 & @xmath910.689 & 71.1 & 3.66 & & & + fe i & 4216.184 & 0.00 & @xmath913.357 & 36.0 & 3.90 & & & + fe i & 4222.213 & 2.45 & @xmath910.914 & 24.2 & 4.00 & & & + fe i & 4233.603 & 2.48 & @xmath910.579 & 29.0 & 3.81 & & & + fe i & 4250.119 & 2.47 & @xmath910.380 & 32.3 & 3.67 & & & + fe i & 4250.787 & 1.56 & @xmath910.713 & 79.6 & 3.97 & & 17.3 & 3.07 + fe i & 4260.474 & 2.40 & 0.077 & 69.0 & 3.86 & & & + fe i & 4271.154 & 2.45 & @xmath910.337 & 40.5 & 3.77 & & & + fe i & 4271.760 & 1.49 & @xmath910.173 & 91.8 & 3.66 & & 50.9 & 3.27 + fe i & 4282.403 & 2.18 & @xmath910.779 & 29.5 & 3.68 & & & + fe i & 4325.762 & 1.61 & 0.006 & 94.2 & 3.67 & & 52.7 & 3.26 + fe i & 4337.046 & 1.56 & @xmath911.695 & 34.6 & 4.00 & & & + fe i & 4375.930 & 0.00 & @xmath913.005 & 46.8 & 3.73 & & & + fe i & 4383.545 & 1.48 & 0.200 & 115.6 & 3.86 & & 58.4 & 3.07 + fe i & 4404.750 & 1.56 & @xmath910.147 & 92.6 & 3.71 & & 45.1 & 3.18 + fe i & 4415.122 & 1.61 & @xmath910.621 & 74.6 & 3.78 & & & + fe i & 4427.310 & 0.05 & @xmath912.924 & 49.3 & 3.75 & & & + fe i & 4447.717 & 2.22 & @xmath911.339 & 15.2 & 3.90 & & & + fe i & 4459.118 & 2.18 & @xmath911.279 & 22.4 & 4.00 & & & + fe i & 4461.653 & 0.09 & @xmath913.194 & 38.1 & 3.85 & & & + fe i & 4476.019 & 2.85 & @xmath910.820 & 11.3 & 3.94 & & & + fe i & 4489.739 & 0.12 & @xmath913.899 & 14.4 & 4.02 & & & + fe i & 4494.563 & 2.20 & @xmath911.143 & 23.7 & 3.91 & & & + fe i & 4528.614 & 2.18 & @xmath910.822 & 41.5 & 3.94 & & & + fe i & 4531.148 & 1.48 & @xmath912.101 & 18.3 & 3.91 & & & + fe i & 4871.318 & 2.87 & @xmath910.362 & 20.5 & 3.79 & & & + fe i & 4872.137 & 2.88 & @xmath910.567 & 12.0 & 3.73 & & & + fe i & 4890.755 & 2.88 & @xmath910.394 & 21.6 & 3.86 & & & + fe i & 4891.492 & 2.85 & @xmath910.111 & 29.3 & 3.73 & & & + fe i & 4918.994 & 2.85 & @xmath910.342 & 19.3 & 3.71 & & & + fe i & 4920.503 & 2.83 & 0.068 & 36.5 & 3.67 & & & + fe i & 5006.119 & 2.83 & @xmath910.615 & 21.9 & 4.03 & & & + fe i & 5012.068 & 0.86 & @xmath912.642 & 32.6 & 4.04 & & & + fe i & 5051.634 & 0.92 & @xmath912.764 & 26.1 & 4.09 & & & + fe i & 5269.537 & 0.86 & @xmath911.333 & 98.2 & 4.11 & & 32.6 & 3.29 + fe i & 5328.039 & 0.92 & @xmath911.466 & 90.3 & 4.12 & & & + fe i & 5328.531 & 1.56 & @xmath911.850 & 18.0 & 3.70 & & & + fe i & 5371.489 & 0.96 & @xmath911.644 & 80.7 & 4.10 & & & + fe i & 5397.128 & 0.92 & @xmath911.982 & 65.6 & 4.06 & & & + fe i & 5405.775 & 0.99 & @xmath911.852 & 58.9 & 3.88 & & & + fe i & 5429.696 & 0.96 & @xmath911.881 & 62.0 & 3.94 & & & + fe i & 5434.524 & 1.01 & @xmath912.126 & 36.6 & 3.76 & & & + fe i & 5446.917 & 0.99 & @xmath911.910 & 54.9 & 3.86 & & & + fe i & 5455.609 & 1.01 & @xmath912.090 & 45.2 & 3.88 & & & + fe i & 5506.779 & 0.99 & @xmath912.789 & 16.0 & 3.91 & & & + fe ii & 4233.170 & 2.58 & @xmath911.970 & 27.2 & 3.85 & & & + fe ii & 4522.630 & 2.84 & @xmath912.250 & 9.3 & 3.84 & & & + fe ii & 4583.840 & 2.81 & @xmath911.930 & 22.0 & 3.93 & & & + fe ii & 4923.930 & 2.89 & @xmath911.320 & 41.6 & 3.81 & & & + fe ii & 5018.450 & 2.89 & @xmath911.220 & 49.2 & 3.85 & & & + co i & 3845.468 & 0.92 & 0.010 & 46.0 & 1.68 & & & + co i & 3873.120 & 0.43 & @xmath910.660 & 55.3 & 1.97 & & & + co i & 3881.869 & 0.58 & @xmath911.130 & 28.4 & 2.07 & & & + co i & 3995.306 & 0.92 & @xmath910.220 & 28.9 & 1.55 & & & + co i & 4121.318 & 0.92 & @xmath910.320 & 35.3 & 1.77 & & & + ni i & 3452.880 & 0.11 & @xmath910.900 & 76.8 & 2.45 & & & + ni i & 3483.770 & 0.28 & @xmath911.120 & 64.5 & 2.48 & & & + ni i & 3492.960 & 0.11 & @xmath910.265 & 92.8 & 2.30 & & & + ni i & 3519.770 & 0.28 & @xmath911.422 & 58.9 & 2.63 & & & + ni i & 3524.540 & 0.03 & 0.007 & 110.7 & 2.40 & & & + ni i & 3566.370 & 0.42 & @xmath910.251 & & & & 68.9 & 2.61 + ni i & 3597.710 & 0.21 & @xmath911.115 & 68.8 & 2.48 & & & + ni i & 3783.520 & 0.42 & @xmath911.420 & 54.2 & 2.56 & & & + ni i & 3807.140 & 0.42 & @xmath911.220 & 59.6 & 2.47 & & & + ni i & 3858.301 & 0.42 & @xmath910.951 & 71.5 & 2.47 & & & + ni i & 5476.900 & 1.83 & @xmath910.890 & 18.3 & 2.75 & & & + lrrrrrrrrrr li & 1.05 & & & & & & 1.70 & & 0.15 & 1 + c & 8.43 & 5.28 & @xmath190.49 & 0.20 & 1 & & @xmath1485.54 & @xmath1491.45 & & 1 + na i & 6.24 & 2.44 & @xmath910.16 & 0.10 & 2 & & 2.34 & @xmath190.44 & 0.05 & 2 + mg i & 7.60 & 4.21 & @xmath190.25 & 0.05 & 7 & & 3.66 & @xmath190.40 & 0.05 & 3 + al i & 6.45 & 1.49 & @xmath911.32 & & 1 & & & & & + si i & 7.51 & 4.14 & @xmath190.27 & & 1 & & 3.88 & @xmath190.71 & & 1 + ca i & 6.34 & 2.93 & @xmath190.23 & 0.08 & 6 & & 2.42 & @xmath190.42 & & 1 + ca ii & 6.34 & & & & & & 2.35 & @xmath190.35 & & 1 + sc ii & 3.15 & @xmath910.79 & @xmath910.30 & 0.05 & 2 & & @xmath910.78 & @xmath190.41 & & 1 + ti ii & 4.95 & 1.42 & @xmath190.11 & 0.05 & 17 & & 1.48 & @xmath190.87 & & 1 + cr i & 5.64 & 1.44 & @xmath910.56 & 0.08 & 5 & & 1.89 & @xmath190.59 & & 1 + mn i & 5.43 & 1.23 & @xmath910.56 & 0.07 & 3 & & & & & + fe i & 7.50 & 3.86 & @xmath913.64 & 0.01 & 88 & & 3.16 & @xmath914.34 & 0.02 & 21 + fe ii & 7.50 & 3.86 & @xmath913.64 & 0.02 & 5 & & & & & + co i & 4.99 & 1.81 & @xmath190.46 & 0.08 & 6 & & & & & + ni i & 6.22 & 2.50 & @xmath910.08 & 0.05 & 10 & & 2.61 & @xmath190.73 & & 1 + sr ii & 2.87 & @xmath912.01 & @xmath911.24 & 0.20 & 2 & & @xmath1501.55 & @xmath1500.08 & & 1 + ba ii & 2.18 & @xmath912.76 & @xmath911.30 & 0.20 & 1 & & @xmath1501.54 & @xmath1490.62 & & 1 + lrrrr na i & 0.15 & @xmath910.01 & @xmath910.03 & 0.18 + mg i & 0.12 & @xmath910.03 & @xmath910.05 & 0.14 + al i & 0.15 & @xmath910.02 & @xmath910.02 & 0.18 + si i & 0.17 & @xmath910.03 & @xmath910.11 & 0.23 + ca i & 0.11 & @xmath910.02 & @xmath910.03 & 0.14 + sc ii & 0.10 & 0.06 & @xmath910.02 & 0.14 + ti ii & 0.09 & 0.06 & @xmath910.03 & 0.12 + cr i & 0.17 & @xmath910.02 & @xmath910.02 & 0.19 + mn i & 0.20 & @xmath910.02 & @xmath910.03 & 0.21 + fe i & 0.18 & @xmath910.02 & @xmath910.07 & 0.19 + fe ii & 0.03 & 0.07 & @xmath910.02 & 0.09 + co i & 0.19 & @xmath910.02 & @xmath910.02 & 0.23 + ni i & 0.22 & @xmath910.03 & @xmath910.09 & 0.24 + sr ii & 0.11 & 0.07 & @xmath910.03 & 0.16 + ba ii & 0.12 & 0.05 & @xmath910.01 & 0.16 @rlrrrrrc@c@r@ 1 & sdss j2209@xmath910028 & @xmath914.00 & @xmath192.56 & 6.99 & & 6.39 & 27.0 & 0.3 & @xcite + 2 & he 2139@xmath915432 & @xmath914.02 & @xmath192.60 & 7.01 & 5.89 & & 28.0 & 0.6 & @xcite + 3 & g77@xmath9161 & @xmath914.03 & @xmath192.49 & 7.00 & 6.40 & & 27.0 & 0.3 & @xcite + 4 & cs 30336@xmath91049 & @xmath914.03 & @xmath190.09 & 4.85 & 4.70 & & 21.5 & 0.3 & @xcite + 5 & he 1424@xmath910241 & @xmath914.05 & @xmath190.63 & 5.01 & & 4.41 & 21.5 & 0.3 & @xcite + 6 & he 0057@xmath915959 & @xmath914.08 & @xmath190.86 & 5.21 & 5.90 & & 27.0 & 0.3 & @xcite + 7 & sdss j0140@xmath1923 & @xmath914.09 & @xmath191.57 & 5.91 & & 5.31 & 27.0 & 0.3 & @xcite + 8 & he 2239@xmath915019 & @xmath914.15 & @xmath191.80 & @xmath1515.98 & @xmath1516.38 & & 15.0 & 10.0 & @xcite + 9 & he 1310@xmath910536 & @xmath914.15 & @xmath192.53 & 6.72 & @xmath1516.88 & & 10.9 & 0.3 & @xcite + 10 & cd@xmath9138 245 & @xmath914.15 & @xmath910.09 & 4.19 & 4.75 & & 21.5 & 0.3 & @xcite + 11 & sdss j1204@xmath191201 & @xmath914.34 & @xmath1491.45 & @xmath1515.54 & & @xmath1514.94 & 10.6 & 0.9 & this work + 12 & cs 22949@xmath91037 & @xmath914.38 & @xmath191.73 & 5.78 & 5.95 & & 27.0 & 0.3 & @xcite + 13 & he 0233@xmath910343 & @xmath914.68 & @xmath193.32 & 7.23 & @xmath1515.95 & & 11.9 & 0.3 & @xcite + 14 & he 0557@xmath914840 & @xmath914.75 & @xmath191.66 & 5.30 & @xmath1515.40 & & 10.9 & 0.6 & @xcite + 15 & sdss j1742@xmath192531 & @xmath914.80 & @xmath193.63 & 7.26 & & 6.66 & 21.5 & 0.3 & @xcite + 16 & sdss j1029@xmath191729 & @xmath914.99 & @xmath1490.70 & @xmath1514.20 & @xmath1513.10 & & 10.6 & 0.9 & @xcite + 17 & sdss j1313@xmath190019 & @xmath915.00 & @xmath192.96 & 6.39 & 6.29 & & 27.0 & 0.3 & @xcite + 18 & sdss j1035@xmath190641 & @xmath1505.07 & @xmath1523.55 & 6.90 & & 6.30 & 23.0 & 0.6 & @xcite + 19 & he 0107@xmath915240 & @xmath915.54 & @xmath192.69 & 5.58 & 3.80 & & 20.5 & 0.6 & @xcite + 20 & he 1327@xmath912326 & @xmath915.65 & @xmath193.48 & 6.26 & 5.93 & & 21.5 & 0.3 & @xcite + 21 & smss j0313@xmath916708 & @xmath1507.80 & @xmath1525.39 & 6.02 & @xmath1513.63 & & 41.0 & 1.2 & @xcite + \(i ) & @xmath54}}}-0.5 $ ] + ( ii ) & @xmath54}}}-1.0 $ ] + ( iii ) & @xmath61}}}-0.5 $ ] + ( iv ) & @xmath61}}}-1.0 $ ] + ( v ) & @xmath61}}}+0.5 $ ] + ( vi ) & @xmath61}}}+1.0 $ ] + ( vii ) & @xmath61}}}+2.0 $ ] + ( viii ) & no carbon / nitrogen + ( ix ) & no carbon + ( x ) & no nitrogen +

isovector and isoscalar dipole excitations inbe andbe are investigated in the framework of antisymmetrized molecular dynamics , in which angular - momentum and parity projections are performed . in the present method , 1p-1h excitations on the ground state and large amplitude-cluster mode are incorporated . the isovector giant dipole resonance ( gdr ) in mev shows the two peak structure which is understood by the dipole excitation in the 2 core part with the prolate deformation . because of valence neutron modes against the core , low - energy e1 resonances appear in mev exhausting about of the thomas - reiche - kuhn sum rule and of the calculated energy - weighted sum . the dipole resonance at mev inbe can be interpreted as the parity partner of the ground state having ahe+ structure and has the remarkable e1 strength because of coherent contribution of two valence neutrons . the isd strength for some low - energy resonances are significantly enhanced by the coupling with the-cluster mode . the calculated e1 strength ofbe reasonably describes the global feature of experimental photonuclear cross sections consisting of the low - energy strength in mev and the gdr in mev .

1511.08530

in neutron - rich nuclei , various exotic phenomena appear because of excess neutrons . one of the current issues concerning exotic excitation modes in neutron - rich nuclei is low - energy dipole excitations @xcite . for stable nuclei , isovector giant dipole resonances ( gdr ) have been systematically observed in various nuclei by measurements of photonuclear cross sections ( for example , ref . @xcite and references therein ) . the gdr is understood as opposite oscillation between protons and neutrons and microscopically described by coherent 1p-1h excitations . peak structure of strength function of the gdr has been often discussed in relation with nuclear deformation . in neutron - rich nuclei , low - energy dipole resonances have been suggested to appear because of excess neutron motion against a core . the so - called soft dipole resonance , which is the enhanced e1 strength observed in the extremely low - energy ( @xmath10 mev ) region in neutron halo nuclei such as @xmath9he and @xmath11li , has been intensively studied by experimental and theoretical groups and often discussed in relation with two - neutron correlation @xcite . more generally the low - energy e1 strength typically in @xmath12 mev region is predicted in various nuclei in a wide mass region , and the one decoupling from the gdr is called pigmy dipole resonance . the role of excess neutrons in and the collectivity of these low - energy dipole excitations are topics of interest in theoretical studies . the low - energy e1 strength has been also discussed in relation with neutron skin thickness and density dependence of the symmetry energy , though the correlation between pigmy dipole strength , neutron skin thickness , and the symmetry energy is under debate @xcite . the experimental measurements of the low - energy e1 strength in @xmath0be by photodisintegration have been performed mainly in astrophysical interests @xcite . in these years , precise data of the e1 strength for low - lying positive - parity states of @xmath0be have been reported @xcite . moreover , the recently measured photodisintegration cross sections of @xmath0be indicate the significant low - energy e1 strength around @xmath13 mev exhausting about 10% of the thomas - reiche - kuhn ( trk ) sum rule @xcite , consistently with the bremsstrahlung data @xcite . the experimental data of the e1 strength of @xmath0be are available in a wide energy region from low energy to high energy : the e1 strength for the positive - parity states in @xmath14 mev @xcite , the significant e1 strength around @xmath13 mev @xcite , and the e1 strength in @xmath3 mev for the gdr @xcite . our aim is to investigate isovector and isoscalar dipole strengths in @xmath0be and @xmath1be to understand the low - energy dipole modes . i try to clarify the role of excess neutrons and the decoupling mechanism of the low - energy dipole modes from the gdr . the low - lying states of @xmath0be are understood by @xmath15 cluster structures as discussed in cluster models @xcite . the photodisintegration cross sections in very low - energy region of @xmath0be have been theoretically investigated by two - body ( @xmath16be+@xmath17 ) and three - body ( @xmath15 ) cluster models with continuum states @xcite . however , such cluster models are not able to describe the high - energy dipole strength of the gdr , to which coherent 1p-1h excitations may contribute . to investigate the pigmy and giant dipole resonances in general nuclei , shell model and mean - field approaches have been applied . the former may not be suitable for such largely deformed nuclei as be isotopes having cluster structures . the latter is usually based on the random phase approximation ( rpa ) with and without continuum . although the rpa calculation is successful for a variety of collective excitations in heavy mass nuclei , it is a small amplitude approximation neglecting large amplitude motion . moreover , in most of current mean - field approaches , the rpa calculation is based a parity - symmetric mean field in a strong coupling picture without the angular - momentum and parity projections , and the coupling of single - particle excitations in the mean - field with rotation and parity transformation is not taken into account microscopically . to take into account the coherent 1p-1h excitations and the large amplitude cluster mode as well as the angular - momentum and parity projections , we develop a method of antisymmetrized molecular dynamics ( amd ) @xcite . the time - dependent amd , which was originally developed for study of heavy - ion reactions @xcite , was applied to investigate e1 and monopole excitations @xcite . however , in the time - dependent amd approach , the angular - momentum and parity projections are not performed , and therefore , ground state structures and the coupling of single - particle excitations with the rotational motion are not sufficiently described . instead of the time - dependent amd , we superpose the angular - momentum and parity projected wave functions of various configurations including the 1p-1h and cluster excitations . we first perform the variation after the angular - momentum and parity projections in the amd framework ( amd+vap ) @xcite to obtain the ground state wave function . then we describe small amplitude motions by taking into account 1p-1h excitations on the obtained ground state wave function with the shifted amd method as done in ref . @xcite for monopole excitations of @xmath18o . to incorporate the large amplitude cluster motion , we combine the the generator coordinate method ( gcm ) with the shifted amd by superposing @xmath19 cluster wave functions . the angular - momentum and parity projections are performed in the present framework . applying the present method , i investigate the dipole transitions @xmath20 , @xmath21 , and @xmath22 . this paper is organized as follows . the present method of amd is formulated in section [ sec : formulation ] , and section [ sec : results ] discusses the ground state structures and the e1 and isoscalar dipole ( isd ) excitations in be isotopes . the paper concludes with a summary in section [ sec : summary ] . i apply the amd+vap method to obtain @xmath23-nucleon wave functions for the ground states of @xmath16be , @xmath0be , and @xmath1be . to investigate dipole excitations , i incorporate 1p-1h excitations on the ground state wave function with the shifted amd method . i also perform the @xmath2-cluster gcm calculation combined with the shifted amd to see how the @xmath2-cluster mode affects the dipole excitations . in this section , i explain the formulation of the amd+vap , the shifted amd , and the @xmath2-cluster gcm calculations , and also describe the definition of the dipole strengths . an amd wave function is given by a slater determinant , @xmath24 where @xmath25 is the antisymmetrizer . the @xmath26th single - particle wave function @xmath27 is written by a product of spatial , spin , and isospin wave functions as @xmath28 @xmath29 and @xmath30 are the spatial and spin functions , respectively , and @xmath31 is the isospin function fixed to be up ( proton ) or down ( neutron ) . the width parameter @xmath32 is fixed to be the optimized value for each nucleus . to separate the center of mass motion from the total wave function @xmath33 , the following condition should be satisfied , @xmath34 in the present calculation , i keep this condition and exactly remove the contribution of the center of mass motion . accordingly , an amd wave function is expressed by a set of variational parameters , @xmath35 , which specify centroids of single - nucleon gaussian wave packets and spin orientations for all nucleons . in the amd framework , existence of clusters is not assumed _ a priori _ because gaussian centroids , @xmath36 of all single - nucleon wave packets are independently treated as variational parameters . nevertheless , a multi - center cluster wave function can be described by the amd wave function with the corresponding configuration of gaussian centroids . it should be commented that the amd wave function is similar to the wave function used in fermionic molecular dynamics calculations @xcite . in the amd+vap method , the parameters @xmath37 in the amd wave function are determined by the energy variation after the angular - momentum and pariy projections ( vap ) . it means that , @xmath38 and @xmath39 for the lowest @xmath40 state are determined so as to minimize the energy expectation value of the hamiltonian for the @xmath40-projected amd wave function ; @xmath41 where @xmath42 is the angular - momentum and parity projection operator . after the vap calculation , the optimized parameters @xmath43 for the lowest @xmath40 state are obtained . for the ground state , the vap with @xmath44 and @xmath45 is performed for @xmath16be and @xmath1be , and that with @xmath46 and @xmath47 is done for @xmath0be . i rewrite the parameters @xmath43 obtained by the vap for the ground state as @xmath48 . to taken into account 1p-1h excitations , i consider small variation of single - particle wave functions in the ground state wave function @xmath49 by shifting the position of the gaussian centroid of the @xmath26th single - particle wave function , @xmath50 , where @xmath51 is a small constant and @xmath52 is an unit vector with the label @xmath53 . in the present calculation , @xmath54 for 8 directions are adopted to obtain the approximately converged result for the e1 and isd strengths . details of the adopted unit vectors @xmath52 ( @xmath55 ) are described in section [ sec : results ] . for the spin part , i consider the spin - nonflip single - particle state @xmath30 and the spin - flip state @xmath56 ( @xmath57 ) , @xmath58 where @xmath59 . for all single - particle wave functions , i consider spin - nonflip and spin - flip states shifted to eight directions independently and prepare @xmath60 amd wave functions , @xmath61 and @xmath62 , with the shifted parameters @xmath63 here @xmath64 is chosen to be @xmath65 to take into account the recoil effect so that the center of mass motion is separated exactly . those shifted amd wave functions @xmath66 and the original wave function @xmath67 are superposed to obtain the final wave functions for the ground and excited states , @xmath68 where the coefficients @xmath69 , @xmath70 , and @xmath71 are determined by diagonalization of the norm and hamiltonian matrices . i call this method `` the shifted amd '' ( samd ) . the spin - nonflip version of the samd has been applied to investigate monopole excitations of @xmath18o in ref . @xcite . i choose an enough small value of the spatial shift @xmath51 , typically @xmath72 fm , so as to obtain @xmath51-independent results . the model space of the samd contains the 1p-1h excitations that are written by a small shift of a single - nucleon gaussian wave function of the ground state wave function . in the intrinsic frame before the angular - momentum and parity projections , the ground state amd wave function is expressed by a slater determinant , and therefore , the samd method corresponds to the rpa in the restricted model space of the linear combination of shifted gaussian wave functions . however , since the projected states are superposed in the samd , the coupling of the 1p-1h excitations with the rotation and parity transformation is properly taken into account . therefore , the samd contains , in principle , higher correlations beyond the rpa in mean - field approximation . note that , the shift @xmath73 of the gaussian wave packet can be expressed by a linear combination of harmonic oscillator ( h.o . ) orbits at @xmath74 . for instance , when three vectors @xmath75 , @xmath76 , and @xmath77 are chosen for @xmath78 , the linear combination of @xmath29 , @xmath79 , @xmath80 , and @xmath81 in the small @xmath51 limit is equivalent to that of the @xmath82 and @xmath83 orbits around @xmath74 . it means that , in the case that the recoil effect is omitted , the samd can be regarded as an extended amd method , in which higher h.o . orbits are incorporated in addition to the default @xmath82 orbit at @xmath74 for the @xmath26th single - particle wave function . in the particular case of @xmath84 , it can be called `` @xmath85-wave amd '' . in the ground state wave functions obtained by the amd+vap for @xmath16be , @xmath0be , and @xmath1be , an @xmath2 cluster is formed even though any clusters are not _ a priori _ assumed in the framework . consequently , gaussian centroids @xmath86 for two protons and two neutrons are located at almost the same . the inter - cluster motion of @xmath87he+@xmath2 and @xmath9he+@xmath2 structures in @xmath0be and @xmath1be can be excited by the dipole operators . to incorporate the large amplitude @xmath2-cluster mode , we perform the @xmath2-cluster gcm ( @xmath2gcm ) calculation with respect to the inter - cluster distance . for simplicity , we label four nucleons composing the @xmath2 cluster as @xmath88 and other nucleons as @xmath89 . the center of mass position of the @xmath2 cluster is localized around @xmath90 $ ] . the inter - cluster distance @xmath91 is written as @xmath92\right|=\frac{a}{a-4}r_{\alpha}\ ] ] with @xmath93 . to perform the @xmath2gcm calculation based on the ground state wave function @xmath94 , i change the inter - cluster distance @xmath95 by shifting positions of single - nucleon gaussian centroids @xmath96 by hand as @xmath97 and superpose the wave functions with different @xmath98 values . i combine the @xmath2gcm with the samd and express the total wave function as @xmath99 where @xmath100 . the coefficients are determined by diagonalization of the norm and hamiltonian matrices . the e1 operator @xmath101 is given by the isovector dipole operator as @xmath102 the isd operator @xmath103 is defined as @xmath104 which excites the compressive dipole mode . the e1 and isd strengths for the transition @xmath105 are given by the matrix elements of the dipole operators as @xmath106 is the ground state angular momentum . the energy - weighted sum ( ews ) of the e1 and isd strengths is defined as @xmath107 where @xmath108 is the energy of the @xmath109 state . if the interaction commutes with the e1 operator , @xmath110 is identical to the trk sum rule : @xmath111 since the nuclear interaction does not commute with the e1 operator , @xmath110 is usually enhanced from @xmath112 . in the present framework , all excited states are discrete states without escaping widths because the out - going condition in the asymptotic region is not taken into account . i calculate the e1 and isd strengths for discrete states and smear the strengths with a gaussian by hand to obtain dipole strength functions as @xmath113 where @xmath114 is the smearing width . the photonuclear cross section is dominated by e1 transitions and related to the e1 strength function as @xmath115 i use an effective nuclear interaction consisting of the central force of the mv1 force@xcite and the spin - orbit force of the g3rs force @xcite , and the coulomb force . the mv1 force is given by a two - range gaussian two - body term and a zero - range three - body term . the g3rs spin - orbit force is a two - range gaussian force . the bartlett , heisenberg and majorana parameters for the case 3 of the mv1 force are @xmath116 and @xmath117 , and the strengths of the g3rs spin - orbit force are @xmath118 mev . these interaction parameters are the same as those used in refs . @xcite , in which the amd+vap calculation describes well properties of the ground and excited states of @xmath1be and @xmath119c . i perform the amd+vap calculation to obtain the ground state wave functions for @xmath16be(@xmath120 ) , @xmath0be(@xmath121 ) , and @xmath1be(@xmath120 ) . the width parameter is chosen to be @xmath122 @xmath123 for @xmath16be and @xmath0be and @xmath124 @xmath123 for @xmath1be to minimize the ground state energy . figure [ fig : beiso - dense](a ) shows intrinsic density distribution of the obtained wave functions @xmath49 for the ground states . as seen in the density , the @xmath125 , @xmath126 , and @xmath127 cluster structures are developed in @xmath16be , @xmath0be , and @xmath1be , respectively . considering that the @xmath128 and @xmath129 clusters have @xmath130 and @xmath131 structures , the ground states of @xmath0be and @xmath1be are regarded as the @xmath4 cluster core with valence neutrons , @xmath15 and @xmath132 , in which the valence neutrons are localized around one of the 2@xmath2 . as given in eq . , an amd wave function is expressed by a single slater determinant . however , the projected state @xmath133 for the ground state wave function contains higher correlations beyond mean - field approximations , which is efficiently incorporated by the vap calculation . indeed , cluster structures are remarkable in the present vap result but they are relatively suppressed in calculations without the projections . for comparison with the present result obtained by the vap ( variation after the angular - momentum and parity projections ) , the result obtained by the variation without the angular - momentum and parity projections and that after the parity projection without the angular - momentum projection are also demonstrated in fig . [ fig : beiso - dense](b ) and ( c ) , respectively . it is clearly seen that the result of @xmath1be obtained by the variation without the projections shows weak clustering with a parity - symmetric intrinsic structure ( see the right panels of fig . [ fig : beiso - dense](b ) and ( c ) ) . this indicates that the angular - momentum and parity projections in the energy variation is essential to obtain the parity - asymmetric structure with the @xmath9he+@xmath2 correlation in @xmath1be . the root mean square radii of point - proton distribution of @xmath16be , @xmath0be , and @xmath1be calculated by the amd+vap are 2.73 fm , 2.69 fm , and 2.43 fm , which are slightly larger than the experimental values , 2.39 fm and 2.22 fm , of @xmath0be and @xmath1be reduced from charge radii . the calculated magnetic and electric quadrupole moments of @xmath0be are @xmath134 @xmath135 and @xmath136 e @xmath137 , which reasonably agree to the experimental values , @xmath138 @xmath135 and @xmath139 e @xmath137 . be , @xmath0be , and @xmath1be obtained by ( a ) the amd+vap ( variation after the angular - momentum and parity projections ) , ( b ) the variation after the parity projection without the angular - momentum projection , and ( c ) the variation without the angular - momentum and parity projections . [ fig : beiso - dense],width=264 ] to investigate dipole excitations , i calculate @xmath140 states of @xmath16be and @xmath1be , and @xmath141 , @xmath142 , and @xmath143 states of @xmath0be by applying the samd and the samd+@xmath2gcm based on the obtained ground state wave functions . for the samd , the shift parameter @xmath51 is taken to be @xmath72 fm which is small enough to give the @xmath51-independent result . for unit vectors i choose three sets , @xmath144 , @xmath145 , and @xmath146 , and check the convergence of the dipole strengths . here , the set of 8 vectors are @xmath147 , and that of 14 vectors are @xmath148 , @xmath149 , @xmath150 , and @xmath151 . the @xmath152 , @xmath153 , and @xmath154 axes are taken to be the principle axes of the inertia of the intrinsic state that satisfy @xmath155 and @xmath156 . for the e1 strength , the samd model space with @xmath144 can cover @xmath157 configurations excited by the e1 operator . however , for the isd strength , a larger number of unit vectors ( @xmath158 ) are necessary for the samd model space to cover @xmath159 configurations excited by the isd operator . the e1 and isd strengths of @xmath160be and @xmath1be calculated by the samd in three cases , @xmath144 , @xmath145 , and @xmath146 , are shown in fig . [ fig : be - sp14 ] . as expected , the set @xmath144 is enough only for the e1 strength but not for the isd strength . it is found that the set @xmath145 is practically enough to get a qualitatively converged result for both the e1 and isd strengths , and therefore this set is adopted in the present calculation of the dipole strengths . for the @xmath2gcm calculation , the distance parameter is taken to be @xmath161 fm . this means that @xmath2-cluster continuum states are treated as discretized states in the box boundary @xmath162 fm . be and @xmath1be obtained by the samd in three cases of @xmath144 , @xmath145 , and @xmath146 . the smearing width is @xmath163 mev . [ fig : be - sp14],width=302 ] the model space of the samd+@xmath2gcm wave function given in eq . covers 1p-1h excitations and @xmath2-cluster excitations from the ground state wave function . for the detailed description of the low - lying energy spectra and their dipole transition strengths , i mix additional configurations optimized for the low - lying levels @xmath0be@xmath164 , @xmath0be@xmath165 , @xmath0be@xmath166 and @xmath1be@xmath167 , which are obtained by the amd+vap with @xmath141 , @xmath142 , and @xmath143 for @xmath0be and that with @xmath140 for @xmath1be . the final wave function with these additional vap configurations is given as @xmath168 the dipole strengths are calculated by the following matrix elements , @xmath169 the calculated dipole strengths with and without the additional vap configurations are found to be almost consistent with each other except for quantitative details of the energy position and the strengths in @xmath170 mev . in this paper , i mainly discuss the dipole strengths calculated by the samd and the samd+@xmath2gcm+cfg wave functions , which i call cal - i and cal - ii , respectively . the former corresponds to the small amplitude calculation containing 1p-1h excitations . the latter contains the large amplitude @xmath2-cluster mode in addition to the 1p-1h excitations described by the samd model space . namely , the samd+@xmath2gcm+cfg includes the higher correlation than the samd in both the ground and excited states . the energy - weighted e1 strength of @xmath0be and @xmath1be obtained by the samd ( cal - i ) and the samd+@xmath2gcm+cfg ( cal - ii ) is shown in fig . [ fig : beiso - e1-gcm ] . the strength functions of two calculations ( i ) and ( ii ) are qualitatively similar to each other except for broadening of the low - energy strength in @xmath171 mev of @xmath0be in the cal - ii . the calculated ews of the e1 strength is enhanced from the trk sum rule value by a factor @xmath172 . figure [ fig : be9-exp ] shows the comparison of the calculated e1 cross section with the experimental photonuclear cross sections of @xmath0be . the calculation reasonably describes the global feature of the experimental cross sections consisting of the low - energy strength in @xmath5 mev and the gdr in @xmath3 mev , though it somewhat overestimates the gdr peak energy and strength . be and @xmath1be obtained by the samd ( cal - i ) and the samd+@xmath2gcm+cfg ( cal - ii ) . the smearing width is @xmath163 mev . [ fig : beiso - e1-gcm],width=226 ] be with the experimental photonuclear cross sections . the calculated values are those obtained with the samd+@xmath2gcm+cfg ( cal - ii ) , smeared by @xmath163 mev . the experimental data are taken from the photonuclear cross sections by ahrens _ et al._@xcite , the bremsstrahlung data by goryachev _ et al._@xcite , and the photodisintegration cross sections by utsunomiya _ et al._@xcite . [ fig : be9-exp],width=226 ] be , @xmath0be , and @xmath1be obtained by the samd ( @xmath55 ) and the samd - z ( longitudinal mode : @xmath173 ) calculations . the samd and samd - z results for ( a)@xmath16be , ( b)@xmath0be , and ( d)@xmath1be are shown by solid and dashed lines , respectively . the decomposition of the strengths for @xmath141 , @xmath142 , and @xmath143 states of @xmath0be is also shown as well as the total strength in the panel ( c ) . the smearing width is @xmath163 mev . [ fig : beiso - e1],width=226 ] as shown in the comparison of the samd ( cal - i ) and the samd+@xmath2gcm+cfg ( cal - ii ) in fig . [ fig : beiso - e1-gcm ] , the e1 strength of @xmath0be and @xmath1be is not affected so much by the coupling with the large amplitude @xmath2-cluster motion . in the following , we give detailed analysis of the e1 strength of @xmath16be , @xmath0be , and @xmath1be based on the samd to discuss effects of excess neutrons on the e1 strength in @xmath0be and @xmath1be . figure [ fig : beiso - e1 ] shows the e1 strength of @xmath16be , @xmath0be , and @xmath1be calculated by the samd . for @xmath0be , decomposition of the transition strength to @xmath174 , @xmath142 , and @xmath143 states is also shown . the gdr in @xmath16be shows a two peak structure in @xmath175 mev . also in @xmath0be , the two peak structure of the gdr is seen but it somewhat broadens . in addition to the gdr , low - lying e1 strength appears in @xmath176 mev of @xmath0be . in @xmath1be , the lower peak of the gdr exists at @xmath177 mev , whereas the higher peak of the gdr is largely fragmented . below the gdr , an e1 resonance appears at @xmath8 mev . the origin of the two - peak structure of the gdr in be isotopes is the prolate deformation of the @xmath4 core . to distinguish the longitudinal mode in the intrinsic frame , we calculate the e1 strength in the truncated samd model space by using wave functions shifted only to the longitudinal ( @xmath154 ) direction , that is , the samd with the fixed @xmath173 as @xmath178 where the coefficients @xmath69 , @xmath70 , and @xmath71 are determined by diagonalization of the norm and hamiltonian matrices . here i omit the @xmath179-mixing and fix @xmath45 for @xmath180be and @xmath1be , and @xmath47 for @xmath160be to take into account only the @xmath181 mode in the intrinsic frame . the samd with @xmath173 , which i call `` samd - z '' , is approximately regarded as the calculation containing the longitudinal mode but no transverse mode , though two modes do not exactly decouple from each other because of the angular - momentum projection . the e1 strength obtained by the samd - z is shown by dashed lines in fig . [ fig : beiso - e1](a ) , ( b ) , and ( d ) . in comparison of the samd and samd - z results , it is found that the lower peak of the gdr at @xmath182 mev is contributed by the longitudinal mode of the @xmath4 core , whereas the higher peak of the gdr comes from the transverse mode . the higher peak broadens in @xmath0be and it is largely fragmented in @xmath1be indicating that the transverse mode is affected by excess neutrons . for the low - lying e1 resonances below the gdr , the strength at @xmath183 mev in @xmath160be and that at @xmath184 mev in @xmath1be are mainly contributed by the longitudinal mode . these low - energy dipole resonances in @xmath160be and @xmath1be are understood by the longitudinal motion of valence neutrons against the @xmath4 core . from the above analysis of the e1 strength of @xmath0be and @xmath1be compared with that of @xmath16be , the effects of excess neutrons on the e1 strength is understood as follows . the longitudinal and transverse dipole modes in the @xmath4 core part contribute to the gdr with the two peak structure . the valence neutron modes couple with the transverse dipole mode of the @xmath4 core and they broaden the higher peak of the gdr . moreover , the valence neutron modes against the @xmath4 core contribute to the low - energy e1 strength . more details of the low - energy dipole excitations are discussed later . as previously mentioned , the samd ( cal - i ) corresponds to the small amplitude calculation , whereas the samd+@xmath2gcm+cfg ( cal - ii ) contains the large amplitude @xmath2-cluster mode . a possible enhancement of the isd strength in the cal - ii relative to the cal - i can be a good probe for the dipole excitation that couples with the @xmath2-cluster mode , because the @xmath2-cluster excitation in @xmath0be and @xmath1be involves the compressive dipole mode the energy - weighted isd strength of @xmath0be and @xmath1be calculated by the cal - i and cal - ii is shown in fig . [ fig : be - isd ] . the strength of the isoscalar gdr in @xmath185 mev is not affected by the @xmath2-cluster mode , whereas the isd strength for some low - energy resonances are significantly enhanced in the cal - ii as a result of the coupling with the @xmath2-cluster mode . in @xmath0be , the isd strength in @xmath186 mev is remarkably enhanced in the cal - ii , whereas the resonance at @xmath187 mev has the weak isd strength in both the cal - i and cal - ii . in @xmath1be , the isd strength around @xmath188 mev is enhanced . be and @xmath1be obtained by the samd ( cal - i ) and the samd+@xmath2gcm+cfg ( cal - ii ) . e1 strength obtained by the the samd+@xmath2gcm+cfg ( cal - ii ) is also shown for comparison . the smearing width is @xmath163 mev . [ fig : be - isd],width=226 ] ccccc & ( i ) & ( ii ) & ( iii ) & ( iv ) + @xmath2 mode in initial & w / o & w & w & w / o + @xmath2 mode in final & w / o & w & w / o & w + + @xmath189 & 58 & 56 & 58 & 56 + @xmath190 & 0.13 & 0.43 & 0.39 & 0.16 + @xmath191 & 3.3 & 2.4 & 2.6 & 3.5 + @xmath192 & 2.7 & 3.3 & 3.3 & 2.6 + @xmath193 & @xmath194 & @xmath195 & @xmath194 & @xmath195 + @xmath196 & 93 & 410 & 58 & 124 + @xmath197 & 108 & 230 & 200 & 131 + @xmath198 & 490 & 560 & 640 & 520 + + @xmath189 & 63 & 62 & 63 & 62 + @xmath199 & 0.09 & 0.08 & 0.06 & 0.10 + @xmath200 & 10.4 & 8.7 & 7.6 & 10.9 + @xmath193 & @xmath201 & @xmath202 & @xmath201 & @xmath202 + @xmath203 & 162 & 157 & 115 & 187 + @xmath204 & 56 & 187 & 83 & 74 + from the analysis of the e1 and isd strengths , the low - energy dipole excitations below the gdr in @xmath0be can be categorized as three resonances in @xmath205 mev , @xmath206 mev , and @xmath207 mev , which i label a1 , a2 , and a3 resonances , respectively . the ews in the corresponding energy regions , @xmath208 , @xmath209 , @xmath210 , is listed in table [ tab : ews ] as well as the total ews value . in addition to the ews obtained by the samd ( cal - i ) and samd+@xmath2gcm+cfg ( cal - ii ) , the ews calculated by the matrix elements @xmath211 for the transitions from the samd+@xmath2gcm+cfg initial states to the samd final states ( cal - iii ) and that by the matrix elements @xmath212 for the transitions from the samd initial states to the samd+@xmath2gcm+cfg final states ( cal - iv ) are also shown in the table . the cal - iii contains the @xmath2-cluster mode in the initial states as the ground state correlation but not in the final states , whereas the cal - iv contains the @xmath2-cluster mode only in the final states but not in the initial states . in all the calculations ( i ) , ( ii ) , ( iii ) , and ( iv ) , @xmath190 is very small , whereas @xmath191 and @xmath192 are significantly large as @xmath21310% of the trk sum rule . consequently , the ews of the e1 strength in @xmath5 mev exhausts @xmath21320% of the trk sum rule and it is @xmath21310% of the calculated total ews . the @xmath2-cluster mode does not affect so much the e1 and isd strengths of a2 and those of a3 , but it gives significant enhancement of the dipole strengths of a1 . in particular , @xmath196 is remarkably enhanced by the coupling with the @xmath2-cluster mode . the enhancement is found only in the cal - ii but not in other calculations , cal - i , cal - iii , and cal - iv . it indicates that the coupling with the @xmath2-cluster mode in the ground state and that in the a1 resonance coherently enhance @xmath196 . the @xmath2-cluster mode also makes @xmath190 three times larger in the cal - ii than the cal - i , though it is still less than 2% of the trk sum rule . as seen in the ews of the samd+@xmath2gcm+cfg ( cal - ii ) in table [ tab : ews ] , the a1 resonance shows the relatively strong isd and weak e1 transitions , whereas the a2 resonance shows the relatively weak isd and strong e1 transitions . these characteristics of the a1 and a2 resonances can be understood by the @xmath15 picture as follows . the ground and low - lying states of @xmath0be are approximately described by the molecular orbital structure , where the valence neutron occupies molecular orbitals formed by the linear combination of @xmath85 orbits around @xmath2 clusters @xcite . let me consider two @xmath2 clusters at the left and right along the @xmath154 axis ( see fig . [ fig : orbital ] ) . i call the left(right ) @xmath2 `` @xmath214 '' , and label single - particle orbits ( atomic orbitals ) around each @xmath2 cluster as @xmath215_{\alpha_{l(r)}}$ ] . here @xmath216 is the total quantum ( node ) number , @xmath217 is the quantum number for the @xmath154-axis , and @xmath218 and @xmath219 are the @xmath154-components of the orbital- and total - angular momenta , respectively . the @xmath220 and @xmath221 molecular orbitals are given by the linear combination of the atomic orbitals @xmath222_{\alpha_{l , r}}$ ] as @xmath223_\textrm{mo}=[101\frac{3}{2}]_{\alpha_{l}}+ [ 101\frac{3}{2}]_{\alpha_{r}}\\ & & \pi^+_{3/2}\equiv[211\frac{3}{2}]_\textrm{mo}=[101\frac{3}{2}]_{\alpha_{l}}- [ 101\frac{3}{2}]_{\alpha_{r}}\\\end{aligned}\ ] ] where @xmath224_\textrm{mo}$ ] is the label indicating the quantum numbers @xmath216 , @xmath217 , @xmath218 , and @xmath219 of the molecular orbital around the @xmath4 . another molecular orbital is the longitudinal orbital @xmath225 given by the linear combination of @xmath226_{\alpha_{l , r}}$ ] as @xmath227_\textrm{mo } = [ 110\frac{1}{2}]_{\alpha_{l}}-[110\frac{1}{2}]_{\alpha_{r}}.\end{aligned}\ ] ] in the case that the @xmath2-@xmath2 distance is not so large , the molecular orbital @xmath224_\textrm{mo}$ ] approximately corresponds to the nilsson ( deformed shell - model ) orbit @xmath228_\omega$ ] with @xmath229 , @xmath230 , and @xmath231 in the prolate deformation . @xmath220 is the negative - parity orbital with no node ( @xmath232 ) , @xmath221 is the positive - parity orbital with one node ( @xmath233 ) , and @xmath225 is the positive - parity orbital with two nodes ( @xmath234 ) along the @xmath154-axis . for the valence neutron around the @xmath4 , @xmath220 is the lowest negative - parity molecular orbital , whereas @xmath225 is the lowest positive - parity orbital . the ground state of @xmath0be dominantly has the @xmath15 structure with the @xmath220 configuration . the a1 resonance is approximately described by the @xmath225 configuration , whereas the a2 resonance is dominated by the @xmath221 configuration ( see fig . [ fig : orbital ] ) . the e1 transition for @xmath235 , i.e. , @xmath222_\textrm{mo}\to [ 211\frac{3}{2}]_\textrm{mo}$ ] is possible because the @xmath181 operator changes @xmath236 and @xmath237 . however , the e1 transition for @xmath238 , i.e. , @xmath222_\textrm{mo}\to [ 220\frac{1}{2}]_\textrm{mo}$ ] is forbidden because the change @xmath239 is not possible for the e1 operator . this is the reason why the e1 strength is large for a2 but it is suppressed for a1 . because of the @xmath240_\textrm{mo}$ ] configuration of the a2 resonance , the e1 strength of a2 shows the @xmath47 band feature that the contribution from transitions to @xmath241 and @xmath143 states is dominant as seen in fig . [ fig : beiso - e1](c ) . the a1 resonance has the large node number @xmath234 along the @xmath4 direction than the a2 resonance ( @xmath233 ) , and therefore , the spatial development of the @xmath4 clustering is more prominent in the a1 resonance . as a result of the developed clustering , the a1 resonance couples rather strongly with the @xmath2-cluster mode . the coupling with the @xmath2-cluster mode , namely , the @xmath87he-@xmath2 relative motion in a1 enhances the isd strength as discussed previously . let me discuss low - energy dipole resonances in @xmath1be . the low - energy dipole strength below the gdr can be categorized as two resonances in @xmath242 mev and @xmath243 mev , which i label b1 and b2 , respectively . the ews of the dipole strengths for the corresponding energy regions are listed in table [ tab : ews ] . the b2 resonance shows the strong e1 transition exhausting more than 20% of the trk sum rule and 10% of the calculated total ews . it also shows the significant isd strength enhanced by the @xmath2-cluster mode in the cal - ii ( samd+@xmath2gcm+cfg ) . the significant e1 strength and the strong coupling with the @xmath2-cluster mode of the b2 resonance can be understood by two neutron correlation in the @xmath132 picture as shown in the schematic figures of fig . [ fig : orbital ] . the configuration in the ground state of @xmath1be is approximately described by the positive - parity projected state of the atomic orbital configuration as @xmath244_{\alpha_{l}}[10 - 1-\frac{3}{2}]_{\alpha_{l}}+ [ 101\frac{3}{2}]_{\alpha_{r } } [ 10 - 1-\frac{3}{2}]_{\alpha_{r}}\\ & & = \frac{1}{2}\left\ { ( \pi^-_{3/2})^2+(\pi^+_{3/2})^2 ) \right\},\end{aligned}\ ] ] which corresponds to the @xmath9he+@xmath2 cluster structure in the intrinsic state of the @xmath1be ground state . the b2 resonance is interpreted as the parity partner of the ground state as @xmath244_{\alpha_{l}}[10 - 1-\frac{3}{2}]_{\alpha_{l } } - [ 101\frac{3}{2}]_{\alpha_{r } } [ 10 - 1-\frac{3}{2}]_{\alpha_{r}}\\ & & = \frac{1}{2}\left\ { \pi^-_{3/2}\pi^+_{3/2 } + \pi^+_{3/2}\pi^-_{3/2}\right\}.\end{aligned}\ ] ] in the transition from the ground state to the b2 resonance , the coherent contribution of two neutrons enhances the e1 strength . the b2 resonance has a large overlap with negative - parity @xmath9he+@xmath2 cluster wave functions , for instance , 60% overlap with @xmath245 . as a result of the strongly coupling with the @xmath2-cluster mode , the isd strength of the b2 resonance is enhanced . in other words , the b2 resonance is regarded as the @xmath2-cluster excitation on the ground state which already contains the @xmath9he+@xmath2 cluster structure . in contrast to the b2 resonance , the b1 resonance is regarded as a single - particle excitation with the molecular orbital configuration @xmath246 and has no coherent contribution of two neutrons to the e1 strength . moreover , the @xmath246 configuration contains the @xmath87he+@xmath87he and @xmath9he@xmath247+@xmath2 components instead of the @xmath9he+@xmath2 component , and therefore , it does not couple with the @xmath2-cluster mode . finally , i discuss the calculated @xmath248 of the @xmath249 , @xmath250 , and @xmath251 states of @xmath0be , which contribute to the dipole strengths of the a1 resonance . in fig . [ fig : be9-e1 ] , i show the e1 strength in @xmath252 mev of @xmath0be . here the smearing width is chosen to be @xmath253 mev to resolve discrete states . the @xmath249 , @xmath250 , and @xmath251 states are obtained as discrete states in the samd ( cal - i ) . in the samd+@xmath2gcm+cfg ( cal - ii ) , the @xmath249 and @xmath251 states are still discrete states , however , the @xmath250 shows a resonance behavior coupling with the discretized continuum states in the box boundary at @xmath254 fm . we evaluate the @xmath248 of the @xmath250 resonance by a sum of the e1 strength in @xmath255 mev and estimate the excitation energy by the @xmath248 weighted averaged energy of @xmath142 states in this energy region . in table [ tab : be1 ] , i compare the calculated e1 strength of the @xmath249 , @xmath250 , and @xmath251 states with the experimental values measured by the @xmath256 cross sections in ref . @xcite . the excitation energies and @xmath248 of the @xmath0be(@xmath251 ) and @xmath0be(@xmath250 ) obtained by the samd+@xmath2gcm+cfg ( cal - ii ) reproduce reasonably the experimental data . however , the calculated @xmath248 of the @xmath0be(@xmath249 ) is quite small inconsistently to the experimental data . the @xmath249 state has been suggested to be a virtual or resonance state of a @xmath257-wave neutron @xcite . however , the present model is insufficient to describe such the virtual or @xmath257-wave resonance state because the valence neutron motion far from the @xmath4 is not taken into account in the model space . , @xmath142 , and @xmath143 states of @xmath0be in @xmath258 mev obtained by ( a ) the samd ( cal - i ) and ( b ) the samd+@xmath2gcm+cfg ( cal - ii ) . the smearing width is @xmath253 mev . [ fig : be9-e1],width=226 ] . [ tab : be1 ] excitation energy and e1 strength of @xmath0be(@xmath249 ) , @xmath0be(@xmath250 ) , @xmath0be(@xmath251 ) , and @xmath1be(@xmath259 ) . the energy ( mev ) and @xmath248 ( @xmath137 ) calculated by the samd ( cal - i ) and the samd+@xmath2gcm+cgf ( cal - ii ) are listed compared with the experimental data . the experimental values of @xmath0be are data measured by the @xmath256 cross sections in ref . @xcite . the experimental excitation energy of @xmath1be(@xmath260 ) is taken from refs . @xcite . [ cols="^,^,^,^,^,^,^ " , ] and valence neutrons for the ground state and the a1 and a2 resonances of @xmath0be , and those for the ground state and the b1 and b2 resonances of @xmath1be . [ fig : orbital],width=283 ] i investigated the isovector and isoscalar dipole excitations in @xmath0be and @xmath1be with the shifted amd combined with the @xmath2-cluster gcm , in which the 1p-1h excitations on the ground state and the large amplitude @xmath2-cluster mode are incorporated . since the angular - momentum and parity projections are done , the coupling of excitations in the intrinsic frame with the rotation and parity transformation is taken into account microscopically . the low - energy e1 resonances appear in @xmath5 mev because of valence neutron modes against the @xmath4 core . they exhaust about @xmath6 of the trk sum rule and @xmath7 of the calculated ews . the gdr shows the two peak structure which is understood by the e1 excitations in the 2@xmath2 core part with the prolate deformation . the higher peak of the gdr for the transverse mode broadens in @xmath0be and it is largely fragmented in @xmath1be because of excess neutrons . by comparing the results of the shifted amd combined with and without the @xmath2-cluster gcm , i investigated how the e1 and isd strengths in @xmath0be and @xmath1be are affected by the large amplitude @xmath2-cluster mode . the isd strength is a good probe to identify the dipole resonances that couples with the @xmath2-cluster mode because the @xmath2-cluster mode in @xmath0be and @xmath1be involves the compressive dipole mode . it was found that the isd strength for some low - energy resonances in @xmath0be and @xmath1be are enhanced by the coupling with the @xmath2-cluster mode , whereas the e1 strength is not so sensitive to the coupling with the @xmath2-cluster mode . in @xmath0be , the isd strength of the low - energy resonance in @xmath186 mev is remarkable . in @xmath1be , the isd strength at @xmath8 mev is enhanced by the coupling with the @xmath2-cluster mode . this resonance at @xmath8 mev in @xmath1be is regarded as the @xmath2-cluster excitation on the ground state having the @xmath9he+@xmath2 structure and can be interpreted as the parity partner of the ground state . the e1 transition of this resonance is also strong because of the coherent contribution of two valence neutrons . the calculated e1 strength of @xmath0be reasonably describes the global feature of experimental photonuclear cross sections consisting of the low - energy strength in @xmath5 mev and the gdr in @xmath3 mev , though it somewhat overestimates the gdr peak energy and its strength . for the low - lying positive - parity states of @xmath0be , the calculated excitation energies and @xmath248 of the @xmath0be(@xmath251 ) and @xmath0be(@xmath250 ) reasonably agree to the experimental data . however , the calculation fails to reproduce the experimental @xmath248 of the @xmath0be(@xmath249 ) because the present model is insufficient to describe the detailed asymptotic behavior of the @xmath257-wave neutron in the @xmath249 state . the author would like to thank dr . utsunomiya and dr . kikuchi for fruitful discussions . the computational calculations of this work were performed by using the supercomputer in the yukawa institute for theoretical physics , kyoto university . this work was supported by jsps kakenhi grant number 26400270 . t. kobayashi _ . lett . * b232 * , 51 ( 1989 ) . k. ieki _ et al . _ , phys . lett . * 70 * , 730 ( 1993 ) . d. sackett _ et al . _ , phys . c * 48 * , 118 ( 1993 ) . s. shimoura _ et al . _ , phys . b * 348 * , 29 ( 1995 ) . m. zinser _ et al . _ , nucl . a * 619 * , 151 ( 1997 ) . t. aumann _ et al . _ , phys . c * 59 * , 1252 ( 1999 ) . t. nakamura _ et al . _ , phys . lett . * 96 * , 252502 ( 2006 ) . r. kanungo _ et al . _ , phys . * 114 * , no . 19 , 192502 ( 2015 ) . g. bertsch and j. foxwell , phys . c * 41 * , 1300 ( 1990 ) [ phys . c * 42 * , 1159 ( 1990 ) ] . y. suzuki and y. tosaka , nucl . phys . a * 517 * , 599 ( 1990 ) . m.honma , h.sagawa prog . . phys . * 84 * , 494 ( 1990 ) . g. f. bertsch and h. esbensen , annals phys . * 209 * , 327 ( 1991 ) . h.sagawa , n.takigawa , nguyen van giai nucl.phys . * a543 * , 575 ( 1992 ) . e. garrido , d. v. fedorov and a. s. jensen , nucl . a * 708 * , 277 ( 2002 ) . t. myo , s. aoyama , k. kato and k. ikeda , phys . lett . b * 576 * ( 2003 ) 281 . l. v. chulkov et al . , nucl . phys . a 759 , 23 ( 2005 ) . c. a. bertulani and m. s. hussein , phys . c * 76 * , 051602 ( 2007 ) . k. hagino and h. sagawa , phys . c * 76 * , 047302 ( 2007 ) . k. hagino , h. sagawa , t. nakamura and s. shimoura , phys . c * 80 * , 031301 ( 2009 ) . d. baye , p. capel , p. descouvemont and y. suzuki , phys . c * 79 * , 024607 ( 2009 ) . y. kikuchi , k. kato , t. myo , m. takashina and k. ikeda , phys . c * 81 * , 044308 ( 2010 ) . e. c. pinilla , p. descouvemont and d. baye , phys . c * 85 * , 054610 ( 2012 ) . y. kikuchi , t. myo , k. kato and k. ikeda , phys . c * 87 * , no . 3 , 034606 ( 2013 ) . m. tohyama , phys.lett . * 323b * , 257 ( 1994 ) . i.hamamoto , h.sagawa and x.z.zhang , phys.rev . c * 53 * , 765 ( 1996 ) . i.hamamoto and h.sagawa , phys.rev . c * 53 * , r1492 ( 1996 ) . h. sagawa and t. suzuki , phys . c * 59 * , 3116 ( 1999 ) . g. colo and p. f. bortignon , nucl . phys . a * 696 * , 427 ( 2001 ) . t. nakatsukasa and k. yabana , phys . c * 71 * , 024301 ( 2005 ) . y. kanada - enyo and m. kimura , phys . c * 72 * , 064301 ( 2005 ) . m. matsuo , nucl . a * 696 * , 371 ( 2001 ) . d. vretenar , n. paar , p. ring and g. a. lalazissis , nucl . a * 692 * , 496 ( 2001 ) . s. goriely and e. khan , nucl . a * 706 * , 217 ( 2002 ) . n. paar , p. ring , t. niksic and d. vretenar , phys . c * 67 * , 034312 ( 2003 ) . n. tsoneva , h. lenske and c. stoyanov , phys . b * 586 * , 213 ( 2004 ) . m. matsuo , k. mizuyama and y. serizawa , phys . c * 71 * , 064326 ( 2005 ) . j. piekarewicz , phys . c * 73 * , 044325 ( 2006 ) . j. terasaki and j. engel , phys . c * 74 * , 044301 ( 2006 ) . j. liang , l. g. cao and z. y. ma , phys . c * 75 * , 054320 ( 2007 ) . n. tsoneva and h. lenske , phys . c * 77 * , 024321 ( 2008 ) . n. paar , d. vretenar , e. khan and g. colo , rept . phys . * 70 * , 691 ( 2007 ) . k. yoshida and n. van giai , phys . c * 78 * , 014305 ( 2008 ) . g. co , v. de donno , c. maieron , m. anguiano and a. m. lallena , phys . rev . c * 80 * , 014308 ( 2009 ) . m. martini , s. peru and m. dupuis , phys . c * 83 * , 034309 ( 2011 ) . t. inakura , t. nakatsukasa and k. yabana , phys . c * 84 * , 021302 ( 2011 ) . x. roca - maza , g. pozzi , m. brenna , k. mizuyama and g. colo , phys . c * 85 * , 024601 ( 2012 ) . s. ebata , t. nakatsukasa and t. inakura , phys . c * 90 * , no . 2 , 024303 ( 2014 ) . s. bacca , n. barnea , g. hagen , m. miorelli , g. orlandini and t. papenbrock , phys . c * 90 * , no . 6 , 064619 ( 2014 ) . j. piekarewicz , phys . c * 83 * , 034319 ( 2011 ) . a. carbone , g. colo , a. bracco , l. g. cao , p. f. bortignon , f. camera and o. wieland , phys . c * 81 * , 041301 ( 2010 ) . reinhard and w. nazarewicz , phys . c * 87 * , no . 1 , 014324 ( 2013 ) . t. inakura , t. nakatsukasa and k. yabana , phys . c * 88 * , no . 5 , 051305 ( 2013 ) . k. govaert , f. bauwens , j. bryssinck , d. de frenne , e. jacobs , w. mondelaers , l. govor and v. y. ponomarev , phys . c * 57 * , 2229 ( 1998 ) . herzberg _ et al . _ , phys . c * 60 * , 051307 ( 1999 ) . a. leistenschneider _ et al . _ , phys . lett . * 86 * , 5442 ( 2001 ) . n. ryezayeva _ et al . _ , phys . lett . * 89 * , 272502 ( 2002 ) . e. tryggestad _ , phys . c * 67 * , 064309 ( 2003 ) . p. adrich _ et al . _ , phys . lett . * 95 * , 132501 ( 2005 ) . j. gibelin _ et al . _ , phys . * 101 * , 212503 ( 2008 ) . a. klimkiewicz _ et al . _ , c * 76 * , 051603 ( 2007 ) . r. schwengner _ et al . _ , c * 78 * , 064314 ( 2008 ) . o. wieland _ et al . _ , phys . lett . * 102 * , 092502 ( 2009 ) . j. endres _ et al . _ , phys . c * 80 * , 034302 ( 2009 ) . j. endres _ et al . _ , phys . lett . * 105 * , 212503 ( 2010 ) . a. tamii _ et al . _ , phys . * 107 * , 062502 ( 2011 ) . b. l. berman and s. c. fultz , rev . phys . * 47 * , 713 ( 1975 ) . h. utsunomiya , y. yonezawa , h. akimune , t. yamagata , m. ohta , m. fujishiro , h. toyokawa and h. ohgaki , phys . c * 63 * , 018801 ( 2001 ) . c. w. arnold , t. b. clegg , c. iliadis , h. j. karwowski , g. c. rich , j. r. tompkins and c. r. howell , phys . c * 85 * , 044605 ( 2012 ) . h. ustunomiya _ et al . c , submitted . s. okabe , y. abe and h. tanaka , prog . * 57 * , 866 ( 1977 ) . s. okabe and y. abe , prog . . phys . * 59 * , 315 ( 1978 ) . a. c. fonseca , j. revai , and a. matveenko , nucl a326 * , 182 ( 1979 ) . p. descouvemont , phys . c * 39 * , 1557 ( 1989 ) . k. arai , y. ogawa , y. suzuki and k. varga , phys . c * 54 * , 132 ( 1996 ) . v. d. efros and j. m. bang , eur . j. a * 4 * , 33 ( 1999 ) . k. arai , p. descouvemont , d. baye and w. n. catford , phys . c * 68 * , 014310 ( 2003 ) . o. burda , p. von neumann - cosel , a. richter , c. forssen and b. a. brown , phys . c * 82 * , 015808 ( 2010 ) . e. garrido , d. v. fedorov and a. s. jensen , phys . b * 684 * , 132 ( 2010 ) . r. alvarez - rodriguez , a. s. jensen , e. garrido and d. v. fedorov , phys . c * 82 * , 034001 ( 2010 ) . v. d. efros _ et al . _ [ european centre for theoretical studies in nuclear physics and related areas collaboration ] , phys . c * 89 * , no . 2 , 027301 ( 2014 ) . m. odsuren , y. kikuchi , t. myo , m. aikawa and k. kato , phys . c * 92 * , no . 1 , 014322 ( 2015 ) . a. ono , h. horiuchi , t. maruyama and a. ohnishi , phys . lett . * 68 * , 2898 ( 1992 ) . a. ono , h. horiuchi , t. maruyama and a. ohnishi , prog . * 87 * , 1185 ( 1992 ) . y. kanada - enyo , h. horiuchi and a. ono , phys . c * 52 * , 628 ( 1995 ) . y. kanada - enyo and h. horiuchi , phys . c * 52 * , 647 ( 1995 ) . y. kanada - enyo and h. horiuchi , prog . . suppl . * 142 * , 205 ( 2001 ) . t. furuta , k. h. o. hasnaoui , f. gulminelli , c. leclercq and a. ono , phys . c * 82 * ( 2010 ) 034307 . y. kanada - enyo , phys . lett . * 81 * , 5291 ( 1998 ) . y. kanada - enyo , h. horiuchi and a. dote , phys . c * 60 * , 064304 ( 1999 ) . y. kanada - enyo , prog . phys . * 117 * , 655 ( 2007 ) [ prog . phys . * 121 * , 895 ( 2009 ) ] . y. kanada - enyo , phys . rev . c * 89 * , 024302 ( 2014 ) . h. feldmeier , k. bieler and j. schnack , nucl . a * 586 * , 493 ( 1995 ) . t. neff and h. feldmeier , nucl . a * 713 * , 311 ( 2003 ) .

we compute the dominant two - loop electroweak correction , of , to the partial width of the decay of an intermediate - mass higgs boson into a pair of photons . we use the asymptotic - expansion technique in order to extract the leading dependence on the top - quark mass plus four expansion terms that describe the dependence on the- and higgs - boson masses . this correction reduces the born result by approximately 2.5% . as a by - product of our analysis , we also recover the correction to the partial width of the higgs - boson decay to two gluon jets . pacs numbers : 12.15.ji , 12.15.lk , 14.80.bn

hep-ph0405232

among the main tasks of the current experiments at the fermilab tevatron and the future experiments at the cern large hadron collider ( lhc ) is the search for the higgs boson , which is the only missing particle in the standard model ( sm ) . the electroweak precision data mainly collected at cern lep and slac slc in combination with the direct top - quark mass measurement at the tevatron favour a light higgs boson with mass @xmath2 gev with an upper bound of about 237 gev at the 95% confidence level @xcite . the allowed mass range is compatible with the so - called intermediate mass range , defined by @xmath3 . in this mass range , the decay into two photons represents one of the most useful detection modes at hadron colliders . since there is no direct coupling of the higgs boson to photons , the process @xmath4 is loop - induced . in the limit of vanishing bottom - quark mass , one distinguishes at lowest order the contributions from virtual top quarks and @xmath1 bosons , where , in covariant gauge , the latter are accompanied by charged goldstone bosons ( @xmath5 ) and faddeev - popov ghosts ( @xmath6 ) . some sample feynman diagrams are depicted in fig . [ 1loop ] . the corresponding contributions have been evaluated for the first time in ref . @xcite ( for reviews , see also ref . qcd corrections , which only affect the top - quark diagrams are known at the two- @xcite and three - loop @xcite orders . recently , also the two - loop electroweak correction induced by light - fermion loops has been evaluated @xcite . in this paper , we compute the two - loop electroweak correction that is enhanced by @xmath7 . for this purpose , we consider the ( formal ) hierarchy @xmath8 and apply the method of asymptotic expansion @xcite , which allows us to also evaluate four expansion terms in the ratio @xmath9 aside from the leading term in @xmath10 . due to electromagnetic gauge invariance , the amputated transition - matrix element of @xmath11 possesses the structure @xmath12 where @xmath13 and @xmath14 are the lorentz indices of the external photons with four - momenta @xmath15 and @xmath16 , respectively . thus , the decay rate of the higgs boson into two photons is given by @xmath17 the form factor @xmath18 is evaluated in perturbation theory as @xmath19 where @xmath20 and @xmath21 denote the one - loop contributions induced by virtual top quarks and @xmath1 bosons , respectively , @xmath22 stands for the two - loop electroweak correction involving virtual top quarks , and the ellipsis represents the residual one- and two - loop contributions as well as all contributions involving more than two loops . in the practical calculation , it is convenient to project out the scalar amplitudes that multiply the basic lorentz tensors @xmath23 , @xmath24 , and @xmath25 . the corresponding projectors can be obtained by in turn contracting @xmath26 with these lorentz tensors and solving the resulting system of linear equations . we separately project out the coefficients of the tensors @xmath27 and @xmath28 , and thus have a strong check on our calculation . furthermore , we adopt a general @xmath29 gauge in our calculation and verify that the gauge parameter drops out in the final result . for simplicity , the element @xmath30 of the cabibbo - kobayashi - maskawa quark mixing matrix is set to unity , so that the quarks of the third fermion generation decouple from those of the first two , which they actually do to very good approximation @xcite . ( 370,180)(4,2 ) ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/straight1 ( 9.20801,13.1807)[br]@xmath33 ( 6.5,10.)(13.,6.)(0.,)/straight-1 ( 9.20801,6.81927)[tr]@xmath33 ( 13.,14.)(13.,6.)(0.,)/straight1 ( 14.274,10.)[l]@xmath33 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/scalardash1 ( 9.20801,13.1807)[br]@xmath5 ( 6.5,10.)(13.,6.)(0.,)/scalardash-1 ( 9.20801,6.81927)[tr]@xmath5 ( 13.,14.)(13.,6.)(0.,)/scalardash1 ( 14.274,10.)[l]@xmath5 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/ghostdash1 ( 9.20801,13.1807)[br]@xmath34 ( 6.5,10.)(13.,6.)(0.,)/ghostdash-1 ( 9.20801,6.81927)[tr]@xmath34 ( 13.,14.)(13.,6.)(0.,)/ghostdash1 ( 14.274,10.)[l]@xmath34 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/sine1 ( 9.20801,13.1807)[br]@xmath1 ( 6.5,10.)(13.,6.)(0.,)/sine-1 ( 9.20801,6.81927)[tr]@xmath1 ( 13.,14.)(13.,6.)(0.,)/sine1 ( 14.274,10.)[l]@xmath1 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/scalardash-1 ( 9.20801,13.1807)[br]@xmath5 ( 6.5,10.)(13.,6.)(0.,)/scalardash1 ( 9.20801,6.81927)[tr]@xmath5 ( 13.,14.)(13.,6.)(0.,)/sine-1 ( 14.274,10.)[l]@xmath1 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 0.,10.)(7.,10.)(0.,)/scalardash0 ( 3.5,9.18)[t]@xmath31 ( 20.,15.)(13.,10.)(0.,)/sine0 ( 16.0791,13.2813)[br]@xmath32 ( 20.,5.)(13.,10.)(0.,)/sine0 ( 16.9209,8.28129)[bl]@xmath32 ( 7.,10.)(13.,10.)(0.8,)/scalardash-1 ( 10.,6.53)[t]@xmath5 ( 7.,10.)(13.,10.)(-0.8,)/scalardash1 ( 10.,13.47)[b]@xmath5 ( 7.,10.)0 ( 13.,10.)0 ( 0.,10.)(7.,10.)(0.,)/scalardash0 ( 3.5,9.18)[t]@xmath31 ( 20.,15.)(13.,10.)(0.,)/sine0 ( 16.0791,13.2813)[br]@xmath32 ( 20.,5.)(13.,10.)(0.,)/sine0 ( 16.9209,8.28129)[bl]@xmath32 ( 7.,10.)(13.,10.)(0.8,)/sine-1 ( 10.,6.53)[t]@xmath1 ( 7.,10.)(13.,10.)(-0.8,)/sine1 ( 10.,13.47)[b]@xmath1 ( 7.,10.)0 ( 13.,10.)0 ( 0.,10.)(10.5,8.)(0.,)/scalardash0 ( 5.00675,8.20296)[t]@xmath31 ( 20.,15.)(12.5,13.5)(0.,)/sine0 ( 15.946,15.2899)[b]@xmath32 ( 20.,5.)(10.5,8.)(0.,)/sine0 ( 14.7832,5.50195)[t]@xmath32 ( 12.5,13.5)(10.5,8.)(0.8,)/scalardash-1 ( 8.32332,12.0797)[r]@xmath5 ( 12.5,13.5)(10.5,8.)(-0.8,)/sine1 ( 14.6767,9.4203)[l]@xmath1 ( 12.5,13.5)0 ( 10.5,8.)0 our paper is organized as follows . in section [ sec::1loop ] , we illustrate the usefulness of the asymptotic - expansion technique by redoing the one - loop calculation . in section [ sec::2loop ] , we discuss the two - loop calculation . section [ sec::numerics ] contains the discussion of the numerical results . we conclude with a summary in section [ sec::summary ] . the application of the asymptotic - expansion technique to the one - loop diagrams , some of which are shown in fig . [ 1loop ] , leads to a naive taylor expansion in the external momenta @xmath35 and @xmath36 . nevertheless , we use already here a completely automated set - up , which consists in the successive use of the computer programs ` qgraf ` @xcite , ` q2e ` @xcite , ` exp ` @xcite , and ` matad ` @xcite . first , ` qgraf ` is used to generate the feynman diagrams . its output is then rewritten by ` q2e ` to be understandable by ` exp ` . in the two - loop case , the latter performs the asymptotic expansion and generates the relevant subdiagrams according to the rules of the so - called hard - mass procedure @xcite . ` form ` @xcite files are generated , which can be read by ` matad ` @xcite , which performs the very calculation of the diagrams . the analytic expression for the born result can be found in refs . @xcite . for completeness , we list it here in closed form and as an expansion for @xmath37 , which we reproduce . one has @xmath38 \right\ } \nonumber\\ & = & \hat{\mathcal{a } } n_c q_t^2 \left ( \frac{2}{3 } + \frac{7}{45 } \tau_{t } + \frac{4}{63 } \tau_{t}^{2 } + \frac{52}{1575 } \tau_{t}^{3 } + \frac{1024}{51975 } \tau_{t}^{4 } + \frac{2432}{189189 } \tau_{t}^{5 } + \ldots \right ) , \\ \mathcal{a}_{w}^{(0 ) } & = & \hat{\mathcal{a } } \left\ { - \frac{1}{2 } \left [ 2 + \frac{3}{\tau_{w } } + \frac{3}{\tau_{w } } \left ( 2 - \frac{1}{\tau_{w } } \right ) \arcsin^{2}\sqrt{\tau_{w } } \right ] \right\ } \nonumber\\ & = & \hat{\mathcal{a } } \left ( - \frac{7}{2 } - \frac{11}{15 } \tau_{w } - \frac{38}{105 } \tau_{w}^{2 } - \frac{116}{525 } \tau_{w}^{3 } - \frac{2624}{17325 } \tau_{w}^{4 } - \frac{640}{5733 } \tau_{w}^{5 } + \ldots \right ) , \label{eq::born}\end{aligned}\ ] ] where @xmath39 , @xmath40 , and @xmath41 is defined in section [ sec::intro ] . here , @xmath42 is sommerfeld s fine - structure constant , @xmath43 is fermi s constant , @xmath44 is the number of quark colours , and @xmath45 is the electric charge of the top quark in units of the positron charge . the higgs - boson mass entering the expansion parameters @xmath46 and @xmath41 partly arises from the couplings involving one higgs and two goldstone bosons , but also from the expansion in the external momenta due to the kinematic relation @xmath47 . we wish to note that the result for the process @xmath48 is simply obtained by setting @xmath49 and performing the substitution @xmath50 , where @xmath51 is the strong - coupling constant . in the remainder of this section , let us discuss the counterterm contributions needed for our two - loop analysis . we adopt the on - mass - shell scheme and regularize the ultraviolet divergences by means of dimensional regularization , with @xmath52 space - time dimensions and t hooft mass scale @xmath13 . we use the anti - commuting definition of @xmath53 . as will become clearer in the next section , it is important to treat the tadpole contributions properly in our calculation . for this reason , in the following , we list the corresponding contributions separately and mark them by the superscript `` tad '' . inspection of the one - loop diagrams reveals that , to @xmath54 , we have to renormalize the higgs - boson wave function and mass , the @xmath1-boson mass , and the top - quark mass . the corresponding counterterms are defined through @xmath55 note that @xmath56 is obtained from the derivative of the higgs - boson self - energy and thus has no tadpole contribution . the mass counterterms are obtained from the corresponding two - point functions , where only the @xmath57-enhanced contributions have to be considered . up to and including quadratic terms in @xmath57 , we have @xmath58 , \nonumber\\ \frac{\delta m_{h}^{2,tad}}{m_{h}^{2 } } & = & 12 n_{c } x_t \frac{m_t^2}{m_{h}^{2 } } \left ( \delta - \ln\frac{m_{t}^{2}}{\mu^{2 } } + 1\right ) , \nonumber\\ \frac{\delta m_{w}^{2}}{m_{w}^{2 } } & = & -2 n_c x_t \left ( \delta - \ln\frac{m_{t}^{2}}{\mu^{2 } } + \frac{1}{2 } \right ) , \nonumber\\ \frac{\delta m_{w}^{2,tad}}{m_{w}^{2 } } & = & 8 n_c x_t \frac{m_{t}^{2}}{m_{h}^{2 } } \left ( \delta - \ln\frac{m_{t}^{2}}{\mu^{2 } } + 1 \right ) , \nonumber\\ \frac{\delta m_{t}}{m_{t } } & = & \frac{3}{2 } x_t \left ( \delta - \ln\frac{m_{t}^{2}}{\mu^{2 } } + \frac{8}{3 } \right ) , \nonumber\\ \frac{\delta m_{t}^{tad}}{m_{t } } & = & 4 n_c x_t \frac{m_{t}^{2}}{m_{h}^{2 } } \left ( \delta - \ln\frac{m_{t}^{2}}{\mu^{2}}+ 1 \right ) , \label{eq : ct}\end{aligned}\ ] ] with @xmath59 and @xmath60 , where @xmath61 is euler s constant . the @xmath1-boson mass renormalization is needed for the @xmath1 , @xmath5 , and @xmath6 propagators , where @xmath62 enters as a parameter . furthermore , also the @xmath63 , @xmath64 , @xmath65 , and @xmath66 vertices contain @xmath62 . the only vertex involving @xmath67 is @xmath68 , which induces two - loop contributions via @xmath69 . finally , @xmath57 occurs in the top - quark propagator and in the @xmath70 vertex . the wave function renormalization and the renormalization of a factor @xmath71 is common to all one - loop diagrams . this allows for the definition of a universal factor @xcite , which is finite in our calculation . it is given by @xmath72 the contributions of @xmath54 are obtained by considering all two - loop electroweak diagrams involving a virtual top quark . this includes also the tadpole diagrams with a closed top - quark loop , which are proportional to @xmath73 . for arbitrary gauge parameter , this leads us to consider a total of @xmath74 diagrams . some of them are depicted in fig . [ 2loop ] . these diagrams naturally split into two classes . the first class consists of those diagrams where a neutral boson , i.e. a higgs boson or a neutral goldstone boson ( @xmath75 ) , is added to the one - loop top - quark diagrams . the exchange of a @xmath76 boson does not produce quadratic contributions in @xmath57 . the application of the asymptotic - expansion technique to these diagrams leads to a simple taylor expansion in the external momenta . this is different for the second class of diagrams , which , next to the top quark , also contain a @xmath1 or @xmath5 boson and , as a consequence , also the bottom quark , which we take to be massless throughout the calculation . due to the presence of cuts through light - particle lines , the asymptotic - expansion technique applied to these diagrams also yields nontrivial terms , as is exemplified in fig . [ mt4asex ] . the first contribution on the right - hand side of the equation in fig . [ mt4asex ] symbolizes the naive taylor expansion in the external momenta . in the second contribution , the subdiagram to the right of the star has to be expanded in its external momenta , which also includes the loop momentum of the co - subgraph to the left of the star . the result of the expansion is inserted as an effective vertex , and the remaining integration over the second loop - momentum is performed after a further expansion in @xmath35 and @xmath36 . the latter is allowed , since we work in the limit @xmath77 . we wish to mention that such contributions develop @xmath73 terms , which cancel in the final result only in combination with the @xmath73 terms from the counterterms of eq . ( [ eq : ct ] ) and the genuine two - loop tadpole diagrams . for this reason , it is crucial to include the latter in our calculation . as a further comment , we note that , unlike the example of fig . [ mt4asex ] , it can also happen that the co - subgraph only involves massless bottom quarks , and no expansion in the external momenta is allowed . in our calculation , the contributions from such co - subgraphs vanish . ( 370,400)(4,4 ) ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 11.,7.1)(5.5,10.)(0.,)/straight1 ( 8.00707,7.65892)[tr]@xmath33 ( 11.,7.1)(15.,5.)(0.,)/scalardash1 ( 12.7595,5.15763)[tr]@xmath5 ( 11.,12.95)(11.,7.1)(0.,)/straight1 ( 12.07,10.025)[l]@xmath78 ( 11.,12.95)(5.5,10.)(0.,)/straight-1 ( 7.99737,12.3609)[br]@xmath33 ( 11.,12.95)(15.,15.)(0.,)/scalardash-1 ( 12.7731,14.8744)[br]@xmath5 ( 15.,15.)(15.,5.)(0.,)/scalardash-1 ( 16.07,10.)[l]@xmath5 ( 11.,12.95)0 ( 11.,7.1)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 11.,7.1)(5.5,10.)(0.,)/straight1 ( 8.00707,7.65892)[tr]@xmath33 ( 11.,7.1)(15.,5.)(0.,)/straight-1 ( 12.7595,5.15763)[tr]@xmath78 ( 11.,12.95)(11.,7.1)(0.,)/scalardash-1 ( 12.07,10.025)[l]@xmath5 ( 11.,12.95)(5.5,10.)(0.,)/straight-1 ( 7.99737,12.3609)[br]@xmath33 ( 11.,12.95)(15.,15.)(0.,)/straight1 ( 12.7731,14.8744)[br]@xmath78 ( 15.,15.)(15.,5.)(0.,)/straight1 ( 16.07,10.)[l]@xmath78 ( 11.,12.95)0 ( 11.,7.1)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 11.,7.1)(5.5,10.)(0.,)/straight-1 ( 8.00707,7.65892)[tr]@xmath33 ( 11.,7.1)(15.,5.)(0.,)/scalardash-1 ( 12.7595,5.15763)[tr]@xmath5 ( 11.,12.95)(11.,7.1)(0.,)/straight-1 ( 12.07,10.025)[l]@xmath78 ( 11.,12.95)(5.5,10.)(0.,)/straight1 ( 7.99737,12.3609)[br]@xmath33 ( 11.,12.95)(15.,15.)(0.,)/scalardash1 ( 12.7731,14.8744)[br]@xmath5 ( 15.,15.)(15.,5.)(0.,)/sine1 ( 16.07,10.)[l]@xmath1 ( 11.,12.95)0 ( 11.,7.1)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 11.,7.1)(5.5,10.)(0.,)/scalardash-1 ( 8.00707,7.65892)[tr]@xmath5 ( 11.,7.1)(15.,5.)(0.,)/straight-1 ( 12.7595,5.15763)[tr]@xmath33 ( 11.,12.95)(11.,7.1)(0.,)/straight-1 ( 12.07,10.025)[l]@xmath78 ( 11.,12.95)(5.5,10.)(0.,)/scalardash1 ( 7.99737,12.3609)[br]@xmath5 ( 11.,12.95)(15.,15.)(0.,)/straight1 ( 12.7731,14.8744)[br]@xmath33 ( 15.,15.)(15.,5.)(0.,)/straight1 ( 16.07,10.)[l]@xmath33 ( 11.,12.95)0 ( 11.,7.1)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 11.,7.1)(5.5,10.)(0.,)/sine-1 ( 8.00707,7.65892)[tr]@xmath1 ( 11.,7.1)(15.,5.)(0.,)/straight-1 ( 12.7595,5.15763)[tr]@xmath33 ( 11.,12.95)(11.,7.1)(0.,)/straight-1 ( 12.07,10.025)[l]@xmath78 ( 11.,12.95)(5.5,10.)(0.,)/scalardash1 ( 7.99737,12.3609)[br]@xmath5 ( 11.,12.95)(15.,15.)(0.,)/straight1 ( 12.7731,14.8744)[br]@xmath33 ( 15.,15.)(15.,5.)(0.,)/straight1 ( 16.07,10.)[l]@xmath33 ( 11.,12.95)0 ( 11.,7.1)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 5.5,10.)(15.,15.)(0.,)/straight1 ( 10.0081,13.3916)[br]@xmath33 ( 9.7,7.75)(5.5,10.)(0.,)/straight1 ( 7.34806,7.98871)[tr]@xmath33 ( 9.7,7.75)(15.,5.)(0.,)/scalardash1 ( 12.1161,5.47918)[tr]@xmath5 ( 15.,10.5)(9.7,7.75)(0.,)/straight1 ( 12.1161,10.0208)[br]@xmath78 ( 15.,10.5)(15.,15.)(0.,)/straight-1 ( 16.07,12.75)[l]@xmath33 ( 15.,10.5)(15.,5.)(0.,)/scalardash-1 ( 16.07,7.75)[l]@xmath5 ( 15.,10.5)0 ( 9.7,7.75)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(15.,15.)(0.,)/sine0 ( 17.5,16.07)[b]@xmath32 ( 20.,5.)(15.,5.)(0.,)/sine0 ( 17.5,3.93)[t]@xmath32 ( 5.5,10.)(15.,15.)(0.,)/scalardash1 ( 10.0081,13.3916)[br]@xmath5 ( 9.7,7.75)(5.5,10.)(0.,)/scalardash1 ( 7.34806,7.98871)[tr]@xmath5 ( 9.7,7.75)(15.,5.)(0.,)/straight-1 ( 12.1161,5.47918)[tr]@xmath78 ( 15.,10.5)(9.7,7.75)(0.,)/straight-1 ( 12.1161,10.0208)[br]@xmath33 ( 15.,10.5)(15.,15.)(0.,)/sine-1 ( 16.07,12.75)[l]@xmath1 ( 15.,10.5)(15.,5.)(0.,)/straight1 ( 16.07,7.75)[l]@xmath78 ( 15.,10.5)0 ( 9.7,7.75)0 ( 5.5,10.)0 ( 15.,15.)0 ( 15.,5.)0 ( 0.,10.)(4.5,10.)(0.,)/scalardash0 ( 2.25,9.18)[t]@xmath31 ( 20.,15.)(14.6,10.)(0.,)/sine0 ( 17.8731,11.8427)[tl]@xmath32 ( 20.,5.)(14.6,10.)(0.,)/sine0 ( 16.7269,6.84267)[tr]@xmath32 ( 7.,13.)(4.5,10.)(0.211476,)/scalardash-1 ( 4.72204,12.2766)[br]@xmath5 ( 12.,13.)(7.,13.)(0.8,)/straight1 ( 9.5,16.07)[b]@xmath33 ( 12.,13.)(7.,13.)(-0.8,)/straight-1 ( 9.5,9.93)[t]@xmath78 ( 12.,13.)(14.6,10.)(-0.182242,)/scalardash1 ( 14.2647,12.2721)[bl]@xmath5 ( 4.5,10.)(14.6,10.)(0.693069,)/scalardash-1 ( 9.55,5.43)[t]@xmath5 ( 12.,13.)0 ( 7.,13.)0 ( 4.5,10.)0 ( 14.6,10.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(14.5,10.)(0.,)/sine0 ( 17.6874,14.1669)[br]@xmath32 ( 20.,5.)(14.5,10.)(0.,)/sine0 ( 17.8126,8.16691)[bl]@xmath32 ( 10.,14.5)(5.5,10.)(0.4,)/straight-1 ( 6.23398,13.766)[br]@xmath33 ( 10.,5.5)(10.,14.5)(0.,)/straight-1 ( 11.07,10.)[l]@xmath78 ( 10.,5.5)(5.5,10.)(-0.4,)/straight1 ( 6.23398,6.23398)[tr]@xmath33 ( 10.,5.5)(14.5,10.)(0.4,)/sine1 ( 13.766,6.23398)[tl]@xmath1 ( 10.,14.5)(14.5,10.)(-0.4,)/sine-1 ( 12.266,14.366)[bl]@xmath1 ( 10.,5.5)0 ( 10.,14.5)0 ( 5.5,10.)0 ( 14.5,10.)0 ( 0.,10.)(8.,6.)(0.,)/scalardash0 ( 3.89862,7.31724)[tr]@xmath31 ( 20.,15.)(13.,15.)(0.,)/sine0 ( 16.5,16.07)[b]@xmath32 ( 20.,5.)(8.,6.)(0.,)/sine0 ( 13.8713,4.43535)[t]@xmath32 ( 9.2,13.9)(13.,15.)(-0.150337,)/scalardash-1 ( 10.5863,15.7445)[b]@xmath5 ( 7.,9.5)(9.2,13.9)(0.8,)/straight1 ( 10.7664,10.6068)[tl]@xmath33 ( 7.,9.5)(9.2,13.9)(-0.8,)/straight-1 ( 5.43364,12.7932)[br]@xmath78 ( 7.,9.5)(8.,6.)(0.222749,)/scalardash1 ( 6.13962,9.00132)[r]@xmath5 ( 13.,15.)(8.,6.)(-0.605305,)/sine-1 ( 14.0988,8.71399)[tl]@xmath1 ( 7.,9.5)0 ( 9.2,13.9)0 ( 13.,15.)0 ( 8.,6.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(10.,10.)(0.,)/sine0 ( 17.1318,12.7564)[tl]@xmath32 ( 20.,5.)(14.5,10.)(0.,)/sine0 ( 17.8126,8.16691)[bl]@xmath32 ( 10.,5.5)(5.5,10.)(-0.4,)/straight1 ( 6.23398,6.23398)[tr]@xmath33 ( 10.,5.5)(10.,10.)(0.,)/straight-1 ( 11.07,7.75)[l]@xmath78 ( 10.,5.5)(14.5,10.)(0.4,)/scalardash1 ( 13.766,6.23398)[tl]@xmath5 ( 10.,14.5)(5.5,10.)(0.4,)/straight-1 ( 6.23398,13.766)[br]@xmath33 ( 10.,14.5)(10.,10.)(0.,)/straight1 ( 8.93,12.25)[r]@xmath78 ( 10.,14.5)(14.5,10.)(-0.4,)/sine-1 ( 13.1968,14.1968)[bl]@xmath1 ( 10.,14.5)0 ( 10.,5.5)0 ( 5.5,10.)0 ( 10.,10.)0 ( 14.5,10.)0 ( 0.,10.)(5.5,10.)(0.,)/scalardash0 ( 2.75,9.18)[t]@xmath31 ( 20.,15.)(14.5,15.)(0.,)/sine0 ( 17.25,16.07)[b]@xmath32 ( 20.,5.)(14.5,5.)(0.,)/sine0 ( 17.25,3.93)[t]@xmath32 ( 5.5,10.)(14.5,15.)(0.,)/sine-1 ( 9.72725,13.3749)[br]@xmath1 ( 5.5,10.)(14.5,5.)(0.,)/sine1 ( 9.72725,6.62506)[tr]@xmath1 ( 14.5,12.5)(14.5,15.)(0.,)/scalardash1 ( 15.57,13.30)[l]@xmath5 ( 14.5,7.5)(14.5,12.5)(0.8,)/straight-1 ( 17.57,10.)[l]@xmath33 ( 14.5,7.5)(14.5,12.5)(-0.8,)/straight1 ( 11.43,10.)[r]@xmath78 ( 14.5,7.5)(14.5,5.)(0.,)/sine-1 ( 15.57,6.50)[l]@xmath1 ( 14.5,7.5)0 ( 14.5,12.5)0 ( 5.5,10.)0 ( 14.5,15.)0 ( 14.5,5.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/scalardash1 ( 9.20801,13.1807)[br]@xmath5 ( 6.5,10.)(13.,6.)(0.,)/scalardash-1 ( 9.20801,6.81927)[tr]@xmath5 ( 13.,14.)(13.,10.)(0.,)/scalardash1 ( 10.274,10.)[l]@xmath5 ( 13.,10.)(13.,6.)(0.,)/scalardash1 ( 13.,10.)(16.,10.)(0.,)/scalardash0 ( 13.500,11.7)[l]@xmath31 ( 16.,10.)(16.,10.)(20.,10.)/straight1 ( 17.674,10.)[l]@xmath33 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 16.,10.)0 ( 13.,10.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/ghostdash1 ( 9.20801,13.1807)[br]@xmath34 ( 6.5,10.)(13.,6.)(0.,)/ghostdash-1 ( 9.20801,6.81927)[tr]@xmath34 ( 13.,14.)(13.,10.)(0.,)/ghostdash1 ( 9.474,9.7)[l]@xmath34 ( 13.,10.)(13.,6.)(0.,)/ghostdash1 ( 13.,10.)(16.,10.)(0.,)/scalardash0 ( 13.500,11.7)[l]@xmath31 ( 16.,10.)(16.,10.)(20.,10.)/straight1 ( 17.674,10.)[l]@xmath33 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 16.,10.)0 ( 13.,10.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,12.5)[t]@xmath31 ( 6.5,10.)(4.5,7.)(0.,)/scalardash0 ( 4.80,8.5)[r]@xmath31 ( 4.5,7.)(4.5,7.)(2.5,4.)/straight1 ( 2.8,5.3)[l]@xmath33 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/sine1 ( 9.20801,13.1807)[br]@xmath1 ( 6.5,10.)(13.,6.)(0.,)/sine-1 ( 9.90801,6.81927)[tr]@xmath1 ( 13.,14.)(13.,6.)(0.,)/sine1 ( 14.274,10.)[l]@xmath1 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 6.5,10.)0 ( 4.5,7.)0 ( 0.,10.)(6.5,10.)(0.,)/scalardash0 ( 3.25,9.18)[t]@xmath31 ( 20.,15.)(13.,14.)(0.,)/sine0 ( 16.2808,15.5544)[b]@xmath32 ( 13.,14.)(10.,16.)(0.,)/scalardash0 ( 11.20,14.)[r]@xmath31 ( 10.,16.)(10.,16.)(7.,18.)/straight1 ( 8.0,17.)[l]@xmath33 ( 20.,5.)(13.,6.)(0.,)/sine0 ( 16.2808,4.44558)[t]@xmath32 ( 6.5,10.)(13.,14.)(0.,)/scalardash-1 ( 7.00801,10.9807)[br]@xmath5 ( 6.5,10.)(13.,6.)(0.,)/scalardash1 ( 9.20801,6.81927)[tr]@xmath5 ( 13.,14.)(13.,6.)(0.,)/sine-1 ( 14.274,10.)[l]@xmath1 ( 6.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 13.,14.)0 ( 10.,16.)0 ( 370,200)(3,2 ) ( 0.,10.)(4.5,10.)(0.,)/scalardash0 ( 16.,15.)(11.,14.)(0.,)/sine0 ( 16.,5.)(11.,6.)(0.,)/sine0 ( 4.5,10.)(7.75,12.)(0.,)/straight0 ( 6.,11.5)[br]@xmath33 ( 4.5,10.)(7.75,8.)(0.,)/straight0 ( 6.,8.5)[tr]@xmath33 ( 7.75,8.)(11.,6.)(0.,)/scalardash0 ( 9.5,6.5)[tr]@xmath5 ( 11.,6.)(11.,14.)(0.,)/scalardash0 ( 12.,10.)[l]@xmath5 ( 7.75,12.)(11.,14.)(0.,)/scalardash0 ( 9.7,13.5)[br]@xmath5 ( 7.75,12.)(7.75,8.)(0.,)/straight0 ( 8.5,10.)[l]@xmath78 ( 4.5,10.)0 ( 11.,14.)0 ( 11.,6.)0 ( 7.75,8.)0 ( 7.75,12.)0 ( 20.,10.)[l ] ( 4.,10.)(8.5,10.)(0.,)/scalardash0 ( 8.5,10.)(13.,13.)(0.,)/sine0 ( 8.5,10.)(13.,7.)(0.,)/sine0 ( 8.5,10.)0 ( 8.5,10 . ) [ ] ( 19,10 . ) [ ] ( 4.,10.)(7.75,12.)(0.,)/straight0 ( 6.,11.5)[br]@xmath33 ( 4.,10.)(7.75,8.)(0.,)/straight0 ( 6.,8.5)[tr]@xmath33 ( 7.75,8)(11.,6.)(0.,)/scalardash0 ( 9.5,6.5)[tr]@xmath5 ( 11.,6.)(11.,14.)(0.,)/scalardash0 ( 12.,10.)[l]@xmath5 ( 7.75,12.)(11.,14.)(0.,)/scalardash0 ( 9.7,13.5)[br]@xmath5 ( 7.75,12.)(7.75,8.)(0.,)/straight0 ( 8.5,10.)[l]@xmath78 ( 7.75,8.)0 ( 7.75,12.)0 ( 4.,10.)0 ( 11.,14.)0 ( 11.,6.)0 ( 20.,10.)[l ] ( 4.,10.)(7.,10.)(0.,)/scalardash0 ( 13.,14.)(17.,15.)(0.,)/sine0 ( 13.,6.)(17.,5.)(0.,)/sine0 ( 8.5,10.)(13.,14.)(0.,)/scalardash0 ( 11.,12.5)[rb]@xmath5 ( 8.5,10.)(13.,6.)(0.,)/scalardash0 ( 11.,7.)[rt]@xmath5 ( 13.,14.)(13.,6.)(0.,)/scalardash0 ( 14.,10.)[l]@xmath5 ( 8.5,10.)0 ( 13.,14.)0 ( 13.,6.)0 ( 8.5,10 . ) [ ] ( 20.,10 . ) [ ] ( 4.5,10.)(7.75,12.)(0.,)/straight0 ( 6.,11.5)[br]@xmath33 ( 4.5,10.)(7.75,8.)(0.,)/straight0 ( 6.,8.5)[tr]@xmath33 ( 7.75,8.)(7.75,12.)(0.,)/straight0 ( 8.75,10.)[l]@xmath78 ( 4.5,10.)0 ( 7.75,12.)0 ( 7.75,8.)0 as already mentioned above , we use in our calculation a general gauge parameter @xmath79 related to the @xmath1 boson and verify that our final result is independent of @xmath79 . we do this in the limit of large and small values of @xmath79 . to this end , we apply the asymptotic - expansion technique in the following four limiting cases @xmath80 where the inequality @xmath81 has to be understood in a formal sense , as , in practice , one has @xmath37 . the result we obtain by asymptotic expansion is an expansion of the exact result for the @xmath0 contribution in powers of @xmath41 . in case @xmath82 , where @xmath83 , which corresponds to t hooft - feynman gauge , we are able to evaluate the first five terms of this expansion . in cases @xmath84@xmath85 , we compute the first two expansion terms and find agreement with the result obtained for @xmath83 . our final result for @xmath22 emerges as the sum @xmath86 where @xmath87 is the universal contribution induced by the one - loop term @xmath88 of eq . ( [ renuniv ] ) , @xmath89 is the two - loop contribution involving virtual @xmath31 and @xmath75 bosons , and @xmath90 the remaining two - loop contribution involving virtual @xmath1 and @xmath5 bosons . in @xmath89 and @xmath90 , also the corresponding counterterm and tadpole contributions are included . for the individual pieces , we find @xmath91 where @xmath92 is defined below eq . ( [ eq::born ] ) and the ellipses indicate terms of @xmath93 . notice that the leading @xmath54 term of @xmath89 is not accompanied by an expansion in @xmath94 , since the contributing diagrams do not involve virtual @xmath1 or @xmath5 bosons . on the other hand , detailed inspection reveals that there is also no expansion in the parameter @xmath95 , contrary to what might be expected at first sight . inserting eq . ( [ eq : terms ] ) into eq . ( [ eq : sum ] ) , we obtain our final result @xmath96 the correction of @xmath54 to @xmath97 was also considered in ref . the expression found in that reference disagrees with our result . one reason is probably that the authors of ref . @xcite only considered virtual @xmath5 bosons , but disregarded virtual @xmath1-bosons . however , our calculation shows that both types of charged bosons have to be taken into account at the same time in order to arrive at an ultraviolet - finite and gauge - parameter - independent result . in ref . @xcite , the dominant two - loop electroweak corrections to the higgs - boson couplings to pairs of gauge bosons and light fermions induced by a sequential isodoublet of ultraheavy quarks @xmath98 were investigated by means of a low - energy theorem @xcite . in that paper , also a result for the @xmath54 correction to @xmath97 is presented , which is obtained by sending the mass of the fourth - generation down quark ( @xmath99 ) to zero at the end of the calculation , which is performed assuming that @xmath100 . this result also deviates from the one obtained above . detailed inspection reveals that this difference may be attributed to the interchange of limits performed in ref . @xcite , which is not justified in the case under consideration , although such a procedure is known to lead to correct results in simpler examples . in fact , reanalyzing the limiting case @xmath101 using the asymptotic - expansion technique , we are able to reproduce the terms of eqs . ( 48)(54 ) in ref . @xcite that survive in this limit and , at the same time , to identify contributions that do not occur if @xmath102 is imposed at the outset of the calculation . for completeness , we also specify our corresponding result for @xmath103 . its evaluation is significantly simpler , since , at one loop , only the top - quark diagrams contribute . consequently , to obtain the leading two - loop correction proportional to @xmath7 , we only have to consider the exchange of @xmath31 , @xmath75 , and @xmath5 bosons . this only requires a naive taylor expansion in @xmath35 and @xmath36 . we obtain @xmath104 with @xmath105 . this result is in agreement with ref . @xcite , where a low - energy theorem @xcite was used . we are now in a position to present our numerical results and to assess the convergence properties of our expansions in @xmath106 and @xmath94 . we use the following numerical values for our input parameters @xcite : @xmath107 gev@xmath108 , @xmath109 gev , and @xmath110 gev . we first consider the one - loop amplitudes @xmath20 and @xmath21 induced by virtual top quarks and @xmath1 bosons , respectively , for which exact results are available . they are shown in figs . [ 1loopt ] and [ 1loopw ] as functions of @xmath46 and @xmath41 , respectively . the solid curves indicate the exact results , while the dashed curves represent the sequences of approximations that are obtained by successively including higher powers of @xmath46 and @xmath41 , respectively , in the expansions . the vertical lines encompass the intermediate - mass range of the higgs boson , @xmath3 . in fig . [ 1loopw ] , the second vertical line coincides with the right edge of the frame . from fig . [ 1loopt ] , we observe that the approximation consisting of the first three terms of the expansion in @xmath46 is practically indistinguishable from the exact result up to @xmath111 . the same is true for the sum of the first five expansion terms up to @xmath112 , which corresponds to @xmath113 gev . in the case of @xmath21 , the convergence is slightly worse , since @xmath114 corresponds to @xmath115 and the exact result behaves like @xmath116 in this limit . nevertheless , for @xmath117 gev , 140 gev , and @xmath118 , the approximation by five expansion terms deviates from the exact result by as little as 0.3% , 1.6% , and 19.9% , respectively . having demonstrated the fast convergence of the expansions in @xmath46 and @xmath41 of the one - loop amplitudes @xmath119 and @xmath120 , respectively , we now proceed to the two - loop electroweak amplitude @xmath121 , for which we found the leading @xmath54 term together with its subleading mass corrections through @xmath122 . this corresponds to a first approximation and four improvements , which are visualized by the five dashed curves in fig . [ 2loopt ] . as in fig . [ 1loopw ] , the dotted vertical line and the right edge of the frame enclose the intermediate - mass range of the higgs boson . we again observe rapid convergence . the goodness of our best approximation may be assessed by considering its relative deviation from the second best one . for @xmath117 gev , 140 gev , and @xmath118 , this amounts to 0.3% , 1.0% , and 2.8% , respectively . the situation is very similar to the one in fig . [ 1loopw ] . in fact , the corresponding figures for @xmath120 are 0.4% , 1.1% , and 3.1% . we thus expect that the goodness of the approximation of @xmath121 by the expansion through @xmath122 is comparable to the case of @xmath120 . for the comparison with future measurements of @xmath97 , all known corrections have to be included in eq . ( [ notation ] ) . in this connection , it is interesting to compare the size of the new @xmath54 electroweak correction with the well - known @xmath123 qcd one @xcite . this is done in fig . ( [ fig::lo_nlo ] ) , where the respective corrections to @xmath97 are displayed as functions of @xmath67 . as in figs . [ 1loopw ] and [ 2loopt ] , the dotted vertical line and the right edge of the frame margin the intermediate - mass range of the higgs boson . we observe that , within the latter , the @xmath54 correction slightly exceeds the @xmath123 one in magnitude , a rather surprising finding . due to the sign difference , the two corrections practically compensate each other . the two - loop electroweak correction induced by light - fermion loops , which has become available recently @xcite , is also negative , but has a slightly smaller size than the @xmath54 correction . we calculated the dominant two - loop electroweak correction , of @xmath0 , to the partial width of the decay into two photons of the sm higgs boson in the intermediate mass range , @xmath3 , where this process is of great phenomenological relevance for the searches at hadron colliders of this elusive missing link of the sm . we evaluated the relevant feynman diagrams by application of the asymptotic - expansion technique exploiting the mass hierarchy @xmath124 . in this way , we obtained an expansion of the full @xmath0 result in the mass ratio @xmath125 through @xmath126 . the convergence property of this expansion and our experience with the analogue expansion at the born level , where the exact result is available for reference , lead us to believe that these five terms should provide a very good approximation to the exact result for @xmath127 gev . by the same token , the deviation of this approximation for the @xmath0 amplitude @xmath121 from the unknown exact result for this quantity is likely to range from 2% to 20% as the value of @xmath67 runs from 140 gev to @xmath128 . in the intermediate higgs - boson mass range , the @xmath0 electroweak correction reduces the size of @xmath97 by approximately 2.5% and thus fully cancels the positive shift due to the well - known @xmath123 qcd correction @xcite . as a by - product of our analysis , we also recovered the @xmath0 correction to the partial width of the decay into two gluon jets of the intermediate - mass higgs boson , in agreement with the literature @xcite . we thank paolo gambino for a useful communication . this work was supported in part by the bundesministerium fr bildung und forschung through grant no . 05 ht4gua/4 , by the deutsche forschungsgemeinschaft through grant no . kn 365/1 - 1 , by the helmholtz - gemeinschaft deutscher forschungszentren through grant no . vh - ng-008 , and by sun microsystems through academic equipment grant no . edud-7832 - 000332-ger . h. zheng , d. wu , phys . d 42 ( 1990 ) 3760 ; + a. djouadi , m. spira , j.j . van der bij , p.m. zerwas , phys . b 257 ( 1991 ) 187 ; + s. dawson , r.p . kauffman , phys . d 47 ( 1993 ) 1264 ; + a. djouadi , m. spira , p.m. zerwas , phys . b 311 ( 1993 ) 255 ; + k. melnikov , o.i . yakovlev , phys . b 312 ( 1993 ) 179 ; + m. inoue , r. najima , t. oka , j. saito , mod . lett . a 9 ( 1994 ) 1189 ; + j. fleischer , o.v . tarasov , z. phys . c 64 ( 1994 ) 413 ; + j. fleischer , o.v . tarasov , v.o . tarasov , phys . b 584 ( 2004 ) 294 . m. steinhauser , in : b.a . kniehl ( ed . ) , proceedings of the ringberg workshop on the higgs puzzle what can we learn from lep2 , lhc , nlc , and fmc ? , ringberg castle , germany , 813 december 1996 , world scientific , singapore , 1997 , p. 177 , report no . hep - ph/9612395 . t. seidensticker , in : g. athanasiou ( ed . ) , proceedings of the 6th international workshop on new computing techniques in physics research : software engineering , artificial intelligence , neural nets , genetic algorithms , symbolic algebra , automatic calculation ( aihenp99 ) , heraklion , greece , 12 - 16 april 1999 , spires conference no c99/04/12 , report no . hep - ph/9905298 ; + r. harlander , t. seidensticker , m. steinhauser , phys . b 426 ( 1998 ) 125 . a. djouadi , p. gambino , phys . rev . lett . 73 ( 1994 ) 2528 ; chetyrkin , b.a . kniehl , m. steinhauser , phys . rev . lett . 78 ( 1997 ) 594 ; + k.g . chetyrkin , b.a . kniehl , m. steinhauser , nucl . b 490 ( 1997 ) 19 .

by means of molecular dynamics computer simulations we investigate the out of equilibrium relaxation dynamics of a simple glass former , a binary lennard - jones system , after a quench to low temperatures . we find that one - time quantities , such as the energy or the structure factor , show only a weak time dependence . by comparing the out of equilibrium structure factor with equilibrium data we find evidence that during the aging process the system remains in that part of phase space that mode - coupling theory classifies as liquid like . two - times correlation functions show a strong time and waiting time dependence . for large and times corresponding to the early-relaxation regime the correlators approach the edwards - anderson value by means of a power - law in time . for large but fixed values of the relaxation dynamics in the-relaxation regime seems to be independent of the observable and temperature . the-relaxation shows a power - law dependence on time with an exponent which is independent of but depends on the observable . we find that at long times the correlation functions can be expressed as and compute the function . this function is found to show a-dependence which is a bit stronger than a logarithm and to depend on the observable considered . if the system is quenched to very low temperatures the relaxation dynamics at long times shows fast drops as a function of time . we relate these drops to relatively local rearangements in which part of the sample relaxes its stress by a collective motion of 50 - 100 particles . finally we discuss our measurements of the time dependent response function . we find that at long times the correlation functions and the response are not related by the usual fluctuation dissipation theorem but that this relation is similar to the one found for spin glasses with one step replica symmetry breaking . pacs numbers : 61.20.lc , 61.20.ja , 02.70.ns , 64.70.pf

cond-mat9905248

it is well known from experiments that the most prominent feature of supercooled liquids and glass forming systems is the rapid increase of their relaxation time as temperature is decreased @xcite . since the early eighties , computer simulations have become a powerful tool to investigate such systems and hence to increase our understanding of their extraordinary dynamical behavior @xcite . however , because of the rapid increase of the relaxation times these simulations are faced with the intrinsic problem that , over a large range of temperature , systems can not be equilibrated , as the simulation time , typically @xmath7s for an atomic system , is smaller than the relaxation time . this means that on a computer , any system with a longer relaxation time can only be studied under nonequilibrium conditions . in other words , in computer simulations the `` glass transition '' always takes place at temperatures higher than in real experiments . despite this drawback , it has been shown that computer simulations can provide useful information on the mechanisms than underly the rapid increase of the time scales in supercooled liquids . one of the most valuable contributions in this direction was the proof that the so called mode - coupling theory @xcite does indeed provide a quantitative description of this slowing down within the time window explored in the simulation , at least for those simple atomic liquids that are `` fragile glass formers '' , i.e. show a strongly non - arrhenius dependence of the relaxation times as a function of temperature . the findings from computer simulations @xcite and experiments @xcite have largely confirmed that this theory provides a consistent picture for the dynamical behaviour of such liquids over a relaxation time range that covers several decades , typically @xmath8s . for larger relaxation times ( lower temperatures ) , equilibrium simulations are no longer possible . on the other hand , one can take advantage of the flexibility of the simulation to explore the _ nonequilibrium _ properties on times scales that would be rather difficult to access in experiments . this possibility stems from the fact that the simulations allows instantaneous quenches to low temperatures , and a monitoring of the subsequent evolution on a fast time scale . experimentally , the study of nonequilibrium evolution in glassy systems following a fast quench is a field that has been explored since a long time in the so called `` aging '' experiments @xcite , performed on a time scale of hours or even days . more specifically , `` physical aging '' consists in a slow evolution of the characteristic properties of glassy systems that do not undergo any changes in their chemical composition . of particular interest is the evolution of response functions ( typically the elastic compliance ) that are usually found , e.g. in polymeric systems , to exhibit a strong slowing down as the _ waiting time _ ( i.e. the time elapsed since the quench ) increases . the recent interest of the physics community in this behavior was aroused when a number of experiments performed on spin glasses showed that a very similar behavior could be observed in these disordered magnetic systems . in particular it was found that the response to a magnetic field of a system that has been quenched into its low temperature , nonequilibrium spin glass phase , becomes more and more sluggish as the waiting time increases @xcite . subsequently , a number of theories were put forward to explain this aging behavior of spin glass systems from first principles statistical physics . a comprehensive review of these theories can be found in reference @xcite . essentially , one is lead to distinguish between phenomenological `` trap models '' , domain growth theories , and mode - coupling or mean field theories . `` trap models '' describe the evolution of systems as a random walk in a complex phase space , which can under certain conditions give rise to aging . the domain growth models assume that the aging results from a coarsening process somewhat similar to what can be observed in a ferromagnetic system quenched below its critical temperature . mean field or mode - coupling theories account exactly for the aging behaviour of some disordered models in the limit of high dimension ( e.g. a particle in a random potential in a high dimensional space ) . interestingly , these mean field theories give rise to a mathematical structure that is very similar to that of the mode - coupling theory commonly used to describe the dynamical properties of supercooled liquids @xcite . a major difference , however , is that high dimensional systems can be shown to have a true transition towards a non - ergodic behavior as the temperature is lowered , whereas it is well known that the transition predicted by the mode - coupling theory of liquids is smeared out and does not give rise to a true singularity . if , however , a true transition is present , a natural consequence is that a system quenched below the transition temperature will display aging behavior . theoretically , this aging behavior of mean field systems was studied in great detail and is now quite well understood @xcite . the question arises , however , to what extent this description is relevant to the aging behaviour of real , three dimensional systems . in particular , in view of the mathematical similarity mentioned above , it is quite natural to inquire whether the nonequilibrium dynamics of those supercooled liquids that are known to be well described by mode - coupling theory can be accounted for by the mean field description of aging . in an attempt to partially answer this question , we present molecular dynamics simulations of a simple glass forming liquid , whose properties under equilibrium conditions have been extensively studied in previous work @xcite . we will concentrate on the characterization of the `` aging '' behaviour of such a system on the time scales that can be investigated in molecular dynamics simulations , i.e. a few nanoseconds . once again , we emphasize that these time scales are very different from those usually investigated in aging experiments on glasses ( although some recent dielectric spectroscopy experiments @xcite investigate relatively high frequency behaviour ) . moreover , the systems under study would , from an experimental viewpoint , be considered as supercooled liquids rather than glasses . glassy behavior is observed only on the restricted time scale of the computer simulation . the possibility of exploring such relatively short time scales is nevertheless interesting , as it is precisely on such time scales that mode - coupling theory is successful . moreover , the simulation offers the opportunity to simultaneously compute time dependent correlation functions , static quantities and response functions . relating these quantities to each other is an essential achievement in mean field theories of aging behaviour , whose predictions can then be tested in detail . the paper is organized as follows . in section ii , we recall the main features of our model and describe some technical details of the simulations . sections iii and iv describe the aging behavior of static properties and of correlation functions , respectively . section v deals with the aging behaviour of response functions . preliminary accounts of these results can be found in @xcite . some closely related studies , on a slightly different system , have been reported in @xcite . the model we use is a binary ( 80:20 ) mixture of particles , which in the following we will call a and b particles , that interact via a lennard - jones potential , @xmath9 , with @xmath10 . the constants @xmath11 and @xmath12 are given by @xmath13 , @xmath14 , @xmath15 , @xmath16 , @xmath17 , and @xmath18 and the potential is cut off and shifted at a distance @xmath19 . in the following we will report the results in reduced units , with @xmath20 and @xmath21 the unit of length and energy , respectively ( setting the boltzmann constant @xmath22 ) . time will be measured in units of @xmath23 , where @xmath24 is the mass of the particles . the simulation was done with @xmath25 particles and at constant volume @xmath26 , with a box length @xmath27 . at this density the _ equilibrium _ dynamics of the system has been investigated intensively @xcite and it has been found that at low temperatures , @xmath28 , this dynamics is described very well by mode - coupling theory @xcite with a critical temperature @xmath29 . although there is evidence that for temperatures @xmath30 the increase of the relaxation times , or the inverse of the diffusion constant , is less strong than the power - law predicted by mct , @xmath31 , with @xmath32 @xcite , the increase is still so strong that at temperatures less than @xmath33 the relaxation time of the system exceeds by orders of magnitude the time scale accessible to a present state of the art computer simulation ( @xmath34 time units ) . hence for a computer simulation @xmath33 plays the role of the glass transition temperature @xmath35 . the equations of motion were integrated with the velocity form of the verlet algorithm , with a step size of 0.02 . the system was equilibrated in the high temperature phase at a temperature @xmath36 and then quenched at time @xmath37 to a temperature @xmath38 . this quench was done by means of a stochastic heat bath , i.e. every 50 time steps we replaced the velocities of all the particles with ones that were drawn from a maxwell - boltzmann distribution corresponding to a temperature @xmath39 . in order to study the dependence of the relaxation behavior on the initial and final temperature we considered several values of @xmath40 and @xmath39 : @xmath41 and @xmath42 . after the quench we propagated the system for a waiting time @xmath0 after which the measurements of the quantities of interest were started . in the course of the simulations we realized that there are appreciable sample to sample variations of the relaxation curves . therefore it was necessary to average for each combination of @xmath40 and @xmath39 over 8 - 10 independent runs . one - time quantities are observables which in equilibrium are constants , such as the total energy of the system , the pressure or , in a magnetic system , the magnetization . in the non - equilibrium situation these observables depend on the time which has elapsed since the quench and thus are commonly called `` one - time '' quantities . it has been shown before , see e.g. refs . @xcite , that the time dependence of such quantities is relatively weak . in fig . [ fig1 ] we show the time dependence of @xmath43 , the potential energy per particle , for various combinations of @xmath40 and @xmath39 . from the figure it becomes evident that this time dependence is rather weak . in ref . @xcite we have shown that it can be approximated well by a power - law with a small exponent ( 0.14 ) , or alternatively by a logarithmic time dependence . qualitatively this result agrees with the one of monte carlo simulations of a soft sphere system @xcite and a polymer model @xcite in which also power - law dependences have been observed for one - time quantities . however , the exponents found in these simulations ( 0.7 and 1/3 ) are significantly larger than our , thus indicating that they are not universal quantities . the large exponent of reference @xcite might , however , be also due to the fact that in that simulation @xmath40 was infinity and that the runs were rather short , i.e. that the dependence of @xmath44 on time was not the one valid for very long times . from fig . [ fig1]a we conclude that at long times the curves do not depend significantly on the value of the initial temperature ( see the curves with @xmath45 ) . for short times , however , such a dependence can be seen , in that , e.g. , the curve with @xmath46 is almost constant and shows only at long times a time dependence which is similar to the one for higher values of @xmath40 ( see also ref . this time dependence starts to be observable only for times that are comparable with the @xmath2-relaxation time of the system at @xmath40 , which for @xmath46 is on the order of @xmath47 time units @xcite , since this is the typical time scale that the system needs to find a significantly different configuration . this estimate is , however , only a lower bound , since the dynamics of the system is given by the temperature @xmath48 , and is hence slower than the one at @xmath40 . see also ref . @xcite for a further discussion of this point . the figure also shows that there is a significant dependence of the curves on @xmath39 in that the values of @xmath43 decrease with decreasing @xmath39 . in a harmonic solid at a temperature @xmath39 one expects that the ( equilibrium ! ) potential energy varies like @xmath49 and thus it is reasonable to subtract this `` trivial contribution '' from the curves . the result of this subtraction is shown in fig . [ fig1]b from which we see that this procedure does indeed make the curves collapse reasonably well . the most significant exceptions are the curves ( @xmath46 , @xmath45 ) and ( @xmath50 , @xmath51 ) . these deviations can be understood by realizing that during the aging process the system is looking for configurations with low lying energies . if this search is started at a relatively low temperature , such as @xmath46 , the system will typically be able to find configurations that have a lower potential energy than in the case when @xmath40 is large thus explaining why the curve for @xmath46 is below the other ones . in addition to this , such a search will be more efficient at a higher temperature @xmath39 than at a low one , since then the system has a better chance to overcome small local barriers . this explains why the curve for ( @xmath50 , @xmath51 ) is above the other ones with the same value of @xmath40 . since also the radial distribution function @xmath52 is a one - time quantity we expect that also its dependence on time is weak . that this is indeed the case is shown in the main part of fig . [ fig2 ] where we show @xmath53 , i.e. the partial radial distribution function for the a particles ( see ref . @xcite for its definition ) for the times @xmath37 ( i.e. before the quench ) , @xmath54 , 100 , 1000 , 10000 , and 63100 time units . from the figure we see that immediately after the quench @xmath53 changes its shape very quickly in that , e.g. , the first nearest neighbor peak which in the high temperature phase ( curve labeled with @xmath37 ) is not very high , becomes much higher and narrower . for times larger than 10 , the shape becomes essentially independent of time , as expected for a one - time quantity and in agreement with the results of ref . @xcite . that the form of @xmath53 _ at long times _ has a significant dependence on @xmath39 is demonstrated in the inset of fig . [ fig2 ] , where we show the main peak for different values of @xmath39 at long times ( @xmath55 ) . as expected we find that the height of this peak increases with decreasing temperature and that it becomes narrower . thus we conclude that the typical configurations in which the system is stuck after long times does depend on the final temperature . very recently latz made the interesting proposition that the typical configurations in which the system is stuck at long times during the aging process share a common property in that all of them are very close to the so - called critical surface of mode - coupling theory ( mct ) @xcite which divides the liquid like phase of the system from its glass like phase @xcite . mct predicts that this critical surface can be calculated by using as input the radial distribution function , i.e. a purely _ static _ quantity @xcite . furthermore it has been shown that , for simple liquids , the relevant part of this input is related to the area under the first peak in the structure factor @xcite , or alternatively , to the area under the first peak in @xmath52 weighted with @xmath56 . in order to test the validity of the proposition by latz we have therefore calculated the integral of @xmath57 between zero and 1.406 , the location of the first minimum in @xmath53 . note that this integral , which we call @xmath58 , is the partial coordination number . the time and @xmath39 dependence of @xmath58 are shown in the main part of fig . [ fig3 ] ( open symbols ) . ( in order to expand the axis at low temperatures we show this data versus the logarithm of @xmath39 . ) . the times @xmath6 shown are spaced essentially equidistant on a logarithmic time axis and are 0 , 10 , 40 , 60 , 100 , 160 , 250 , 400 , 630 , 1000 , 1580 , 2510 , 3980 , 6310 , 10000 , 15850 , 25120 , 39810 , and 63100 . ( in order to show the time dependence we plot the data for @xmath55 at @xmath59 and with decreasing @xmath6 at a temperature which is a factor of 1.003 higher than the previous data point . ) from this figure we recognize that with increasing time the value of @xmath58 increases rapidly and then becomes constant to within the precision of our data . the value of this limiting constant , which we call @xmath60 , increases weakly with increasing @xmath39 but to a first approximation it can be considered as independent of @xmath39 , thus supporting the prediction of latz . in order to see whether the @xmath39 dependence of @xmath60 is indeed weak it is useful to compare it with the temperature dependence of the area under the first peak @xmath61 _ in equilibrium_. for this we have included in the figure also the value of @xmath58 for temperatures @xmath62 ( filled symbols ) . ( the @xmath53 for the various temperatures are results from other simulations @xcite . ) in this temperature range the equilibrium value of @xmath58 shows an appreciable temperatures dependence ( which is close to a logarithmic dependence ) which is indeed much more pronounced than the @xmath39 dependence of @xmath60 of the non - equilibrium simulations . to a first approximation the value of @xmath60 is given by the value of @xmath58 of the equilibrium curve at @xmath63 , where @xmath33 is the so - called critical temperature of mct @xcite . for the present system this critical temperature has been estimated to be around @xmath64 @xcite . due to the problem to equilibrate the system at temperatures close to @xmath33 it is currently not possible to determine the equilibrium value of @xmath58 at @xmath33 and therefore to compare this value with the @xmath60 from the out - of - equilibrium simulations . however , since the equilibrium value of @xmath58 is expected to show a smooth temperature dependence , it can be estimated quite reliably from the equilibrium data at a bit higher temperatures . in the inset of fig . [ fig3 ] we show an enlargement of the region which is marked by a box in the main figure and which encloses the temperature range around @xmath33 . from this inset we see that the values of @xmath60 for @xmath65 and @xmath45 are indeed very close to an extrapolation of the equilibrium curve to @xmath33 . hence we conclude that it is indeed possible to calculate from the equilibrium data with reasonable accuracy also certain quantities for the out - of - equilibrium situation . this point is discussed in more detail in ref . in equilibrium , time correlation functions between any two observables @xmath66 and @xmath67 , @xmath68 depend only on the time difference @xmath3 , i.e. @xmath69 , for all waiting times @xmath0 . for the out of equilibrium situation this equality no longer holds and therefore such time correlation functions depend on two quantities , the time difference @xmath3 and @xmath0 , the time passed since the generation of the non - equilibrium situation . therefore such correlation functions are called two - times quantities and in this subsection we will demonstrate that such quantities are very well suited to investigate the aging dynamics of out of equilibrium systems . for liquids in _ equilibrium _ the dynamics is often studied by means of the intermediate scattering function @xmath70 which is defined by @xcite @xmath71 and by the its so - called self part , @xmath72 given by @xmath73 here @xmath74 are the positions of the particles at time @xmath6 and @xmath75 is the wave vector . for the non - equilibrium situation these functions are generalized to @xmath76 and @xmath77 ( note that these last equations are trivially generalized to multi - component systems . ) for the present system the dependence of @xmath70 and @xmath72 on time and temperature have been discussed extensively in refs . @xcite . in fig . [ fig4 ] we show the time dependence of the corresponding quantities @xmath78 and @xmath79 for different waiting times @xmath0 . the values of @xmath0 are 0 , i.e. immediately after the quench , and @xmath80 , 100 , 1000 , 10000 , and 63100 time units and the final temperature @xmath45 ( solid lines ) . in fig . [ fig4]a we show @xmath78 for @xmath81 , the location of the first peak in the partial structure factor of the a - a correlation @xcite . we see that the correlation function decays quite quickly as a function of @xmath3 , which in view of the fact that we are at a very low temperature might be surprising . however , one should realize that the decay of the correlation function to zero does _ not _ imply that the system has relaxed to equilibrium but only that the particle configuration at the end of the measurement is uncorrelated with the configuration at the start . from the figure we see that with increasing value of the waiting time the decay of the curves occurs at longer and longer times . this observation can be rationalized by realizing that the driving force which leads to the decorrelation of the state at time @xmath3 from the state at time zero decreases with increasing waiting time @xmath0 since in the time between zero and @xmath0 the system had already the possibility to relax . this @xmath0 dependence of the curves is , however , only observed at long times . for short times the curves fall onto a master curve and leave this master curve only at a time which increases with @xmath0 ( and is roughly proportional to @xmath0 @xcite ) . the time regime in which this master curve is observed is usually called `` short time regime '' whereas the time regime in which the curves show a significant @xmath0 dependence is called the `` aging regime '' . in the short time regime the particles rattle in the cages formed by their nearest neighbors and this type of motion is thus not sensitive on the value of @xmath0 ( if @xmath0 is not too small ) . only for longer times the particles are able to leave this cage and the typical time scale for this process depends strongly on @xmath0 . we also mention that the oscillation seen in the curves at @xmath82 originates from the coupling of the system to the heat bath . since at this time the velocities of all particles are swapped with the ones drawn from a maxwell - boltzmann distribution at temperature @xmath39 , the motion of the particles is , on average , slowed down or even reversed . hence for a brief period the relaxation is slower than expected and thus the correlation curve decays slower . this effect is , however , only seen at short times and thus can be disregarded for large values of @xmath3 . it is also interesting to compare the relaxation behavior of the system in this non - equilibrium situation with the one in equilibrium . in fig . [ fig4]a we have therefore also included the incoherent intermediate scattering function @xmath72 in equilibrium ( dashed lines ) which was obtained in a simulation of gleim _ et al . _ @xcite . the dashed line is a curve that corresponds to the equilibrium relaxation dynamics of the system at @xmath83 ( this was the lowest temperature at which the system could be equilibrated ) . for very short times , @xmath84 , this curve essentially coincides with the above discussed master curve in the short time regime . for times @xmath3 larger than 1.0 deviations are observed , which are _ not _ due to the mentioned oscillations . a careful inspection of the curves shows that the approach of the master curve and the dashed curve to the ( quasi ) plateau at intermediate times is quite different from each other in that the former shows a very slow approach whereas the one of the latter is quite fast . also at long times we find differences in the relaxation behavior . in equilibrium it has been demonstrated that the relaxation curves can be fitted very well by the so - called kohlrausch - williams - watts ( kww ) function , @xmath85 , where @xmath86 is the relaxation time and an exponent @xmath87 @xcite . from the figure we see that at long times the shape of the equilibrium and non - equilibrium curves are quite different and a more detailed analysis shows that a kww function does not give a good fit to the data . what seems to be the same , however , ( or at least very similar ) is the height of the plateau , i.e. the so - called edwards - anderson parameter @xcite or non - ergodicity parameter . in fig . [ fig4]b we show the relaxation curves for the collective function , i.e. @xmath88 for the same waiting times . the wave - vector is now @xmath89 , the location of the minimum in the partial structure factor of the a - a correlation @xcite . due to the collective nature of this correlation function the statistics is worse than the one for the self part but from the figure we see that also this observable can be studied with satisfactory accuracy . since at this wave - vector the ( equilibrium ) value of the non - ergodicity parameter has a local minimum @xcite it is expected that also the one for the non - equilibrium case is relatively small and from the figure we see that this is indeed the case . in the following we will now analyze the relaxation behavior of the system in more detail . for this we start first with the dynamics at short times . mean field calculations predict that for times at which the correlators are in the vicinity of the plateau , a time regime which in the equilibrium dynamics of glasses is called @xmath1-relaxation regime , two power - laws are observed @xcite . denoting by @xmath90 the value of the edwards - anderson parameter , it is predicted that any time correlation function @xmath91 should behave like @xmath92 and @xmath93 here @xmath94 is the typical relaxation time at long times and the exponents @xmath95 and @xmath96 are related by @xmath97 the quantity @xmath24 in the last equation is the so - called fluctuation - dissipation - violation factor and will be discussed later in more detail . for the moment it is sufficient to know that it is expected to be a system universal number equal or less than 1.0 . we also mention that the two power - laws of eqs . ( [ eq5 ] ) and ( [ eq6 ] ) have been discussed for a long time and in great detail in the mct of supercooled liquids and glasses @xcite . in order to see whether it is possible to see these power - laws in our data we have estimated the value of @xmath90 for @xmath98 for @xmath99 and show in fig . [ fig5 ] @xmath100 in a double logarithmic plot for different waiting times . this value of @xmath101 is relatively large so that the value of @xmath90 is relatively small , 0.47 . a small value of @xmath90 is useful if one wants to check for the presence of the power - law given in eq . ( [ eq5 ] ) whereas for the check of eq . ( [ eq6 ] ) a large value of @xmath90 is better . ( these conclusions are reached from the analysis of equilibrium dynamics @xcite but it can be expected that they hold also for the nonequilibrium case . ) from the figure we recognize that it is indeed possible to see a power - law in the short time regime with an exponent around @xmath102 ( bold solid line ) . in the discussion of fig . [ fig4 ] we said that in the late part of the @xmath1-relaxation regime the equilibrium curves and the non - equilibrium curves show a very similar time dependence and that the one of equilibrium is given by a power - law . the exponent of this power - law is around 0.63 @xcite . if we assume the relation ( [ eq7 ] ) to hold true one therefore obtains @xmath103 . due to the relatively large error of the exponents @xmath95 and @xmath96 this number also has a significant error . we have also checked for the presence of the power - law for other values of @xmath101 , as well as for the collective correlation function @xmath79 and found that all of them show such a time dependence with an exponent which is compatible with 0.42 . unfortunately , however , the exact determination of the exponent is rather difficult , since a change of the ( unknown ) value of @xmath104 will lead to a change in the exponents also . in order to avoid this problem to some extend we use a trick which has proved to be useful in the context of the analysis of equilibrium data ( see , e.g. , @xcite ) . for the equilibrium case mct predicts that _ in the @xmath1-relaxation regime _ the so - called factorization property holds @xcite . this means that any time correlation function @xmath105 can be written as @xmath106 , where @xmath107 is a constant and the whole time dependence is given by the @xmath108-independent , i.e. _ system universal _ , function @xmath109 . if we assume that a similar relation holds also in the non - equilibrium situation we have @xmath110 we see that if the exponents @xmath95 and @xmath96 as well as the quantity @xmath24 are independent of the observable that then eqs . ( [ eq5 ] ) and ( [ eq6 ] ) are indeed compatible with such an ansatz . from eq . ( [ eq8 ] ) it follows immediately that if @xmath111 and @xmath112 are two arbitrary times in the @xmath1-relaxation regime , the ratio @xmath113 is independent of @xmath108 if @xmath3 is in the @xmath1-relaxation regime also . in fig . [ fig6 ] we plot this ratio for different choices of @xmath108 . these are @xmath79 for @xmath114 , 7.23 , 9.6 , and 12.53 as well as @xmath78 for the same wave - vectors and also @xmath115 and @xmath116 . the value of @xmath0 is kept fixed at @xmath117 . from the figure we see that in the @xmath1-relaxation regime the different curves collapse nicely onto a master curve , thus demonstrating the validity of the factorization property . that this result is by no means trivial can be recognized from the fact that for very short and very long times the different curves do not fall onto a master curve at all . from the existence of the master curve in the @xmath1-relaxation regime we thus come to the conclusion that , within the accuracy of our data , in this time regime the time dependence of the relaxation is governed by one single system universal function @xmath118 . from the results shown in fig . [ fig5 ] we see that this function is compatible with power - law of the form given by eq . ( [ eq5 ] ) . as we will see later , it is expected that the quantity @xmath24 depends on the value of @xmath39 @xcite . therefore one might conclude that also the exponent @xmath95 and @xmath96 depend on temperature . from making fits to the master curves in the @xmath1-relaxation regime of our data , see fig . [ fig5 ] , it is hard to conclude whether or not such a dependence is indeed present since the choice of the non - ergodicity parameter affects the values of the exponents also . therefore we have calculated the ratios @xmath119 also for lower values of @xmath39 , @xmath120 and 0.1 , and found that the curves for the different observables do indeed fall onto a master curve also . more important is the observation that within the accuracy of our data this master curve does not depend on @xmath39 , thus giving evidence that the function @xmath118 is only a weak function of @xmath39 . we also mention that mean - field calculations lead to the expectation that the quantities @xmath95 , @xmath96 and @xmath24 in eq . ( [ eq7 ] ) do depend on the final temperature . ( this is in contrast to the equilibrium mct where @xmath121 and @xmath95 and @xmath96 are independent of temperature . ) in sec . [ secvc ] we will show that @xmath24 shows a significant dependence on the final temperature in that it becomes smaller with decreasing @xmath39 . thus that result might seem to be in contradiction with the fact that @xmath118 seems to depend only weakly on @xmath39 . the resolution of this apparent problem is that if we assume @xmath96 to be constant that a decrease of @xmath24 can be compensated by an increase of @xmath95 . it is , however , simple to see that if @xmath24 is small @xmath95 has to be close to 0.5 and that a variation of @xmath24 can be compensated by a very small change in @xmath95 , i.e. without leading to an appreciable change in @xmath118 . having analyzed the relaxation behavior of the system in the @xmath1-relaxation regime we now investigate the one of the @xmath2-regime , i.e. the relaxation of the system at long times . when we discussed the time correlation functions in fig . [ fig4]a we already mentioned that in this regime the relaxation differs from the one in the equilibrium situation in that the time correlation functions can not be described well by the kww - law . in fig . [ fig7]a we show the same correlation functions as in fig . [ fig4]a , but this time in a double log plot . from this presentation of the data we see that at long times the out - of - equilibrium curves show a power - law . the exponent is around 0.4 and is , within the accuracy of our data , independent of @xmath0 . a similar time dependence has also been found in simulations of spin - glasses @xcite , although in that case the exponent was significantly smaller , and is thus not unusual for the aging dynamics . the curves in fig . [ fig7]a are for the wave - vector @xmath122 , the location of the main peak in the static structure factor . in fig . [ fig7]b we show the same type of correlation function for different wave - vectors and @xmath123 . from the main figure we see that the dependence of @xmath124 on @xmath101 is quite pronounced , in that e.g. the height of the plateau increases quickly with decreasing wave - vector . such a dependence can be understood at least qualitatively by recalling that _ in equilibrium _ the wave - vector dependence of the edwards - anderson parameter is very similar to a gaussian , and that at short times and long waiting times the time correlation functions do not depend on @xmath0 ( see fig . [ fig4 ] ) , i.e. show a quasi - equilibrium behavior . from fig . [ fig7]b we see that a rough estimate of @xmath125 is given by the value of @xmath78 at @xmath126 and we find that this quantity does indeed show a gaussian like dependence on @xmath101 . in the inset of the figure we show the same correlation functions in a double logarithmic plot . from this graph we recognize that for all wave - vectors the long time dependence of the functions are compatible with a power - law and that the exponent depends on the wave - vector in that it decreases with decreasing @xmath101 . since at long times @xmath124 shows a power - law dependence it is not possible to define a relaxation time . however , from the figure it becomes clear that the relaxation of the system is much faster on small length scales than on large ones . it is instructive to recall that for a diffusion process the relaxation time depend on the wave - vector like @xmath127 . a comparison of the curves for @xmath116 and @xmath128 shows that the time needed to decay to 50% of the initial value differs by almost a factor of 100 , thus much more than the factor of 4 expected for a diffusive process , or the factor found in the relaxation dynamics of supercooled liquids _ in equilibrium_. this shows that during the aging process the relaxation at long times is indeed very different from the one in supercooled liquids . we now turn our attention to the time and waiting time dependence of the two - time correlation functions at long times . within mean - field theory it is expected that for systems which show a discontinuous dependence of the non - ergodicity parameter on temperature , which is the case for the structural glass studied here , this dependence can be written ( in the limit of long waiting times ) as @xmath129 here @xmath130 is the time dependence of @xmath91 at short times which is supposed to be independent of @xmath0 ( in agreement with our findings , see fig . [ fig4 ] ) and to decay quickly to zero . @xmath131 is a function whose form depends on @xmath91 and @xmath5 is a monotonously increasing function of the argument . the interesting point of this equation is that the whole @xmath3 and @xmath0 dependence of the aging regime enters only through the combination @xmath132 . apart from eq . ( [ eq10 ] ) , not much more is known about the @xmath3 and @xmath0 dependence of @xmath91 . in order to gain more insight into this dependence we investigate whether the function @xmath5 is independent of the correlation function @xmath91 . the presence of such a dependence can , e.g. , be checked by plotting a time correlation function @xmath133 versus a different time correlation function , e.g. , @xmath78 , i.e. by making a parametric plot with time @xmath3 as the running parameter . if @xmath5 is independent of the correlation function considered , it is easy to see that in such a parametric plot the curves corresponding to different waiting times will fall on top of each other . ( this holds true only if the time dependence of @xmath134 can be neglected , i.e. at long times . ) in fig . [ fig8 ] we show such parametric plots for the waiting times @xmath135 , 10 , 100 , 1000 , 10000 , and 63100 time units . the independent variable ( abscissa ) is @xmath78 for @xmath81 and the dependent variables ( the ordinate ) are the same correlation function for different values of @xmath101 . let us focus first on @xmath115 . from the graph we recognize that at short times ( corresponding to large values of @xmath124 and @xmath136 ) the parametric plot is independent of @xmath0 , as it would be expected for the equilibrium case . however , for times at which the correlation functions have fallen below their edwards - anderson parameters a waiting time dependence is seen . qualitatively the same behavior is found for the other values of @xmath137 . from this observation we hence conclude that if the ansatz in eq . ( [ eq10 ] ) is correct , then the function @xmath5 does depend on the observable , i.e. in this case on the wave - vector . we also mention that qualitatively the same results are obtained for the different values of @xmath39 . the time dependence of the function @xmath5 can be used to distinguish between different theoretical models to describe the aging dynamics . e.g. the droplet model of fisher and huse @xcite predicts @xmath5 to be of the form @xmath138 . in ref . @xcite it was argued that the present lennard - jones model showed this type of aging dynamics , a conclusion which was not confirmed by the present authors @xcite . the reason for this discrepancy remains still unresolved . in order to settle this issue we have attempted to determine the function @xmath5 from our data without making reference to any model , i.e. functional form of @xmath5 . for this we assumed that at long times the correlation functions are given by the second term in eq . [ eq10 ] . starting from the initial guess @xmath138 we plotted the correlation function versus @xmath132 for all values of @xmath0 , which gave at long times a clustering of the curves . by iteratively bending @xmath5 in a continuous way by small amounts and subsequently making the scaling plot we attempted to generate a collapse of the curves . the outcome of this procedure is shown in fig . [ fig9 ] for the case @xmath124 with @xmath81 . ( for the sake of clarity only the curves for @xmath135 , 10 , 100 , 1000 , 10000 , and 63100 time units are shown , although more waiting times were considered to determine the master curve . also , in order to expand the abscissa we have subtracted 1.0 from @xmath132 and plot its logarithm . ) . from this figure we see that the it is indeed possible to obtain a satisfactory scaling of the curves at long times . the function @xmath5 which was obtained by the procedure described above is shown in the inset . we see that to a first approximation it is close to a logarithm , but that significant deviations are present . also included in the figure are the curves for different values of the wave - vector . we see that in these cases the curves for the different waiting times do not collapse nicely at long times . thus this is further evidence that the function @xmath5 depends on the observable considered . we also mention that although we have determined the optimal shape of @xmath139 also for the other wave - vectors , it is difficult to compare these different functions with each other . the problem is , that if @xmath5 is an optimal function that for example also @xmath140 ( with arbitrary @xmath2 ) is an optimal function , i.e. there is no unique representation of @xmath5 . essentially all the results discussed so far were obtained for a final temperature @xmath45 . this is only slightly ( 10% ) less that the mode coupling critical temperature for the system , @xmath29 @xcite . a markedly different behavior in the relaxation is observed for temperatures much lower than @xmath33 , as illustrated in figure [ fig10 ] , which shows the results for @xmath78 after a quench to @xmath51 . the values of the waiting times and of the wave - vector @xmath101 are the same as in figure [ fig4 ] ( @xmath141 . at short times @xmath3 , the dependences are qualitatively similar for the two final temperatures . the only difference is the increase in the plateau value when the temperature is lowered . this increase is easily understood from the fact that the amplitude of the vibrational motion about an equilibrium position decreases when @xmath142 decreases . from harmonic theory a linear dependence of @xmath143 on @xmath142 at low temperatures can be expected , and such a dependence is indeed compatible with our results . for large values of @xmath0 and @xmath3 , the behaviour for @xmath51 is strikingly different from what was observed at @xmath45 . instead of decaying rapidly to zero for @xmath144 , the correlators level off and appear to display an additional plateau at a value of @xmath145 smaller than @xmath146 . in order to get some insight into this surprising behavior , we have studied separately the correlation functions obtained for various samples prior to averaging . looking at the data shown in fig . [ fig10 ] for some of the samples , it appears that the decay to this second plateau is triggered by a large amplitude and rather sudden drop of the correlation function , that , depending on the sample , takes place typically @xmath147 to @xmath47 time units after the quench ( see inset of fig . [ fig10 ] ) . an analysis of the configurations shows that this sudden drop is related to a very collective dynamical event , in which a substantial fraction of the particles ( 10% ) move by a rather small amount , typically 0.1 - 0.5@xmath148 . such motions can be understood from the fact that a deep quench leaves the system in a highly stressed configuration , so that the relaxation first proceeds through large `` earthquake like '' events , that release the local stress significantly . only at longer times the aging dynamics crosses over to the very smooth process observed at higher values of @xmath39 . we emphasize that these events do not seem to be related to the hopping like motion of the particles which is observed in the deeply supercooled state _ above _ the glass transition temperature @xcite . in that case only very few ( 2 - 3 ) particles are involved and during the jump these particles move on the order of one nearest neighbor distance . in contrast to this , the `` catastrophic events '' during the aging are very collective ( 50 - 100 particles ) and the particles move only about 10 - 30 % of the typical nearest neighbor distance ( as can be inferred from the self part of the van hove correlation function ) . a typical event is shown in fig . [ fig11 ] where we show the particles just before the event ( spheres ) and their location after the event ( tip of arrows ) . from this figure we recognize that in such an event the stress is released along a surface or even a line through the sample and not through the motion of the particles in a ( bulk - like ) three dimensional blob . obviously , the catastrophic events that cause the decay of the correlations in this situation are rather difficult to average over , and a very large number of samples would be required to obtain a reasonable statistical accuracy . moreover , it is quite likely that these events correspond to a transient , so that the second plateau we observe is not really representative of the asymptotic behavior . interestingly , however , rather similar shapes of the time correlation functions were observed by bonn and coworkers @xcite in their dynamical light scattering studies of aging in clay ( laponite ) suspensions . in that case the height of the second plateau was found to steadily increase with waiting time , confirming the transient nature of the effect . we also mention that the occurence of these events are related to the fact that the typical configuration at high temperature ( @xmath50 ) are quite different from the ones towards the system evolves to at @xmath39 , thus giving rise to large stresses . since in _ strong _ glass formers ( e.g. @xmath149 ) the structure has a much weaker temperature dependance than in fragil eglassformers such as the present system , it can be expected that in the aging dynamics of the former no such `` catastrophic events '' should be observed . one of the crucial points that was derived from the solution of the dynamical equations in mean field models of spin glasses is the fact that , in a nonequilibrium system , the fluctuation dissipation theorem ( fdt ) is violated in a nontrivial way . we will first briefly recall the main results from mean - field theories of spin glasses , and then discuss how the violation of fdt can be quantified in our system . let us consider an observable @xmath66 whose normalized autocorrelation function is denoted by @xmath145 . if @xmath150 denotes a field conjugate to @xmath66 , the response of @xmath66 to @xmath150 is defined as @xmath151 . in an equilibrium system , this response function is invariant under time translation , i.e. @xmath152 , and is related to the correlation function by the fluctuation dissipation theorem , @xmath153 . out of equilibrium , this relation does not hold any more , and a `` fluctuation dissipation ratio '' ( fdr ) @xmath154 can be defined as @xmath155 much attention has been devoted to the asymptotic behavior of this fluctuation dissipation ratio . in particular , it has been shown @xcite that in the asymptotic limit @xmath156 , the fluctuation dissipation ratio in mean field models of spin glasses , as well as in coarsening systems , becomes a non - singular function of @xmath157 , i.e. @xmath158 . a direct consequence from this is that the more easily accessible integrated response @xmath159 can be written as a continuous function of @xmath145 @xmath160 from a practical point of view , the nontrivial consequence of these statements is that a parametric plot of @xmath161 versus @xmath91 ( with @xmath3 as the parameter ) should , for long times , converge towards a master curve , independent of the waiting time . such a behavior has been observed in a number of spin glass simulations @xcite . in order to test whether the same property holds in structural glasses , we have to devise a way of measuring the response function associated to @xmath162 , which is the only correlation function that can be obtained with a reasonable accuracy in the aging regime . the procedure we use is the following . a fictive charge " @xmath163 is assigned randomly to each particle . an additional term of the form @xmath164 , where @xmath165 is a small ( @xmath166 external potential , is then added to the hamiltonian . it is then easy to check that , _ if one averages over several realizations of the random charge distribution _ , the time - correlation function of the observable @xmath167 is the incoherent scattering function @xmath124 . the procedure to generate the response function associated to @xmath124 is thus straightforward : for a given realization of the random charge distribution , the system is equilibrated at a high temperature ( @xmath50 ) , and quenched at @xmath37 to the desired final temperature @xmath39 . the evolution is followed with the field @xmath168 off until a waiting time @xmath0 , then the field is switched on and the response @xmath169 is monitored . the same procedure is repeated for several ( 7 to 10 ) realizations of the charge distribution , in order to get the response function . the quantity we obtain by this procedure is then an integrated response function @xmath161 ( eq . [ eq18 ] ) , since @xmath170 we have checked that this procedure indeed yields the correct response function by checking the fluctuation dissipation theorem in an equilibrium system . in that case , a slightly different procedure was used , in the sense that we monitored the decay of the response after switching off the field . the fluctuation dissipation theorem implies that this decay is directly proportional to the correlation function , a result that is illustrated in fig . [ fig12 ] . from this figure we see that the quality of the response data is significantly poorer than the one for the correlation data . this is due to the fact that the latter can be averaged over time origins and wave - vector directions , hence improving drastically the statistical accuracy . for the nonequilibrium case typical results for the integrated response function are shown in fig . [ fig13 ] , for @xmath45 , @xmath171 and two values of the waiting time . as expected , the response , like the correlation , gets slower as the waiting time increases . as in the equilibrium case , the quality of the response data is poorer than that of the correlation data . in that case , the average over wave - vector directions in obtaining the correlation data probably explains this difference . despite this noise it can be seen from the figure that at long times the curve for the response lies above the one for the correlation function , i.e. that the response is smaller than expected from the fdt . as explained above , mean field theories of spin glasses suggest that interesting information can be obtained from a parametric plot of @xmath172 versus @xmath173 at fixed @xmath0 . such a plot is shown in figures [ fig14 ] ( a - c ) for three values of the final temperature , @xmath45 , @xmath120 and @xmath51 . the wave - vector is @xmath171 . for each temperature , results obtained at different waiting times are shown . in spite of the scatter in the data , the figures are clearly compatible with the existence of a master curve independent of the waiting time . obviously , the quality of the data does not allow a quantitative analysis of this master curve . we can nevertheless argue that two different regimes can be distinguished . for high values of the correlation ( @xmath124 close to unity ) , the parametric plot is essentially a straight line with slope @xmath174 . a slope of @xmath174 would be observed in an equilibrium system obeying fdt . at lower values of the correlation , a clear deviation from this fdt slope is observed . although other fits are certainly possible , we can describe the parametric plot as consisting of two straight lines , one with slope @xmath174 and one with a negative slope @xmath175 . the reason for choosing such a description and the corresponding interpretation will be discussed in the next section . a similar analysis was performed for another value of the wave - vector ( @xmath176 ) . very similar results are obtained , and the slope of @xmath24 of the non - fdt part in the parametric plot seems to be independent of the wave - vector . before discussing the results , we mention that the results for this parametric plot are very sensitive to the linearity of the response . hence we always checked carefully that our results were independent of the amplitude of the applied field . in practice , applying a stronger field does not affect the fdt part of the parametric plot , but tends to affect strongly the second part , in that it leads to a decrease in the parameter @xmath24 that describes the non - fdt part of the curve . the importance of the fluctuation dissipation ratio @xmath177 in glassy systems was first recognized in the context of the mean field theory of spin glasses @xcite , when it appeared that this fdr was intimately related to the nature of ergodicity breaking in the system . the nature of this relationship was recently clarified by franz _ @xcite , who showed that the asymptotic value of the fdr could be quite generally expressed in terms of the probability distribution @xmath178 of overlaps between replicas introduced by parisi @xcite through @xmath179 this relation between a purely static quantity , @xmath178 , and a dynamical one , @xmath177 , is extremely powerful , since it implies that a nontrivial information on the nature of phase space may be encoded in the nonequilibrium time dependent behavior . if we accept equation ( [ eq20 ] ) , a small number of possible scenarios are documented in the literature @xcite . for systems with continuous step replica symmetry breaking , @xmath178 is a continuous function , and so is @xmath177 . for systems with one step replica symmetry breaking like @xmath180-spins systems ( @xmath181 ) , @xmath178 consists of two @xmath182-functions at @xmath183 and @xmath184 , so that @xmath185 for @xmath186 and @xmath187 for @xmath188 . finally simple coarsening systems , like the ising system below its ferromagnetic transition , have an essentially trivial @xmath178 , @xmath189 , where @xmath190 is the magnetization . therefore , in that case the fdr is 1 if @xmath191 , and 0 if @xmath192 @xcite . within this theoretical framework , we have to find which scenario is compatible with the results presented in the last section . from figs . [ fig14 ] , it would appear that the most likely scenario is that of a system with one step replica symmetry breaking , for which @xmath177 is a stepwise constant function . in that case the parametric plot consists in two straight lines , one with slope @xmath174 and one with slope @xmath193 ( bold straight lines ) . obviously the asymptotic nature of the results can be questioned . however , the fact that we obtained an essentially @xmath0 independent plot is already an indication that we are approaching the asymptotic limit . it could also be that preasymptotic effects cancel out when the parametric plot is used . indeed , such a plot is even obtained for @xmath51 , where we have seen that strong `` catastrophic '' events influence the relaxation . finally , we mention that in systems in which preasymptotic effects have been observed and studied ( mostly domain growth models ) @xcite , they very clearly show up in the parametric plot as a @xmath0 dependence of the crossover region between the fdt part and the non - fdt region . the overall shape of the parametric plot is not affected much . such a dependence could then explain the fact that in our data , the crossover between the fdt part and the non - fdt part takes place at a value of @xmath145 _ smaller _ than the plateau value @xmath146 in the correlation function , a feature which contradicts theoretical expectations . the tendency of the curves to `` overshoot '' in the crossover region , perceptible in fig . [ fig14 ] , is reminiscent of the observations made in reference @xcite , and could also be a transient effect . in that paper it was argued that , e.g. , in a coarsening system with some defects the de - pinning of a domain wall from a defect will give rise to a strong enhancement of the response just after the event . since the `` catastrophic '' events correspond to a violent and , most likely , pre - asymptotic release of the stress , it is reasonable to assume that also in this case the response will be larger than expected . if we accept that our parametric plots correspond to the one step replica symmetry breaking case , we can extract from the data an estimate for the slope @xmath194 of the non - fdt part . the resulting slopes are given by : @xmath45 , @xmath195 ; @xmath120 , @xmath196 ; @xmath51 , @xmath197 . as mentioned above , these results appear to be independent of the wave - vector . within the accuracy of our data , they are compatible with a linear dependence of @xmath24 on @xmath39 , quite similar to that found by parisi @xcite for a soft - sphere system . we also note that the value found for @xmath45 is compatible with the result @xmath198 that was extracted from a scaling analysis of the correlation functions alone . it is interesting to analyze our results for @xmath24 using the `` effective temperature '' concept introduced in reference @xcite . ( for a different approach to the effective temperature see the papers of nieuwenhuizen @xcite ) . in this approach , an effective `` fluctuation dissipation temperature '' @xmath199 is defined as @xmath200 , where @xmath142 is the actual external temperature . crudely speaking , the `` fast '' degrees of freedom ( those that correspond to the rapid decay of @xmath145 to its plateau value ) are at equilibrium with the thermostat at temperature @xmath142 , while the slow degrees of freedom , which govern the aging behavior , are at equilibrium at a higher temperature . then the concept of effective temperature can be thought as a rationalization of the older `` fictive temperature '' concept @xcite . within this interpretation , a linear dependence of @xmath201 on @xmath39 corresponds to a constant effective temperature . it is therefore quite natural to expect first that @xmath201 should be equal to @xmath202 , where @xmath33 is the mode coupling critical temperature , since systems cooled below @xmath33 fall out of equilibrium at @xmath33 . this result was proposed by parisi @xcite on the basis of his simulations in a soft sphere system . our results do not corroborate such a simple interpretation . the `` effective temperature '' in our aging system is substantially larger than @xmath33 , typically @xmath203 . our results here are similar those of alvarez _ et al . _ @xcite for the @xmath180-spin model in that these authors found for a temperature a bit below @xmath33 a value of @xmath24 which is significantly smaller than 1 . indeed , the `` ideal '' result @xmath204 can only be expected to hold for systems that would be cooled infinitely slowly , so that they would remain in equilibrium down to @xmath33 , and for which ergodicity restoring hopping processes can be completely neglected . for a system that is cooled with a finite cooling rate , it is not surprising to find an effective temperature which is above @xmath33 . in this paper , we have presented a detailed numerical study of the out of equilibrium relaxation ( `` aging '' ) of a simple glass forming system . the time scales we have investigated are those allowed by the current possibilities of molecular dynamics simulations , typically @xmath205s . we will now try to briefly summarize the main conclusions that can be inferred from these numerical results . first , it appears that static quantities are , on these time scales , only very weakly dependent of time . by static ( or `` one time '' ) quantities we mean , as usual , those quantities that can be obtained from the knowledge of a single configuration of the system . in particular , we have shown that the pair correlation functions ( and hence all the derived quantities like the energy or the pressure ) appear to equilibrate very quickly after the quench . these one - time quantities show some sensitivity to the quench history and to the final temperature and we find that to a first approximation they correspond to configurations which are very close to the critical surface of ( equilibrium ) mode - coupling theory . hence we have evidence that for this model and within the time span of the simulation the system is not able to penetrate this surface and hence remains in the liquid like part of configuration space . in contrast to this weak dependence , the two - time correlation functions are very sensitive to the aging time . when the quench temperature is relatively high ( i.e. still close to the mode - coupling critical temperature that was identified for our system ) , the waiting time dependence of these functions can be described in rather simple terms . functions obtained for various @xmath0 can be rescaled on a master curve , and an analysis of this curve yields results that are compatible with the predictions of mean field ( `` mode - coupling '' ) theories of the glass transition formulated for the `` @xmath180-spin '' spin glass models . at lower final temperatures , the behaviour of the two - time correlation functions is much more complex . the relaxation is largely dominated by `` catastrophic events '' that involve a small , but collective , displacement of a large number of particles . these infrequent and large events are very difficult to average over , so that the statistical quality of the data for the correlation functions is rather poor . due to the limited time range of the simulations , we do not know whether these events are transients effects releasing some initial stresses due to the quench , or constitute a genuine characteristic of low temperature relaxation . the flexibility of md simulations allows an independent measurement of the two - time response functions , which at equilibrium would be related to the correlation functions through the fluctuation dissipation theorem ( fdt ) . a parametric representation of the response versus the correlation allows to clearly distinguish between a `` quasi - equilibrium '' region in which the fdt holds , and a second regime in which it is violated . an essential feature is that this parametric plot does not ( or , at most , very weakly ) depend on waiting time , so that the `` fdt '' and `` non - fdt '' regimes do not correspond to _ time windows _ but rather to _ correlation _ windows . this remarkable feature , predicted by mean field theories of spin glasses to hold in the asymptotic limit , is observed here for structural glasses at finite times . this is likely related to the formal similarities between the mode - coupling theories of structural glasses , which are known to describe quite well the equilibrium behaviour of our system , and these mean field models of spin glasses . as the response / correlation relationship observed for our model resembles the one observed in spin systems with one step replica symmetry breaking , one may speculate that the phase space structures in both systems are also similar . if this is the case , a quite appealing scenario for the glass transition can be devised , that reconciles some of the old ideas of the adams - gibbs approach with the more modern mode - coupling scenario , as discussed in detail in reference @xcite . unfortunately , we presently do not have at our disposal a quantitative theory of nonequilibrium behaviour in structural glasses . the theoretical framework currently provided by mean - field theories of spin glasses can be considered as a schematic model , and hence does not allow comparisons beyond the qualitative level . at this level , we find that the general behaviour of our system is in quite good agreement with this schematic model . the agreement ( and disagreements ) are somewhat reminiscent of what is observed when comparing equilibrium data with predictions of schematic mode - coupling theories . finally , it is natural to inquire about the relevance of a work on aging phenomena based on a method that is limited to time scales smaller than @xmath7s to the aging effects observed on much longer time scales . the general picture we obtain provides , however , some support for a scenario @xcite that may be of general relevance for much longer time scales . in this scenario , the critical temperature @xmath33 of mode coupling theory is associated to the dynamical freezing temperature of mean field spin models . in a purely mean field ( or , in other words , ideal mode - coupling ) situation , the system would be out of equilibrium at all temperatures below @xmath33 . of course , real systems can still be equilibrated below @xmath33 , as `` hopping processes '' allow a relaxation that is not described by mode - coupling theory ( at least in its ideal version ) . hence in real experiments ( or in our system , if we could allow for longer simulation times ) these systems will show only interrupted aging , i.e. aging for short waiting times and no aging at long waiting times . however , we have shown that if the system was artificially driven into a nonequilibrium situation , its `` short time aging '' can be described reasonably well within the framework of mean - field / mode - coupling theories . we may here draw a parallel with the fact that , in equilibrium , below @xmath33 , when the @xmath2 ( terminal ) relaxation becomes dominated by hopping processes , the @xmath1 ( intermediate times ) relaxation is still very well accounted for by ideal mode coupling theory @xcite . it is then quite tempting to speculate that for any `` fragile '' system ( i.e. a system in which hopping processes are not too strong , so that mode coupling theory accounts well for the dynamics close to @xmath33 ) quenched below its glass transition , the mean - field / mode - coupling description will be relevant for a large fraction of the aging process . this should be particularly relevant in systems where hopping processes are known to be weak , like colloidal suspensions . we acknoweledge fruitful discussions with l. cugliandolo , j. kurchan and a. latz . this work was supported by the ple scientifique de modlisation numrique at ens - lyon , and by the deutsche forschungsgemeinschaft under sfb262 . for a recent review see the series of papers in science , * 267 * , 1924 ( 1995 ) and the proceedings of _ third international discussion meeting on relaxations in complex systems _ , eds . k. l. ngai , e. riande , m. d. ingram , j. non - crystal . solids * 235 - 237 * ( 1998 ) . see , e.g. , j .- barrat and m. l. klein , annu . phys . chem . * 42 * , 23 ( 1991 ) ; w. kob , p. 1 in _ annual reviews of computational physics _ vol iii , ed . d. stauffer ( world scientific , singapore , 1995 ) ; w. kob , j. phys . : condens . matter * 11 * r85 ( 1999 ) . w. gtze , p. 287 in _ liquids , freezing and the glass transition _ eds . : j. p. hansen , d. levesque and j. zinn - justin , les houches . session li , 1989 , ( north - holland , amsterdam , 1991 ) ; w. gtze and l. sjgren , rep . . phys . * 55 * , 241 ( 1992 ) ; w. kob , p. 28 in _ experimental and theoretical approaches to supercooled liquids : advances and novel applications _ eds . : j. fourkas , d. kivelson , u. mohanty , and k. nelson ( acs books , washington , 1997 ) . l. c. e. struik , _ physical aging in amorphous polymers and other materials _ ( elsevier , amsterdam , 1978 ) ; g. b. mckenna , p. 311 in _ comprehensive polymer science , 12 , polymer properties _ , eds . : c. booth and c. price ( pergamon , oxford , 1989 ) . bouchaud , l.f . cugliandolo , j. kurchan , and m. mzard , physica a * 226 * , 243 ( 1996 ) ; j .- bouchaud , l.f . cugliandolo , j. kurchan and m. mzard in _ spin glasses and random fields _ , ed . young ( world scientific , singapore , 1998 ) ; preprint condmat/9511042 .

we have determined the fundamental properties and distance of a fourth eclipsing binary system ( eb ) in the large magellanic cloud , hv 5936 (b0.5 v +b2 iii ) . as in our previous studies , we combine `` classical '' eb light curve and radial velocity curve analyses with modeling of the uv - through - optical spectral energy distribution of hv 5936 to produce a detailed characterization of the system . in this paper , we also include an analysis of the high - resolution optical absorption line spectra of the binary components . we find hv 5936 to be an algol - class system , in which the masses of the primary and secondary stars have evolved via mass transfer to their current values of 11.6 and 4.7 , respectively . the properties of the primary star are indistinguishable from those of a `` normal '' star of the same current mass . the secondary is found to be overluminous for its current mass and exhibits a factor - of-2 enhancement in its surface he abundance . these results are compatible with `` case a '' mass exchange occurring during the core hydrogen burning phase of the current secondary . the distance derived to the system , kpc , implies a distance of.3 to the optical center of the lmc . this is several kpc closer than found in our analyses of other systems and we suggest that hv 5936 lies `` above '' the lmc disk . this is supported by the very low interstellar h i column density and low found for the system both of which are consistent with expected milky way foreground material and may be associated with hv 5936 s location near the lmc supergiant shell lmc 4 .

astro-ph0301296

this is the fourth in a series of papers presenting results from detailed analyses of b - type eclipsing binary ( eb ) systems in the large magellanic cloud ( lmc ) . our primary scientific goals are : 1 ) to determine an essentially complete description of the stellar properties of each system , thus providing tests and constraints for stellar evolution theory ; and 2 ) to measure precise individual distance for each system , from which the general distance to the lmc can be derived . because of its role as a fundamental calibrator for distance indicators , the lmc s distance is particularly important for determining the size scale of the universe , and its current uncertainty of 10 - 15% contributes considerably to the uncertainty in the hubble constant ( e.g. , mould et al . 2000 ) . in our previous studies , we examined the lmc eb systems hv 2274 ( guinan et al . 1998 , hereafter `` paper i '' ; ribas et al . 2000a ) , hv 982 ( fitzpatrick et al . 2002 , `` paper ii '' ) , and eros 1044 ( ribas et al . 2002 , `` paper iii '' ) . the apparent locations of these systems in the lmc can be seen in figure [ figlmc ] . the results of these analyses are beautifully consistent with expectations from stellar structure theory and provide strong constraints on the distance to the lmc . as discussed in paper iii , the three individual distances are all consistent with a mean of @xmath4 kpc , although there is a suggestion that hv 982 and , by association , perhaps the general 30 doradus region , may lie behind the lmc s `` bar '' , by several kpc . conclusions about the lmc s distance are somewhat dependent on assumptions about its spatial orientation and need to be strengthened with additional measurements from the remaining eclipsing binaries in our program . our approach is ideally suited to pursuing the issues of the spatial orientation , structure , and general distance to the lmc , since we measure precise distances to individuals systems which are widely spread across the face of the lmc . in this paper , we apply our analysis to a fourth lmc b - type eb , hv 5936 , and derive its stellar properties and distance . this system , with @xmath5 , stands in some contrast to those in our previous studies in both its composition and its location . hv 5936 is a semidetached system , in which the cooler , less massive component fills completely its roche lobe . thus , the currently more massive ( and more luminous ) component began its life as the junior , lower mass member of the binary . this provides an excellent opportunity to examine the characteristics of the massive component , which due to the rapid dynamical relaxation expected to follow mass transfer should be indistinguishable from a `` normal '' star of the same current mass . in addition to this feature , hv 5936 is located in a very different part of the lmc from our previous targets ( see fig . [ figlmc ] ) , lying several degrees north of the lmc s bar and superimposed on a well known `` hole '' in the lmc s h i distribution ( mcgee & milton 1966 ) , corresponding to the supergiant shell lmc 4 ( meaburn 1980 ) . its distance should reflect strongly the spatial orientation of the lmc , i.e. , its inclination angle and line of nodes orientation , and could provide constraints on these factors . the structure of this paper is similar to our earlier works . in 2 we describe the data included in the study . in 3 we describe , and present the results from , our standard analysis , which incorporates the binary s light curve , radial velocity curve , and uv - through - optical spectral energy distribution . a study of the high - resolution optical spectrum of the hv 5936 components is presented in 4 . we combine all our results and give a detailed characterization of the properties and likely history of the hv 5936 system in 5 . some aspects of our results relating to the interstellar medium towards hv 5936 are described in 6 , including an indication of the relative location of hv 5936 within the lmc . finally , we derive the distance to the system and compare it with our previous results in 7 and provide some summary comments in 8 . three distinct datasets are required to carry out our analyses of the lmc eb systems : precise differential photometry ( yielding light curves ) , medium - resolution spectroscopy ( yielding radial velocity curves ) , and multiwavelength spectrophotometry ( yielding temperature and reddening information ) . each of these three datasets are described briefly below . as in our previous papers , the primary ( @xmath6 ) and secondary ( @xmath7 ) components of the hv 5936 system are defined photometrically and refer to the hotter and cooler components , respectively . ccd differential photometric observations of hv 5936 were reported by jensen , clausen , & gimnez ( 1988 ) . these data were obtained between 1983 and 1984 with a 1.54-m telescope at the european southern observatory ( la silla , chile ) . the published light curves , obtained in the johnson @xmath8 and @xmath9 passbands , have fairly good phase coverage , with 44 and 144 measurements , respectively . according to jensen et al . , the precision of the individual differential photometric measurements is better than 0.010 mag . in this study , we adopt the orbital ephemeris determined by jensen et al . : @xmath10 radial velocity curves for hv 5936 , and a number of other lmc ebs , were derived from optical echelle spectra obtained by us during 6-night and 8-night observing runs in january and december 2000 , respectively , with the blanco 4-m telescope at cerro tololo inter - american observatory in chile . the seeing conditions during the two runs ranged between 0.7 and 1.8 arcsec . we secured eighteen spectra of hv 5936 near orbital quadratures covering the wavelength range 36005500 , with a spectral resolution of @xmath11 , and a s / n of @xmath020:1 . the plate scale of the data is 0.08 pix@xmath12 ( 5.3 km s@xmath12 pix@xmath12 ) and there are 2.6 pixels per resolution element . identical instrumental setups were used for both observing runs . the exposure time per spectrum was 1800 sec , sufficiently short to avoid significant radial velocity shifts during the integrations . all the hv 5936 observations were bracketed with thar comparison spectra for proper wavelength calibration . the raw images were reduced using standard noao / iraf tasks ( including bias subtraction , flat field correction , sky - background subtraction , cosmic ray removal , extraction of the orders , dispersion correction , merging , and continuum normalization ) . a typical spectrum is shown in figure [ figspec ] . the h i balmer lines and the strongest he i features are labeled , with arrows marking the expected line positions for the two components of the system ( according to the radial velocity curve solution described in 3.2 ) . this spectrum was obtained at orbital phase 0.30 and illustrates the clean velocity separation of the two components of the binary . to determine the radial velocities of hv 5936 s two components from the echelle spectra , we followed the procedure described in paper ii utilizing the korel program ( hadrava 1995 , 1997 ) . korel is based on the `` spectral disentangling '' technique , which assumes that an observed double - lined spectrum is a simple linear combination of two single - lined spectra whose velocities reflect the orbital properties of the system . when applied to a set of spectra obtained over a range of orbital phases korel yields component velocities ( relative to the system barycenter ) for each individual spectrum and `` disentangled '' spectra for each of the two binary members , combining information from the whole ensemble of spectra . we applied the korel analysis to our echelle data in the 40005000 wavelength region . the h@xmath13 , h@xmath14 and h@xmath15 lines were excluded by setting the normalized flux to unity in a window around their central wavelength . to translate the velocities to the heliocentric system , we determined the systemic velocity of hv 5936 from the individual disentangled spectra . cross - correlation of these extracted spectra with a high s / n ( @xmath0250 ) observed spectrum ( hr 1443 , b2 iv - v ) and with a synthetic template yielded values consistent with a systemic velocity of @xmath16 km s@xmath12 . table [ tabrv ] lists the heliocentric radial velocities derived from all the ctio spectra using the procedure outlined above ( `` rv@xmath17 '' and `` rv@xmath18 '' ) . also listed are the dates of observation and the corresponding phases . a large number of korel runs from different starting points were carried out to explore the parameter space and make realistic estimations of the uncertainties . a detailed discussion of these is left for 3.2 . the individual disentangled spectra of the primary and secondary provide valuable insight into the nature of the hv 5936 system , as well as confirmation of the results from our general analysis . these spectra are discussed in 4 . we obtained spectrophotometric observations of hv 5936 at uv and optical wavelengths with the _ hubble space telescope _ using both the faint object spectrograph ( fos ) and the space telescope imaging spectrograph ( stis ) . these data are summarized in table [ tabhst ] . the fos observations utilized the @xmath19 aperture , yielding a spectral resolution of @xmath20 . the four individual spectra were processed and calibrated using the standard pipeline processing software for the fos and then merged to form a single spectrum which covers the range 1145 to 4790 . the stis observations utilized the @xmath21 aperture , yielding a spectral resolution of @xmath22 . these data were processed and calibrated using the standard pipeline processing software . cosmic ray blemishes were cleaned `` by hand '' and the g430l and g750l spectra were trimmed to the regions 35105690 and 54107490 , respectively . because of concerns about the photometric zeropoints and stability , the two stis spectra were not merged ( see 3.3 ) . our study of hv 5936 proceeds from three separate but interdependent analyses . these involve the radial velocity curve , the light curve , and the spectral energy distribution ( sed ) . the combined results provide essentially a complete description of the gross physical properties of the hv 5936 system and a precise measurement of its distance . each of the three analyses is described below . the fits to the light curves were carried out using an improved version of the wilson - devinney ( w - d ) program ( wilson & devinney 1971 ) that includes a model atmosphere routine developed by milone , stagg & kurucz ( 1992 ) for the computation of the stellar radiative parameters . both detailed reflection model ( mref=2 , nref=1 ) and proximity effect corrections were taken into account when fitting the light curves . the bolometric albedo and the gravity - brightening coefficients were both set at the canonical value of 1.0 for stars with radiative envelopes . for the limb darkening we used a logarithmic law as defined in klinglesmith & sobieski ( 1970 ) , with first- and second - order coefficients interpolated at each iteration for the exact @xmath23 and @xmath24 of each component from a set of tables computed in advance using a grid of atlas9 model atmospheres . a mass ratio of @xmath25 was adopted from the spectroscopic solution ( 3.2 ) , and the temperature of the primary component was set to @xmath26 k , as discussed in [ secsed ] . we have adopted a circular orbit as suggested by the equal width of the eclipses and the occurrence of the secondary eclipse at exactly phase 0.5 . further support for this comes from the fact that the system is semidetached ( see below ) and orbital circularization takes place over a very short timescale . finally , the rotational velocity of the primary star was set to 1.25 times the synchronous rate and the secondary star was adopted to rotate synchronously . a discussion of the component s rotational velocities is provided in 4 . our initial light curve fits were run with a detached configuration ( w - d mode 2 ) . however , preliminary tests indicated a rapid decrease of the surface gravitational potential of the secondary component until reaching its critical value . several runs from different starting values confirmed this behavior . therefore , all further w - d solutions were run in mode 5 , i.e. , secondary component filling its roche critical surface . according to this , hv 5936 is a semidetached binary where the cooler , less massive component appears to be more evolved and fills its roche lobe . this is the classical configuration of post - mass transfer algol - class systems . in our final analysis we solved for the following light curve parameters : the orbital inclination ( @xmath27 ) , the temperature of the secondary ( @xmath28 ) , the gravitational potential of the primary ( @xmath29 ) , the luminosity of the primary ( @xmath30 ) , and a phase offset ( @xmath31 ) which accounts for possible inaccuracy in the adopted ephemeris reference epoch . the iterations with the w - d code were carried out automatically until convergence , and a solution was defined as the set of parameters for which the differential corrections suggested by the program were smaller than the internal probable errors on three consecutive iterations . as a general rule , several runs with different starting parameters are used to make realistic estimates of the uncertainties and to test the uniqueness of the solution . because of the relatively small number of free parameters and the constraint provided by the semidetached configuration , convergence was achieved rapidly in the light curve fits . the r.m.s . residuals were determined to be 0.010 , and 0.009 mag for the @xmath8 and @xmath9 light curves , respectively . these values are approximately equal to the observed scatter of the observations . the resulting orbital and physical parameters are well - defined and the best - fitting model light curves , together with the o c residuals , are shown in figure [ figlc ] . as can be seen in the figure , a small systematic departure of the fit arises in the first quadrature ( phase 0.20.4 ) of the @xmath9 light curve . this can not be confirmed in the @xmath8 light curve because of the lack of photometric coverage . it is uncertain at this point whether this is a real effect or just an artifact of the photometric reduction . if it were real , the asymmetry of the quadrature maxima could arise from the presence of a hot spot on the primary component . this is not unexpected in an algol system because active mass transfer could be taking place and a hot spot in the atmosphere of the hotter component might arise from the impact of the accreting material and subsequent kinetic heating . we explored this scenario by running w - d solutions with an area on the primary component 10% hotter than the rest of the atmosphere . the systematic trend in the residuals did indeed disappear when the spot was included but this is not surprising because we added two new free parameters in the analysis ( spot radius and location ) . the new r.m.s . residuals were found to be 0.009 and 0.008 mag for the @xmath8 and @xmath9 light curves , respectively . the solution with the hot spot is very similar but yields a radius for the primary component about 2% smaller and a temperature ratio about 5% larger . in the absence of conclusive evidence , we decided to adopt the `` unperturbed '' solution without a hot spot . besides , the resulting spot location is not consistent with the expected impact site of a putative stream of matter using the model of lubow & shu ( 1975 ) . ( note that the fractional radius of the primary component is sufficiently large to prevent the formation of a stable accretion disk ( e.g. albright & richards 1993 ) . ) as is evident from figure [ figlc ] , the light curves of hv 5936 display rather prominent out - of - eclipse variability . this arises chiefly from variations in the effective radiating area of the secondary star . due to its non - spherical shape ( filling its roche lobe ) , the surface area of the secondary star changes with the orbital phase and so does its total brightness . the stellar cross - section reaches a maximum at the orbital quadratures and a minimum at the eclipses . another factor contributing to the out - of - eclipse variation is irradiation . this is a rather significant effect in algols because the secondary star is often larger but cooler that the primary . thus , when the primary component is in front ( near the phase of the secondary eclipse ) , we see its light reflected off the atmosphere of the secondary star . this causes the ingress and egress of the secondary eclipse to be brighter than the ingress and egress of the primary eclipse . both effects discussed here are fully accounted for by the physical model upon which w - d is based , as proven by the excellent fit to the light curves . the final orbital and stellar parameters adopted from the light curve analysis are listed in table [ tabparms ] . the uncertainties given in this table were not adopted from the formal probable errors provided by the w - d code , but instead from numerical simulations and other considerations . several sets of starting parameters were tried in order to explore the full extent of the parameter space . in addition , the w - d iterations were not stopped after a solution was found , instead , the program was kept running to test the stability of the solution and the geometry of the @xmath32 function near the minimum . the scatter in the resulting parameters from numerous additional solutions yielded estimated uncertainties that we consider to be more realistic , and are generally several times larger than the internal statistical errors . as an internal consistency check , we re - analyzed the light curves with the mass ratio @xmath33 left as a free parameter , rather than fixing it to the spectroscopically determined value . this test yielded a `` photometric mass ratio '' of @xmath34 which is in excellent with the spectroscopic result of @xmath35 ( see 3.2 ) . photometric estimates of q are strongly dependent on outside - of - eclipse light variations which , in the case of hv 5936 , arise primarily from the changing aspects of the tidally distorted stars ( as well as the reflection effect ) . the good agreement of @xmath36 with the directly determined spectroscopic mass ratio indicates that the light curves are essentially free of significant perturbing effects from gas flows and accretion heating consistent with our conclusion above from examining the residuals to the light curve fits . this agreement reaffirms that the orbital and stellar properties determined from the combined solutions of light and radial velocity curves are both self - consistent and robust . the same light curves as analyzed here ( i.e. , from jensen et al . 1988 ) have also been studied by bell et al . those authors employed two different light - curve synthesis programs , one of which was a 1983 version of the w - d code . their solutions were run with a mass ratio quite different from ours ( @xmath37 ) and this resulted in a larger fractional radius for the secondary component than we find . also , their adopted temperature for the primary was about 3000 k larger than our value . apart from those , the rest of the light curve parameters obtained by bell et al . are compatible with the ones listed in table [ tabparms ] . it should be mentioned , however , that our fits display significantly smaller residuals , which probably result from a better - determined mass ratio and the more sophisticated fitting program that we have employed . the radial velocity curve was fit using the same version of the w - d program as described above . the free parameters were : the orbital semi - major axis ( @xmath38 ) , the mass ratio ( @xmath33 ) , and a velocity zero point ( the systemic radial velocity @xmath39 ) . the rest of the parameters were set to those resulting from the light curve solutions discussed in the previous section . the best fit to the radial velocity curve is shown in figure [ figrv ] . the fit residuals correspond to r.m.s . internal errors of 1.2 and 2.8 km s@xmath12 for the primary and secondary components , respectively . the relatively large difference between these residuals is a consequence of the secondary star being significantly less luminous , and thus its velocities have intrinsically larger errors . the parameters resulting from the radial velocity curve fit are listed in table [ tabparms ] . the uncertainties given in the table are not taken directly from the w - d output , since they fail to account for any systematic effects that could be present in the velocity data . instead , we estimated more realistic errors by considering the scatter of the velocities derived from the disentangling analysis of separate spectral regions . thus , we divided our entire spectrum into four wavelength intervals and analyzed these separately with korel . the standard deviation of the resulting velocities was found to be 3.2 and 7.5 km s@xmath12 for the primary and secondary components , respectively . we conservatively adopted these values as the uncertainty of the velocity semiamplitudes and scaled the rest of the parameter errors listed in table [ tabparms ] accordingly . independent radial velocity observations were secured and analyzed by bell et al . according to these authors , the observations were acquired in less - than - perfect conditions , with atmospheric seeing of 36 arcsec and intermittent clouds . their resulting radial velocity curves have large scatter , with `` o c '' residuals up to 80 km s@xmath12 . the better quality of our radial velocity data is due to both the excellent atmospheric conditions at ctio and the use of the spectral disentangling technique , which has been proved to be superior to classical cross - correlation . when comparing the solutions , the value of the velocity that bell et al . report for the primary ( more luminous ) component is in excellent agreement with our determination listed in table [ tabparms ] . however , the velocity semiamplitude of the less luminous component , more severely affected by poor - quality observations , found by bell et al . is much smaller ( @xmath050 km s@xmath12 ; 3.5@xmath40 ) than our value . in general , the observed sed @xmath41 of a binary system can be expressed as : @xmath42 \times 10^{-0.4 e(b - v ) [ k(\lambda - v ) + r(v)]}\end{aligned}\ ] ] where @xmath43 @xmath44 are the surface fluxes of the primary and secondary stars , the @xmath45 are the absolute radii of the components , and @xmath46 is the distance to the binary . the last term carries the extinction information , including @xmath3 , the normalized extinction curve @xmath47 , and the ratio of selective - to - total extinction in the @xmath9 band @xmath48 . in our studies , we represent the stellar surface fluxes with r.l . atlas9 atmospheres and use a parameterized representation of uv - through - ir extinction based on the work of fitzpatrick & massa ( 1990 ) and fitzpatrick ( 1999 ; hereafter f99 ) . the kurucz models are each functions of four parameters ( @xmath23 , @xmath24 , [ m / h ] , and microturbulence velocity @xmath49 ) , and the extinction curves are functions of six parameters ( see f99 ) . note that , for the purposes of eq . [ basic1 ] , it does not matter which star in a system is identified as the primary . we model the observed sed by performing a non - linear least squares fit to determine the best - fit values of all parameters which contribute to the right side of equation [ basic1 ] . for hv 5936 ( and as for our previous studies ) , we can make several simplifications which reduce the number of free parameters in the problem : ( 1 ) the temperature ratio of the two stars is known from the light curve analysis ; ( 2 ) the surface gravities can be determined by combining results from the light and radial velocity curve analyses and are log g = 3.98 and 3.49 for the primary and secondary stars , respectively ( see 5 ) ; ( 3 ) the values of [ m / h ] and @xmath49 can be assumed to be identical for both components ; ( 4 ) the ratio @xmath50 is known ; and ( 5 ) the standard mean value of @xmath51 found for the milky way can reasonably be assumed given the existing lmc measurements ( e.g. , koornneef 1982 ; morgan & nandy 1982 ; see 6 ) . we prepared the spectrophotometric datasets for the sed analysis by ( 1 ) velocity - shifting to bring the centroids of the stellar features to rest velocity ; ( 2 ) correcting for the presence of a strong interstellar h i ly@xmath52 absorption feature in the fos spectrum at 1215.7 ( see 6 ) ; and ( 3 ) binning to match the atlas9 wavelength grid . the statistical errors assigned to each bin were computed from the statistical errors of the original data , i.e. , @xmath53 , where the @xmath54 are the statistical errors of the individual spectrophotometric data points within a bin . for all the spectra , these binned uncertainties typically lie in the range 0.5% to 1.5% of the binned fluxes . the weighting factor for each bin in the least squares procedure is given by @xmath55 . we exclude a number of individual bins from the fit ( i.e. , set the weight to zero ) for the reasons discussed by fm99 ( mainly due to the presence of interstellar gas absorption features ) . as discussed in paper ii , we do not merge the fos and stis data into a single spectrum , but rather perform the fit on the three binned spectra simultaneously and independently . we assume that the fos fluxes represent the `` true '' flux level and account for zero point uncertainties in the stis data by incorporating two zeropoint corrections ( one for each stis spectrum ) in the fitting procedure . we later explicitly determine the uncertainties in the results introduced by zeropoint uncertainties in fos . as noted in 2.1 , the out - of - eclipse variations seen in hv 5936 s light curve result primarily from changes in the apparent size ( i.e. , as presented toward the earth ) of the secondary due to its mild non - sphericity and also from phase - modulated reflection of light from the primary off the secondary . these effects must be taken into account in the sed analysis because they affect the relative contributions of primary and secondary light to the observed sed . we incorporate the effects through a simple modification of eq . [ basic1 ] : @xmath56 \times 10^{-0.4 e(b - v ) [ k(\lambda - v ) + r(v)]}\end{aligned}\ ] ] where @xmath57 and @xmath58 are phase - dependent correction factors accounting for additional reflected light from the primary and the varying apparent size of the secondary , respectively . the values of @xmath57 and @xmath58 for each of the spectrophotometric observations are listed in table [ tabhst ] and were computed from the results of the w - d light curve analysis . the @xmath57 are always greater than 1 , since reflection always adds otherwise - unseen light from the primary . the value of @xmath58 for the fos / g130h spectrum indicates that , at the phase of this observation , the apparent size of the secondary was slightly smaller than its mean value ( as given by the `` volume radius '' computed by the w - d program ) . for all the other spectra , the apparent size of the secondary was larger than its mean value . note that all the corrections for the fos datasets which provide the photometric zeropoint for the sed analysis are very close to 1.0 . we computed the final fit to hv 5936 s sed utilizing eq . [ basic2 ] with the appropriate values of @xmath57 and @xmath58 inserted for each dataset . as in previous papers , we adjusted the weights in the fitting procedure to yield a final value of @xmath59 since the statistical errors of the data under - represent the total uncertainties ( see the discussion in paper ii ) . this was accomplished by quadratically adding an uncertainty equivalent to 1.9% of the local binned flux to the statistical uncertainty of each flux point . ( essentially identical results occur if the statistical errors are simply scaled by a factor of 2.2 to yield @xmath59 . ) this value of 1.9@xmath60 gives an indication of the general quality of the fit to hv 5936 s sed , excluding the effect of statistical noise . it is comparable the quality level we have seen in the previous analyses . the best - fitting values of the energy distribution parameters for hv 5936 and their associated 1@xmath40 uncertainties ( `` internal errors '' ) are listed in tables [ tabparms ] ( stellar properties ) , [ tabstis ] ( stis offsets ) , and [ tabext ] ( extinction curve parameters ) . a comparison between the observed spectra and the best - fitting model is shown in figure [ figsed ] . the three binned spectra are plotted separately in the figure for clarity ( small filled circles ) . the zeropoint offset corrections ( see table [ tabstis ] ) were applied to all stis spectra in figure [ figsed ] . note that we show the quantity @xmath61f@xmath62 as the ordinate in figure [ figsed ] ( rather than f@xmath62 ) strictly for plotting purposes , to `` flatten out '' the energy distributions . the correction factors of 10.2% and 5.9% are required to rectify the stis g430l and g750l spectra , respectively , are similar to the results found in papers ii and iii . this apparently systematic effect probably results from small light losses in the stis @xmath63 slit . this will be tested by using a wider slit in future stis observations . in 2.2 , we utilized the korel program primarily to determine the radial velocities of hv 5936 s component stars . however , byproducts of this program , i.e. , a high - resolution , `` disentangled '' optical spectrum for each star , can provide valuable additional information on the binary system which is _ entirely independent _ of the analysis described in the preceding section . this information can be tapped by modeling these absorption line spectra with synthetic spectra . the potential value of such an analysis is threefold , since it can : 1 ) add new information on the system , 2 ) independently verify results of the preceding analysis , or 3 ) identify problems with the preceding analysis . the disentangled spectra of hv 5936 s two components , as produced by the korel program are shown in figures [ figprim ] and [ figsecon ] . each spectrum has a resolution of @xmath00.2 and consists of 12,453 data points . the most prominent features in both are lines of h i and he i. the strong c ii @xmath614267 and si iii @xmath644553,4568,4574 lines are noted in the primary s spectrum ( fig . [ figprim ] ) . virtually all the other , weaker absorption features in that spectrum are due to lines of o ii . there are no positive identifications of individual metal absorption lines in the noisier spectrum of the secondary . the ripple - like structure in the range 41404170 of both spectra ( near the position of a he i line ) is an artifact of the korel program , and results because the observations were mainly obtained near orbital quadratures and not distributed more uniformly throughout the orbit . the strengths of all the lines in these two spectra are diluted by the presence of continuum light from both binary components . given the phase distribution of the original optical data ( see table 2 ) , we compute that the primary and secondary contribute 63.9% and 36.1% , respectively , of the continuum light in these mean spectra . these values incorporate the reflection and ellipticity effects noted in 3 , and have uncertainties of order @xmath01% . in figures [ figprim ] and [ figsecon ] , we have adjusted the vertical axes so that the lines are in their correct strengths relative to the bottoms of each panel . we model the disentangled spectra by utilizing atlas9 atmospheric structure models , ivan hubeny s spectral synthesis program synspec , and essentially the same @xmath32-minimization technique as described in 3.3 for the sed analysis . in general , finding the best - fitting synthetic spectrum for these stars requires the determination of seven parameters . four of these @xmath23 , @xmath24 , [ m / h ] , and @xmath49 are required to specify the appropriate atlas9 model . two more @xmath65 and @xmath66 ( along with @xmath49 ) are used explicitly by synspec in determining line positions and widths . a final parameter , which we characterize as the percent contribution of each star to the observed continuum , is required to reproduce the dilution of the line strengths . for the case of the hv 5936 stars , we found that the microturbulence velocity was poorly determined and so we simply adopted the value of 2.6 km s@xmath12 derived in the sed analysis in 3.3 . in addition , we simultaneously determined the coefficients of a high - order legendre polynomial to allow the smoothing out of `` bumps and wiggles '' which can be seen in the spectra of both stars . these result from deficiencies in the normalization of the original echelle spectra used by korel . the 41404170 region , noted above , was excluded from all fits . when analyzing the primary s spectrum we found that the results of the fitting procedure depended somewhat on the assumed order of the normalizing function . therefore , we performed 7 independent fits , utilizing 10 , 15 , 20 , 25 , 30 , 35 , or 40 order legendre polynomials . these sample the reasonable range , since a polynomial with fewer than 10 orders can not match the observed structure in the continuum , while those with more than 40 orders have too much freedom and can distort the observed features . the results of these fits are given in column ( 2 ) of table [ tabfine ] where we list the simple means of the parameters derived from the 7 fits . the uncertainties quoted in the table are the quadratic sum of the internal uncertainties in a single fit and the standard deviation of the ensemble of 7 fits . in figure [ figprim ] we show the best fitting model corresponding to the case with a 20-order legendre normalizing function ( thick solid curve ) . the polynomial itself is shown by the dotted line . the results from this particular fit are very close to the ensemble averages in table [ tabfine ] . the fit to the primary s disentangled spectrum is very good . the results for @xmath23 , [ m / h ] , and the contribution factor are all consistent at the 1@xmath40 level with previous determinations ( 26,450 k , @xmath67 , and 63.9% , respectively ) . the value of @xmath24 is 0.12 dex ( @xmath02.2@xmath40 ) higher than that determined from the binary analysis . however , this is not a large discrepancy and , since the gravity determination depends on the wings of the balmer lines , we suspect it arises from deficiencies in the normalization of the original echelle spectra . it is interesting that the metallicity determined here which mainly reflects weak but numerous lines of o ii agrees so well with that derived by the sed analysis based mainly fe ii and iii absorption in the uv . these two results present a very consistent picture of a general factor - of-4 metal underabundance for hv 5936 . note that all of the fits to the primary s spectrum assume a helium abundance of @xmath68 , which is based on the observed metal abundance and standard chemical enrichment laws ( see 5 ) and corresponds to a helium mass fraction of @xmath69 . fits to the secondary s spectrum were performed in a similar manner as above except that since the optical metal lines are all too weak to allow a meaningful determination of [ m / h ] we fixed it s value to @xmath70 = -0.63 $ ] as derived in 3 . also , we found that the 10-order legendre fit was inadequate to model the continuum undulations and so we base our results on 6 fits with normalization polynomials of 15 , 20 , 25 , 30 , 35 , and 40 orders . the resulting mean parameters and 1@xmath40 uncertainties from these fits are listed in column ( 3 ) of table [ tabfine ] . these results are much less satisfying than for the primary . @xmath23 , @xmath24 , and the contribution factor all differ greatly from the previously determined values ( 17,600 k , 3.49 , and 36.1% , respectively ) . in addition , the value of @xmath71 appears inconsistent with the virtually unavoidable requirement that the secondary s rotation be tidally locked to its orbit ( see 5 ) . from detailed examination of the fits to the secondary , we found that the discrepancies arise essentially because the secondary s he i lines are too strong to be well - fit with the temperature of 17,600 k inferred earlier in 3 . it is not likely , however , that the secondary can be significantly hotter than this , since this temperature results from a very well - determined eclipse - based temperature ratio and an apparently well - determined primary star temperature . rather , we suggest that the secondary s he i lines are enhanced in strength arising from a modest enhancement in the he surface abundance . the viability of this suggestion is demonstrated by case ii in column ( 4 ) of table [ tabfine ] . the mean parameters and 1@xmath40 uncertainties were derived in an identical manner to those in case i , except that the helium abundance was approximately doubled to a value of @xmath72 . in addition , the abundances of c and o were depleted by a factor of 10 , compared to the base metallicity of [ m / h ] = -0.626 , and n was enhanced by a factor of 70 . these factors are based on stellar interior models which will be discussed below in 5 . it is clear that these adjustments to the elemental abundances yield results completely consistent with the previous measurements of @xmath73 and @xmath24 for the secondary and the known value of the continuum contribution factor . in addition the value of @xmath71 is well - determined and consistent with synchronous rotation ( see 5 ) . in figure [ figsecon ] we show the best - fitting model corresponding to case ii with a 25-order legendre normalizing function ( thick solid curve ) . the results from this particular fit are very close to the ensemble averages in column ( 4 ) table [ tabfine ] . the value of @xmath74 used in case ii was not arrived at arbitrarily . rather , we ran a set of fits to the secondary s spectrum ( each with a different order legendre polynomial ) with @xmath73 held fixed at the value of 17,600 k found in 3 and with the he abundance as a free parameter . ( @xmath23 and @xmath75 are degenerate in their effects and can not both be determined from fitting the spectrum . ) this set of models resulted in a best - fitting he abundance of @xmath76 , which we rounded off to use in case ii . the significance of this result as an indicator of a true he enhancement can be gauged from a similar analysis of the primary . a set of fits of the primary s spectrum , with @xmath23 fixed at 26,450 k ( from the sed analysis ) and the he abundance as a free parameter , yielded a best fitting he abundance of @xmath77 . this is consistent with that inferred in 5 below , from the metallicity of the system and chemical enrichment laws . our conclusions from analysis of the high - resolution disentangled optical spectra of hv 5936 are twofold : 1 ) the primary s absorption line spectrum is completely consistent with the stellar properties derived in 3 , given the random errors and small systematic effects which may have arisen during the processing and `` disentangling '' of the spectra ; and 2 ) the secondary s spectrum is consistent with the derived properties only if a @xmath0factor - of-2 enhancement is assumed in the surface he abundance , as compared to that of the primary . in 5 below we consider the plausibility of a he enhancement in the secondary . we summarize the basic physical properties of the hv 5936 system in table [ tabstar ] . the notes to the table indicate how the individual stellar properties were derived from the results of the preceding sections . a scale model of the system is shown in figure [ fig3d ] . the locations of the hv 5936 components in the @xmath78 vs. @xmath79 diagram are shown in figure [ fighrd ] , where the skewed rectangular boxes indicate the 1@xmath40 error loci ( recall that errors in @xmath23 and @xmath80 are correlated ) . the position of the zero age main sequence ( zams ) is indicated in figure 9 by the thick curve ; the theoretical evolution tracks shown on the figure will be discussed below . examination of figure [ fighrd ] and table [ tabstar ] suggests a paradox , in that the more evolved star ( i.e. , farthest from the zams ) is the less massive component of the pair . the most plausible explanation for this scenario is that hv 5936 is a post - mass transfer system , as commonly argued to explain the so - called `` algol paradox . '' the originally more massive star evolved beyond the roche critical surface at a certain point and some fraction of its mass was transferred to its companion . after that process , the mass - accreting component became more massive than the mass - donor , and this is the current status of the system . in contrast with our previous studies , the strong interaction between the system components prevents us from using evolutionary models to perform a critical self - consistency check of the results . i.e. , a single isochrone is not expected to fit both components of the system because of their history of mass transfer . however , comparison with theoretical evolution tracks is still instructive . thus , we considered the evolutionary models of claret ( 1995 , 1997 ) and claret & gimnez ( 1995 , 1998 ) ( altogether referred to as the cg models ) . these models cover a wide range in both metallicity ( @xmath81 ) and initial helium abundance ( @xmath82 ) , incorporate the most modern input physics , and use a value of 0.2 h@xmath83 as the convective overshooting parameter . in our particular case , we adopted the metal abundance from the sed fit ( see table [ tabstar ] ) , which results in a value of @xmath84 . using this metal abundance , the empirical chemical law of ribas et al . ( 2000b ) yields a helium abundance of @xmath85 , corresponding to @xmath69 . ( the analysis of the disentangled spectrum of the primary in 4 provides a remarkable confirmation of this result . ) in figure [ fighrd ] we show the cg models for masses of 11.6 m@xmath86 and 4.7 m@xmath86 , corresponding to the masses measured for the hv 5936 components . despite its history of significant mass gain , the primary component has a luminosity in excellent agreement with theoretical expectations for a star of its mass and radius . this is not surprising since it has been shown that the hydrodynamical relaxation time of the accretion process is short enough so that the mass - gaining component in a non - conservative system is expected to behave like a normal star of its mass and radius ( de grve 1991 , 1993 ) . empirical proof of this has been obtained from the study of galactic algol systems ( garca & gimnez 1990 ) . the secondary , on the other hand , is significantly over - luminous for its current mass . this is consistent with the behavior exhibited by the mass - donors in galactic algol systems ( giuricin , mardirossian , & mezzetti 1983 ; hilditch & bell 1987 ) . the rotational velocity determined from the disentangled spectrum for the secondary is consistent with synchronization with the orbit ( compare the values of @xmath71 and @xmath87 in table [ tabstar ] ) . this is expected since the secondary completely fills its roche lobe and tidal forces are very efficient at locking the star s spin rate to follow the orbital motion . this agreement between observations and theoretical expectations provides a consistency check on the measurements of @xmath71 , stellar radius , and orbital properties . the primary component , however , is seen to have a rotational velocity that is @xmath025% larger than the synchronization value . the departure from unity is well beyond the observational uncertainties but is not unprecedented or unexpected . the primaries of numerous galactic algol systems have been determined to rotate faster than their synchronous rate and even close to the centrifugal limit ( see , e.g. , van hamme & wilson 1990 ) . a rotational velocity 25% larger than synchronicity is well within the observed range . angular momentum transfer through mass accretion is probably the most plausible model to explain the spin - up of algol primaries ( huang 1966 ) . the @xmath0factor - of-2 enhancement observed in the secondary s surface he abundance , as discussed in 4 , provides a strong constraint on the original masses of the stars in the hv 5936 system , particularly the secondary . the nature of this constraint is as follows : the initial mass of the current secondary must have been such that , by the time the star evolved to fill its roche lobe ( and commence mass transfer ) , the point in its interior where the he abundance was twice the surface value was at a radius which enclosed 4.7 @xmath1 of material ( i.e. , the current mass of the secondary ) . this interior point now at the surface of the secondary clearly must have been in the outer regions of its original h - burning core . to exploit this constraint , we considered the evolution of 4 different simulated binary systems , constructed with the assumptions of : 1 ) non - conservative mass exchange with 50% efficiency in the transfer of mass from the donor star to the mass gaining star and 2 ) a fractional angular momentum loss proportional to the fractional mass lost to the system . the initial masses of these systems were 12 @xmath1 ( current secondary ) + 8 @xmath1 ( current primary ) , 14 @xmath1 + 7 @xmath1 , 16 @xmath1 + 6 @xmath1 , and 18 @xmath1 + 5 @xmath1 . the calculations were done following the formalism in torres , neuhuser , & wichmann ( 1998 ) but considering the non - conservative mass transfer expression in vanbeveren et al . ( 1979 ) . from these , we estimate initial orbital periods of 1.154 , 1.318 , 1.689 , and 2.449 days , and initial roche lobe radii of 4.8 , 5.2 , 6.1 , 7.8 , and 10.7 @xmath88 , respectively , for the four mass combinations . the results showed that each of the systems experiences case a mass transfer ( i.e. , occurring during the core hydrogen burning phase of the mass donor ) and all could eventually evolve to the configuration observed for hv 5936 but with different resultant surface he abundances for the mass donor . using internal composition profiles computed by one of us ( a.c . ) , we find that the surface he abundances of the initially 12 @xmath1 , 14 @xmath1 , 16 @xmath1 , and 18 @xmath1 mass donors are enhanced by factors of 1.5 , 2.6 , 3.0 , and 3.5 , respectively , by the time donor has been reduced to 4.7 @xmath1 . the observed enhancement of @xmath02.0 is roughly midway between the 12 @xmath1 and 14 @xmath1 cases . taking into account all the results above , we postulate that the hv 5936 system began with initial masses of @xmath013 @xmath1 and @xmath07.5@xmath89 for the current secondary and current primary , respectively . case a mass exchange then resulted in the observed current masses of the components , the spinning up of the primary to a rotational rate faster than the synchronous rate , and the uncovering of processed material in the atmosphere of the secondary . the estimated initial mass of the secondary depends only weakly on our assumption of a 50% mass transfer efficiency ( which is probably an upper limit ) . the primary s initial mass is much more sensitive to this assumption , although it might be possible to constrain it further ( and thus the transfer efficiency ) by considering the spin - up process . along with the he enhancement , the secondary s surface should also exhibit a @xmath070@xmath90 enhancement of @xmath91n and a @xmath010@xmath90 depletion of @xmath92c and @xmath93o , according to our interior composition models . these enhancements and depletions were assumed in the case ii modeling of the secondary s disentangled spectrum as discussed in 4 above . because of the already low metallicity of this lmc star , the modifications in cno abundances do not have striking effect on the secondary s spectrum . in figure [ figsecon ] we indicate the locations of the most prominent n ii lines . even with their enhanced abundances , these lines are too weak to claim a positive detection , although there do appear to be features in the spectrum at their locations . higher quality spectra will be required to confirm their presence and , thus , the surface enhancement of n. our analysis of hv 5936 provides some insight into the conditions of the interstellar medium along the line of sight to the system . in this section we examine this information and also show how these data help constrain the relative position of hv 5936 within the lmc . the column density of interstellar h i in the foreground of hv 5936 was measured by comparing the observed h i ly@xmath52 1216.7 absorption line profile ( using the unbinned fos data ) with theoretical profiles consisting of a synthetic stellar spectrum convolved with an interstellar absorption profile . in general , the interstellar profile is constructed by assuming a component at 0 km s@xmath12 with @xmath94 @xmath95 , corresponding to milky way foreground gas ( see , e.g. , schwering & israel 1991 ) , and a second component with a lmc - like velocity and a column density which is varied to produce the best fit to the data ( as judged by visual inspection ) . in examining the fos data for hv 5936 , we found no evidence for h i absorption at lmc velocities , and a milky way contribution at 0 km s@xmath12 of @xmath96 , slightly less than , but consistent with , the mean value of schwering & israel . the best - fitting ly@xmath52 profile is shown in figure [ fighi ] , where we illustrate the unbinned fos spectrum , the synthetic stellar spectrum ( dotted line ) and the convolution of the synthetic spectrum with the interstellar profile ( thick solid curve ) . note that the bottom of the ly@xmath52 profile is filled in by geocoronal emission . this notable absence of neutral lmc interstellar gas towards hv 5936 is consistent with the interstellar reddening results . the value of @xmath97 found for hv 5936 from the sed analysis is essentially identical to the mean value for 56 milky way foreground stars within 1 of the hv 5936 line - of - sight ( @xmath98 from the data of oestreicher , gochermann , & schmidt - kaler 1995 ) . thus , there is no measurable contribution to @xmath99 from lmc material . the lack of large amounts of lmc gas and dust towards hv 5936 provides an opportunity to examine some properties of the milky way ism along a sightline passing completely through the milky way s halo but free of extragalactic contamination . in particular , the gas - to - dust ratio for this material is @xmath100 @xmath95 mag@xmath12 , which is roughly twice the value found in the galactic disk and suggests a `` cleaner '' ism in the halo . in addition , the wavelength - dependent extinction curve derived from the sed analysis is now recognized to be a measure of the extinction properties of galactic halo dust grains . this curve is shown in figure [ figext ] where the small symbols indicate the normalized ratio of model fluxes to observed fluxes , while the thick solid line shows the parameterized representation of the extinction curve , which was actually determined by the fitting process . the most remarkable feature of this curve is the very weak 2175 bump , which may be the weakest bump ever seen in milky way dust . the curve is consistent with the results of kiszkurno - koziej & lequeux ( 1987 ) , which show a weak trend of weakening bump strength and steepening 15001800 extinction with increasing height above the galactic plane . the shape of the hv 5936 curve is very similar to those seen towards hv 982 and eros 1044 ( papers ii and iii ) , whose total extinctions are dominated by milky way dust . the ism results for hv 5936 provide constraints on the position of the system within the lmc . its apparent location places it within the outlines of lmc-4 , the largest supergiant h ii shell in the lmc ( goudis & meaburn 1978 ; meaburn 1980 ) . the shell has a diameter of @xmath01.1 kpc , and hv 5936 is positioned ne of the shell center and near the inside edge . a hint of the shell can be seen in figure [ figlmc ] . the position of the shell coincides with a `` hole '' in the h i distribution ( mcgee & milton 1966 ) . the h i 21-cm emission line data of rohlfs et al . ( 1984 ) , obtained with a half - power beam width of 15 , reveal a total lmc h i column density of @xmath101 at the position of hv 5936 , with higher values to the east and comparable values in other directions . the minimum value of @xmath102 seen in the general vicinity of hv 5936 is @xmath103 , at a position 04 west of the system . while we have not established a rigorous upper limit for the non - detection of lmc h i ly@xmath52 absorption towards hv 5936 , such a limit is certainly below @xmath104 . thus , while we can not rule out the presence of localized , deep holes in the ism distribution , the simplest explanation of both the hv 5936 gas and dust results is that the system lies at an undetermined distance above the main h i layer at its apparent location on the lmc . the discussion in 5 demonstrates that , despite the interesting evolutionary state of the hv 5936 system and the mass exchange that has occurred , the stellar properties appear well - determined and well - understood . particularly , the properties of the primary star are indistinguishable from those of a `` normal '' single star , which came by its mass through the normal star - formation process . we can thus determine a distance to the system based on the results for the primary star , by combining its absolute radius @xmath105 ( derived from the classical eb analysis ) with the distance attenuation factor @xmath106 ( derived from the sed analysis ) . we find @xmath107 kpc , corresponding to a distance modulus of @xmath108 mag . the uncertainty in the distance is estimated from considering three independent sources of error : ( 1 ) the internal measurement errors in @xmath105 and @xmath106 as given in table [ tabparms ] ; ( 2 ) uncertainty in the appropriate value of the extinction parameter @xmath109 ; and ( 3 ) uncertainty in the fos flux scale zeropoint due to calibration errors and instrument stability . straightforward propagation of errors shows that these three factors yield individual uncertainties of @xmath110 kpc , @xmath111 kpc ( assuming @xmath112 ) , and @xmath113 kpc ( assuming @xmath114% ) , respectively . the overall 1@xmath40 uncertainty quoted above is the quadratic sum of these three errors . it is dominated by the uncertainty in the primary s radius , which alone accounts for 1.7 kpc in the error budget . note that the only `` adjustable '' factor in the analysis is the extinction parameter @xmath109 , for which we have assumed the value 3.1 . because of the very low reddening of the system , our distance result is very insensitive to this assumption . the weak dependence of the distance modulus on @xmath109 is given by : @xmath115 $ ] . we have considered the possibility that this distance result may be tainted due to the advanced evolutionary state of the secondary star . in general , the light curve and radial velocity curve analyses are very insensitive to the natures of the eb s components , i.e. , the temperature ratio of the two stars , their mean radii , surface gravities , and relative contributions to the observed light in the optical spectral region are virtually model - independent and very well - determined . these results tightly constrain the sed analysis and so it would seem that the results could be compromised only if the secondary s sed departs significantly from the shape predicted by an atlas9 model of the derived @xmath23 , @xmath24 , [ m / h ] , and @xmath49 values . the evidence , however , suggests that this is not the case . in support , we note the high quality of the fit to the combined sed of the system , which is comparable to our previous studies of systems containing wholly `` normal '' stars , and the consistency between the sed analysis and the independent modeling of the disentangled optical spectra ( see 4 ) . the only anomaly found is the probable modest enhancement of the secondary s he abundance , which would have only very small effect on the star s sed . we have run a number of tests with the sed analysis , arbitrarily modifying the secondary s sed ( by varying @xmath23 , @xmath24 , [ m / h ] , or @xmath116 ) . even relatively large changes have little effect on the derived system distance , because of the dominance of the primary in the combined sed . we thus conclude , based both on a lack of contrary evidence and on the apparent robustness of the result , that uncertainties arising from the evolutionary state of the secondary are likely significantly smaller than the other sources of error noted above . in table [ tabdist ] we show the individual system distances for hv 5936 and the other ebs in our study . we also list the implied distances to a standard reference point in the lmc , based on the assumption that the ebs are all located within a flat disk - like lmc whose spatial orientation is specified by an inclination angle and a position angle for the line of nodes in the plane of the sky . we choose the optical center of the lmc s bar as the reference point and show two sets of lmc distances , based on two different assumptions about the lmc s spatial orientation . details are given in the notes to the table . the results for hv 5936 in table [ tabdist ] clearly stand out from the others . this is partially explained by projection effects . because hv 5936 lies relatively far from the adopted reference point and in the `` nearside '' of the lmc , it is expected to be somewhat closer than the other systems . however , when the projection is taken into account , the distance implied for the lmc reference point is @xmath117 kpc , corresponding to a distance modulus of @xmath118 ( from taking a simple mean of the two cases shown in table [ tabdist ] ) . this is about 2@xmath40 away from the mean of the other systems and suggests that the assumption that all the systems lie in the same flat disk may be invalid . although the relatively large uncertainty in the hv 5936 result prevents definitive conclusions , we suggest that hv 5936 lies `` above '' the extrapolated position of the lmc s disk . this notion is consistent with , and actually suggested by , the interstellar results , which reveal no evidence for absorption by lmc h i and no measurable extinction by lmc dust along the hv 5936 sightline . this interpretation of the ism data is complicated by the coincidence of hv 5936 s position in the sky with the supergiant shell lmc 4 , towards which a very low h i 21-cm emission column density is observed . depending on the patchiness of the h i , it is conceivable that a star located within or even behind lmc 4 could show an undetectably low h i absorption column density . this seems unlikely for hv 5936 , however , since it is located near the eastern edge of the lmc 4 , where the 21-cm column density is @xmath119 and rises steeply towards the east . the simplest explanation is that hv 5936 just lies in front of most of the lmc s interstellar matter . it is not clear why hv 5936 should occupy such a position , although this might well be related to the formation of shell lmc 4 itself . likewise , it remains to be seen whether hv 5936 is a pathological object , in terms of its location , or is merely representative of a spatially extended stellar population in the ne quadrant of the lmc . alternative explanations for the hv 5936 results include assuming simple error in the spatial orientation of the lmc or perhaps a major structural feature , such as a warp , in the lmc s disk . the latter may arise from strong past interactions between the magellanic clouds and the milky way galaxy . both hypotheses would allow hv 5936 to lie close to the lmc s disk ( but still above , given the ism results ) , but require the disk to be closer to us than currently assumed . unfortunately for these ideas , the required orientation for the lmc would necessitate an inclination angle ( @xmath120 ) much steeper than previolusly measured . moreover , structural studies of the lmc have revealed no evidence of a warp in the vicinity of hv 5936 ( e.g. , see olsen & salyk 2002 ) . the analysis presented here for hv 5936 has differed somewhat from those reported previously in papers i , ii , and iii . here our efforts have been focused more strongly on the physical properties of the stellar components and the evolutionary history of the binary system , rather than on the distance measurement . this arises from the interesting evolutionary state of this semidetached system and from its outlying location in the lmc . because of its location , the implications of hv 5936 s individual distance for the distance to the lmc as a whole is not as firmly established as for our previously analyzed systems . it is worth noting that hv 5936 is the first semidetached eb system to be used as a distance indicator for the lmc . although our current analysis shows that semidetached binaries can be used successfully as such , there are a number of complications that call for the careful use of these systems . among the related problems are : _ 1 ) _ the lack of evolutionary cross checks , _ 2 ) _ the often very unequal component masses , effective temperatures and luminosities that complicate the spectrophotometric analysis , _ 3 ) _ the large out - of - eclipse variations and distortions , and _ 4 ) _ the relative radii that are strongly dependent on the adopted mass ratio our success with hv 5936 springs directly from the high quality observations available , in which the spectroscopic and photometric signatures of both stars are clearly present and separable . future distance studies within our program will be focused primarily on non - interacting systems lying closer to the apparent center of the lmc . we currently have four detached systems under analysis , all of which can be expected to yield accurate distances and a number of cross checks to verify the results . in addition , two of these systems , eros 1066 and macho 053648.7@xmath121691700 ( see figure 1 ) , have been especially selected to provide insight into the possible problem of line - of - sight extension of the lmc . this work was supported by nasa grants nag5 - 7113 , hst go-06683 , hst go-08691 , and nsf / rui ast-0071260 . we are grateful for the skilled assistance of the ctio support staff during our observing runs . e.f . acknowledges support from nasa adp grant nag5 - 7117 to villanova university and thanks michael oestreicher for kindly making his lmc foreground extinction data available . ccccccc fos & g130h & y3fu0803 t & 31 jan 1997 & 0.115 & 1.0059 & 0.9875 + fos & g190h & y3fu0806 t & 31 jan 1997 & 0.137 & 1.0116 & 1.0151 + fos & g270h & y3fu0805 t & 31 jan 1997 & 0.133 & 1.0106 & 1.0101 + fos & g400h & y3fu0804 t & 31 jan 1997 & 0.128 & 1.0093 & 1.0038 + stis & g430l & o665b6030 & 22 april 2001 & 0.654 & 1.0491 & 1.0360 + stis & g750l & o665b6040 & 22 april 2001 & 0.657 & 1.0490 & 1.0396 + [ tabhst ] rrrrrr 51558.7339 & 0.6719 & 411.2 & 84.6 & 1.8 & 2.0 + 51558.7600 & 0.6812 & 413.7 & 77.5 & 1.5 & 1.3 + 51560.7402 & 0.3871 & 240.4 & 486.6 & @xmath1211.9 & @xmath1216.2 + 51560.7659 & 0.3963 & 246.8 & 475.4 & @xmath1210.3 & @xmath1215.8 + 51561.6849 & 0.7239 & 421.1 & 55.8 & @xmath1210.2 & @xmath1211.1 + 51561.7072 & 0.7318 & 421.6 & 53.5 & @xmath1210.5 & @xmath1211.6 + 51562.7743 & 0.1122 & 245.7 & 484.5 & 1.3 & 1.0 + 51562.7930 & 0.1189 & 240.6 & 490.5 & @xmath1210.3 & @xmath1211.2 + 51895.7673 & 0.8234 & 412.8 & 76.3 & 0.2 & @xmath1211.3 + 51895.7906 & 0.8317 & 408.4 & 81.2 & @xmath1211.6 & @xmath1212.4 + 51898.6384 & 0.8470 & 402.7 & 96.0 & @xmath1211.9 & 0.0 + 51898.6616 & 0.8553 & 402.7 & 102.1 & 1.3 & @xmath1211.6 + 51899.6325 & 0.2014 & 211.9 & 565.6 & 1.1 & 2.8 + 51899.7847 & 0.2556 & 206.5 & 578.1 & 1.0 & 3.0 + 51900.8064 & 0.6199 & 387.3 & 133.8 & @xmath1210.3 & 1.3 + 51900.8296 & 0.6281 & 391.9 & 127.4 & 0.2 & 4.5 + 51902.6812 & 0.2882 & 207.6 & 569.7 & @xmath1210.7 & 0.4 + 51902.7046 & 0.2966 & 209.6 & 568.0 & @xmath1210.2 & 1.8 + [ tabrv ] lc + period & @xmath122 days + eccentricity & @xmath123 ( fixed ) + inclination & @xmath124 deg + @xmath125 & @xmath126 + @xmath127_{\rm b}$ ] & @xmath128 + @xmath127_{\rm v}$ ] & @xmath129 + @xmath130 & @xmath131 + @xmath132 & @xmath133 + @xmath134 & @xmath135 + @xmath136 & @xmath137 + @xmath31 & @xmath138 + + @xmath139 & @xmath140 km s@xmath12 + @xmath141 & @xmath142 km s@xmath12 + @xmath33 & @xmath143 + @xmath39 & @xmath144 km s@xmath12 + @xmath38 & @xmath145 r@xmath146 + + @xmath147 & @xmath148 k + @xmath149}_{\rm ps}$ ] & @xmath150 + @xmath151 & @xmath152 km s@xmath12 + e(b@xmath121v ) & @xmath153 mag + @xmath154 & @xmath155 + [ tabparms ] clc @xmath158 & uv bump centroid & @xmath159 @xmath160 + @xmath14 & uv bump fwhm & @xmath161 @xmath160 + @xmath162 & linear offset & @xmath163 + @xmath164 & linear slope & @xmath165 + @xmath166 & uv bump strength & @xmath167 + @xmath168 & fuv curvature & @xmath169 + @xmath109 & @xmath170 & 3.1 ( assumed ) + [ tabext ] lccc n(he)/n(h ) & 0.084 & 0.084 & 0.16 + t@xmath171 ( k ) & @xmath172 & @xmath173 & @xmath174 + @xmath24 ( cgs ) & @xmath175 & @xmath176 & @xmath177 + @xmath178}$ ] & @xmath179 & @xmath67 & @xmath67 + @xmath49 ( km s@xmath12 ) & 2.6 & 2.6 & 2.6 + @xmath71 ( km s@xmath12 ) & @xmath180 & @xmath181 & @xmath182 + @xmath65 ( km s@xmath12 ) & @xmath183 & @xmath184 & @xmath185 + contribution to light ( % ) & @xmath186 & @xmath187 & @xmath188 + [ tabfine ] lcc spectral type & b0.5 v & b2 iii + mass ( m@xmath86 ) & @xmath189 & @xmath190 + radius ( r@xmath86 ) & @xmath191 & @xmath192 + @xmath24 ( cgs ) & @xmath193 & @xmath194 + @xmath23 ( k ) & @xmath148 & @xmath195 + @xmath196 & @xmath197 & @xmath198 + @xmath199m / h@xmath200 $ ] & @xmath201 & @xmath201 + @xmath74 & 0.084 & 0.16 + @xmath71 ( km s@xmath12 ) & @xmath180 & @xmath182 + @xmath202 ( km s@xmath12 ) & @xmath203 & @xmath204 + @xmath205 ( kpc ) & + [ tabstar ] llccc hv 2274 & paper i , ii & @xmath206 kpc & @xmath207 kpc & @xmath208 kpc + hv 982 & paper ii & @xmath209 kpc & @xmath210 kpc & @xmath211 kpc + eros 1044 & paper iii & @xmath212 kpc & @xmath213 kpc & @xmath214 kpc + hv 5936 & this paper & @xmath2 kpc & @xmath215 kpc & @xmath216 kpc +

higher order conformal perturbation theory is studied for theories with and without boundaries . we identify systematically the universal quantities in the beta function equations , and we give explicit formulae for the universal coefficients at next - to - leading order in terms of integrated correlation functions . as an example , we analyse the radius - dependence of the conformal dimension of some boundary operators for the case of a single neumann brane on a circle , and for an intersecting brane configuration on a torus , reproducing in both cases the expected geometrical answer .

0811.3149

perturbations of conformal field theories by relevant operators have been intensively studied starting with the work of zamolodchikov @xcite on integrable perturbations of conformal field theories . numerous examples have been considered , but there are also a number of structural results , in particular the @xmath0-theorem of zamolodchikov @xcite that states that the central charge @xmath0 can not increase along renormalisation group flows , as well as the analogous @xmath1-theorem @xcite for the boundary entropy @xcite . perturbations of conformal field theories also play an important role in string theory , for example for time dependent backgrounds , see _ e.g. _ @xcite . in the context of string theory also _ marginal _ perturbations are of significance . most string theories of interest possess moduli , _ i.e. _ free parameters such as the size and shape of the background or the position of some d - brane , and these correspond to marginal operators in the two - dimensional world - sheet theory . exact conformal field theory solutions are often only available at special points ( in particular the rational points ) in moduli space , and it is important to learn to control the theory away from these special points , _ i.e. _ after perturbations by marginal operators . usually one thinks of the moduli as corresponding to _ exactly marginal _ operators , and then the renormalisation group analysis is , by definition , trivial . however , in the context of bulk and boundary perturbations , the situation can be more subtle . in particular , exactly marginal bulk operators ( describing moduli of the closed string background ) can cease to be exactly marginal in the presence of a boundary . if this is the case , they induce a non - trivial renormalisation group flow on the boundary @xcite . in the analysis of perturbations by relevant operators , a first order analysis is usually sufficient ( see _ e.g. _ @xcite ) . however , in the context of marginal perturbations , it is often necessary to go to higher order in perturbation theory . the simplest example of such a situation is a single neumann brane on a circle , for which the conformal dimension of the momentum eigenstates depends on the radius modulus . from the point of view of the world - sheet , the change in conformal dimension for this boundary operator does not arise at first order , but only appears at next - to - leading order in perturbation theory . higher order conformal perturbation theory also plays a role in proofs of integrability of particular bulk and/or boundary perturbations @xcite . conformal perturbation theory at higher orders was studied for particular models in @xcite ; general aspects of the pure bulk case were also discussed in @xcite . in this paper we make an attempt at a systematic analysis of conformal perturbation theory beyond the leading order . we begin by analysing which rg coefficients are scheme - independent ( or universal ) and thus can have a physical interpretation . ( in particular , we show that this is the case for the coefficient describing the change in conformal dimension of the momentum fields on the neumann brane . ) we then outline a specific scheme the position space minimal subtraction scheme in which higher order rg coefficients can be calculated . this allows us to prove that the combined bulk - boundary perturbation problem is renormalisable at the quadratic order . while the minimal subtraction scheme is conceptually clean , explicit calculations of the rg coefficients are often rather cumbersome . we therefore also consider another , wilsonian type scheme , to which we refer to as the ope scheme since the first nontrivial terms in the beta functions are given by ope coefficients . this scheme has some conceptual shortcomings at higher orders but is computationally somewhat simpler . for the universal coefficients we are interested in , the result is independent of which of the two schemes we use ( as we also verify explicitly ) . we can therefore determine the coefficients of interest ( in particular the formula for the shift in conformal dimension for the momentum fields on the neumann brane ) in the wilsonian approach . in the resulting formulae the universal quantities are expressed as integrals over certain correlation functions . as an illustration we also apply the formulae to an intersecting brane model on a torus , and again reproduce the geometric result . the paper is organised as follows . in section 2 we discuss which rg coefficients in the boundary beta function are universal in the presence of marginal bulk perturbations ( section 2.1 ) . we then introduce the minimal subtraction scheme , both for pure boundary perturbations ( section 2.2.1 ) , as well as for the combined bulk - boundary problem for which we prove renormalisability at the quadratic order ( section 2.2.2 ) . we also introduce the wilsonian scheme and discuss its advantages and shortcomings ( section 2.3 ) . finally , we explain how the discussion can be generalised to include boundary changing operators ( section 2.4 ) . in section 3 these ideas are applied to two examples , the single neumann brane on a circle ( section 3.1 ) , as well as a configuration of two intersecting d1-branes on a 2-torus ( section 3.2 ) . finally , we discuss in section 4 how our techniques for the calculation of higher order rg coefficients can also be applied to pure bulk or pure boundary perturbation theory . let us start with a general discussion of the effect of perturbations by marginal bulk operators on boundary degrees of freedom . generically such a deformation will induce a renormalisation group ( rg ) flow on the space of boundary conditions @xcite . under a certain condition , to be formulated precisely below , the induced boundary deformation is however scale independent to the first order in the bulk deformation parameter . in this case one can study how the set of boundary scaling dimensions changes with the bulk deformation . we will demonstrate that ( despite no occurrence of rg flows ) the rg technique is very useful in addressing this question . in particular we will derive , using the rg methods , general expressions for a first order change in dimensions of boundary operators along a bulk deformation . before we give a detailed discussion we need to introduce some notation . consider a boundary conformal field theory ( bcft ) defined on the upper half plane @xmath2 with complex coordinate @xmath3 . let @xmath4 be bulk primary fields with conformal weights @xmath5 so that their scaling dimensions are @xmath6 . for a single ( fundamental ) conformal boundary condition we denote the boundary primaries by @xmath7 , and their scaling dimensions by @xmath8 . later we will generalise our discussion to superpositions of conformal boundary conditions . we will assume that the two point functions are normalised as @xmath9 in particular , this means that we assume all fields to be self - conjugate ; this is obviously not a real restriction , and our analysis can easily be generalised . the operator product expansion ( ope ) for pairs of bulk and boundary operators has the form @xmath10 finally , when a bulk operator approaches the boundary it can be expanded using the bulk to boundary ope @xmath11 with these preparations , let us now consider a perturbation of the given bcft generated by the euclidean action perturbation @xmath12 here @xmath13 , @xmath14 are the dimensionless coupling constants of the respective operators , and @xmath15 is a renormalisation distance scale . up to second order in the coupling constants , the beta functions have the following general form @xmath16 \beta^{p}&=&y_p\mu^{p } + \sum_i{\mathcal{b}}_{i}^{p } \ , \lambda^i+\sum_{qr}\mathcal{d}_{qr}^{p}\mu^{q } \mu^{r}+\sum_{iq}\mathcal{e}_{iq}^{p}\lambda^i\mu^{q}+\ldots \ . \label{genbdrybeta}\end{aligned}\ ] ] here , @xmath17 , and @xmath18 are the bulk and boundary anomalous dimensions , respectively . the omitted terms stand for higher orders in the coupling constants . a general property of any local rg scheme is that the bulk beta functions are independent of the boundary couplings . it has been known for quite some time @xcite ( see also @xcite ) that in a particular renormalisation scheme the coefficients @xmath19 for bulk theories are given by @xmath20 , where @xmath21 are the bulk ope coefficients from ( [ bulkbulkope ] ) . the same scheme , to be discussed in more detail in section 2.3 , can be easily adapted for theories on the half plane . the coefficients @xmath22 coincide then with the boundary structure constants @xmath23 ( see _ e.g. _ @xcite ) , and for the coefficients @xmath24 we have @xmath25 , where @xmath26 are the bulk to boundary ope coefficients from ( [ bulkboundaryope ] ) , see @xcite . consider now the case where we perturb the bcft by a single bulk field @xmath27 with a coupling constant @xmath28 . furthermore we want to assume that the bulk beta function @xmath29 vanishes . in this case , even in the absence of an initial boundary perturbation @xmath30 , a boundary renormalisation group flow can be triggered by the terms @xmath31 in the boundary beta function . such boundary terms , however , are in general not universal . for example , if the induced boundary fields are all relevant , _ i.e. _ @xmath32 , then the corresponding terms in the boundary beta function can be removed by a coupling constant redefinition @xmath33 the above coupling constant redefinition looks peculiar in that @xmath34 is not proportional to @xmath35 . it has , however , a simple meaning . let @xmath36 be the renormalised partition function of the perturbed theory ( [ deltas ] ) . we have @xmath37 where the partial derivatives on the left hand side are taken with the boundary constants @xmath14 or @xmath34 held fixed . the identities ( [ zredef ] ) mean that , after the redefinition ( [ redef ] ) , the bulk coupling constant @xmath28 couples to a re - defined field @xmath38 part of the renormalisation procedure amounts to defining the operator coupling to @xmath28 on the half plane so that the correlation functions involving that operator are distributions ( and thus integrable in any bounded region on the half plane ) . in the interior of the half - plane the resulting operator must coincide with the bulk operator @xmath27 , but in general extra subtractions may be required at the boundary . the redefinition ( [ redef2 ] ) stemming from the change of scheme ( [ redef ] ) reflects the natural ambiguity in defining such a fully subtracted operator extending @xmath27 . suppose now that all terms linear in @xmath28 can be removed in this manner . then the resulting boundary beta functions have the form @xmath39 where @xmath40 and @xmath41 . now that the boundary beta functions are all proportional to the boundary coupling constants we can treat the boundary perturbations ( at least those of the relevant operators ) infinitesimally to read off the dimensions of the boundary operators in the deformed theory . more specifically , we claim that the eigenvalues of the matrix @xmath42 are to be identified with @xmath43 , where @xmath44 is the scaling dimension of the boundary operator in the deformed theory , and the corresponding boundary primaries are the eigenvectors of @xmath42 . to leading order in @xmath28 , the matrix @xmath42 can be diagonalised by the transformation @xmath45 the corresponding primary fields are$ ] are defined up to adding a multiple of @xmath46 . ] @xmath47= \psi_p - \lambda \sum_{q\ne p}\frac{\tilde { \mathcal{e}}_{\phi p}^{q}}{y_q - y_p } \psi_q \ , \ ] ] and their anomalous dimensions are @xmath48=y_p + \lambda \ , \tilde { \mathcal{e}}_{\phi p}^{p } \ .\ ] ] we further claim that the quantity specifying the dimension shifts @xmath49 is scheme independent . to see this we consider a coupling constant redefinition of the form @xmath50 under this redefinition the coefficients in the beta functions ( [ genbdrybeta ] ) change as @xmath51 it is straightforward to check that under the transformations ( [ coef_shifts ] ) the quantity ( [ dim_shift ] ) is indeed invariant . in the following we want to explain in detail how these coefficients in particular ( [ dim_shift ] ) can be calculated explicitly . we shall first study this question in a minimal subtraction scheme . in order to make sense of the formal perturbation series we shall use a point splitting regularisation . in particular , we require that any two perturbing bulk or boundary fields do not approach each other closer than a cut - off @xmath52 , and that the perturbing bulk fields only approach the boundary up to a distance @xmath53 . before specialising to the bulk and boundary situation discussed in the previous section , let us first discuss some generalities of renormalisation . for brevity we consider only boundary perturbations but that is inessential for the points we want to make . let us consider a perturbed bcft action @xmath54 where @xmath55 are the bare coupling constants . let @xmath15 be an infrared distance scale at which we wish to renormalise the theory . in terms of the renormalised dimensionless coupling constants @xmath14 , the same lagrangian ( [ l1 ] ) can be expressed as @xmath56 where @xmath57 are anomalous dimensions of the fields @xmath58 , and @xmath59 is a counterterm action . perturbation theory generates integrals of the form @xmath60 where we have also introduced an infrared regulator @xmath61 . the above expression is to be understood in the operator sense , _ i.e. _ inside a correlator with arbitrary other insertions . the product of fields in ( [ pert_gen ] ) can be expanded in terms of a complete set of local operators @xmath62 as @xmath63 where we have arbitrarily chosen the point of insertion on the right hand side to be @xmath64 . if the opes of the conformal families of the primaries @xmath65 close on themselves , we can take for @xmath62 the fields @xmath65 and their conformal descendants . in conformal field theory the expansion ( [ multope ] ) always converges @xcite unlike in massive qfts for which the ope may be merely an asymptotic expansion . substituting ( [ multope ] ) into ( [ pert_gen ] ) we obtain expressions of the form @xmath66 where @xmath67 the integrals ( [ ints ] ) are finite because of the cut - offs . in the limit @xmath68 the coefficients @xmath69 diverge with the divergences coming from the regions of integration in which two or more insertion points @xmath70 collide . in fact , if @xmath71 operators @xmath72 come close together ( with the other insertions bounded away from the point of coincidence ) to produce an operator @xmath73 , the leading divergence has the form @xmath74 where @xmath75 is some numerical constant . here we have assumed that the resonance condition @xmath76 does not hold ; otherwise the corresponding divergence is logarithmic . the above reasoning follows essentially from dimensional counting , as well as from the locality of the ope , ensuring the independence of the expansion ( [ multope ] ) from @xmath61 . if we only perturb by marginal or relevant fields , @xmath77 , then divergences can only occur if also @xmath78 is relevant , _ i.e. _ @xmath79 . assuming that the ope is closed , @xmath73 is then one of the perturbing relevant or marginal primary fields @xmath65 . in the ` minimal subtraction scheme ' we are using here , we only introduce counterterms for actually divergent contributions ; the above reasoning then implies that the scheme closes on itself . the divergences arising when @xmath71 operators come together first emerge at order @xmath80 in perturbation theory . one expects that they can be canceled by local ( @xmath61-independent ) counterterms . these counterterms then also cancel the non - local subdivergences ( [ ldivs ] ) that appear at order @xmath81 in perturbation theory . thus we only need to deal with the case when @xmath80 , in which case @xmath78 and @xmath82 must have a non - trivial two - point function , and hence @xmath83 . then the coefficient ( [ ldivs ] ) is independent of @xmath61 , and hence converges when @xmath84 . note that the lower order counterterms may also contribute to the @xmath85 divergence when the counterterm insertion from the order @xmath86 comes close together with @xmath87 fields @xmath88 . the same dimensional reasoning however tells us that the final coefficient must again be independent of @xmath61 , and the remaining divergence can be canceled by a local counterterm . the above discussion should however not be taken to be a recursive proof of renormalisability of conformal perturbation theory . one problem that needs to be tackled is the classical problem of overlapping divergences ; in the case at hand this occurs when @xmath71 points come together with a subset of @xmath89 points coming together much faster than the remaining ones . the associativity of the ope in conformal field theory should be the key property ensuring the consistency in dealing with overlapping divergences , but we have not attempted to work this out in detail . however , we will see in the concrete examples of the next subsections how the above discussion can be made more rigorous . in particular we will prove the renormalisability of conformal perturbation theory at the next - to - leading order using analytic properties of conformal blocks . in order to illustrate these ideas , let us now consider an integral that emerges at second order in perturbation theory , @xmath90 the product of the two boundary fields can be expanded via the ope ( [ boundaryboundaryope ] ) . performing one of the integrals , we see that we get ultraviolet divergences of the form ( in the limit @xmath84 ) @xmath91 where the summation runs only over those indices @xmath92 for which @xmath93 . in particular , @xmath94 , and thus only relevant primary fields @xmath95 contribute . as before , we have also assumed here that there are ` no resonances ' , _ i.e. _ that none of the expressions @xmath96 vanishes . then only power divergences occur at this order . the above divergences can be canceled by adding a minimal action counterterm @xmath97 . equating the two expressions ( [ l1 ] ) and ( [ l2 ] ) we obtain up to second order in the coupling constants @xmath98 \ , \ ] ] where @xmath99 is the set of pairs of indices @xmath100 for which @xmath93 . differentiating both sides of ( [ mub ] ) with respect to @xmath15 with fixed @xmath101 we obtain @xmath102 thus the beta functions are linear in @xmath103 . it is easy to see that this property continues to hold also at higher order in perturbation theory , as long as the divergences are power - like . on the other hand , if we have a non - trivial resonance at lowest order , _ i.e. _ if @xmath104 , then formula ( [ sdiv ] ) takes the form ( we are assuming for simplicity that there are no other divergences at this order ) @xmath105 where we cut off the divergent integral in the infrared region at the renormalisation scale @xmath15 . introducing a counterterm @xmath97 we then obtain a beta function for the coupling @xmath106 @xmath107 more generally , in the minimal subtraction scheme at hand , the nonlinear terms in the beta functions all come from resonances . however , in general not all resonant terms are universal . after this interlude we now return to the case of interest , namely the description of the minimal subtraction scheme for bulk - boundary perturbations . in fact , the above discussion generalises in a straightforward manner to include an additional perturbation by a bulk field . for simplicity of presentation , we shall assume that the bulk field @xmath108 is a spinless relevant or marginal primary field of scaling dimension @xmath109 with @xmath110 . as we shall explain below , the bulk - boundary perturbation at the next to leading order in perturbation theory is then renormalisable . we will specialise to the situation where the bulk field is marginal ( @xmath111 ) later . at the linear order in the bulk coupling @xmath28 the divergences in perturbation theory only arise from singularities as the bulk field approaches the boundary . these are described by the bulk to boundary ope . if there are no boundary fields for which @xmath112 we have power divergences of the form ( [ bulkboundaryope ] ) @xmath113 where @xmath114 in the minimal subtraction scheme the counterterm is then simply @xmath115 . if there is a boundary field for which @xmath116 and the resonance condition @xmath117 is satisfied , we have a logarithmic divergence which results in a universal term linear in @xmath28 in the boundary beta function @xcite @xmath118 in the case when the bulk perturbation is marginal the resonance condition requires that the boundary field is also marginal . we shall , in the following , always assume that the resonance condition is not satisfied , _ i.e. _ that @xmath119 ; this is for example true in the context of section [ sec2.0 ] where @xmath120 and @xmath32 . then the counterterm is given by @xmath115 . at the next order in perturbation theory we encounter the integral @xmath121 \ , \ ] ] where @xmath3 and @xmath122 is an infrared regulator . the quantity in the square brackets in ( [ psum ] ) can be expanded in local boundary fields as in ( [ multope ] ) and ( [ expro ] ) ; the coefficients of these fields can be expressed in terms of certain integrals ( see below ) . by the same arguments as above , only coefficients of ( primary ) relevant fields can be divergent as we send @xmath68 . more precisely , the coefficient with which the primary field @xmath65 will appear in ( [ psum ] ) equals @xmath123 using the mbius symmetry , the correlation function appearing in this formula can be written as @xmath124 and @xmath125 \sum_a b_{\phi}{}^{a } d_{aq}{}^{p } e^{i\frac{\pi}{2}(y_{\phi } - y_a -1 ) } f_{\phi\bar{\phi}qp}^a(\eta)&({\mathrm{re\,}}z<0)\ . \end{array } \right.\ ] ] here the index @xmath126 runs over all conformal primaries whose conformal families appear in the intermediate channel . the conformal blocks @xmath127 have a branch cut along the real @xmath128-axis from @xmath129 to @xmath130 and are normalised so that @xmath131 with coefficient @xmath130 as @xmath132 . the conformal blocks entering the function @xmath133 are defined on opposite sides of the branch cut for @xmath134 and @xmath135 . the analyticity in @xmath136 implies that the values of @xmath133 in the lower half @xmath128-plane are obtained by the analytic continuation in a clockwise direction from the upper half - plane @xcite . passing to polar coordinates @xmath137 and using ( [ 4pty ] ) we can rewrite ( [ ipq ] ) as @xmath138 where @xmath139 since @xmath140 depends only on @xmath141 we can perform the integral over @xmath142 . in the remaining integral it is convenient to change the integration variable @xmath141 to @xmath128 . altogether we then obtain @xmath143 \eta^{\delta+y_{\phi}-2}(1-\eta)^{-(y_{\phi}+\delta)/2}y(\eta ) \ , \ ] ] where @xmath144 and the contour of @xmath128-integration is a segment of the circle of radius 1 centered at @xmath145 and oriented clockwise @xmath146 obviously , this expression only makes sense if @xmath147 . in the resonance case ( _ i.e. _ for @xmath148 ) we have instead @xmath149 the divergences of @xmath150 and @xmath151 in the limit @xmath152 can now be analysed using well - known properties of conformal blocks . there are two kinds of divergences that will be important to us : those that come from the region of integration @xmath153 where the bulk operator approaches the boundary far away from the point of insertion of @xmath154 ; and those that arise when the bulk field approaches the boundary in the vicinity of the boundary field @xmath154 . as we have argued before ( and as will become clear below ) the former divergences are canceled by the contribution from the lower order counterterm @xmath155 , while the remaining divergences have the power @xmath156 . to see this , we use the asymptotics see ( [ y ] ) @xmath157 in ( [ ieta ] ) and ( [ ietar ] ) , and then perform the @xmath128 integrals in the vicinity of @xmath158 , _ i.e. _ from @xmath159 up to some intermediate cut - off @xmath160 . this leads to @xmath161 \ , \label{ias}\ ] ] @xmath162 \ , , \label{las1}\ ] ] where @xmath163 , @xmath164 , @xmath165 and @xmath166 are some constants independent of @xmath52 and @xmath122 ( that come from the evaluation of the primitive function at @xmath160 , as well as from the remaining part of the integral ) . since we are only interested in divergent contributions in @xmath52 , we may restrict the fields @xmath126 to be relevant primary fields in @xmath167 . furthermore , we can ignore all the subleading terms @xmath168 since they vanish in the limit @xmath152 .. in a unitary bcft @xmath169 corresponds always to the identity operator @xmath170 which does not have a descendant operator at level one since @xmath171 . ] now we want to show that the divergent terms in the sum in ( [ ias ] ) are precisely canceled by the lower order counterterm @xmath155 given in ( [ s1 ] ) . at order @xmath172 recall that @xmath173 is the coupling constant corresponding to @xmath174 the counterterm leads to the contribution @xmath175 again this can be expanded in terms of local boundary fields , and the divergence in the coefficient of @xmath176 equals @xmath177\ , .\ ] ] here we have assumed that there is no resonance among the boundary fields , _ i.e. _ that @xmath178 for any @xmath179 ; if there is a resonance , _ i.e. _ @xmath180 , then ( [ ict1 ] ) has to be modified in the obvious manner . in either case , by comparison with ( [ ias ] ) , it is now clear that the contribution @xmath181 cancels precisely the divergent part of the sum in ( [ ias ] ) , and similarly for the resonant case @xmath148 . thus the divergent contribution only comes from the first term in ( [ ias ] ) and ( [ las1 ] ) @xmath182 where @xmath183 and @xmath184 are coefficients that can be obtained by taking the limits @xmath185 using the explicit expressions ( [ ieta ] ) , ( [ ietar ] ) and ( [ ict1 ] ) we finally obtain @xmath186 + \sum_{s\in i^{(1 ) } } \frac{b_{\phi}{}^{s}(d_{qs}{}^{p}+d_{sq}{}^{p})}{2(y_s - y_{\phi } ) ( y_s + y_q - y_p ) } \ , \nonumber \\[4pt ] ( \tilde c_{q}^{p})_{\rm res } & = & \lim_{\epsilon \to 0 } \ , \bigl [ \frac{i}{2}\ ! \int\limits_{c(\epsilon / r)}\!\!d\eta\ , \eta^{y_{\phi } + \delta -2}(1-\eta)^{-(y_{\phi } + \delta)/2}y(\eta ) \nonumber \\[-16pt ] & & \qquad \qquad \qquad \qquad \qquad\qquad + \sum_{s\in i^{(1 ) } } \frac{b_{\phi}{}^{s}(d_{qs}{}^{p}+d_{sq}{}^{p})}{2(y_s - y_{\phi } ) } \left ( \frac{r}{\epsilon}\right)^{y_s - y_{\phi } } \bigr]\ . \label{cres } \end{aligned}\ ] ] it is worth noting that the contours of the @xmath128-integrations in ( [ c_pq ] ) and ( [ cres ] ) can be deformed provided the ends of the contour are held fixed and the cut is not crossed . in particular one can deform @xmath187 to run infinitesimally above the cut to @xmath145 , and then infinitesimally below back to the second endpoint near @xmath158 . given that the conformal blocks only have singularities at @xmath188 with standard asymptotics , the quantities @xmath183 , @xmath189 defined in ( [ c_pq ] ) and ( [ cres ] ) are finite and independent of @xmath122 . this essentially provides a proof that at order @xmath190 the divergences can be canceled by local counterterms which are linear combinations of relevant operators . together with the well known results at orders @xmath191 and @xmath192 we have thus shown that a generic perturbation by relevant or marginal bulk and boundary fields is renormalisable at the quadratic order in the couplings . for any boundary field @xmath193 , and that we only have a single bulk field , are not crucial for the argument . ] it should also be possible to extend the analysis to higher orders in perturbation theory , but we have not attempted to do so . if the resonance condition @xmath148 is not satisfied the divergence of @xmath194 is power like , and in the minimal subtraction scheme there are no terms of order @xmath190 in the beta function @xmath195 . on the other hand , when the resonance condition is satisfied for a pair @xmath100 we have a universal term ( _ cf . _ ( [ ex1 ] ) and ( [ ex2 ] ) ) @xmath196 for a marginal bulk perturbation @xmath120 , the @xmath197 case ( and only that one ) is always resonant so that we have in the notation of section [ sec2.0 ] @xmath198 moreover , assuming as in section [ sec2.0 ] that all @xmath199 , we have in the minimal subtraction scheme @xmath200 and therefore by ( [ dim_shift ] ) @xmath201 where @xmath202 is given by formula ( [ cres ] ) for @xmath120 and @xmath197 . as we proved in section [ sec2.0 ] , the quantity @xmath203 is scheme independent ; we have therefore managed to obtain a description of this universal quantity in terms of conformal blocks this is the main result of this subsection . it is instructive to compute ( [ dim_shift ] ) also in a different , wilsonian type , renormalisation scheme the one of @xcite . we will refer to this scheme as the ` ope scheme ' for the reason that the first nontrivial terms in the beta functions are given by various ope coefficients . this scheme is often employed in conformal perturbation theory at the leading order . one of the advantages of this scheme is that in the presence of a nearby infrared fixed point in theory space , the corresponding coordinates are nonsingular near that fixed point . another attractive feature is that formulae for universal quantities , such as the dimension shift ( [ dim_shift ] ) , can be obtained quite easily in contrast with the minimal subtraction scheme . on the other hand we will see at the end of this section that the scheme has some pitfalls when applied to computing non - universal quantities at higher order in perturbation theory . in the ope scheme the theory is also regulated by a point - splitting cut - off @xmath52 , just as in the minimal subtraction scheme of the previous section . it is however convenient to introduce the infrared regulator slightly differently : we introduce a cut - off whenever two coordinates @xmath204 , @xmath205 are separated by a distance larger than @xmath61 , and whenever there is a bulk operator at a distance @xmath206 . the dimensionless couplings are now introduced by using the uv cut - off scale itself , which is understood as a fundamental uv scale ( lattice spacing , atomic or molecular scale ) . thus we have @xmath207 note that one can also include in ( [ deltas2 ] ) irrelevant operators @xmath208 with @xmath209 . their contributions will be relatively suppressed as @xmath210 but one may worry that in the perturbation expansion they will lead to contributions more singular than this suppression factor . we will see that , although there is no need to introduce irrelevant operators at the leading order , they are sometimes necessary to be taken into account at higher orders in perturbation theory . we will confine ourselves throughout this subsection to the case of a single marginal bulk field @xmath27 ( @xmath211 ) . in this case the terms in the perturbation expansion we are interested in are @xmath212 the cut - off variation @xmath213 can be computed assuming that the coupling constants depend on the cut - off via the couplings themselves according to @xmath214 where @xmath215 are the beta functions ( [ genbdrybeta ] ) . in the ope scheme we now vary @xmath216 with respect to @xmath52 , _ i.e._we compute @xmath213 , and demand that the variation vanishes at the leading order in @xmath52 . this reflects the main principle of the wilsonian renormalisation group approach , namely that the renormalised quantities must be independent of the uv scale . the resulting equations fix order by order the coefficients of the beta functions . the linear terms in the beta functions are always scheme independent with the coefficients given by the anomalous dimensions . it is easy to check that at the linear order in @xmath35 the equation @xmath217 is satisfied automatically . the equation arising at the linear order in @xmath28 fixes @xmath218 where @xmath219 are the bulk - to - boundary ope coefficients ( [ bulkboundaryope ] ) ( see @xcite ) . at the quadratic order in the boundary couplings one obtains the well known expression @xmath220 where @xmath221 are the boundary ope coefficients ( [ boundaryboundaryope ] ) . finally , the equation at order @xmath222 is @xmath223\ .\end{aligned}\ ] ] the first line in the above expression came from applying @xmath224 to the cut - off function @xmath225 while the second line came from the lower order term ( [ calb ] ) . the coefficients @xmath226 are formally obtained by taking correlation functions with the operator @xmath65 inserted at infinity @xmath227\ .\end{aligned}\ ] ] the above expression is formal because the limit may not exist . the integrals in the second line in ( [ exp4pt ] ) can be evaluated explicitly @xmath228 \equiv h(\epsilon / l ) \ . \end{aligned}\ ] ] to study the convergence we rewrite expression ( [ exp4pt ] ) via conformal blocks using ( [ 4pty ] ) @xmath229 \ , \ ] ] where @xmath230 is a segment of a unit circle around @xmath145 oriented clockwise . if we now evaluate @xmath231 of ( [ dim_shift ] ) in the ope scheme , using ( [ expy2 ] ) for @xmath197 , as well as ( [ cald ] ) and ( [ calb ] ) , we obtain the same expression in terms of conformal blocks as given in ( [ cres ] ) and ( [ ec ] ) . thus @xmath231 is indeed scheme - independent , as we have argued before . in terms of correlation functions , it can now be written as @xmath232 \ .\ ] ] writing @xmath233 and using the variable @xmath234 in the integral ( [ expy2 ] ) , we can also obtain another , perhaps more elegant expression for @xmath231 @xmath235 \ , \ ] ] where the bulk field insertion runs over a semicircle of radius one around the boundary insertion @xmath236 . note that there is nothing special about the radius being one , since @xmath237 for any @xmath238 . let us now come back to the expression ( [ expy2 ] ) for @xmath239 . substituting the asymptotic expansion ( [ yas ] ) into ( [ expy2 ] ) we find that although the most dangerous divergences , associated with the leading contributions of relevant primaries in ( [ yas ] ) , cancel out , there may be divergences coming from terms in ( [ yas ] ) associated with irrelevant fields . more precisely there are additional divergences in @xmath240 from the region near @xmath241 whenever there is an irrelevant primary @xmath242 in the theory such that @xmath243 or whenever we have a relevant primary @xmath244 such that @xmath245 in the last case the irrelevant field causing the divergence is a descendant of the primary @xmath246 . from the point of view of the minimal subtraction scheme of the previous subsection , this problem does not arise since the conditions ( [ 1div ] ) and ( [ 2div ] ) imply that there are no divergences from the region where the bulk field approaches the boundary insertion . in fact , the additional @xmath68 divergences in the ope scheme come directly from the extra divergent factors of @xmath52 included in the action ( [ deltas2 ] ) . the situation can be mended if we include in the original perturbed action ( [ deltas2 ] ) also irrelevant fields , and introduce their beta functions by requiring that @xmath247 at the subleading orders in @xmath52 . then the function @xmath248 is modified accordingly to include more divergent terms that cancel out the divergences coming from the integral in ( [ expy2 ] ) . although this resolution looks quite natural from the wilsonian point of view , the whole scheme becomes quite unwieldy for practical applications whenever ( [ 1div ] ) or ( [ 2div ] ) happens . note , however , that these extra divergences do _ not _ appear for universal quantities like @xmath231 . thus , as long as we are only interested in these quantities we can ( and will ) use the technically simpler ope scheme . in particular , we will use this method to compute analogous quantities for pure bulk and pure boundary perturbations in section 4 . it is worth noting that the complications related to ( [ 1div ] ) and ( [ 2div ] ) arise only in the presence of several running coupling constants . although beta function coefficients were studied for some models to a very large order , see _ e.g. _ @xcite , such computations typically involved only a single coupling constant . up to now we have assumed that there is a single ( fundamental ) boundary condition . the whole analysis can be easily generalised to the situation where we have superpositions of boundary conditions ; in that case the set of boundary operators includes also boundary changing operators @xmath249 , where the two boundary conditions are labeled by @xmath250 and @xmath251 with @xmath250 being the boundary condition to the left of @xmath252 and @xmath251 to the right . local excitations of the pure boundary @xmath250 are denoted @xmath253 . the study of renormalisation group flows involving such operators was initiated in @xcite . the opes of a bulk field approaching the boundary with label @xmath250 , and that of two boundary fields have the form @xmath254 \psi_p^{ab}(x)\psi_q^{bc}(y)&=&\sum_{r}d_{pq}^{(abc)r } ( y - x)^{h_r - h_p - h_q}\psi_r(y)+\ldots \qquad ( y > x ) \ . \label{boundaryboundaryope2}\end{aligned}\ ] ] the only difference to the previous analysis is that there are now various superselection rules that demand for example , that products of boundary operators can only be non - zero if the intermediate boundary conditions match , or that the boundary fields that appear in the bulk to boundary ope are always boundary preserving fields . taking this into account , the boundary beta functions then have the following general form @xmath255 where @xmath256 are anomalous dimensions and @xmath257 are the coupling constants of the operators @xmath258 . the expression for the dimension shift ( [ dim_shift ] ) generalises as @xmath259 similarly , the main results of the previous subsection ( [ wilson_e ] ) and ( [ wilsone2 ] ) now become @xmath260 \ , \label{wilson_e2 } \\ \tilde { \mathcal{e}}_{\phi p}^{p(ab ) } & = & \lim_{\delta \to 0 } \biggl [ \int\limits_{\delta}^{\pi-\delta}\!\ ! d\vartheta\ , \langle \phi(e^{i\vartheta } ) \psi_{p}^{ab}(0 ) \psi_{p}^{ba}(\infty)\rangle \nonumber \\ & & \qquad \qquad -\sum_{r } d_{rp}^{(aab)p } \frac { { { } ^{a\ , \!\ ! } b_{\phi}^{\phantom{\phi}\ ! r}}}{2y_{r}^{aa } } \left(\frac{1}{2\delta}\right)^{y_r^{aa } } -\sum_{r}d_{pr}^{(abb)p } \frac{{{}^{b\ , \!\ ! } b_{\phi}^{\phantom{\phi}\ ! r } } } { 2y_{r}^{bb } } \left(\frac{1}{2\delta}\right)^{y_r^{bb } } \label{wilson_e3 } \biggr ] \ . \end{aligned}\ ] ] up to now our analysis has been very general . in this section we want to illustrate these general results with two simple examples . the simplest example is the case of a single neumann brane on a circle of radius @xmath122 . the action of this theory is simply . ] @xmath261 the bulk field that corresponds to changing the radius @xmath122 is @xmath262 , which is an exactly marginal operator in the bulk . more specifically , we shall consider the perturbation @xmath263 that changes the radius @xmath122 as @xmath264 so that to the first order we have @xmath265 . on the neumann brane we have open string momentum states corresponding to the vertex operators @xmath266 , whose conformal dimension is @xmath267 with @xmath268 and @xmath269 . we want to study how the conformal dimension of these operators changes as we change the radius . thus we need to calculate is not a self - conjugate field , and we therefore have to insert the conjugate field at infinity . ] @xmath270 \ .\ ] ] on the neumann boundary we have @xmath271 , and the correlation function equals @xmath272 thus the integral is simply @xmath273 the term that is singular in @xmath274 is subtracted by the last term in ( [ qq1 ] ) . in fact , the only relevant or marginal boundary field that is switched on is the identity field with @xmath275 and @xmath276 , and the corresponding bulk to boundary ope coefficient is @xmath277 since @xmath278 thus we find that @xmath279 in this example , which implies that @xmath280 this then agrees with the geometrical expectation since for @xmath281 with @xmath268 we have @xmath282 a somewhat more interesting example is the configuration of two d1-branes that stretch diagonally across a @xmath283-torus , crossing each other at an angle ( see figure 1 ) . this brane configuration is obviously unstable since the relative open string between the two d1-branes is tachyonic but this will not be important in the following . ( one can imagine that this is only part of a more complicated background involving additional directions , and that the boundary conditions of these d - branes in the other directions are chosen so that the relative open string is not tachyonic . ) * fig . 1 : * _ radius perturbation on the torus : changing the radius of @xmath284 modifies the relative angle between the two branes , and hence the conformal dimension of the corresponding boundary changing field . _ for simplicity consider the situation where the @xmath285 torus is orthogonal with radii @xmath284 and @xmath286 . the bulk operator that changes either radius is an exactly marginal bulk operator , but it does have an important impact on the boundary theory since the ratio of the two radii determines the conformal dimension of the lowest string excitation between the two d1-branes . in the following ( section 3.2.2 ) we shall calculate the change in conformal dimension using the rg formalism we have developed above . as we shall see , this will reproduce the standard formula for the conformal dimension of boundary fields on branes at angles that will be reviewed in section 3.2.1 . the torus theory is described by the action @xmath287 we shall now consider changing the radius @xmath284 by means of the perturbation @xmath288 which to first order gives @xmath289 . the two d - branes stretch diagonally across the torus ; their angle relative to the @xmath290 axis will be denoted by @xmath291 , where @xmath292 satisfies @xmath293 the open string that stretches between the two branes satisfies the @xmath294 boundary condition @xmath295 at one end ( say for @xmath296 on the positive real axis ) , and the @xmath297 boundary condition @xmath298 at the other ( say for @xmath296 on the negative real axis ) . by going to complex variables , _ i.e. _ by writing @xmath299 , @xmath300 , we can write the open string fields as @xmath301 where @xmath302 . the modes @xmath303 satisfy the canonical commutation relations @xmath304=(m-\nu)\delta_{m ,- n}\ , \ ] ] and the virasoro generators can be expressed in terms of them as @xmath305 the conformal dimension of the lowest boundary changing operator @xmath306 is thus @xmath307 according to the analysis of @xcite , the two d - branes will respond to the radius changing bulk perturbation ( [ radiusc ] ) by simply adjusting themselves infinitesimally , so that they continue to stretch diagonally across . to first order in @xmath28 , the angle @xmath292 thus changes via ( [ angle ] ) as @xmath308 with @xmath309 this implies that the conformal dimension of the lowest boundary changing operator changes as @xmath310 this is the result we now want to reproduce using the rg approach explained above . formula ( [ wilson_e3 ] ) applied to the situation at hand reads @xmath311 \ , \end{aligned}\ ] ] where the index @xmath142 runs over all relevant boundary operators in the respective sectors . since the model at hand is gaussian the only relevant operator induced on the boundary by @xmath108 is the identity operator in the respective @xmath294 or @xmath297 sector . the corresponding bulk to boundary ope coefficients can be read off from the expectation values @xmath312 which can be computed using the mode expansions ( [ complex ] ) and ( [ complex1 ] ) for @xmath313 . ( the string fields in the presence of a single boundary are of the same form as ( [ complex ] ) and ( [ complex1 ] ) but with @xmath313 . ) thus @xmath314 the three - point correlator in ( [ e4 ] ) is given by the one - point function of the radius changing operator in the presence of the boundary conditions ( [ bc+ ] ) and ( [ bc- ] ) . a straightforward computation yields @xmath315 viewed as a function on @xmath316 rather than on @xmath317 , the correlator has a logarithmic branch cut along the negative real axis . the integral at hand can be easily evaluated @xmath318 & = & \displaystyle{\frac{\cos(-\theta+\vartheta(2\nu-1))}{4\sin\vartheta}\bigg|^{\vartheta=\pi-\delta}_{\vartheta=\delta } } = -\frac{\cos(-\theta+\delta(2\nu-1))}{2\sin\delta } \ , \nonumber\end{aligned}\ ] ] where in the last step we used @xmath319 . substituting ( [ integral ] ) and ( [ bbs ] ) into ( [ e4 ] ) we obtain @xmath320 = -\frac{1}{2}\ , ( 2\nu -1 ) \ , \sin \theta \ .\ ] ] noting that @xmath321 we finally get the same result as in ( [ geometric_dh ] ) . in section 2 we explained how to calculate higher order coefficients in a theory with bulk and boundary perturbations . actually , the techniques used there can also be easily generalised to the pure bulk or pure boundary case ; this will be sketched in the following . we begin with a discussion about which coefficients of the third order terms in the beta functions contain universal quantities , paralleling the discussion in section 2.1 . in section 4.2 we then describe how to obtain useful formulae for these coefficients . in this subsection we shall only consider the pure bulk theory ; the discussion for the pure boundary case is very similar . consider a conformal field theory perturbed by @xmath322 where as before @xmath323 are the anomalous dimensions , @xmath15 is a renormalisation length scale , and @xmath324 are the dimensionless coupling constants . the beta functions have the general form @xmath325 we take the constants @xmath326 and @xmath327 to be totally symmetric in @xmath328 and @xmath329 , respectively . under a general change of scheme the coupling constants are redefined as @xmath330 where the @xmath331 are , without loss of generality , symmetric in @xmath332 and @xmath333 . the beta functions in the new scheme are @xmath334 + \mathcal{o}(\!\tilde{\lambda}^4)\ . \end{array}\ ] ] we observe that the second order coefficients @xmath326 do not change under this transformation if and only if the second order resonance condition @xmath335 is satisfied . as for the coefficients @xmath327 at the cubic powers of the couplings , it can be seen from ( [ change ] ) that the basic requirement for @xmath327 to be universal is that it satisfies the resonance condition @xmath336 . however , even if the resonance condition is satisfied , the third line in ( [ change ] ) shows that the corresponding coefficient may not be invariant under general scheme changes because of the lower order coefficients @xmath326 . the resulting transformations of the resonant coefficients are parametrised by the tensors @xmath337 . for an @xmath338-dimensional coupling space the dimension of the space of coefficients @xmath327 is @xmath339 while that of the coefficients @xmath337 is @xmath340 . depending on how many coefficients are resonant , there may be some functions defined on these resonant coefficients which are invariant and thus give universal quantities . for example if all couplings are marginal ( @xmath341 for all @xmath332 ) then generically there must be a subspace of scheme independent coefficients of dimension @xmath342 . while in general it is hard to write out explicit expressions for universal quantities in terms of @xmath327 and @xmath326 we can do so in the absence of second order resonances because we can then use a special scheme in which all @xmath326 vanish . given a cubic resonance @xmath336 , the values of the cubic coefficients @xmath343 in that scheme are universal and can be expressed via the coefficients in an arbitrary scheme as @xmath344 the scheme independence of ( [ change_quadrtozero ] ) can be checked directly using ( [ redef4 ] ) and ( [ change ] ) . we can also consider a situation analogous to the one considered in section 2.1 when the universal quantity gives dimension shifts under a truly marginal deformation . let @xmath28 be a coupling constant corresponding to an exactly marginal operator @xmath108 . the beta functions for the other operators @xmath345 have the form @xmath346 we can make the beta functions @xmath347 homogeneous in @xmath348 up to cubic order by a coupling constant redefinition @xmath349 this redefinition is possible because @xmath28 is truly marginal . the anomalous dimensions of the operators @xmath350 are then given by the eigenvalues of the matrix @xmath351 and have the form @xmath352 = y_{i } + \lambda \delta_{i}^{(1 ) } + \lambda^{2}\delta_{i}^{(2 ) } + \dots \ .\ ] ] a straightforward computation yields @xmath353 and @xmath354 the coefficients @xmath355 are resonant and thus universal . one can also check that ( [ d2 ] ) is invariant under an arbitrary coupling constants redefinition of the form @xmath356 where @xmath357 , @xmath358 , @xmath359 are polynomial functions of @xmath28 . finally let us mention the well known fact that if there is a single running coupling constant whose uv dimension is marginal , then both the quadratic and cubic terms in its beta function are universal . now that we have understood which coefficients are universal , we can ask how they can be calculated explicitly . as in the bulk - boundary case discussed in section 2 , we can either use a minimal subtraction scheme ( see section 2.2 ) or the ope scheme of section 2.3 . as before , the minimal subtraction scheme is conceptually clearer since one does not need to introduce beta functions for irrelevant fields . however , the calculation is somewhat unwieldy in this scheme , since one has to isolate the divergences in the uv cut - off @xmath52 for finite ir cut - off @xmath61 . in the following we shall only consider universal quantities for which the calculation in either scheme must give the same answer . since the ope scheme is technically simpler , we shall use it to determine explicit expressions for these coefficients . we have also checked that our result agrees with what would have been obtained in the minimal subtraction scheme ( as must be the case ) . moreover for brevity we will focus on the quantity ( [ change_quadrtozero ] ) . it is straightforward to extend our results to the dimension shifts ( [ d2 ] ) and to a cubic term in a beta function of a single marginal coupling . in the ope scheme the rg equations are determined from the condition that the variation @xmath360 vanishes in the limit @xmath152 . as before we regularise the theory by point splitting , _ i.e. _ we introduce a sharp uv cut - off @xmath52 . in addition we have an ir cut - off @xmath61 . to cubic order in the couplings we have @xmath361 where @xmath362 . the variation @xmath363 of this expression can be computed using ( [ betafunctions_bulk ] ) . setting @xmath364 at second order in the couplings one obtains the well known expression @xmath365 where @xmath366 are the bulk ope coefficients ( [ bulkbulkope ] ) . at the cubic order we have the equation @xmath367 & \displaystyle{+\sum_m\pi c_{ij}{}^{m } \epsilon^{-y_m - y_k}\int d^2z_1d^2z_2\,\theta_{12}\ , \phi_m(z_1,\bar{z}_1)\phi_k(z_2,\bar{z}_2)}\\[18pt ] & \displaystyle{+\sum_l\mathcal{f}_{ijk}^{l } \epsilon^{-y_l}\int d^2z\,\phi_l(z,\bar{z})\,\bigg]}\ . \end{array}\ ] ] as was discussed in section 2.3 the above equation in general may still have divergences . however , in the resonant case , _ i.e. _ if @xmath336 no such complications arise . then we can write @xmath368 & & \qquad-2\pi\sum_m \epsilon^{y_i+y_j - y_m-1 } { c_{i j}^{\phantom{i j}\ ! m}}{\left\langle \phi_m(z_2,\bar{z}_2)\phi_k(0)\phi_l(\infty ) \right\rangle } \bigg]\ , \label{q1}\end{aligned}\ ] ] where @xmath369 , and similarly for @xmath370 . as before , we consider spinless fields , for which we can express the four - point correlator in terms of conformal blocks , @xmath371 with @xmath372 and @xmath373 \sum_mc_{32}{}^{m}c_{m1}{}^{4}f^m_{32,14}(1-\eta)\tilde{f}^m_{32,14}(1-\bar{\eta})\\[4pt ] \sum_mc_{13}{}^{m}c_{m2}{}^{4}f^m_{13,24}(1/\eta)\tilde{f}^m_{13,24}(1/\bar{\eta})\ . \end{array } \right.\ ] ] the conformal blocks are normalised such that @xmath374 replacing the variables in the first part of the square brackets in ( [ coefficient_0 ] ) by an angular variable @xmath375 and the cross ratio @xmath128 , @xmath376 the angular variable can be integrated out by the integral over @xmath377 , and one finds @xmath378 in the second term in the bracket of ( [ q1 ] ) , we changed variables to @xmath379 . the function in the last line in ( [ third_order_1 ] ) can be integrated explicitly , and we obtain @xmath380 } \nonumber \\[5pt ] & & \displaystyle{\qquad+\frac{\pi^2}{3}\sum_{\mathrm{perm}(i , j , k ) } \sum_m\frac{c_{ij}{}^{m } c_{mk}{}^{l}}{y_l - y_k - y_m}\ .}\end{aligned}\ ] ] the universal quantity @xmath381 defined in ( [ change_quadrtozero ] ) is then simply given by the first two lines of ( [ third_order_2 ] ) . it can be checked using the asymptotics ( [ conf_block_asymptotic ] ) and the properties of conformal blocks ( [ bulk_conformal_block ] ) that the integral ( [ third_order_1 ] ) converges in the regions @xmath382 , @xmath383 and @xmath384 . one can thus safely set @xmath385 in ( [ third_order_1 ] ) to obtain an integral expression @xmath386 & & \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \displaystyle{-\sum_m c_{ij}{}^{m } c_{mk}{}^{l}|\eta|^{y_i+y_j - y_m-2}\bigg\}}\ .\end{aligned}\ ] ] note also that expression ( [ third_order_3 ] ) is @xmath61 independent . in particular this means that the infrared divergences that were present in individual summands in ( [ third_order_2 ] ) mutually cancel each other . this agrees with the general results of @xcite . by suitable changes of the integration variable @xmath128 in the terms with permuted indices @xmath329 it is possible to write @xmath327 by means of integrals over three disjoint subsets tiling the whole @xmath128-plane . consider the transformation @xmath387 , for which the cut - off functions in the integral ( [ third_order_3 ] ) become @xmath388 . the asymptotics ( [ bulk_conformal_block ] ) for @xmath389 are such that the divergence of the transformed integrand that arise from the limit @xmath390 is again canceled , once we take the transformed subtractions , _ i.e. _ the second line in ( [ third_order_3 ] ) , into account . the other transformation is @xmath391 . in this case the cut - off functions read @xmath392 after the transformation , and the divergence of the corresponding integrand for @xmath393 is canceled as well . together , the regions carved out by the cut - off functions for the three coordinate choices tile the whole @xmath128 plane . using this we can recast ( [ third_order_3 ] ) as @xmath394 \ , \ ] ] where @xmath395 in this form the integration runs over the whole @xmath128-plane . although the subtraction function @xmath396 still has a piecewise form it is expressed quite explicitly . on the boundary the computation can be done in a similar way as in the bulk . we consider a boundary perturbation of the form @xmath397 where now @xmath398 . up to the third order in the couplings the rg equations take the form @xmath399 as before , we only introduce counterterms at the quadratic order for marginal or relevant fields , and the corresponding coefficients are @xmath400 where @xmath401 is the ope coefficient of two boundary fields ( [ boundaryboundaryope ] ) . in the resonant case where we have @xmath402 , the coefficient @xmath403 can be written as @xmath404 by similar arguments as in the previous subsection we can find @xmath405 where @xmath406 here the conformal blocks @xmath407 have cuts running from @xmath129 to zero , and from @xmath130 to @xmath408 . in addition , their asymptotic behaviour is @xmath409 where @xmath274 is defined as before , _ i.e. _ @xmath410 . finally , the scheme - independent quantity is given by @xmath411 in the case where several irreducible boundary conditions are involved , one has to keep track of their labels , and bear in mind the superselection rules , in particular the order of operators . this leads to additional splittings and recombinations of the integrals over four - point functions and subtractions . apart from this technicality , it is however straightforward to include the boundary labels . we have refrained from writing them explicitly to keep the formulae simpler . in this paper we have studied conformal perturbation theory beyond the leading order . we have shown that , at least up to quadratic order , the combined bulk boundary perturbation problem is renormalisable , using the minimal subtraction scheme . we also discussed the more commonly used ` wilsonian ' ope scheme , and found it to have some shortcomings at higher order in perturbation theory . we identified systematically the universal ( scheme - independent ) quantities , and gave explicit formulae for them at third order in terms of integrals of conformal 4-point functions . finally , we explained how essentially the same analysis works for the pure bulk and pure boundary case . it seems plausible that similar techniques should allow one to prove renormalisability at arbitrary order in perturbation theory , but we have not attempted to do so . our work was originally motivated by the question of how the dependence of the conformal dimension of a boundary changing field upon a bulk modulus can be understood from the world - sheet perspective . our considerations demonstrate that this effect is captured by a certain universal quadratic rg coefficient , for which we gave an explicit formula . this result should also have interesting applications in other contexts ; in particular , it provides a world - sheet method to study the stability of brane setups under arbitrary bulk deformations . the research of mrg and csc has been partially supported by the swiss national science foundation and the marie curie network ` constituents , fundamental forces and symmetries of the universe ' ( mrtn - ct-2004 - 005104 ) . csc thanks the university of heriot - watt , edinburgh , for hospitality , where part of this work was carried out . we thank stefan hohenegger , christoph keller , and andreas ludwig for useful conversations . zamolodchikov , _ renormalization group and perturbation theory near fixed points in two - dimensional field theory _ , sov . j. nucl . * 46 * ( 1987 ) 1090 [ yad . fiz . * 46 * ( 1987 ) 1819 ] . zamolodchikov , _ integrable field theory from conformal field theory _ , adv . stud . pure math . * 19 * ( 1989 ) 641 . zamolodchikov , _ irreversibility of the flux of the renormalization group in a 2d field theory _ , jetp lett . * 43 * ( 1986 ) 730 [ pisma zh . * 43 * ( 1986 ) 565 ] . d. friedan and a. konechny , _ on the boundary entropy of one - dimensional quantum systems at low temperature _ , phys . * 93 * ( 2004 ) 030402 . i. affleck and a.w.w . universal noninteger ground state degeneracy in critical quantum systems _ , phys . * 67 * ( 1991 ) 161 . freedman , m. headrick and a. lawrence , _ on closed string tachyon dynamics _ d * 73 * ( 2006 ) 066015 . k. graham , a. konechny and j. teschner , _ on the time - dependent description for the decay of unstable d - branes _ , jhep * 0702 * ( 2007 ) 011 . s. fredenhagen , m.r . gaberdiel and c.a . keller , _ bulk induced boundary perturbations _ , j. phys . a * 40 * ( 2007 ) f17 . zamolodchikov , _ renormalization group and perturbation theory about fixed points in two - dimensional field theory _ , sov . j. nucl . * 46 * ( 1987 ) 1090 . f. constantinescu and r. flume , _ perturbation theory around two - dimensional critical systems through holomorphic decomposition _ , j. phys . a * 23 * ( 1990 ) 2971 . p. chaselon , f. constantinescu and r. flume , _ deformations along critical lines of field theories in two dimensions _ , b * 257 * ( 1991 ) 63 . ludwig and k.j . the 4-loop beta - function in the 2d non - abelian thirring model , and comparison with its conjectured exact form _ , nucl . b * 661 * ( 2003 ) 577 . k. graham , _ on perturbations of unitary minimal models by boundary condition changing operators _ , jhep * 0203 * ( 2002 ) 028 . s. fredenhagen , m.r . gaberdiel and c.a . keller , _ symmetries of perturbed conformal field theories _ , j. phys . a * 40 * ( 2007 ) 13685 .

we have used the gemini near - infrared integral field spectrograph ( nifs ) to map the gas kinematics of the inner 200pc of the seyfert galaxy ngc4151 in the z , j , h and k bands at a resolving power and spatial resolution ofpc . the ionised gas emission is most extended along the known ionisation bi - cone at position angle pa=60240 , but is observed also along its equatorial plane . this indicates that the agn ionizes gas beyond the borders of the bi - cone , within a sphere with radius around the nucleus . the ionised gas has three kinematic components : ( 1 ) one observed at the systemic velocity and interpreted as originating in the galaxy disk ; ( 2 ) one outflowing along the bi - cone , with line - of - sight velocities between and 600kms and strongest emission atkms ; ( 3 ) and another component due to the interaction of the radio jet with ambient gas . the radio jet ( at pa=75255 ) is not aligned with the nlr , and produces flux enhancements mostly observed at the systemic velocity , suggesting that the jet is launched close to the plane of the galaxy ( plane of the sky ) . the mass outflow rate , estimated to bemyr along each cone , exceeds the inferred black hole accretion rate by a factor of . this can be understood if the narrow - line - region ( nlr ) is formed mostly by entrained gas from the circumnuclear interstellar medium by an outflow probably originating in the accretion disk . this flow represents feedback from the agn , estimated to release a kinetic power ofergs , which is only.3% of the bolometric luminosity of the agn . there is no evidence in our data for the gradual acceleration followed by gradual deceleration proposed by previous modelling of the [ oiii ] emitting gas . our data allow the possibility that the nlr clouds are accelerated close to the nucleus ( within 0 pc ) after which the flow moves at essentially constant velocity (kms ) , being consistent with nir emission arising predominantly from the interaction of the outflow with gas in the galactic disk . the molecular gas exhiibits distinct kinematics relative to the ionised gas . its emission arises in extended regions approximately perpendicular to the axis of the bi - cone and along the axis of the galaxy s stellar bar , avoiding the innermost ionised regions . it does not show an outflowing component , being observed only at velocities very close to systemic , and is thus consistent with an origin in the galaxy plane . this hot molecular gas may only be the tracer of a larger reservoir of colder gas which represents the agn feeding . [ firstpage ] galaxies : nuclei , galaxies : active , galaxies : seyfert , galaxies : kinematics , galaxies : jets , galaxies : individual ( ngc4151 )

0911.2212

this is a continuing study of the narrow - line region ( hereafter nlr ) of the seyfert galaxy ngc4151 using data obtained with the gemini near - infrared integral field spectrograph ( nifs ) . the data comprise spectra of the inner@xmath12pc@xmath13 , at a spatial resolution of@xmath14pc at the galaxy , covering the wavelength range@xmath15@xmath16 m at a spectral resolving power r@xmath17 . in a previous paper ( * ? ? ? * hereafter paperi ) we have used these data to map the nlr intensity distributions of @xmath18 emission lines , as well as their ratios . the main results were the distinct flux distributions and physical properties observed for the ionised , molecular and coronal gas . the ionised gas is co - spatial with the ionisation bi - cone observed in the optical ] emission at position angle ( hereafter pa)@xmath19 @xcite , and seems to trace the outflow along the bi - cone @xcite . in the inner region , the nir ionised - gas emission extends beyond the borders of the cone , which does not have a sharp apex as noted also by @xcite . the h@xmath20 molecular gas intensity distribution , on the other hand , avoids the region of the bi - cone and seems to originate in the galaxy disc , while the coronal gas emission is barely resolved . in another recent work @xcite , we have studied the unresolved nuclear continuum showing that its origin is emission by hot dust within @xmath44pc from the nucleus as expected from the dusty torus postulated by the unified model @xcite . the origin of this structure is probably a dusty wind that originates in the outer parts of the accretion disk . this wind is probably clumpy , as suggested by recent models @xcite , and necessary in order to allow the escape of radiation along the equatorial plane of the bi - cone in order to ionize the gas and produce the observed flux maps of paperi . in the present paper we use the nifs data to map the nlr kinematics of ngc4151 . although many papers have been devoted to such a study , most of them are based on long - slit spectroscopy obtained in the optical with the hubble space telescope ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the present study was performed in the near - ir , a waveband which is less affected by dust , and using integral field spectroscopy , which allows a full two - dimensional ( hereafter 2d ) coverage of the nlr kinematics . with our data and analysis we aim to examine previous claims of acceleration of the gas along the nlr , quantify the mass outflow rate and the corresponding feedback power , as well as investigate the origin of the nlr gas . as ngc4151 harbors the closest bright agn , its nlr is one of the best suited for this type of study . our approach for the analysis of the nlr kinematics of ngc4151 is as follows . first we map the centroid velocity and velocity dispersion of the ionised , molecular and coronal gas , obtained from fits to the emission - line profiles and compare our results with those of previous studies . then we use an alternative approach to map the nlr kinematics , made possible by integral field spectrographs : a `` velocity tomography '' of the emitting gas , obtained by slicing the line profiles in velocity bins of@xmath21kms@xmath6 , producing channel maps which provide a clearer view of the velocity distribution in the different gas phases . we have grouped the channel maps together in a sequence of velocity bins , generating movies which can be recovered at the authors website.thaisa/ifu_movies/ngc4151 ] we adopt in this paper the same distance to ngc4151 as in paperi:@xmath22mpc , corresponding to a scale at the galaxy of@xmath23pcarcsec@xmath6 @xcite . the paper is organised as follows . in [ data ] we provide a quick description of the data , in [ results ] we present the centroid velocity and velocity dispersion measurements , in [ tomography ] we present the emission - line `` tomography '' , in [ discussion ] we discuss our results , in [ feed ] we estimate the mass outflow rate along the nlr and compare with the mass accretion rate and in [ conclusion ] we present our conclusions . the observations comprise three pointings obtained with the gemini north telescope with the nifs instrument @xcite operating with the altair adaptive optics module , each covering a square field of view of@xmath24arcsec@xmath13 . one of the nifs fields was centred on the ngc4151 nucleus and the others were offset by@xmath25arcsec along position angle@xmath26 , resulting in a total field of view of@xmath27arcsec@xmath13 , with the longest extent oriented approximately along the radio jet @xcite . the spectra cover the z , j , h , and k spectral bands at two - pixel resolving powers of @xmath28 , @xmath29 , @xmath30 , and @xmath30 , respectively . the angular resolution , as determined from the fwhm of the psf , ranges from @xmath31arcsec in the h and k spectral bands to @xmath32arcsec in the z and j bands . this corresponds to a spatial resolution at the galaxy of @xmath33 pc at h and k bands , and @xmath34pc in the z and j bands . more details about the data and the data reduction can be found in paperi , as well as flux measurements of 61 emission lines from three characteristic locations along the nlr . in the present paper , we concentrate on the kinematics measured for the strongest emission lines of the ionised and molecular gas , namely : ] @xmath35 m , @xmath36 m , ] @xmath37 m , pa@xmath38@xmath39 m , ] @xmath40 m and h@xmath20@xmath41 m . we have also measured the kinematics for the coronal lines of ] @xmath42 m and ] @xmath43 m . we first present and discuss the centroid velocity and velocity dispersion maps and then present the velocity tomography . centroid velocity and velocity dispersion ( @xmath44 ) maps were obtained by fitting gaussians to the emission - line profiles . the velocities were obtained from the central wavelength of the gaussians ( after subtraction of the systemic velocity of @xmath45kms@xmath6 , and applying a heliocentric correction of @xmath46kms@xmath6 ) while the velocity dispersions were obtained from the full width at half maximum ( hereafter fwhm ) of the gaussians as @xmath47fwhm@xmath48 . in many locations along the nlr of ngc4151 , the strongest ionised gas emission lines those of ] and show two kinematic components , with a mean velocity separation between them ranging from @xmath49 to @xmath50kms@xmath6 . we were successful in fitting two gaussian components to both these lines , although the line has a broad component and a nuclear absorption component which introduces uncertainties in the fitting procedure for regions close to the nucleus . in a few regions , we observe double components also in ] , pa@xmath38 , and br@xmath51 . elsewhere , these lines are broader at locations where double components are observed in ] and . this suggests that the ] , pa@xmath38 , and br@xmath51 emission lines are also double . the lower signal - to - noise ratio ( hereafter s / n ) of these lines , the presence of a broad component for the emission - lines close to the nucleus , and the fact that the ] emission line appears to be intrinsically broader contribute to blend together two distinct kinematic components . double components have been previously observed in the ] @xmath52 emission - line profiles by @xcite , and @xcite . @xcite using higher resolution ( r@xmath53 ) spectra , showed that in some locations there are even three or four components . our lower spectral resolution did not allow us to resolve more than two components . typical fits to the ] emission - line profile are shown in fig . [ f - profile ] , corresponding to locations ( a , b , c , d and e ) identified in the ] radial velocity maps which are shown in fig . [ f - doublevel ] . because the s / n is not high enough to allow the fit of each component with the necessary number of free parameters , we had to constrain the width of the two gaussian components to the same value . the implication of this assumption is that there will be some uncertainty in the gas kinematics derived in this way , mainly in the velocity dispersions , as discussed bellow . although in the next section we derive the kinematics in terms of channel maps , with no assumptions , we decided to include the kinematics in terms of two components in order to be able to compare our results with those of previous studies , in which this was the method used to derive the gas kinematics . in addition , the separation in two components helps in the identification of the origin of the gas emission . in the top panels of fig . [ f - doublevel ] we show velocity maps in the emission lines of ] and of regions with one kinematic component , after subtraction of the systemic velocity . in the middle and bottom panels we show the velocity maps of the regions with two kinematic components : the middle panel shows the high velocity component , corresponding to negative ( blueshifted ) velocities to the sw and positive ( redshifted ) velocities to the ne , while the bottom panel shows the low velocity component . for the ] emission , the lowest velocities are observed along the galaxy minor axis ( pa@xmath54 ) in the upper panel of fig.[f - doublevel ] ( single component ) , and in neighboring locations where two components were fit to the data . this low velocity component , essentially at zero velocity , probably originates in the galaxy disc , while the redshifted and blueshifted velocity components originate in the outflow along the bi - cone . the highest blueshifts are observed in the bottom part of the strip labelled c " in the middle left panel of fig.[f - doublevel ] , while the highest redshifts are observed in the region labelled b " . in order to better illustrate the magnitude of the velocity variations , and to investigate further the origin of the different gas components , we show in fig . [ f - pv ] the velocities as a function of distance from the nucleus along pa@xmath55 . the black dots represent the single components and the red dots represent the double components , while the labels allow the comparison with fig.[f - doublevel ] . [ f - pv ] suggests the presence of a @xmath4 zero velocity component observed between @xmath56 and @xmath57 from the nucleus along pa@xmath58 , whose origin is probably the galaxy plane , as pointed out above . then there is a sequence of blueshifts at @xmath59kms@xmath6 to the sw ( positive distances in fig.[f - pv ] ) which probably originates in the far side of the approaching cone ( see fig.[f - geom ] and discussion in sec.[geom ] ) . a similar sequence of redshifts at somewhat lower velocities , up to @xmath60kms@xmath6 to the ne can be identified with the near side of the receding cone . the highest blueshifts to the sw and highest redshifts to the ne can then be identified with the near side of the approaching cone and the far side of the receding cone , respectively . these highest velocities thus originate in the walls of the bi - cone that are most inclined relative to the line - of - sight . velocity dispersion ( @xmath44 ) maps of the ] and emission lines for the single and two - component fits are shown in fig . [ f - doublesig ] . we note again that the measured velocity dispersions are upper limits to the intrinisic gas velocity dispersion in regions where a single gaussian component has been fit , possibly to multiple unresolved components . for example , in region e in fig . [ f - doublesig ] , the velocity dispersion is @xmath2150kms@xmath6 . but this large velocity dispersion is probably due to the blending of at least two components , as it surrounds a region with two components ( region d ) . similarly , the measured values are only indicative of the gas velocity dispersion in regions where two components have been fit , as their widths have been constrained to the same value . this is the case of region e , where there are a component at zero velocity and a blueshifted component , which do not necessary have the same velocity dispersion . nevertheless , these @xmath44 values should not be too far off because they are of the order of the @xmath44 value of the zero velocity single component along the minor axis of the galaxy . the lowest @xmath44 values for the single components are @xmath0 100kms@xmath6 . these are seen mostly along the minor axis of the galaxy . the low velocity dispersion suggests that the emitting gas in these regions is not disturbed and probably lies close to the galaxy plane . the velocity dispersion is larger in other regions this could be due to turbulence caused by the outflow or to multiple velocity components . the @xmath44 values for the double components are in some regions ( e.g. region d " ) almost as low as the lowest values of the single components , suggesting that at least one of the components originates in the galaxy plane . regions of higher @xmath44 values in the two components are those labelled b " in the ne outflow and the bottom part of the strip c " , where the highest blueshifts are seen in the centroid velocity maps , supporting an origin in the front side of the sw cone , as discussed above . the centroid velocity distribution of h@xmath20 emission lines is dominated by velocities close to systemic . this emission is concentrated along the minor axis of the galaxy , whose orientation approximately coincides with that of the galaxy bar . there are some redshifts to the ne mainly along the galaxy major axis and some blueshifts to the sw along both the galaxy major axis and the cone axis ( see fig.[f - chmh2 ] ) . the velocity dispersions are mostly constant at @xmath480kms@xmath6 supporting an origin for this gas in the galaxy plane . the coronal line flux distributions are just barely resolved and their centroid velocities are also close to systemic , while their velocity dispersions are @xmath4100kms@xmath6 . we have also mapped the gas kinematics by integrating the fluxes within wavelength slices extracted along the emission - line profiles corresponding to velocity bins of @xmath61kms@xmath6 . these channel maps " display the distributions of the gas emission at different velocities , which occur at different depths along the line of sight . in this sense , they can be described as velocities tomography " of the nlr using emission - line scanning . the channel maps are shown in figs . [ f - chmsiii ] , [ f - chmfeii ] , [ f - chmbrg ] and [ f - chmh2 ] for the emission lines of ] , ] , br@xmath51 and h@xmath20 , respectively . channel maps in the other ionised gas emission lines , such as and pa@xmath38 , are similar to those in ] and thus are not shown here . the lowest flux shown in the channel maps is @xmath62 times the root mean square variation of the surrounding sky flux . the bi - cone axis at pa@xmath55 as determined by @xcite is represented by a dashed line in the figures , while the major axis of the galaxy at pa@xmath63 is represented by a continuous line . the contours of a radio merlin image ( kindly made available by c. mundell ) have been overplotted on the central velocity bins ( velocities close to systemic ) . in the channel maps of the ] emission line , shown in fig . [ f - chmsiii ] , the most blueshifted gas at @xmath64kms@xmath6 lies very close to the nucleus . for blueshifts of @xmath65kms@xmath6 the emission extends further out , and for lower velocities broadens in the shape of a projected cone extending from the bi - cone axis towards the galaxy major axis . at the slice corresponding to @xmath66kms@xmath6 there seems to be a spatial association between the line emission and the radio structure . in particular , both of them broaden towards the major axis . at velocities close to zero , and up to redshifts @xmath67kms@xmath6 , the emission is extended in all directions around the nucleus , evidencing the absence of a sharp cone apex . the largest elongation of this emission occurs along the horizontal axis , which is approximately the orientation of the radio axis . the overplotted radio contours show that , although there is not a detailed correspondence between the ] and radio image , the brightest regions of the ] image coincide with the two inner knots of the radio image . for redshifts larger than @xmath68kms@xmath6 , the gas is observed mostly to the ne in the opposite direction to that of the blueshifted gas , showing also a similar flux distribution : for the lowest redshifts the emission is distributed in a fan structure extending from the major axis of the galaxy to the bi - cone axis , while for the intermediate and the highest redshifts the emission is more elongated and concentrated closer to the bi - cone axis to the ne . channel maps for the ] emission line are shown in fig . [ f - chmfeii ] . in the high - velocity blueshifted channels , this emission seems to be more collimated along the bi - cone axis than for other emission lines . some blueshifts from @xmath69kms@xmath6 to zero velocity are observed also to the ne , unlike the slices in the other emission lines . the flux distributions for velocities close to zero and redshifted gas are similar to those in ] , although the highest redshifts are as high as the highest blueshifts and are observed up to similar distances from the nucleus . in all velocity channels , there seems to be more structure in ] than in ] , in the form of knots of emission in closer association with the radio knots . this is particularly clear at the velocity bins from @xmath69 to @xmath70kms@xmath6 : at @xmath66kms@xmath6 the strongest emission traces the `` flaring '' of the radio structure to sw towards the galaxy major axis ; between velocities of @xmath71 and @xmath72kms@xmath6 , the strongest emission approximately traces the two central radio knots and , for higher redshifts , there is an enhancement of the ] emission at @xmath73arcsec ne , at the location of another radio knot . fig.[f - chmbrg ] shows the channel maps in the br@xmath51 emission line . these show similar morphologies to those seen in ] , but with more structure . this can be attributed , at least in part , to the better spatial resolution achieved by the adaptive optics system in the k band relative to the z band . the channel maps in the h@xmath20 emission line , shown in fig . [ f - chmh2 ] , are completely different from those of the ionised gas , in support of a distinct origin for the molecular gas , as already pointed out in . most of the emission occurs in the @xmath0 zero velocity channels . a small excess of blueshifts to the sw ( from @xmath74kms@xmath6 ) and similar redshifts to the ne suggest rotation around the minor axis . the flux distributions show a hollow core around the nucleus , and beyond this region the emission surrounds the minor axis of the galaxy in the shape of two arcs extending up to @xmath75arcsec to nw and se of the nucleus . a comparison with the channel maps in the ionised gas at the same velocity bins ( around zero velocity ) shows that the molecular gas emission avoids the innermost region of ionised gas , and the hollow structure can thus be explained by the presence of ionised gas surrounding the nucleus at zero velocity . in the centroid velocity maps for the ionised gas emission ( fig.[f - doublevel ] ) , the lowest velocities are observed at the nucleus , while the highest are observed away from the nucleus . this is the case for ngc4151 and also for 6 other seyfert galaxies for which we have recently obtained kinematic data with the ifu of the gemini multi - objetct spectrograph ( gmos ) @xcite . the channel maps , on the other hand , show a different behavior : the highest velocities are usually observed at the nucleus @xcite . in the case of ngc4151 , blueshifts of @xmath76kms@xmath6 in ] and ] are observed at the nucleus and in a narrow region along the bi - cone axis extending from the nucleus almost to the outer edge of the nlr ( figs . [ f - chmsiii ] and [ f - chmfeii ] ) . lower velocity blueshifted gas appears to fill the bi - cone and extend into the region of the radio jet near the systemic velocity . a similar structure is seen in redshifted gas with the emission at moderate redshifts appearing to fill the receding bi - cone . emission at the highest redshifts of @xmath77kms@xmath6 occurs at the nucleus and in a localised region @xmath0 0@xmath109 to the ne of the nucleus that is offset from the bi - cone axis . this morphology differs from that of the highest velocity blueshifted gas and may simply reflect the clumpy nature of the nlr emission . we attribute the difference between the kinematics probed by the centroid velocity and channel maps to the fact that the centroid velocity probes the brightest emission , while the channel maps also probe fainter emission ( in the wings of the line profiles ) . in the vicinity of the nucleus the brightest component is the one originating in the galactic disc , while away from the nucleus the brightest component is the outflowing one . as a consequence , the centroid velocities show an apparent increase in velocity outwards , that mimics acceleration along the nlr . in the channel maps we see high velocity gas at the nucleus , showing that the outflow does not leave the nucleus at zero velocity . we interpret this high velocity gas as an outflowing wind from the agn or ambient gas more directly interacting with this wind . as this highest velocity gas moves away from the nucleus , it pushes and accelerates the gas from the disc . the small amount of high velocity gas thus transfers momentum to a large reservoir of low velocity gas from the disc , causing the observed emission - line profiles and velocity structure along the bi - cone . this scenario is consistent with the one proposed by @xcite who , after verifying the failure of various wind models to reproduce the observed outflows in ngc4151 , suggested that the interaction of an outflowing nuclear wind with the ism could explain the observed velocity profile . some of the previous studies of the nlr of ngc4151 have suggested that there is no relation between the radio jet and the bi - conical outflow @xcite . the distinct orientations of the two outflows do support distinct origins for them : the radio jet ( oriented at pa@xmath475@xmath3 ) originates in the funnel of the accretion disc , while the bi - conical outflow originates further out , probably still in the accretion disc . the disc should also be warped , in order to produce a wind with a distinct orientation ( pa=60@xmath3 ) from that of the radio jet . the channel maps of figs . [ f - chmsiii ] , [ f - chmfeii ] and [ f - chmbrg ] confirm that the flux distributions at high velocities show no relation with the radio jet . but at velocities close to systemic the flux distributions become oriented along the radio axis instead of along the bi - cone . in addition , a comparison between the channel maps for velocities close to systemic and the contours of the radio map reveals line flux enhancements at the locations of radio knots . our proposed interpretation is that the radio jet has been launched close to the plane of the galaxy , and does interact with the circumnuclear gas pushing it out approximately along the plane and introducing the flux enhancements mostly due to gas density enhancement which are observed at velocities close to systemic . the observed flux enhancements could also be due to extra ionization by shock - induced x - rays in the nlr clouds interacting with the jet . recent chandra x - ray observations of the nlr of ngc4151 @xcite have shown extended emission along the nlr with x - ray flux enhancements in association with radio knots . they show that the radio - jet parameters are only consistent with the x - ray flux beeing due to thermal emission associated with the radio jet . the presence of shock heating is supported by the results of paperi , namely the increase of the [ feii]1.257/[pii ] line ratio to values larger than @xmath42 in association with the radio knots . values larger than @xmath42 imply the presence of fast shocks to destroy dust grains and release the fe . we also find , in paperi , a high temperature ( 16 000k ) in the flux - enhanced regions , as derived from the [ feii ] line ratios , in agreement with this scenario . in the channel maps , the relation of the gas emission with the radio jet seems also to be strongest for ] , as can be observed in fig . [ f - chmfeii ] : at velocities from @xmath78 to @xmath79kms@xmath6 , there is enhanced ] emission at the locations of most radio knots . there seems also to be an association between a `` flared structure '' to the sw observed in the radio map with the line emission at velocities of up to @xmath80kms@xmath6 . this suggests that the radio jet has been deflected towards the line - of - sight at this location . some spatial association between optical and radio emission in the nlr of ngc4151 has also been observed by @xcite . comparing high resolution radio images to an hst ] image , they found a number of bright ] clouds that seem to bound a number of knots observed in the radio image , which they interpret are due to shocks of the radio jet with the circumnuclear ism as the jet moves through the nlr and encounters gas clouds . @xcite studied the ] kinematics of the nlr of the prototypical seyfert 2 galaxy , ngc1068 , using a long - slit spectrum obtained at pa 202 with stis on hst . they found a distribution of line centroid velocities along the slit that resembled a `` figure of eight '' , with prominent high - velocity redshifted and blueshifted emission clumps and relatively few low - velocity clouds at intermediate distances from the nucleus . this distribution led them to infer a highly inclined conical geometry for the nlr of ngc1068 , with the emission arising from clouds outflowing radially along the cone walls . the cloud outflow velocities increase approximately linearly with radial distance on both sides of the nucleus to @xmath81kms@xmath6 at distances of @xmath82arcsec from the nucleus , and then decline back to zero by @xmath83arcsec . @xcite were forced to hypothesise a gradual acceleration of the nlr clouds over @xmath84pc followed by gradual deceleration to @xmath85pc to account for this velocity structure . @xcite used a more extensive set of stis spectra to develop this model for ngc1068 further . @xcite applied the same model to stis spectra of the nlr of ngc4151 . the nlr of ngc4151 is far less inclined to the line of sight than ngc1068 @xcite infer @xmath86 compared to @xmath87 for ngc1068 so radial flows along the cone walls do not separate as cleanly into high - velocity redshifted and blueshifted structures . consequently , the fit to the ngc4151 data is less compelling than for ngc1068 . @xcite revisited the ngc4151 nlr kinematics by fitting a similar model to spectra from five adjacent stis slit positions . their comprehensive analysis provides strong evidence that the nlr of ngc4151 is inclined by @xmath88 to the line of sight , with a hollow bi - conical outflow occupying cone half - angles between @xmath89 and @xmath90 . @xcite conclude that the outflow along the walls of the bi - cone is best reproduced with a velocity law : @xmath91 valid for distances @xmath92 from the nucleus up to a turnover point at @xmath93pc ( @xmath94 ) , from where the gas starts to decelerate and reaches the systemic velocity at @xmath95pc ( @xmath96 ) . the sw side of the bi - cone is tilted toward us , as illustrated in fig . [ f - geom ] . in this geometry and orientation , we view the nucleus just outside the outer wall , at an angle of @xmath97 with respect to the nearest part of the sw cone . this geometry and orientation has led to the idea that we see ngc4151 as a seyfert 1 galaxy because the blr can be viewed directly through a clumpy medium defining the cone walls . the angle between the bi - cone axis and the normal to the galaxy disk is @xmath98 @xcite , and thus , for a total opening angle of the bi - cone of @xmath99 , one side of the cone exits the galactic disk steeply , with the outer wall making an angle of 87@xmath3 with the galaxy plane , while the other outer wall leaves the plane at only 21@xmath3 , impacting a larger area of the disk ( for a typical scale height for the disc of several tens of pc ) , as illustrated in fig.[f - geom ] . @xcite argue that the intersection of the cone with the disc produces the observed geometry of the extended narrow - line region as shown by @xcite . if we now inspect our centroid velocity maps of fig . [ f - doublevel ] and fig.[f - pv ] in light of the geometry proposed by @xcite , we conclude that , in most regions where we see two velocity components , there is always one component at the systemic velocity plus an outflowing component in blueshift to the sw and in redshift to the ne . we attribute the systemic velocity component to emitting gas from the galaxy disc , in which case we are led to the conclusion that the agn ionizes a region of the galaxy disc comparable in extent to that of the outflowing gas emission we detect along the bi - cone axis ( @xmath415 , or @xmath4100pc ) and somewhat less extended perpendicular to it ( @xmath41 , or @xmath465pc ) . blueshifted components are observed to the sw at @xmath100kms@xmath6 and at higher velocities of up to @xmath5kms@xmath6 ( fig.[f - doublevel ] for ] and fig.[f - pv ] ) . we interpret the component with blueshifts @xmath100kms@xmath6 at the position labelled _ d _ in fig.[f - doublevel ] , together with the single component surrounding it ( region _ e _ in fig.[f - doublevel ] ) , as originating in the back surface of the sw cone . in the bottom part of the region we labelled _ c _ in fig.[f - doublevel ] , we observe a few points with higher blueshifts of up to @xmath5kms@xmath6 ( see also fig.[f - pv ] ) . these components probably originate in the front part of the sw cone , which we see at a small angle ( @xmath97 ) and thus with a large velocity component along the line - of - sight . there is less emission from this front part of the sw cone than from the back , and our hypothesis to explain this is that the cone walls are formed by gas entrained from the galaxy disk . the front wall is thus likely to be comprised of more tenuous gas than the back wall , as there is less gas to be entrained at high angles to the galactic disk . redshifted components are observed to the ne at somewhat lower velocities than to the sw ( region labelled _ b _ in fig.[f - doublevel ] ) . this can be seen more clearly in fig.[f - pv ] : there is a sequence of redshifts of up to @xmath60kms@xmath6 , which can be attributed to the front of the ne cone and then some higher redshifts of up to 400500kms@xmath6 , which can be attributed to the back of the ne cone . in fig.[f - doublevel ] , there is more emission from gas at the highest redshifts than emission from gas at the highest blueshifts . this result , together with the lower redshifts to the ne than blueshifts to the sw , suggest that the ne cone has its axis somewhat tilted towards the line of sight , instead of being aligned with the axis of the sw cone . this tilt would allow us to see more of the back wall of the ne cone and would result in lower observed velocities for both its front and back walls . in the model proposed by @xcite , the nlr gas first accelerates up to @xmath4100pc from the nucleus and then decelerates , analogous to their successful model for ngc1068 . while it is appealing in terms of agn unification for ngc1068 and ngc4151 to be fit with a similar model , neither our centroid velocity maps of figs.[f - doublevel ] and [ f - pv ] nor the ] kinematic data for ngc4151 unambiguously point to this interpretation . the gradual acceleration over the inner @xmath101pc and then deceleration required by the model remain unexplained . @xcite noted that radiative and magnetic acceleration from the region of the accretion disk both accelerate gas on length scales of the order of the launch radius , so winds driven by these mechanisms should reach their terminal velocities by @xmath102pc from the central source , not the @xmath103pc required by the model fits to the data . thermal winds also fail in this regard @xcite . dynamical models including various acceleration and deceleration mechanisms have been developed @xcite . while these models fail to identify the acceleration mechanism , they do demonstrate that gravitational forces alone can not produce the required deceleration , while drag forces from the interaction of nlr clouds with a diffuse , hot , x - ray - emitting ambient medium could be responsible . the difficulty in identifying a suitable wind acceleration mechanism leads to the question of whether some of the nlr kinematics are due to gradual global wind acceleration or whether they result from the local interaction of an already accelerated wind with its ambient medium , as we have proposed above on the basis of the channel maps . this scenario has been proposed by @xcite , who realised that the interaction of an outflowing nuclear wind with the ism could reproduce the apparent acceleration of the nlr following the works of @xcite and @xcite . these authors showed that a protostellar hydromagnetic wind expanding into the surrounding gas produces a bipolar outflow with velocity proportional to the distance to the star until a maximum radius beyond which the gas starts to decelerate . in order to be able to compare our data with those of previous works ( e.g. @xcite , @xcite ) , we have built position - velocity diagrams ( hereafter pv diagrams ) , which are shown in fig . [ f - pv2siii ] for the ] emission line . these diagrams were obtained from the @xmath104-band nifs data cube by extracting @xmath105arcsec wide pseudo - slit spectra oriented at pa@xmath55 ( which is the orientation of the bi - cone axis ) , offset from the nucleus perpendicular to the pseudo - slit direction by between @xmath106 ( along pa@xmath107 ) and @xmath108 ( along pa@xmath109 ) in steps of @xmath110 . the ] diagrams are plotted both on a logarithmic scale ( top of fig . [ f - pv2siii ] ) to reveal the faint emission structure and on a linear scale ( bottom ) to show the regions with strongest emission . while qualitatively similar to previous long - slit ] spectra ( e.g. fig . 2 of * ? ? ? * ) , the ] diagrams seem to show less emission at velocities exceeding @xmath111kms@xmath6 . in particular , high - velocity ] clouds a and b in @xcite ( near @xmath112kms@xmath6 and @xmath113 ) are not apparent in our data . in part , this could be due to the poorer spatial resolution obtained with nifs+altair at short near - infrared wavelengths . as with the ] line , most of the ] emission occurs at velocities below @xmath114kms@xmath6 . the central diagrams of fig . [ f - pv2siii ] can be compared directly with the ] slit spectrum shown in fig . 2 of @xcite . the ] spectra are much less structured than the ] spectra , presumably due to poorer spatial resolution and consequent lower sensitivity to faint emission . nevertheless , the two emission lines show a similar overall appearance along the central slit with blueshifted emission near @xmath115kms@xmath6 extending from @xmath116 to @xmath1173 from the nucleus towards pa@xmath118 ( sw , bottom of fig.[f - pv2siii ] ) , emission at the systemic velocity near the nucleus , and some redshifted emission near @xmath119kms@xmath6 extending to @xmath120arcsec from the nucleus towards pa@xmath55 ( ne , top of fig.[f - pv2siii ] ) . the off - nucleus pv diagrams in fig . [ f - pv2siii ] have a similar overall structure , but show that the @xmath115kms@xmath6 component is present in all negative offsets up to the slit at @xmath1218 . of note is the fact that the velocity of this emission remains approximately constant over its full extent . the brightest @xmath122kms@xmath6 emission is only clearly apparent towards pa@xmath55 along the cone axis ( i.e. , in the upper part of the central panel of fig . [ f - pv2siii ] ) . some aditional emission from this component can be observed along the slits with positive offsets at fainter levels and at larger distances from the nucleus . the systemic velocity component is prominent in the upper section ( ne ) of pseudo - slits with negative offsets and in the lower sections ( sw ) of pseudo - slits having positive offsets , and originates in circumnuclear gas from the galaxy plane . [ f - pvfeii ] shows the corresponding pv diagrams for the ] 1.644 line . the emission in the central pseudo - slit corresponds closely to the brightest ] emission in fig . 2 of @xcite , but the high - velocity extensions of the ] clumps are much less prominent in our ] data . although the major velocity components apparent in the ] line are also apparent in ] , the higher spatial resolution achieved at this longer wavelength helps to reveal sharper structures than in fig . [ f - pv2siii ] . on the other hand , the structures seen in the ] emission can also in part be attributed to enhancements associated with the radio jet , as observed in the channel maps . the component near @xmath115kms@xmath6 is comprised of several clumps and extends to only @xmath1235 along the central pseudo - slit towards pa@xmath118 , but reaches @xmath124arcsec at large negative , and possibly also positive , slit offsets i.e. , in the extreme left and right panels of fig . [ f - pvfeii ] . clouds with velocities @xmath125kms@xmath6 are observed at @xmath126arcsec along the central slit and closer to the nucleus along the @xmath110 and @xmath127 offset slits . emission with velocities up to @xmath128kms@xmath6 is apparent at slit offsets of @xmath129 and @xmath130 . the systemic - velocity component is restricted to the nucleus in the central slit , but extends to nearly @xmath131arcsec along the @xmath130 to @xmath132 offset slits and to nearly @xmath133arcsec along the @xmath129 and @xmath134 offset slits . however , the systemic - velocity emission at negative slit offsets to the left in fig . [ f - pvfeii ] is dominated by a further bright clump near @xmath133arcsec approximately east of nucleus , in the vicinity of the receding radio lobe . overall the impression from the ] data is of roughly constant redshifted velocities towards pa@xmath55 and blueshifted velocities towards pa@xmath118 with an additional systemic - velocity component originating in the galaxy disc and associated with the radio jet . the role of the geometry and projection effects on the observations can best be evaluated via modelling . in this section , we compare our data with models with and without acceleration along the nlr . we begin by testing the model proposed by @xcite on the basis of the optical ] emission obtained with stis , namely of an inclined hollow cone with radial outflow along the cone walls that is characterised by linear acceleration followed by linear deceleration @xcite . our 3d kinematic data allow us to test this model against both channel maps and pv diagrams through multiple pseudo - slits in ways that were not available to previous authors . we begin by presenting both these views for the model parameters of @xcite in figs . [ f - daspv ] and [ f - daschm ] . as our data cover essentially the entire region of acceleration in the models , emission from decelerating gas beyond the velocity maximum has been suppressed in fig [ f - daschm ] . . ] the @xcite model predictions differ in several ways from the near - infrared emission - line data . first , the high velocity gas just to the sw and ne of the nucleus predicted by the models ( fig . [ f - daspv ] ) and seen in ] is less conspicuous in our data . it is seen in our data only in the log plots for ] , where the faint emission gets enhanced ( top panel of fig.[f - pv2siii ] ) . we also see little evidence of acceleration and deceleration in the near - infrared data , which mostly show sequences at constant velocities from the nucleus all the way out . even in the deep ] pv diagrams , gas up to @xmath135arcsec from the nucleus in the lower half of these diagrams extending to pa@xmath118 is still moving at blueshifted velocities of @xmath100kms@xmath6 , with similar but smaller redshifted velocities in the upper half of the diagrams . another significant difference between the model and the near - infrared pv data is that the model predicts that pseudo - slits offset from the nucleus will show no emission near the slit centre fig . [ f - daspv ] . this is the equatorial region away from the nucleus and between the two oppositely - directed cone `` lobes '' . most of the actual emission at off - nucleus slit positions arises in this region that is excluded by the model c.f . , fig . [ f - pvfeii ] . there is also emission at zero velocity mainly observed in the ] pv diagram which is not included in the model . finally , the model predicts that the channel maps should show ridges of emission perpendicular to the cone axis at high positive and negative velocities fig . [ f - daschm ] . this morphology arises because outflowing material in the hollow cone essentially emits from an annular region perpendicular to the cone axis convolved with other orientation factors . in contrast , the channel maps in figs . [ f - chmsiii ] and [ f - chmfeii ] show that the highest - velocity emission is confined to positions close to the cone axis , and is possibly extended along the cone axis towards the nucleus . we now investigate modifications to the hollow - cone , radial - outflow model that attempt to address the above differences . the simplest modification is to hypothesise that the cone wall that is steeply inclined to the plane of the sky does not emit strongly in near - infrared lines . this could occur because this cone wall is more inclined to the galactic disk . as discussed in sec . [ centroid ] , under the hypothesis that the cone walls are formed by gas entrained from the galaxy disk , this wall is likely to be comprised of more tenuous gas , as there is less gas to be entrained at high angles to the galactic disk . this modification removes the high - velocity components in fig.[f - daspv ] . to this , we then add an illustrative systemic - velocity component that is aligned with the radio jet and an arbitrary galactic disk component . the radio - jet component is modeled as a central nuclear component plus two clumps offset e - w from the nucleus by 1 . each clump is assumed to be at the systemic velocity of the galaxy . the galaxy component has a gaussian intensity profile and is assumed to be at the systemic velocity . the presence of this component is suggested both by the zero velocity component in the centroid velocity maps fig . [ f - doublevel ] and by the channel maps , which show strong emission at zero velocities ( figs . [ f - chmsiii ] and [ f - chmfeii ] ) . the resulting pv diagrams are shown in fig.[f - modpv ] . the pv diagrams in fig.[f - modpv ] do not provide a perfect fit to the data ( in particular to the highest spatial resolution ] data in fig.[f - pvfeii ] ) . however , they serve to illustrate the following points : firstly , the emission from the cone ( blue regions in fig.[f - modpv ] ) is a better fit to the data in the linear plots ( showing the strongest emitting gas ) if only the section of the cone wall closest to the plane of the sky , and hence closest to the galactic disk , is included . this may be due to interaction of the nlr gas in the cone wall with the galactic disk , possibly by the entrainment of disk material into the nlr outflow . given that this material is moving in a direction closer to the plane of the sky , we are less sensitive to its true velocity so can not make strong claims about the length scale on which it is accelerated . fig.[f - modpv ] is drawn for the gradual acceleration and subsequent deceleration proposed by @xcite . however , the data permit virtually any velocity law , including constant velocity outflow , on the length scales to which they are sensitive . the second point illustrated by fig.[f - modpv ] is that emission from gas interaction with the radio jet is significant . this is modeled as nuclear and off - nuclear clumps , but in reality other emitting gas forms a bridge between the main clumps and the nucleus . this is seen in the channel maps shown in fig.[f - chmfeii ] . finally , the galaxy component at systemic velocity reproduces the data at pa=60@xmath3 for pseudo - slits at negative offsets and at pa@xmath136for pseudo - slits at positive offsets , supporting the presence of a disc component , as concluded also from the centroid velocity maps of fig.[f - doublevel ] . simulated channel maps are shown in fig.[f - modchm ] . these only approximately reproduce the data presented in fig.[f - chmfeii ] . in particular , two separated emission clumps aligned perpendicular to the cone axis are predicted from the cone walls at high - redshifted and high - blueshifted velocities while only a single clump near the cone axis is actually seen in the data . .[f - modchm ] ] finally , we have considered a truncated hollow conical outflow at a constant velocity of @xmath137kms@xmath6 , as we could not see acceleration or deceleration in our data . as in the previous model , we add the radio jet and galactic disk component . the resulting position - velocity diagrams are shown in fig . [ f - modconstpv ] . the fit to the observed constant velocity sequence at @xmath100kms@xmath6 for ] ( fig.[f - pv2siii ] ) along the slit passing through the nucleus and the ones with negative offsets are improved by this model . the disc component at zero velocity reproduces emission observed at the nucleus and along pa@xmath138 ( positive position in the panels of fig . [ f - pv2siii ] ) and pa@xmath139 ( to the left in fig . [ f - pv2siii ] ) , as well as less extended emission along pa@xmath140 ( negative positions in fig . [ f - pv2siii ] ) and pa@xmath141 ( to the right in fig . [ f - pv2siii ] ) . the high velocity sequences predicted by the models although not conspicuous in the linear pv diagrams of ] ( bottom panel of fig.[f - pv2siii ] ) are observable in the log pv diagrams ( upper panel of fig.[f - pv2siii ] ) , due to its faint emission . these high velocity sequences and are due to the front wall of the approaching cone and far wall of the receeding cone . we point out that this short high velocity sequence as predicted by the constant velocity model seems to be adequate to reproduce also the high velocity gas observed in ] ( fig . 2 of @xcite ) . nevertheless , this model did not improve the fit for pseudo - slits offset by more than @xmath127 from the nucleus , which still do not show emission near the slit centre in the model . in addition , these pseudo - slits show velocity sequences that decrease with increasing distance from the nucleus . in order to improve the fit , new geometries should be considered . for example , a flow that is not conical , but in the form of a bowl as observed in planetary nebulae and in recent nifs observations of ngc1068 @xcite . simulated channel maps are shown in fig . [ f - modconstchm ] . these maps are an improvement relative to the previous model for velocity channels between @xmath142kms@xmath6and @xmath143kms@xmath6 , but show double structures for velocities between @xmath144kms@xmath6and @xmath145kms@xmath6 , which are not seen in the data . for the highest velocities , a small triangular structure is seen , while the data show only clumps near the cone axis . this discrepancy can also be attributed to possible deviations from a bi - conical geometry and to very weak emission from the wall of the bi - cone that is most inclined . ( see text ) . the orientation and velocity bins are as for fig.[f - pv2siii ] . [ f - modconstchm ] ] the modelled integrated emission - line map is similar for the three models discussed above . we show in fig . [ f : mod_int ] the map obtained for the modified @xcite model . a comparison with fig . [ f : feii_int ] shows that it provides a reasonable reproduction of the data , taking into account that one should add the contribution of gas from the galaxy plane to the modelled emission . @xcite identified blueshifted ] emission with velocities up to @xmath146kms@xmath6 in the ngc1068 nlr @xmath147pc ne of the nucleus . this is the region attributed to the highest radial outflow velocities in the gradual acceleration model . many of these knots are linked kinematically to more massive and slower - moving clumps located closer to the nucleus . @xcite identify these knots as ablation flows from disintegrating molecular clouds that are being photoionised and radiatively accelerated by the agn . @xcite detected faint , high - velocity ] emission with radial velocities up to @xmath148kms@xmath6 in stis slitless spectra of ngc4151 . the high - velocity emission was shown to arise in high - excitation ionised gas , and has velocity dispersions in excess of @xmath149kms@xmath6 @xcite , which is also consistent with an origin in an ablation flow . near - infrared ] emission arises in conditions similar to the optical ] emission , and can thus be formed in tenuous fully ionised gas experiencing a high local ionisation parameter . but then , why do we see less ] than ] emission ? either we lack the sensitivity to detect faint high - velocity emission , or this emission is not present . in order to investigate this , version 8.0 of the photoionisation code cloudy , last described by @xcite , was used to calculate the ratio of ] to ] for nlr clouds with a range of densities . the ionisation parameter was set to @xmath150 for a gas density of @xmath151@xmath152 and then varied inversely with the gas density to simulate clouds with different densities being illuminated by the same radiation field . such a situation would occur if the low - velocity ] is emitted at the front face of a dense nlr cloud and the high - velocity ] emission arises in an associated low - density ablation flow . the resulting emission - line ratios are plotted in fig . [ f - ratio ] where it can be seen that ] / ] increases from @xmath153 at densities typical of nlr clouds to @xmath154 at densities @xmath155 @xmath152 that might be typical of an ablation flow . the values at high densities are typical of the integrated line ratios observed in seyfert galaxies nlrs ( e.g. , * ? ? ? the strength of ] falls sharply to lower densities while the strength of ] is largely unchanged . the reason for this dependence on ionisation parameter is that sulphur exists predominantly as ] at typical nlr densities , but as ] and higher species in more highly excited low - density regions . in contrast , the higher ionisation potential of ] ( 54.9 ev compared to 35.1 ev for ] and 47.2 ev for ] ) means that ] survives better in highly excited regions . the lower intensity of ] in low - density irradiated regions makes it more difficult to detect a low - density ablation flow in ] emission than in ] emission . this combined with the lower strehl ratio and consequent poorer image quality achieved by adaptive optics systems at short near - infrared wavelengths probably account for our inability to detect more high - velocity gas in ngc4151 . there is little evidence in our near - infrared data for the gradual acceleration followed by gradual deceleration proposed by @xcite based on ] kinematics . no viable physical mechanism has been identified for such a gradual acceleration . a possible explanation is that the high - velocity ] emission does not represent the bulk motion of the outflow . rather , this could be low - density material ablated from nlr clouds that we observe outflowing at @xmath4250kms@xmath6 and are accelerated much closer to the agn . this is consistent with the low surface brightness of the high - velocity , low - density emission in ] and with the prominence of near - infrared emission formed on the sides of the outflow closer to high - density gas in the galactic disk . while the kinematics of the ] , , , and ] emitting gas is dominated by outflows along the bi - cone , the h@xmath20 radial velocity field in fig . [ f - chmh2 ] is dominated by velocities close to zero around the galaxy minor axis . although the velocity range is small , it can be observed that , from the blueshifted to the redshifted channels , the flux distribution moves around the minor axis of the galaxy , from the sw to the ne , as expected for gas rotating in the galaxy plane . although we see a kinematic component at systemic velocity also in the ionised gas , we do not see the outflowing component in h@xmath20 . the velocity dispersion in the h@xmath20 gas is also the smallest among the emission - line profiles . distinct kinematics for the h@xmath20 emitting gas relative to the emission lines of ionised gas has also been observed in previous studies of other objects @xcite . the h@xmath20 emission lines usually have small velocity dispersions and their kinematics seem to be dominated by rotation in the galaxy plane , while the other emission lines have larger velocity dispersions and other velocity components besides rotation , as observed also in ngc4151 . in paperi we have argued that the flux distributions , physical properties and location suggest that the h@xmath20 emission gas traces the agn feeding , while the ionised gas traces its feedback . although we do not see the h@xmath20 emitting gas actually flowing down into the nucleus , it does present a distinct kinematic signature relative to the ionised gas . the inflow may be occurring further in , in non - emitting colder molecular gas . the hot molecular gas is estimated to be a fraction of only @xmath156 to @xmath157 of the total mass of molecular gas @xcite in agn . in the case of ngc4151 , for which we obtained a mass of hot h@xmath20 of @xmath158m@xmath8 , we can thus estimate a total molecular gas mass of@xmath159m@xmath8 . such a large molecular gas reservoir around the nucleus of active galaxies is supported by radio observations of the central region of active galaxies in general , by , for example , the `` nuga '' group , which report molecular gas masses in the range @xmath160@xmath161 in the inner few hundred parsecs of active galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? if such a reservoir of molecular gas has accumulated by a flow that was the trigger of the present nuclear activity , it can be concluded that a steady flow would have to bring gas inwards at a rate in the range @xmath162@xmath163m@xmath164yr@xmath6 . in ngc4151 , @xcite have measured inflows in radio observations of that occur along narrow channels interpreted as shocks along the leading edges of the large scale bar ( at pa@xmath165 ) . these inflows seem to end at the inner part of the bar in the nuclear region . the orientation of the flow is approximately perpendicular to the `` arcs '' seen in the h@xmath20 flux distribution see fig . [ f - chmh2 ] and one possibility is that the radio observations are tracing the gas inflows that accumulate in the nuclear region and contribute to build up the molecular gas , part of which we observe due to excitation of the h@xmath20 molecule by x - rays emitted by the nucleus . the large amounts of molecular gas observed around the nuclei of active galaxies can only lead to the conclusion that inflows must be occurring @xcite . nevertheless , the actual kinematic observation of inflowing gas in active galaxies is difficult to obtain . this is due to the fact that the emission of the cold , inflowing gas is much weaker than the emission of the hotter , outflowing gas . in @xcite and @xcite we were successful in mapping the inflow in low - excitation ionised gas because the targets were low - activity agn the liners ngc1097 and ngc6951 for which the nuclear radiation and outflows are much weaker than in seyfert galaxies , thus allowing the observation of inflows in the disc in ionised gas . we are probably seeing just the ionised `` skin '' of a flow of cold gas on its way to feed the agn . we were also successful in mapping an inflow observed in hot molecular gas in the galaxy ngc4051 @xcite . in this latter case , the x - ray emission of the agn may have worked as a flashlight which `` illuminated '' the inflowing molecular gas . we obtained a mass inflow rate of @xmath166m@xmath164yr@xmath6 . if we use the same ratio of hot to total molecular gas as above , the inflow rate becomes of the order of a few m@xmath164yr@xmath6 . another recent case of inflow seen in molecular gas was reported by @xcite , using near - ir ifu observations of ngc1068 obtained with sinfoni . the feedback from the agn in the form of the mass outflow rate along the nlr and its kinetic power can be calculated using the ionised gas luminosity and geometry of the outflow through the expression : @xmath167 where @xmath168 is the mass of the proton , @xmath169 is the ionised gas density , @xmath170 is the outflow velocity , @xmath171 is the gas filling factor and @xmath172 is the cross section area of the outflow . the filling factor was estimated from : @xmath173 where @xmath174 is the br@xmath51 luminosity emitted by a volume @xmath175 and @xmath176erg@xmath152s@xmath6 for a temperature t=10000k and gas density @xmath177@xmath152 @xcite . as in but got as a result f@xmath1781 , which is unphysical , supporting a higher average gas density . ] the volume @xmath175 was calculated assuming the geometry proposed by @xcite and used in the modelling described in the previous section , namely a hollow bi - cone with a height of 130pc for each cone ( corresponding to 2 ) , and inner and outer opening angles of @xmath179 and @xmath180 , respectively . we obtained @xmath181@xmath152 . the br@xmath51 luminosity was obtained from paperi . we have further divided this by a factor of 2 , because part of the ionised gas emission comes from galactic disc gas that is not outflowing ( see previous section ) . we estimate from the channel maps that approximately half of the total ionised gas luminosity arises in this galactic disc gas . substituting these estimates in the expression above we obtain a filling factor value of @xmath182 . filling factor values reported in the literature are usually smaller , but they usually consider an isotropic ( spherical ) outflow . assuming an isotropic outflow instead of a bi - conical one , we obtain @xmath183 which is similar to the typical values in the literature . the cross section area @xmath172 of the hollow cone was calculated at a distance of 65pc ( 1 ) from the cone apex for the hollow cone geometry and resulted in @xmath184cm@xmath13 . adopting the velocity @xmath186kms@xmath6 for the outflow ( which is the one assumed for the constant velocity outflow model ) , @xmath187@xmath152 , @xmath182 we finally obtain a mass outflow rate of @xmath188m@xmath164yr@xmath6 in each cone . this value is of the order of the mass outflow rates reported for the nlr of active galaxies in neutral and ionised gas @xcite , and @xmath06 times larger than the mass outflow rate estimated by @xcite from uv absorbers in the spectra . the mass outflow rate can be compared with the accretion rate to feed the supermassive black hole . the accretion rate can be estimated from : @xmath189 where @xmath190 is the speed of light , and the bolometric luminosity is @xmath191ergs@xmath6 @xcite , which is 0.012 times the eddington luminosity for a black hole mass of @xmath192m@xmath8 @xcite . for this eddington ratio we adopt an efficiency factor for the conversion of rest mass energy into radiation of @xmath193 , a typical value for an optically thin geometrically thick accretion disc @xcite . the result is @xmath194yr@xmath6 , which is two orders of magnitude smaller than the mass outflow rate , indicating that the origin of the outflowing gas is not the agn but could be surrouding gas from the galaxy ism accelerated by a nuclear outflow , as discussed in the previous section . our data also allows us to estimate the kinetic power of the outflow , including both the radial and turbulent components of the gas motion . assuming that the velocity dispersion of the hii emitting gas is due to turbulence , the kinetic power can be estimated as : @xmath195 from fig . [ f - doublesig ] we obtain @xmath196kms@xmath6 , and adopting as above , @xmath197kms@xmath6and using @xmath198 ( considering the outflows in both cones ) , we obtain @xmath9ergs@xmath6 , which is @xmath199l@xmath200 . thus , from the power that can be extracted from the accretion rate of @xmath194yr@xmath6 , @xmath40.3% is transformed in kinetic power in the nlr outflows . we note that this fraction is of the order of the agn feedback derived in simulations for co - evolution of black holes and galaxies ( 0.5% ) , in order to match the m-@xmath44 relation @xcite . this kinetic power is somewhat larger than those we have recently derived for a sample of 6 nearby seyfert galaxies showing nlr outflows @xcite , which are in the range @xmath201 . in order to gauge the impact that the above kinetic power may have in the galaxy , we estimate the total kinetic energy released if the outflow as now observed lasts for at least @xmath160yrs ( as an estimated duration for the activity phase of the galaxy ) to be @xmath202erg . this energy can then be equated to the binding energy of the gas mass that could be pushed away from the central region of the galaxy ( following the reasoning in @xcite ) @xmath203 . for @xmath204erg , @xmath205m@xmath8 ( estimated from the galaxy absolute magnitude ) and @xmath206pc ( just beyond the observed outflow ) , we obtain @xmath207m@xmath8 , which is @xmath4 5 times the mass which outflows through the nlr in 10@xmath208yrs . it is interesting to note that this mass is of the order of the estimated molecular gas mass present in the nuclear region of ngc4151 , as discussed above , being also the typical value observed in the nuclear region of active galaxies in general @xcite . thus , it appears that in the case of ngc4151 the molecular gas reservoir that is needed to trigger the nuclear activity can be blown away by the nlr outflow by the end of the activity cycle . we have measured the kinematics of the ionised and molecular gas surrounding the nucleus of ngc4151 using integral - field spectroscopy obtained with the nifs instrument at the gemini north telescope . the main results of this paper are : antonucci , r. r. & miller , j. s. , 1985 , apj , 297 , 621

search for the statistical anisotropy in the cmb data is a powerful tool for constraining models of the early universe . in this paper we focus on the recently proposed cosmological scenario with conformal rolling . we consider two sub - scenarios , one of which involves a long intermediate stage between conformal rolling and conventional hot epoch . primordial scalar perturbations generated within these sub - scenarios have different direction - dependent power spectra , both characterized by a single parameter . we search for the signatures of this anisotropy in the seven - year wmap data using quadratic maximum likelihood method , first applied for similar purposes by hanson and lewis . we confirm the large quadrupole anisotropy detected in and bands , which has been argued to originate from systematic effects rather than from cosmology . we construct an estimator for the parameter . in the case of the sub - scenario with the intermediate stage we set an upper limit at the 95% confidence level . the constraint on is much weaker in the case of another sub - scenario , where the intermediate stage is absent .

1202.4357

recent advances in the observational cosmology make it real to start probing the most intriguing aspects of the universe . in particular , it is of importance to inquire whether the statistical isotropy of the scalar perturbations is exact or only approximate . this issue is of special interest because the statistical isotropy is one of the key assumptions of the six - parametric @xmath4cdm model and is favored by inflation . thus , the violation of this property in the observed cmb would imply a highly non - trivial extension of the now standard cosmological model . an additional motivation to search for the statistical anisotropy is the possible presence of various anomalies in the cmb data , such as alignment of low multipoles , axis of evil , power assymetries , cold spots and others @xcite@xcite . in the statistically anisotropic but spatially homogeneous universe , the power spectrum of the primordial scalar perturbations @xmath5 depends on the direction of the wave vector @xmath6 . the power spectrum can then be written as follows , @xmath7 \ ; , \ ] ] where @xmath8 . the coefficients @xmath9 parametrize the direction - dependent part , which one expands in spherical harmonics @xmath10 . unlike in ref . @xcite , we assume here that the dependence on the wavenumber @xmath11 may be absorbed into one function @xmath12 . commutativity of the classical field @xmath13 yields @xmath14 and hence @xmath15 for odd @xmath16 . the generic prediction of the inflationary theory is that the power spectrum is isotropic , @xmath17 . however , the statistical anisotropy can be generated in models of inflation involving vector fields @xcite or scalar fields with non - minimal kinetic terms @xcite ; for reviews see also refs . @xcite and @xcite . somewhat more exotic examples are given by introducing bianchi i geometry @xcite or noncommutative field theory @xcite . the most common feature of these models is the statistical anisotropy of a special quadrupole type ( the only non - vanishing coefficient in a certain reference frame on the celestial sphere is @xmath18 ) . this prediction arises , e.g. , in the model with the rotational invariance broken by a space - like vector @xcite , or in the hybrid inflation incorporating a vector field coupled to the waterfall scalar @xcite . however , higher multipoles @xmath9 can also emerge within the inflationary framework , see , e.g. , ref . @xcite . in inflationary theory , the observed approximate flatness of the scalar power spectrum is due to the approximate de sitter symmetry of inflating space - time . it has been suggested some time ago @xcite that , alternatively , the flatness of the power spectrum may be a consequence of conformal symmetry . concrete models of this sort have been proposed in refs . @xcite and further developed in refs . it is this class of models that we focus on in this paper . in models of refs . @xcite and similar ones , it is assumed that the cosmological evolution starts from or passes through a conformally invariant state with effectively flat geometry . this state is unstable , and conformal symmetry @xmath19 gets broken down to @xmath20 by a time - dependent ( rolling ) scalar field . during this conformal rolling stage , another field of zero conformal weight develops perturbations which automatically have flat power spectrum . these perturbations are reprocessed into adiabatic perturbations by one or another mechanism ( e.g. , of refs . @xcite ) at some later epoch , after the end of conformal rolling . it is then natural that the adiabatic perturbations inherit the properties of the original field perturbations , modulo possible additional non - gaussianity . at the conformal rolling stage , the properties of perturbations to both linear and leading non - linear orders are uniquely determined by the underlying conformal symmetry @xcite , modulo the overall amplitude and a single dimensionless parameter which we call @xmath0 ( in the model of ref . @xcite , the amplitude is also determined by @xmath0 , see section [ sec : fields ] ) . this parameter governs the non - gaussianity and statistical anisotropy . the statistical anisotropy generated in the conformal rolling scenario is quite different from the predictions of inflation . in particular , the coefficients @xmath9 parametrizing the power spectrum are the random variables rather than fixed parameters of the model . there are two sub - scenarios of the conformal rolling scenario which differ by the behaviour of the cosmologically interesting modes after the end of the conformal rolling stage . one possibility is that these modes are already superhorizon by the end of conformal rolling and remain frozen until the late hot epoch . the second sub - scenario assumes that there is an intermediate stage at which the field perturbations evolve in a non - trivial way before crossing out the cosmological horizon and getting finally frozen out . in the latter case , the evolution of the field perturbations results in the statistical anisotropy of all even multipoles . notably , it does not depend on the magnitude of the momentum @xmath6 , i.e. , @xmath12 is independent of @xmath11 @xcite . the predictions of the former sub - scenario for the statistical anisotropy are considerably different . to the linear order in the parameter @xmath21 , one obtains the statistical anisotropy of the general quadrupole type . it is characterized by the amplitude which decreases with the wavenumber @xmath11 , i.e. @xmath22 @xcite . this means that the corresponding effects in the cmb sky are suppressed at high cmb multipoles @xmath23 . the statistical anisotropy of the special quadrupole type also appears in the next - to - linear order . though the effect is suppressed by the additional power of the parameter @xmath21 , now the amplitude @xmath12 does not depend on the wavenumber @xmath11 hence , the subleading contribution may make stronger imprint on the cmb . signatures of the statistical anisotropy in the cmb temperature maps have been searched for in refs . @xcite@xcite . motivated by the model of ref . @xcite , groeneboom and eriksen @xcite discovered the evidence for the quadrupole statistical anisotropy in the five - year wmap data . however , it was found to be nearly aligned with the ecliptic poles . using quadratic maximum likelihood estimator , hanson and lewis @xcite extended the analysis to higher multipoles . they also included the relevant prefactor in the covariance neglected in refs . @xcite and @xcite . hanson and lewis @xcite confirmed the result on the large quadrupole @xmath24 lying nearly in the ecliptic plane . the strongest indication of the statistical isotropy violation , non - zero at the @xmath25 confidence level , was found in the @xmath2 band of the five - year wmap data in ref . these findings have been confirmed by the wmap team @xcite in their analysis of the seven - year data . one possible explanation of the anomalous quadrupole is the systematics inherent in the wmap data . as argued in ref . @xcite , large observed statistical anisotropy may result from beam asymmetries rather than have the cosmological origin . the purpose of this paper is to constrain the parameter @xmath0 of the conformal rolling scenario from the non - observation of the cosmological statistical anisotropy in the seven - year wmap data . we follow the general method proposed by hirata and seljak for the purpose of studying cmb lensing and known as the quadratic maximum likelihood ( qml ) estimation @xcite . as discussed in ref . @xcite , the same idea can be applied to the study of the statistically anisotropic properties of cmb . in this case one assumes that the coefficients @xmath9 are small and expands the log - likelihood of the observed cmb to the second order in these parameters . by maximizing the log - likelihood with respect to the coefficients @xmath9 , one obtains the estimator . results derived within the qml approximation are in a good agreement with the exact likelihood methods . we apply this method to construct the estimator for the parameter @xmath0 . in view of the results quoted , the estimated values are expected to be inconsistent with the statistical isotropy because of the alleged systematics present in the wmap data . assuming that the interpretation in terms of systematics is correct , we set the upper limits on the parameter @xmath0 in the following way . for each value of @xmath0 , we simulate the parameter sets @xmath26 , and then generate a number of anisotropic maps for each set @xmath26 . from the maps generated , we estimate the values of the parameter @xmath0 . we require that in 95% cases they should not exceed the value estimated from the observed cmb . in this way we constrain the conformal rolling sub - scenario with the intermediate stage , @xmath27 at the 95% confidence level . the constraint is much weaker in the framework of the alternative sub - scenario . the reason is that the amplitude of the leading order quadrupole decreases as @xmath28 . this translates into the suppression of the statistical anisotropy effects at high cmb multipoles . thus , the data useful for the analysis are effectively limited , statistical errors are large and the constraint is @xmath29 at the 95% confidence level . the constraint is improved significantly , once we take into account the subleading contribution to the statistical anisotropy . this contribution is of the special quadrupole type , and has the amplitude @xmath12 , which is independent of the wavenumber @xmath11 . thus , the number of cmb multipoles useful in the analysis is much larger . this somewhat compensates the smallness of the constant @xmath21 , and we obtain the stronger constraint , @xmath30 at the 95% confidence level . here @xmath31 is the present value of the hubble parameter , which plays the role of the ultraviolet cutoff , and @xmath4 is the infrared cutoff . without going into speculation on the value of the constant @xmath4 , we point out that this constraint is still very weak , in view of the fact that the conformal rolling scenario is self - consistent only at @xmath32 anyway . we conclude that the statistical anisotropy is the relevant signature of the conformal rolling with the intermediate stage . it is of particular interest in view of the upcoming planck data . hopefully , the latter will be free of the quadrupole anomaly . the other expected advantage of the planck data is the larger range of the cmb multipoles , which translates into smaller statistical errors . these two factors are expected to improve the sensitivity of the data to the parameter @xmath0 by more than an order of magnitude . on the other hand , statistical anisotropy appears to be a weak signature of the alternative sub - scenario , and the planck data are not expected to improve the situation significantly . thus , it makes sense to focus on the other prediction of this sub - scenario , the non - gaussianity @xcite . at the level of bispectrum , the non - gaussianities in the conformal scenario are not particularly special . the shape of the intrinsic bispectrum is dictated @xcite by the symmetry breaking pattern @xmath33 and coincides with the bispectrum of a spectator massless scalar field in inflationary theory @xcite ( in fact , the intrinsic bispectrum may vanish for symmetry reasons , see section [ sec : fields ] ) . the non - gaussianity generated at the conversion epoch is not specific to the conformal scenario either . so , bispectrum alone can not discriminate between the conformal scenario and , say , inflation equipped with the curvaton mechanism . on the other hand , the non - gaussianity of a rather peculiar form arises in the trispectrum . existing constraints ( see , e.g. ref . @xcite ) are model - dependent and can not be directly applied to our model . we leave for the future the analysis of the cmb data aiming at the search for the non - gaussianity characteristic of the conformal scenario . this paper is organized as follows . in section [ sec : fields ] we review the properties of the field perturbations at the conformal rolling stage . the power spectrum of relevant perturbations generated by the end of their evolution posseses directional dependence . we show this explicitly in section [ statan ] . we consider two sub - scenarios : one in which the cosmologically interesting modes are superhorizon by the end of the conformal rolling stage ( section [ sec : scenarioa ] ) and another , with long intermediate stage ( section [ sec : scenariob ] ) . in section [ sec : estimator ] we review the main ideas of the qml method and construct model - independent estimators for the coefficients @xmath9 . we also construct an estimator for the parameter @xmath0 . section [ sec : results ] contains our main results . we implement the estimators to the seven - year wmap data and constrain the parameter @xmath0 of the two sub - scenarios . the main ingredient of the conformal rolling scenario @xcite is the conformal stage preceding the conventional hot epoch . as we point out below , the properties of the conformal rolling stage are quite general @xcite , as they are unambiguously determined by conformal symmetry . yet it is instructive to review a simple explicit model @xcite illustrating this scenario . it involves a complex scalar field @xmath34 , conformally coupled to gravity , which rolls down the negative quartic potential @xmath35 , where @xmath21 is a small parameter . at large field values , the potential is assumed to change and have a minimum at @xmath36 where conformal symmetry is explicitly broken . the field @xmath34 is a spectator at the conformal rolling stage and somewhat later , i.e. , the evolution of the universe is dominated by some other matter ( see ref . @xcite for an alternative version where @xmath34 is the dominant matter component ) . the central object in the model is the phase @xmath37 defined by @xmath38 . its perturbations @xmath39 start off as vacuum fluctuations and eventually freeze out . by the end of conformal rolling , perturbations @xmath39 have flat power spectrum . once the radial field @xmath40 settles down to @xmath41 and conformal symmetry gets broken , what remains are the perturbations of the phase , which at this point are isocurvature perturbations . they get reprocessed into adiabatic perturbations at much later epoch . in this way the flat power spectrum of the adiabatic perturbations is obtained . it can be slightly tilted if there is small explicit breaking of conformal symmetry at the conformal rolling stage @xcite . let us discuss conformal rolling in more detail . at this stage , the theory is described by the action @xmath42 where @xmath43 is the action for gravity and some matter that dominates the evolution of the universe , and @xmath44\ ] ] is the action for the complex scalar field . assuming that the background metric is homogeneous , isotropic and spatially flat , one introduces the field @xmath45 , where @xmath46 is the scale factor , and obtains its action in conformal coordinates in the minkowskian form , @xmath47 = \int d^3 x d\eta \left(\eta^{\mu \nu}\partial_{\mu } \chi^{\star } \partial_{\nu } \chi + h^2 |\chi|^4\right ) \ ; .\ ] ] here @xmath48 is the minkowski metric and @xmath49 is the conformal time . since the scalar potential is negative , the conformally invariant state @xmath50 is unstable , and the field rolls down its potential . one assumes that the background @xmath51 is spatially homogeneous and without loss of generality chooses the background solution real . it is given by @xmath52 where @xmath53 is a free parameter , which is interpreted as the end - of - roll time . let us consider perturbations in this background . to the leading order in @xmath21 , perturbations @xmath54 and @xmath55 decouple from each other . we begin with the radial perturbations @xmath56 . they obey the linearized field equation , in momentum representation , @xmath57 the properly normalized solution is @xmath58 b_{{\bf p}}+h.c.\ ; , \ ] ] where @xmath59 and @xmath60 are annihilation and creation operators obeying the canonical commutational relations ; @xmath61 is the hankel function . at late times the solution approaches the asymptotics @xmath62 the behaviour @xmath63 is interpreted as a shift of the end - of - roll time @xmath53 , which now becomes a random field . indeed , with perturbations included , the radial field @xmath64 can be written at late times as follows @xmath65 } \ ; , \ ] ] where @xmath66 as it stands , the expression ( [ feb15 - 1 ] ) involves all powers of the small parameter @xmath21 , i.e. , it appears to imply some resummation . however , we will be primarily interested in the first non - trivial order in @xmath21 , so we understand eq . ( [ feb15 - 1 ] ) merely as a convenient book - keeping tool : to the first non - trivial order in @xmath21 the right hand side of eq . ( [ feb15 - 1 ] ) is equivalent to @xmath67 . it is natural to assume that the fields are initially in their vacuum state . then the field @xmath68 is a time - independent gaussian field with red power spectrum . clearly , the overall spatially homogeneous shift of the end - of - roll time is irrelevant , since it can be absorbed into the redefinition of @xmath53 . what is important is the gradient of @xmath69 , @xmath70 it has flat power spectrum , while the higher derivatives have blue spectra . let us turn to the object of primary interest , namely , the perturbations of the phase @xmath37 or , equivalently , perturbations @xmath71 of the imaginary part of the field . we are interested in the leading and subleading orders in the small parameter @xmath21 , so we take the ( real ) background in the form ( [ feb15 - 1 ] ) . the linearized field equation for @xmath71 reads @xmath72 ^ 2 } \delta \chi_2 = 0\ ; .\ ] ] at early times , when @xmath73 , we get back to the minkowskian massless equation . the solution to eq . has the following form , @xmath74 where @xmath75 tends to @xmath76 as @xmath77 and @xmath78 , @xmath79 is another set of annihilation and creation operators . modulo corrections proportional to @xmath80 and @xmath81 , the solution with this initial condition is @xcite @xmath82 } \ ; , \ ] ] where @xmath83 and @xmath84 is given by eq . . at late times one has @xmath85^{-1}$ ] . as a result , the phase perturbations @xmath86 freeze out . in this regime , the phase perturbations , including corrections of order @xmath87 and @xmath81 , have the following form @xcite : @xmath88 where @xmath89 is the standard lorentz factor ; @xmath90 and @xmath91 denote the momenta in the direction of the `` velocity '' @xmath84 and in the orthogonal direction , respectively . these are the phase perturbations by the end of conformal rolling . as promised , they do not depend on the conformal time @xmath49 , @xmath92 to the leading order in @xmath21 , the parameter @xmath93 is independent of @xmath94 , so that @xmath95 , @xmath96 and the phase perturbations are gaussian random field with the flat power spectrum and amplitude of order @xmath21 . to the subleading order , the expression involves another time - independent gaussian field @xmath97 . this leads to both non - gaussianity and statistical anisotropy of @xmath98 and hence @xmath99 . note that the symmetry @xmath100 ensures that the intrinsic bispectrum vanishes in the concrete model we review . an important remark is in order . even though we illustrated the conformal rolling mechanism by making use of the concrete model , the results are characteristic of the entire class of conformal models . as an example , the above formulas are valid @xcite , modulo field redefinition , in the galilean genesis model @xcite based on conformal galileon field with higher derivative action @xcite . in fact , these formulas hold @xcite , provided that the theory has the following general properties at the conformal rolling stage : ( i ) space - time is effectively minkowskian ; ( ii ) conformal symmetry @xmath19 is spontaneously broken down to @xmath20 by a homogeneously rolling scalar field of non - zero conformal weight ( @xmath101 in the above example ) ; ( iii ) there is another scalar field of zero conformal weight ( the phase field @xmath37 above ) whose perturbations are in the end converted into the adiabatic perturbations . the only qualification is that the latter field need not be compact , so the amplitude of its canonically normalized perturbations can be arbitrary , whereas the overall amplitude of @xmath39 in the model is proportional to the same parameter @xmath21 that determines the amplitude of the perturbations @xmath97 and hence non - linear terms in @xmath39 , see eqs . and . to avoid confusion , we note that the model ( [ feb15 - 2 ] ) is different from the galilean genesis model @xcite and the pseudo - conformal model of ref . @xcite in that in the latter models , the cosmological background itself is driven by the rolling scalar field ( @xmath101 in our notations ) . therefore , the perturbation of this field ( mode @xmath56 ) is gauge - dependent , while the cuvature perturbation @xmath102 has blue power spectrum . this , however , does not invalidate the discussion in the previous paragraph . indeed , the time shift ( analogous to our @xmath97 ) can be given a gauge - invariant definition @xcite ; this field is related to , but different from @xmath102 . furthermore , at early times , when the conformal mechanism operates in models of refs . @xcite , the energy density and pressure of the rolling field are small , so the effects due to gravity are negligible . this is particularly clear in the newtonian gauge @xcite , where the metric perturbations are small at early times , while the perturbations in the rolling field have the form equivalent to eq . . in this regime , the curvature perturbation @xmath102 is irrelevant , as long as one is interested in the perturbations of the field similar to our phase @xmath37 , and the formulas of this section remain valid . this has been shown expicitly @xcite in the pseudo - conformal model of ref . @xcite , both in the newtonian gauge and in the gauge @xmath103 . we continue to use the terminology borrowed from the model . at the conformal rolling stage , the phase perturbations @xmath39 get frozen out because of the fast evolving background . once the conformal rolling stage ends and the radial field settles down to a constant value @xmath41 , the phase becomes a scalar field minimally coupled to gravity . the behavior of its perturbations at later times depends on whether they are super- or subhorizon in the conventional sense by the end of conformal rolling . accordingly , there are two sub - scenarios which give different predictions for the statistical anisotropy ( and non - gaussinaity as well ) @xcite . in this sub - scenario , the phase perturbations are already superhorizon by the end of conformal rolling . so , they remain frozen out , and there is a simple relation between the adiabatic and phase perturbations , @xmath104 plus possible non - linear terms , where @xmath98 are the phase perturbations late at the conformal rolling stage . so , the properties of @xmath102 can be read off from eq . , modulo possible non - gaussianity generated when @xmath39 is converted into @xmath102 . the interaction of the phase perturbations with the radial ones at the conformal rolling stage leads to non - trivial effects in the spectrum of the primordial perturbations . in particular , it gives rise to the statistical anisotropy . indeed , let us consider the two - point product @xmath105 and average it over the realizations of the operators @xmath78 and @xmath79 . to the leading order , we obtain the flat and isotropic power spectrum . the directional dependence appears once we take into account corrections coming from the derivatives of the end - of - roll time @xmath106 and keep only those modes of @xmath107 which are still superhorizon today ( shorter modes of @xmath107 give rise to the non - gaussianity rather than statistical anisotropy @xcite ) . for so long modes of @xmath107 , it does not make sense to average over the realizations of the operators @xmath59 , @xmath108 at this stage . in this way one obtains the power spectrum of the primordial perturbations @xmath5 @xcite : @xmath109 the directional dependence is encoded in the functions @xmath110 and @xmath111 , which originate from the corrections to the linear and next - to - linear orders in the parameter @xmath21 , respectively , @xmath112 @xmath113 where @xmath114 . first , let us consider the leading order contribution @xmath110 . we expand it in spherical harmonics , @xmath115 where @xmath116 by comparing with , one concludes that the anisotropic coefficients @xmath9 are gaussian variables , since they are linearly related to the derivatives of the end - of - roll time @xmath69 , which is the gaussian field . we keep very long modes of @xmath107 with @xmath117 , where @xmath31 is the present value of the hubble parameter . at shorter wavelengths the field @xmath68 gets averaged out . therefore , the expression in parenthesis in should be treated as a constant tensor throughout our part of the universe ; retaining its dependence on @xmath94 would result in effects suppressed by @xmath118 . for this reason , only the quadrupole of the general type survives in eq . . neither its direction nor precise magnitude can be predicted because of the cosmic variance . yet its variance in the ensemble of universes like ours is calculable . one makes use of eq . and finds @xmath119 for similar reason , the second contribution @xmath120 also represents the quadrupole statistical anisotropy , but of the special type . it can be expanded in the same fashion as in . as compared to the previous case , the amplitude @xmath12 is independent of the wavenumber @xmath11 . we will see that this fact is crucial from the viewpoint of the cmb observations . the other important distinction is that the quantities @xmath24 are not gaussian now . therefore , it will be convenient to work with the components of the `` velocity '' @xmath84 , which are gaussian variables with zero means and variances @xmath121 here the present value of the hubble parameter and the constant @xmath4 appear as the ultraviolet and infrared cutoffs , respectively . the quantities @xmath24 are then given by @xmath122 where @xmath123 is the unit vector in the direction of the `` velocity '' @xmath84 . equations , , and are the starting point of our analysis of the statistical anisotropy in the cmb within the sub - scenario a. let us now turn to the second sub - scenario , in which the phase perturbations are still subhorizon at the end of the conformal rolling stage and exit the cosmological horizon later @xcite . this case is rather non - trivial . first , we need to make certain assumptions about the intermediate stage following conformal rolling . barring fine - tuning , we consider the intermediate stage very long as compared to the interesting cosmological scales , namely , @xmath124 where @xmath125 is the end - time of the intermediate stage , at which the modes @xmath39 exit the cosmological horizon . in order not to modify the flat power spectrum generated by the end of conformal rolling , one assumes that the dynamics of the phase perturbations @xmath39 is nearly minkowskian at the intermediate stage , so that they obey the field equation @xmath126 the latter condition appears very restrictive . however , there are at least two cosmological scenarios where it is obeyed . one is the bouncing universe with the super - stiff equation of state , @xmath127 @xcite ; interestingly , cosmological contraction with stiff equation of state is inherent also in the pseudo - conformal universe model of ref . another is genesis @xcite , where the universe stays static for a very long period of time before it starts to expand rapidly . under the above assumptions , the solution to eq . supplemented with the initial conditions and is given by @xcite @xmath128 where @xmath129 is the sum of two incoherent waves coming from the direction @xmath130 and from the opposite direction , @xmath131 + e^{i\psi_{- } } \left(1-\hat{{\bf k } } { \bf v}^{(-\hat{{\bf k } } ) } \right ) \right ) \ ; . \label{feb15 - 13}\ ] ] here @xmath132 upper labels @xmath133 and @xmath134 denote quanities calculated at the points @xmath135 and @xmath136 , respectively . note that eq . is valid at the horizon exit , so it determines , in fact , the properties of the adiabatic perturbations ( again modulo possible additional non - gaussianity ) . note also that under the assumption , the modes of @xmath97 relevant in eq . are indeed longer than the size of the visible universe . proceeding as in section [ sec : scenarioa ] , we arrive at the power spectrum , but now with @xmath137 in this case all coefficients @xmath9 are generically different from zero . their variances in an ensemble of universes are calculated by making use of eq . and are given by @xcite @xmath138 notably , the function @xmath12 does not depend on the momentum , i.e. @xmath139 equations and determine the statistical anisotropy of the primordial adiabatic perturbations in the sub - scenario b in terms of a single unknown parameter @xmath0 . let us first apply the quadratic maximum likelihood ( qml ) method to construct the estimators for the coefficients @xmath140 . here we define the latter in a model - independent way , assuming only that the dependence on the wavenumber @xmath11 can be factorized as in eq . . we closely follow the technique developed in ref . @xcite . in section [ sec : parameter ] we use the same ideas when constructing the estimator for the parameter @xmath0 . in what follows we use the harmonic representation for the temperature fluctuations and their covariances unless the opposite stated . the log - likelihood of the observed cmb temperature map @xmath141 is given by @xmath142 where @xmath143 is the vector of coefficients @xmath9 ; the covariance matrix @xmath144 incorporates the theoretical covariance corresponding to the signal as well as the instrumental noise , @xmath145 . the theoretical covariance is given by @xmath146 here @xmath147 are the theoretical temperature fluctuations of the cmb sky @xmath148 in the harmonic representation , @xmath149 and @xmath150 is the power spectrum of the primordial perturbations ; @xmath151 are transfer functions . under the assumption of the statistical anisotropy , we write the theoretical covariance as follows , @xmath152 where the leading contribution @xmath153 comes from the isotropic signal well fitted by the @xmath4cdm model ; the effects of the statistical isotropy violation are incorporated into @xmath154 . the matrix @xmath153 is diagonal in the harmonic representation , @xmath155 where @xmath156 is the standard cmb angular spectrum . the matrix @xmath154 is given by @xmath157 where @xmath158 the integral of three spherical harmonics reads @xmath159 where @xmath160 are the clebsch - gordan coefficients and @xmath161 normally , the estimators for the coefficients @xmath9 are determined by equating the derivative of the log - likelihood to zero , @xmath162 however , the covariance matrix @xmath144 is not sparse and direct calculations are too costly . thus , we need an appropriate approximation to work with . at this point we make use of the qml approach . assuming that the statistical anisotropy is weak , we expand the log - likelihood derivative to the linear order in @xmath143 , @xmath163 we replace the second derivative of the log - likelihood in this expansion by its expectation value @xcite , @xmath164 where @xmath165 is the fisher matrix . the first equality in eq . follows from the normalization condition @xmath166 in what follows we use the derivatives of the log - likelihood calculated under the assumption of the statistical isotropy unless the opposite stated , and omit the subscript `` 0 '' . the first derivative of the log - likelihood is then given by @xmath167 where @xmath168 is the statistically isotropic covariance incorporating the noise , @xmath169 . the second term in the right hand side of this equation is , in fact , the average of the first term over the realizations of cmb . this follows from the identity @xmath170 , where @xmath171 is any matrix and @xmath94 is a vector of gaussian random variables with the covariance @xmath144 . thus , one writes @xmath172 where @xmath173 and the quantities @xmath174 are the inverse - variance filtered cmb harmonics calculated in the absence of the statistical anisotropy , @xmath175 by substituting eqs . and into eq . and equating the result to zero , we obtain the qml estimator , @xmath176 in what follows we use the fisher matrix calculated in the full sky and homogeneous noise approximation @xcite . it has only diagonal elements , which do not depend on @xmath177 , @xmath178 where @xmath179 is the sum of the standard cmb spectrum @xmath156 and the noise spectrum @xmath180 . the constant @xmath181 denotes the uncut fraction of the sky . we include this factor to achieve better agreement between the approximate fisher matrix and the exact one defined as the average over the ensemble of simulated maps with the real sky coverage and inhomogeneous noise . in the framework of the conformal rolling scenario , the coefficients @xmath9 are random variables with zero expectation values . to the linear order in @xmath21 they are gaussian and have variances given by eq . or . the variances depend on the constant @xmath0 , which is the only parameter of the model . this makes it possible to constrain the conformal rolling scenario from the non - observation of the cosmological statistical anisotropy . we again use the qml method to construct the estimator for the parameter @xmath0 . we do that starting from the gaussian hypothesis about the coefficients @xmath9 . this hypothesis is particularly appropriate in the context of the sub - scenario b. the non - gaussian @xmath24 s appear in the sub - scenario a to the subleading order ; we comment on this case in the end of this section . the probability density of the coefficients @xmath9 for a given value of @xmath0 is @xmath182 here the matrix @xmath183 is the covariance of the anisotropy parameters , @xmath184 . to obtain the likelihood of the observed cmb with respect to the parameter @xmath0 , one integrates the product of two probability densities over the set of the parameters @xmath9 , @xmath185 where @xmath186 $ ] and @xmath187 is the log - likelihood introduced in eq . . following the main idea of the qml estimation , we expand the log - likelihood to the second order in @xmath143 , @xmath188 where we again replaced the second derivative by its expectation value over the cmb isotropic realizations . now the integral in eq . takes a simple gaussian form and can be straightforwardly evaluated , @xmath189 maximizing with respect to the parameter @xmath0 , @xmath190 we obtain the equation for the estimator of @xmath0 , @xmath191 in the full sky and homogeneous noise approximation , the fisher matrix is diagonal , @xmath192 the matrix @xmath193 has the same property , @xmath194 where we introduce the quantities @xmath195 which do not depend on the parameter @xmath0 . then the equation determining the estimator takes the form @xmath196 where we use the same notation @xmath0 for the estimator as for the parameter of the model . the quantities @xmath197 entering eq . are given by @xmath198 where the coefficients @xmath9 are defined by eq . . note that it follows from eq . that if the predicted statistical anisotropy is of the quadrupole form , i.e. , with non - zero @xmath24 s only , then the parameter @xmath0 can be estimated simply as @xmath199 , modulo obvious additive and multiplicative constants . this is also clear on general grounds . indeed , the rotational invariance requires that the estimator should be some function of @xmath200 , i.e. @xmath201 . in the small statistical anisotropy approximation , we keep only linear terms in the taylor expansion of the function @xmath202 . this immediately implies the quoted relationship between @xmath0 and @xmath200 . since no assumptions about the properties of the random quantities @xmath24 have been used in the latter argument , it holds for non - gaussian @xmath24 s , which describe the quadrupole of the special type , see eq . . the only qualification is that the statistical anisotropy is of the order @xmath203 in that case . hence , the corresponding estimator reads @xmath204 , up to multiplicative and additive constants . we search for the statistical anisotropy using wmap seven - year maps @xcite . we study the @xmath1 and @xmath2 band data at 61 and 94 ghz . the first step is to implement inverse - variance filtering defined by eq . . we write that equation in the form appropriate for applying the conjugate gradient technique , @xmath205{\bf s}^i\bar{{\bf \theta}}= \tilde{{\bf y}}^{\dagger } { \bf n}^{-1 } \hat{{\bf \theta } } \ ; .\ ] ] here we use the pixel representation for the noise covariance @xmath206 and the observed cmb temperature @xmath141 . the matrix @xmath207 relates the harmonic space covariance and the observed map , @xmath208 where @xmath209 are the beam transfer functions and @xmath210 labels pixels . we use the foreground reduced @xmath1 and @xmath2 seven - year maps @xcite provided in healpix format @xcite with @xmath211 . for the beam transfer function we use the average of @xmath212 and @xmath213 functions for @xmath1 band and the average of @xmath214 , @xmath215 , @xmath216 and @xmath217 for @xmath2 band . we consider the noise of the pixels uncorrelated and having the variance @xmath218 , where @xmath219 is @xmath220 mk and @xmath221 mk for @xmath1 and @xmath2 bands , respectively , and @xmath222 is specific to each pixel and tabulated in the maps . to remove foreground contaminated pixels we use the wmap temperature analysis mask which leaves us with @xmath223 of the sky . we take the noise covariance to be infinite ( inverse noise is zero ) for masked pixels . the noise model @xmath224 is constructed using noise covariance and template maps for removing monopole and dipole contributions . to evaluate the confidence intervals , inverse filtering should be performed on both data and large number of simulated maps . thus , the system must be well preconditioned . following ref . @xcite , we make use of the multigrid preconditioner , first proposed by smith et . al . in ref . it is known to be the fastest to date and has a typical cost of ten minutes when evaluated to @xmath225 . next , we compute the quantities @xmath226 given by eq . . using eqs . and , we write them as follows , @xmath227 we calculate the clebsch - gordan coefficients using the gsl @xcite and slatec @xcite libraries . the summation in and is performed up to @xmath228 . we use the publicly available boltzman code ( camb ) @xcite to compute the quantities @xmath229 . we have simulated large number of statistically isotropic monte - carlo ( mc ) realizations of the field @xmath230 using wmap noise covariance and beam transfer functions . we store the mc maps in the same format as the original map , and the analysis procedure explained in this section is applied to both data and mc maps on equal footing . now we can check the consistency of the observed cmb with the hypothesis of the statistical isotropy . we begin with the model - independent analysis , as outlined in section [ sec : modelindependent ] . we reconstruct coefficients @xmath231 , defined by eq . , from the seven - year wmap data as well as from the mc maps . the results are presented in fig . they are in a good agreement with the results obtained by hanson and lewis @xcite for the five - year maps . in particular , we confirm the result on the large quadrupole for the @xmath1 and @xmath2 bands . as discussed in refs . @xcite@xcite , the preferred quadrupole direction lies very close to the ecliptic poles . another suspicious thing is the frequency dependence of the signal . namely , it is non - zero in the @xmath2 band at much higher confidence level than in the @xmath1 band . this indicates a systematic effect rather than the cosmological origin . as discussed in ref . @xcite , the account of beam asymmetries can provide a complete explanation of the anomaly . let us turn to the conformal rolling scenario . first , we consider the version of the model with the intermediate stage ( sub - scenario b ) . the statistical anisotropy is determined by eqs . and . having the set of the coefficients @xmath231 reconstructed from the observed cmb , we solve eq . and estimate the value of @xmath0 . we perform the analysis for the multipole numbers starting from @xmath232 and ranging up to @xmath233 . the results are presented in fig . [ h2 ] . to evaluate the statistical errors we use about one hundred mc simulated isotropic maps . we see that the isotropic model is ruled out at more than @xmath234 confidence level even in the @xmath1 band . however , the large value of @xmath0 ( e.g. , @xmath235 at @xmath236 ) is due to the anomalous quadrupole anisotropy , which is argued to have non - cosmological origin . aiming at constraining the parameter @xmath0 , we simulate a large number of anisotropic maps for each value of this parameter . we adapt the approach of ref . @xcite , and use the following procedure , adequate in the case of small statistical anisotropy . we first simulate a seed map @xmath237 with a covariance @xmath153 given by eq . . then we generate a set of coefficients @xmath238 based on the value of @xmath0 . the map @xmath239^{-1 } \right)^{1/2 } { \bf \theta}^i \ ; , \ ] ] has covariance @xmath240 , where @xmath241 is given by eq . . to the linear order in the anisotropic effects we have @xmath242^{-1 } { \bf \theta}^{i } \ ; .\ ] ] finally , we multiply the map by the beam transfer function in the harmonic space , convert it to coordinate space and add pixel noise to get statistically anisotropic simulated map @xmath243 similar to that observed by wmap . to set an upper limit , we allow @xmath0 to be so large that for @xmath244 of simulated anisotropic maps the value of the estimated parameter exceeds the value estimated from the observed cmb map . in this way we obtain the upper limit , which reads @xmath245 at the 95% confidence level . in view of the likely non - cosmological origin of the anomalous quadrupole in the statistical anisotropy of the wmap data , one would like to constrain the parameter @xmath0 from the non - observation of higher multipoles only . one way to do that would be to follow the same procedure as discussed in section [ sec : parameter ] but keep the set @xmath246 of the quadrupole coefficients fixed and taken from the observational data . in practice , things are simpler . indeed , the effects of the statistical anisotropy corresponding to different multipole numbers @xmath16 , @xmath177 do not interfere with each other , at least in the approximation of small coefficients @xmath9 . to see this , we note that the theoretical reconstruction of the fisher matrix is diagonal . the covariances of the quantities @xmath9 are diagonal as well . as a consequence , it is straightforward to neglect the effect of the quadrupole modulation by using the estimator for the parameter @xmath0 as in but with the summation starting from @xmath247 . the values of @xmath0 estimated in this way are plotted in fig . we restrict our analysis to @xmath236 and obtain that @xmath0 is consistent with zero for the v band . making use of the statistical uncertainty inferred from isotropic mc maps , we obtain the upper limit @xmath248 at the @xmath244 confidence level . even though omitting the anomalous quadrupole makes the situation cleaner ( at least in the @xmath1 band ) , this constraint is similar to eq . . the reason is twofold . first , according to eq . , the predicted statistical anisotropy spectrum @xmath231 decreases with @xmath16 as @xmath249 second , the errors grow with the multipole number roughly as @xmath16 , see fig . [ cl ] . with the planck data available , we expect substantial improvement of the constraint . the reason is twofold . hopefully , the quadrupole anomaly will be absent in the planck data . also , the range of the cmb multipoles useful in the analysis will be considerably extended . the error bars , whch can be roughly estimated by making use of the inverse fisher matrix , scale with the number of multipoles as @xmath250 . this is clear from the eq . . taking , e.g. , @xmath251 , one would be able to reduce the error bars by about an order of magnitude . hence , the non - observation of the statistical anisotropy will give the constraint as strong as @xmath252 . we conclude that the statistical anisotropy is a promising signature from the viewpoint of the cmb observations in the case of the sub - scenario b. finally , we consider the sub - scenario a. to the linear order in constant @xmath21 , the statistical anisotropy is of the general quadrupole type with decreasing amplitude , @xmath22 . this fact is crucial for the search for the statistical anisotropy in the cmb sky . indeed , the contribution to the signal @xmath154 is additionally suppressed by the cmb multipole number @xmath23 . this suppression is due to the fact that the integral in eq . is saturated , roughly speaking , at @xmath253 . effectively , it results in low statistics of the relevant cmb multipoles and large statistical errors , which severely restrict the opportunity to observe the ( cosmological ) statistical anisotropy of the type predicted . somewhat loosely we apply the qml estimator to the seven - year wmap data . in fig . [ k1 ] we show the results for @xmath197 of the wmap reconstructed coefficients @xmath9 , assuming @xmath254 , but not restricting yet to the quadrupole - only @xmath9 . we apply the procedure used in the case of the sub - scenario with the intermediate stage , to constrain the sub - scenario a ; to this end , the quadrupole point @xmath255 in fig . [ k1 ] is relevant only . the limit on the parameter @xmath0 then reads @xmath256 at the 95% confidence level . note , however , that for large values of @xmath0 , the qml procedure is questionable . this limit can be viewed merely as an indication that the leading order contribution to the statistical anisotropy is in fact negligible . the stronger constraint comes from the subleading contribution encoded in . the reason is that the amplitude @xmath12 is independent of the wavenumber @xmath11 in this case . thus , the suppression at high cmb multipoles is absent , and the range of relevant @xmath23 s is extended up to @xmath228 . since the quantities @xmath24 are non - gaussian in this particular case , the constraining procedure is somewhat different . first , we generate the components of the `` velocity '' @xmath84 starting from a given value of the effective constant @xmath203 . then , using , we calculate the coefficients @xmath24 . the quantity @xmath257 constructed out of these @xmath24 is compared with the one estimated from the seven - year wmap data . in this way we obtain the constraint , which reads @xmath258 at the 95% confidence level . assuming that the logarithmic enhancement is not particularly strong , we conclude that this constraint is still weak . note also that the statistical anisotropy predicted by the sub - scenario a is of the same type as in some inflationary models @xcite @xcite . thus , it can not be used to descriminate our model from other models of the generation of primordial perturbations . fortunately , the sub - scenario a gives rise to the non - gaussianity in the trispectrum @xcite , which is in the sharp contrast with the inflationary predictions . we leave for the future search for the corresponding signatures in the cmb sky . we are indebted to v. rubakov for numerous helpful suggestions at all stages . we are grateful to d. hanson for kindly providing the code for inverse variance filtering . we would like to thank a. lewis , p. naselsky , m. sazhin and o. verkhodanov for fruitful discussions . the work is supported in part by the rfbr grants 11 - 02 - 01528a and 12 - 02 - 00653a ( gr ) , by the grants of the president of the russian federation ns-5590.2012.2 , mk-1632.2011.2 ( gr ) , by the grant of the government of russian federation 11.g34.31.0047 ( gr ) . the numerical part of the work has been done at the cluster of the theoretical division of inr ras . 99 m. tegmark , a. de oliveira - costa and a. hamilton , phys . d * 68 * ( 2003 ) 123523 [ arxiv : astro - ph/0302496 ] . p. bielewicz , h. k. eriksen , a. j. banday , k. m. gorski and p. b. lilje , astrophys . j. * 635 * ( 2005 ) 750 [ arxiv : astro - ph/0507186 ] . c. j. copi , d. huterer , d. j. schwarz and g. d. starkman , mon . not . * 367 * ( 2006 ) 79 [ arxiv : astro - ph/0508047 ] . k. land and j. magueijo , phys . lett . * 95 * ( 2005 ) 071301 [ arxiv : astro - ph/0502237 ] . f. k. hansen , a. j. banday , k. m. gorski , h. k. eriksen and p. b. lilje , astrophys . j. * 704 * ( 2009 ) 1448 [ arxiv:0812.3795 [ astro - ph ] ] . h. k. eriksen , f. k. hansen , a. j. banday , k. m. gorski , and p. b. lilje , astrophys . j. * 605 * , 14 ( 2004 ) , [ arxiv:0307507 [ astro - ph ] ] . c. l. bennett , r. s. hill , g. hinshaw , d. larson , k. m. smith , j. dunkley , b. gold and m. halpern _ et al . _ , astrophys . j. suppl . * 192 * , 17 ( 2011 ) [ arxiv:1001.4758 [ astro-ph.co ] ] . m. cruz , l. cayon , e. martinez - gonzalez , p. vielva and j. jin , astrophys . j. * 655 * ( 2007 ) 11 [ arxiv : astro - ph/0603859 ] . verkhodanov , to appear in physics uspekhi . j. kim and p. naselsky , astrophys . j. * 714 * ( 2010 ) l265 [ arxiv:1001.4613 [ astro-ph.co ] ] . j. kim and p. naselsky , phys . d * 82 * ( 2010 ) 063002 [ arxiv:1002.0148 [ astro-ph.co ] ] . c. rath , p. schuecker and a. j. banday , phys . * 102 * ( 2009 ) 131301 [ arxiv : astro - ph/0702163 ] . a. r. pullen and m. kamionkowski , phys . d * 76 * ( 2007 ) 103529 [ arxiv:0709.1144 [ astro - ph ] ] . l. ackerman , s. m. carroll and m. b. wise , phys . d * 75 * ( 2007 ) 083502 [ erratum - ibid . d * 80 * ( 2009 ) 069901 ] [ arxiv : astro - ph/0701357 ] . e. akofor , a. p. balachandran , s. g. jo , a. joseph and b. a. qureshi , `` direction - dependent cmb power spectrum and statistical anisotropy from noncommutative geometry , '' jhep * 0805 * , 092 ( 2008 ) [ arxiv:0710.5897 [ astro - ph ] ] . i. antoniadis , p. o. mazur and e. mottola , phys . lett . * 79 * ( 1997 ) 14 [ arxiv : astro - ph/9611208 ] . v. a. rubakov , jcap * 0909 * ( 2009 ) 030 [ arxiv:0906.3693 [ hep - th ] ] . creminelli , a. nicolis and e. trincherini , jcap * 1011 * ( 2010 ) 021 [ arxiv:1007.0027 [ hep - th ] ] . m. osipov and v. rubakov , jetp lett . * 93 * ( 2011 ) 52 [ arxiv:1007.3417 [ hep - th ] ] . m. libanov and v. rubakov , jcap * 1011 * ( 2010 ) 045 [ arxiv:1007.4949 [ hep - th ] ] . m. libanov , s. mironov and v. rubakov , prog . * 190 * ( 2011 ) 120 [ arxiv:1012.5737 [ hep - th ] ] . m. libanov , s. mironov and v. rubakov , phys . d * 84 * ( 2011 ) 083502 [ arxiv:1105.6230 [ astro-ph.co ] ] . m. libanov , s. ramazanov and v. rubakov , jcap * 1106 * ( 2011 ) 010 [ arxiv:1102.1390 [ hep - th ] ] . k. hinterbichler and j. khoury , arxiv:1106.1428 [ hep - th ] . s. mollerach , phys . d * 42 * , 313 ( 1990 ) ; + a. d. linde and v. f. mukhanov , phys . d * 56 * ( 1997 ) , 535 [ arxiv:9610219 [ astro - ph ] ] ; + k. enqvist and m. s. sloth , nucl . b * 626 * ( 2002 ) , 395 [ arxiv:0109214 [ hep - ph ] ] ; + d. h. lyth and d. wands , phys . b * 524 * ( 2002 ) , 5 [ arxiv:0110002 [ hep - ph ] ] ; + t. moroi and t. takahashi , phys . lett . b * 522 * ( 2001 ) , 215 [ erratum - ibid . b * 539 * ( 2002 ) , 303 ] [ arxiv:0110096 [ hep - ph ] ] ; + k. dimopoulos , d. h. lyth , a. notari and a. riotto , jhep * 0307 * , ( 2003 ) , 053 [ arxiv:0304050 [ hep - ph ] ] . g. dvali , a. gruzinov and m. zaldarriaga , phys . d * 69 * ( 2004 ) , 023505 [ arxiv:0303591 [ astro - ph ] ] ; + l. kofman , [ arxiv:0303614 [ astro - ph ] ] ; + g. dvali , a. gruzinov and m. zaldarriaga , phys . d * 69 * ( 2004 ) , 083505 [ arxiv:0305548 [ astro - ph ] ] . n. e. groeneboom and h. k. eriksen , astrophys . j. * 690 * ( 2009 ) 1807 [ arxiv:0807.2242 [ astro - ph ] ] . d. hanson and a. lewis , phys . d * 80 * ( 2009 ) 063004 [ arxiv:0908.0963 [ astro-ph.co ] ] . n. e. groeneboom , l. ackerman , i. k. wehus and h. k. eriksen , astrophys . j. * 722 * ( 2010 ) 452 [ arxiv:0911.0150 [ astro-ph.co ] ] . d. hanson , a. lewis and a. challinor , phys . d * 81 * ( 2010 ) 103003 [ arxiv:1003.0198 [ astro-ph.co ] ] . k. hinterbichler , a. joyce and j. khoury , arxiv:1202.6056 [ hep - th ] . m. zaldarriaga , phys . d * 69 * ( 2004 ) 043508 [ astro - ph/0306006 ] . p. creminelli , phys . d * 85 * ( 2012 ) 041302 [ arxiv:1108.0874 [ hep - th ] ] . c. m. hirata and u. seljak , phys . d * 67 * ( 2003 ) 043001 [ arxiv : astro - ph/0209489 ] . a. nicolis , r. rattazzi and e. trincherini , phys . d * 79 * ( 2009 ) 064036 [ arxiv:0811.2197 [ hep - th ] ] . m. libanov and v. rubakov , arxiv:1107.1036 [ hep - th ] . j. k. erickson , d. h. wesley , p. j. steinhardt and n. turok , phys . rev . d * 69 * ( 2004 ) 063514 ; [ arxiv:0312009 [ hep - th ] ] . + d. garfinkle , w. c. lim , f. pretorius and p. j. steinhardt , phys . d * 78 * ( 2008 ) 083537 ; [ arxiv:0808.0542 [ hep - th ] ] . e. i. buchbinder , j. khoury and b. a. ovrut , phys . d * 76 * ( 2007 ) 123503 ; [ arxiv:0702154 [ hep - th ] ] ; + p. creminelli and l. senatore , jcap * 0711 * ( 2007 ) 010 ; [ arxiv:0702165 [ hep - th ] ] . j. l. lehners , phys . * 465 * ( 2008 ) 223 ; [ arxiv:0806.1245 [ astro - ph ] ] . n. jarosik , c. l. bennett , j. dunkley , b. gold , m. r. greason , m. halpern , r. s. hill and g. hinshaw _ et al . _ , astrophys . j. suppl . * 192 * , 14 ( 2011 ) [ arxiv:1001.4744 [ astro-ph.co ] ] . e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 192 * , 18 ( 2011 ) [ arxiv:1001.4538 [ astro-ph.co ] ] . http://lambda.gsfc.nasa.gov/ k. m. gorski , e. hivon , a. j. banday , b. d. wandelt , f. k. hansen , m. reinecke and m. bartelman , astrophys . j. * 622 * , 759 ( 2005 ) [ arxiv:0409513 [ astro - ph ] ] . oh , d. n. spergel , and g. hinshaw , astrophys . j. * 510 * , 551 ( 1999 ) , [ arxiv:9805339 [ astro - ph ] ] . k. m. smith , o. zahn , and o. dore , phys . d * 76 * , 043510 ( 2007 ) , [ arxiv:0705.3980 [ astro - ph ] ] . a. lewis , a. challinor and a. lasenby , astrophys . j. * 538 * ( 2000 ) 473 [ arxiv : astro - ph/9911177 ] .

we investigate the properties of optically passive spirals and dusty red galaxies in the a901/2 cluster complex at redshift using restframe near - uv - optical seds , 24 m ir data and hst morphologies from the stages dataset . the cluster sample is based on combo-17 redshifts with an rms precision of km / sec . we find that dusty red galaxies and optically passive spirals in a901/2 are largely the same phenomenon , and that they form stars at a substantial rate , which is only lower than that in blue spirals at fixed mass . this star formation is more obscured than in blue galaxies and its optical signatures are weak . they appear predominantly in the stellar mass range of $ ] where they constitute over half of the star - forming galaxies in the cluster ; they are thus a vital ingredient for understanding the overall picture of star formation quenching in clusters . we find that the mean specific sfr of star - forming galaxies in the cluster is clearly lower than in the field , in contrast to the specific sfr properties of blue galaxies alone , which appear similar in cluster and field . such a rich red spiral population is best explained if quenching is a slow process and morphological transformation is delayed even more . at , such galaxies are rare , suggesting that their quenching is fast and accompanied by morphological change . we note , that edge - on spirals play a minor role ; despite being dust - reddened they form only a small fraction of spirals independent of environment . surveys ; galaxies : evolution ; galaxies : spiral ; galaxies : clusters : general ; stars : formation ; infrared : galaxies

0811.3873

we still do not understand the origin of the morphology - density - relation and the star - formation ( sf)-density relation . the local morphology - density relation ( dressler 1980 ) appears to persist out to @xmath6 in the field ( guzzo et al 2007 ) and to @xmath7 in clusters @xcite , while the sf - density relation appears to invert its slope between high redshift where star formation first commences at highest densities ( see * ? ? ? * ; * ? ? ? * for results at @xmath7 ) and low - redshift ( lewis et al 2002 ; gomez et al 2003 ) where it avoids overdense regions . this evolution could be explained by continuous formation of non - starforming early - type galaxies throughout cosmic history ( faber et al 2007 ) , where the higher - density regions would be leading the process ( smith et al 2005 ) . as a result , we also find a colour - density relation at low redshift , whereby dense regions are inhabited mostly by red galaxies with old stellar populations ; @xcite and @xcite have shown that the colour - density relation is the primary trend , while superimposed residual trends in morphology are relatively weak ; mostly morphology appears to be driven by colour whatever the density . in clusters , however , the situation is more controversial : the above picture suggests the presence of an age - density relation at fixed morphology : thomas et al ( 2005 ) claim a strong dependence of early - type ages with environmental density at all masses ; while no such evidence is seen in the rich local coma cluster ( mehlert et al 2003 ; trager et al 2008 ) , wolf et al ( 2007 ) see an age - density relation at fixed morphology and low luminosity in the young a901/2 cluster complex . @xcite again do not see an age trend at higher luminosity in @xmath8 clusters and find that the sf - d and morphology - density relation are equivalent and thus equally fundamental . but if there is no age - density relation at fixed morphology , either close synchronisation between sf quenching and morphological transformation is required , or all transformation producing the patterns must have happened in the distant past such that memories of time scales are wiped out . while star formation rate ( sfr ) is a unique number , morphology is a complex multi - parametric variable : an important distinction must be made between the mass concentration of a galaxy , which is driven by its dynamical history , and its clumpiness , which is driven by star formation and correlates with colour : over three decades ago _ anemic spirals _ with smooth arms and no signs of clumpy star formation were discovered @xcite , which could provide a missing link by being a phase in the transition from a blue star - forming field spiral to a red , non - starforming cluster s0 . also , poggianti et al . ( 1999 , 2004 ) and goto et al . ( 2003 ) found spirals without optical signs of star formation in outskirts of clusters , and dubbed them optically - passive spirals. using an sed - based approach , wolf , gray & meisenheimer ( 2005 , hereafter wgm05 ) have described a population of dusty red galaxies dominating the outskirts of the clusters a901/2 . detailed seds from the combo-17 survey measure dust reddening independent of age - induced reddening . their sample has red colours due to both dust and intermediate age rather than old age as for regular early - types . ground - based combo-17 morphologies show them to be mostly early spirals @xcite . at least four interesting questions arise here : \(1 ) what are the properties of these dusty red spirals ? both passive spirals and the wgm05 dusty red galaxies show no strong optical signs of star formation . the presence of dust and absence of significant optical emission lines , however , motivates a study of obscured star formation with far - infrared data in this paper . \(2 ) where are the cluster galaxies undergoing change , such as a decline in star formation ? truncated star - formation disks are common in virgo spirals and mark a slowly transforming population @xcite . but most studies of large data sets find that the average star formation rate in star - forming galaxies does not change with environment ; instead it appears as if only the fraction of star - forming galaxies declines towards higher densities @xcite . in this paper , we argue that the role of star formation in red spirals has often been ignored ( see also * ? ? ? * ) , while it is an important aspect in clusters : in fact , we find most star - forming galaxies in a901/2 to be red , while a blue galaxy sample restricts attention to only the most intense tail of star formation . \(3 ) are they credible progenitors for cluster s0s ? it has been argued that disk fading in field spirals does not produce s0s of the highest known luminosities @xcite and correct distribution of bulge / disk ratios @xcite , so that the progenitors of bright cluster s0s are still at large . instead , bulge enhancement @xcite or mergers among infalling galaxies @xcite may be required . there is indeed evidence for centrally concentrated star formation in cluster spirals , both at low @xcite and intermediate redshift @xcite that will increase concentration over time . recent work on higher - quality data also suggests that the concentration differences between spirals and s0s are not as large as once claimed @xcite . \(4 ) which processes initiate the quenching of star formation and the morphological transformation ? both internal and environmental factors may be involved . internal factors include feedback from supernovae and agn @xcite as well as primordial differences in the formation conditions of massive galaxies at high redshift propagating into their present - day behaviour . they are expected to depend on galaxy mass , and upper galaxy mass limits for star formation may evolve with redshift @xcite . environmental mechanisms include ram - pressure stripping by hot icm ( gunn & gott 1972 ) ; star formation enhanced by major mergers or tidal excitation from the cluster potential ( barnes 1992 ; bekki 1999 ) ; loss of gas by suffocation , harassment or strangulation ( larson et al 1980 ; moore et al 1996 ; balogh et al 2000 ) . e.g. , @xcite find gradual star formation decline and red spirals in the outskirts of clusters , but truncated star formation and e / s0 galaxies with smooth profiles in the core regions filled by hot gas , presumably due to ram - pressure stripping . it was also speculated , that gravitational heating inherent in structure growth on a cosmic scale ( khochfar & ostriker 2007 ) may quench cosmic star formation as effectively as ad - hoc agn feedback in semi - analytic models . any progress in identifying the mechanism producing red spirals requires the measurement of masses as well as environmental indicators . a low - redshift study by @xcite revealed that concentration correlates with mass , while clumpiness correlates with environment . to what extent are then red spirals a mass - specific or a cluster - specific phenomenon ? agn feedback and environment - driven gas depletion are both expected to suppress star formation , perhaps in different mass regimes and over different timescales . at this point , it is not yet clear what their relative role is in suppressing star formation as galaxies grow and structure is built up . to learn more , we need larger samples of galaxies ideally showing their transformation in a few discernible stages . also , it is important to address evolution in a mass - resolved picture and not with mass - limited samples , as in the latter a mass mix changing with environment produces purely apparent environmental trends . the intricate problem described above requires a multi - faceted data set of a region in our universe where galaxies are presently undergoing major change . we have previously identified the supercluster region abell 901ab/902 ( @xmath9 ) as a complex environment that is particularly rich in dusty red galaxies and suggestive of strong galaxy infall activity . the cluster redshift is ideal for obtaining observations that have both high signal - to - noise ratio and complete wide - field multi-@xmath10 coverage . such a data set has been obtained by the stages team ( * ? ? ? * hereafter g08 ) and provides a high - quality description of a few environmental components coupled to different physics , as well as masses , seds , morphologies and star - formation rates for over 1000 cluster members . this effort started with the combo-17 survey delivering 17-filter seds and high - fidelity photometric redshifts ( rms scatter among 800 brightest members is 0.006 , see wgm05 ) . the large hst survey at the core of the stages initiative added morphologies with @xmath11pc resolution for all cluster members over a @xmath12 mpc@xmath13 ( @xmath14 ) area and the highest resolution dark matter map of large - scale structure @xcite . xmm and gmrt observations reveal intracluster gas and agn @xcite . we found that a901/2 is made up of four distinct cluster cores with different masses and geometries , icm , dark matter and galaxy properties . the role of environment in the stages data has already been subject to some studies : @xcite find that mergers and interactions are found predominantly outside the cluster cores and can be accounted for by the large galaxy velocity dispersion in the cores and the accretion of groups and field galaxies within coherent flows . @xcite find no region of enhancement in overall star formation , whether or not obscured star formation is included . with no clear indication of enhanced mergers or triggered star formation , the main impact of environment appears to be a suppression in star formation as already seen in red spirals . in this paper , we focus on the rich sample of optically passive , red spiral galaxies , which appear to undergo transformation as we observe them , possibly descending from blue spirals and evolving into s0 galaxies . while their sfr is lower than in blue galaxies by a sufficient degree to give them red colours , we choose to relabel them semi - passive given their persistent levels of star formation ; as our spitzer / mips data show , these galaxies still harbour significant star formation , which is more obscured than in the average cluster population . we present the stages data relevant to this subject in sect . 2 , and define our environment measures . in sect . 3 we present relations between ir and optical properties , confirming earlier inferences from combo-17 seds . we demonstrate the differences between splitting galaxies by blue vs. red and by star - forming vs. non - star - forming . sect . 4 presents our morphological results and the correspondence of red spirals with dusty red galaxies. in sect . 5 , we discuss galaxy mass functions by type , motivating a mass - resolved investigation of any environmental trends . sect . 6 shows galaxy properties and type fractions by mass and environment , and finally we discuss a possible evolutionary sequence involving red spirals in sect . this study is entirely based on a data set assembled by the stages project and recently published in catalogue form ( g08 ) . it encompasses high - fidelity photometric redshifts and observed-/rest - frame seds from combo-17 ( wolf et al 2003 ) , imaging with the advanced camera for surveys ( acs ) on board of hst and spitzer / mips ir data , all overlapping on an area of almost @xmath15 . further data have been obtained with xmm , gmrt and galex , but are neither part of the recent data release nor used in this study . we select cluster members on the basis of their photometric redshifts . throughout the paper , we use cosmological parameters of @xmath16 km / s / mpc@xmath17 . the combo-17 survey was observed with the wide field imager ( wfi ) at the mpg / eso 2.2m - telescope on la silla , chile . it used five broad - band filters ubvri and 12 medium - band filters with resolution @xmath18 , altogether covering wavelengths from 350 to 930 nm , to define detailed optical seds for objects of @xmath19 . one of its four fields observed from 1999 to 2001 covers an area of @xmath20 around the a901/2 clusters ; see g08 for all technical details of object search , photometry , classification , redshifts and further derived parameters . the literature contains several combo-17 studies of galaxy evolution using samples defined by photometric redshift . all these restrict themselves to galaxies brighter than @xmath21 to make sure only reliable redshifts are being used . the g08 catalogue of the a901 field contains @xmath22 galaxies with redshift estimates at @xmath23 . however , photometric redshift ( photo - z ) accuracy is a function of photometric noise and declines with fainter magnitudes and field contamination in a photo - z - selected cluster sample increases directly with photo - z scatter . g08 present a detailed discussion of the photo - z scatter and of the completeness and contamination in a901 cluster samples as a function of depth . in sect . 2.5 we specify the galaxy samples and photo - z accuracies drawn for the purposes of this paper ; given the low redshift of a901 , we can choose a mass - limited sample while including only the brightest objects with the most accurate photo - z s . here , we only repeat numbers characterizing the overall combo-17 galaxy sample . generally , at bright magnitudes of @xmath24 the combo-17 photo - z s have an rms in @xmath25 of @xmath26 , which has been confirmed with spectroscopy on all three combo-17 fields with a full data set ( cdfs , a901 , s11 ) . using a spectroscopy of 404 galaxies including 249 members of the a901 cluster , @xcite have shown that the cluster photo - z s show an rms scatter of @xmath27 corresponding to @xmath28 km / s . @xmath29 of these galaxies have photo - z deviations from the true redshift of @xmath30 . only three objects ( @xmath31% ) are confirmed to deviate by more than 0.04 from the true redshift and are typically galaxies with unusual seds due to contributions from an active nucleus . a deeper comparison is currently only available on the cdfs field , where @xcite find an increase of photo - z scatter to @xmath32 at @xmath33 increasing further to fainter magnitudes . on the a901 field low - resolution prism spectroscopy of a faint sample was obtained with imacs at magellan , which will shed light on the error behaviour at fainter magnitudes when analysed . the detailed observed - frame seds and accurate redshifts have allowed the derivation of robust rest - frame magnitudes for some wavebands including johnson and sdss filters and a pseudo - continuum filter at 280 nm intended as a proxy for unobscured star formation rate . at the cluster redshift , the wfi filters used for combo-17 cover the rest - frame spectrum from 300 to 800 nm . the combo-17 sed fit involves an age parameter and an estimate of the dust reddening @xmath34 . stellar mass estimates are based on sed fitting as well and assume a kroupa ( 1993 ) imf , while a kroupa ( 2001 ) or chabrier ( 2003 ) imf would yield values within @xmath35% of those ( see g08 for all details ) . note , that a @xcite imf increases stellar masses by a factor of 1.8 . the combo-17 survey mostly targetted galaxies at @xmath36 $ ] @xcite with much smaller angular extent than the most massive cluster galaxies in a901 . thus its processing was optimised to obtain high signal - to - noise photometry of faint objects from identical physical footprints ( after psf deconvolution ) in all filters . these seds represent a central gaussian aperture with @xmath37 fwhm on a hypothetical infinitely sharp image , and are the basis for the estimation of redshifts , mass - to - light ( m / l ) ratios and stellar masses , as they have been published now for the a901 field by g08 . the sed is normalised with the total object photometry in the deep ( 20 ksec ) r - band image , in which the total light is measured with optimal surface brightness sensitivity , but the colours outside the aperture are ignored . for small objects or particularly large objects without colour gradients this has no consequence . but if large size , low concentration and strong colour gradients are combined , then the seds constructed from total magnitudes may be different from the seds based on aperture magnitudes . since our stellar mass estimates depend on m / l ratios from these seds , aperture - based seds that are not representative of the total galaxy will yield a stellar mass estimate that is incorrect . at the cluster distance the aperture has 4.25 kpc fwhm , and is broadly comparable to the footprint of the fiber aperture in sdss spectroscopic data at @xmath38 . this implies that any sed types for galaxies that we discuss later in the paper are defined by the stars seen in the central footprint of 4.25 kpc fwhm . the total seds could be different if substantial light with a different sed originated outside this aperture , which may change the colour class to which an object is allocated . we have investigated the total colours of a few dozen selected galaxies in a separate procedure using sextractor total magnitudes of all broad bands , although for general use , these colours would be too noisy and object matching is too unreliable , especially given the low sensitivity of the u - band imaging . this has revealed , that aperture values are similar to the total ones across a wide parameter space in our sample . masses of spirals are reliable at @xmath39 and masses of e / s0 galaxies are reliable everywhere . but there is an issue with cluster spirals estimated currently at @xmath40 : their total observed - frame @xmath41 colours are on average 0.3 mag bluer than the aperture colours , and this translates into their @xmath42 being overestimated by @xmath43 dex . the @xmath44 corrections range from 0 to @xmath45 mag , and in the most extreme object ( 42497 ) the mass is overestimated by 0.5 dex as a result . this is in contrast to s0 galaxies at @xmath46 , whose total @xmath41 is only 0.04 mag bluer than in the aperture . for this reason , most of the 16 spirals in the sample with @xmath40 are truly just below 11 . almost all galaxies in the field or cluster sample that truly have @xmath40 appear to be old , red and of e or s0 type . in the absence of a consistent fix for the mass values of high - mass spirals we will omit them later when focussing on the properties of spirals and limit ourselves to the mass range of @xmath47 $ ] then . the primary hst observations of the stages project cover an 80-tile mosaic with acs covering an area of @xmath48 in the v606w filter during cycle 13 , thus covering @xmath49% of the combo-17 area . various parallel observations of other hst instruments are not part of this study . the mean psf on these images after data reduction has @xmath50 fwhm which corresponds to a resolution of 300 pc at the cluster redshift . while the surface - brightness sensitivity of the hst images is lower than that in the 20 ksec r - band image of the ground - based combo-17 observations , structural features are seen much more clearly in the acs images ( with few exceptions of lsb tidal arms ) due to the @xmath51 better resolution of hst . the published catalogue contains morphological parameters from sersic profile fitting with galfit @xcite , but here we have used mostly hubble types . all galaxies have been visually classified by seven co - authors ( aas , fb , mg , kj , kl , dm , cw ) into hubble types , including flags on asymmetries and interactions . we distinguish between the types ( e , s0,sa , sb , sc , sd , irr ) and their barred cousins , and define an s0 as a disk galaxy with a visible bulge but no spiral arms . when we notice spiral arms , we put the object into the sa - sd range on the basis of our perceived bulge - to - disk ratio , but we suspect that smooth disks and arms as in anemic spirals may bias our judgement towards earlier types @xcite . the classifiers agree generally well , but small or faint galaxies can be ambiguous when they show little features . for this paper , we use weighted average estimates of the hubble types , ignoring bars and degrees of asymmetry . we also include all mergers in the irregular class , since they are not of separate interest to this paper ( with the exception of fig . [ rfcols ] ) . [ cls ] shows examples of cluster galaxies from all morphological types we differentiate in this work and covers the entire mass range of our sample . barred or interacting galaxies are studied in separate works @xcite . .properties of cluster galaxy sample . [ sample ] [ cols="<,>,>,>,>",options="header " , ] * the specific sfr and the dust extinction ( row 1 and 2 ) of the overall stellar population both change gradually with density . it appears as if there is already a change to less star formation and less extinction from the field to our outermost infall points ; while the error bars overlap , a formal decline is seen in mass bin . this may correspond to the group pre - processing much discussed in the literature ( e.g * ? ? ? * ) . * there is no difference in the morphological properties ( row 3 and 4 ) between field galaxies and those in the outermost infall bin at @xmath52 and there appears to be a more sudden change further into the infall region . below 10 , however , the morphology appears to change gradually with density again . * at @xmath53 all trends appear much weaker , because field galaxies are already mostly of early type and have little star formation . below 10.5 , the overall changes from field to cores are virtually identical between the two mass bins despite possible differences in the infall region . unfortunately , we could not sensibly plot trends of the mean extinction in the star - forming regions of star - forming galaxies due to the shallow ir sensitivity limits in the field sample . however , in sect . 3 we have already established broadly that dusty red galaxies dominate the sf sample in the cluster while having little relevance in the field , and have more obscured star formation regions than blue galaxies . fig . [ tdrel ] shows trends in several galaxy type fractions with environment . the first two rows plot hubble type fractions ; the third row compares the fractions of the three sed types , blue cloud vs. old red and dusty red galaxies . the last row shows in dark points the fraction of spirals is redwards of the cmr cut ( old red plus dusty red ) ; the light points show the fraction of dusty red galaxies out of all red galaxies , which we take to be the fraction of star - forming galaxies among red galaxies ( sf / red ) . we draw the following conclusions : * the strongest morphological changes ( row 1 and 2 ) are seen within the infall region of the mass bin @xmath54 $ ] and are associated particularly with the demise of all kinds of spirals and a steep increase in the s0 fraction . scd and irr galaxies decline first , while sa / sb spirals only decline closer to the cores . at @xmath55 $ ] s0s dominate over e s everywhere in the cluster , but at higher masses ( not shown here ) ellipticals are clearly dominant . * there is little change in the hubble type fractions ( row 1 and 2 ) between the field and the lowest - density infall point that is plotted , especially above @xmath56 . * dusty red galaxies ( row 3 ) are most dominant at intermediate mass and density , but they play virtually no role at @xmath57 ; by implication the same applies to red spirals ( row 4 ) . at @xmath58 the rise of red spirals in the infall region is due to sfr decline that is _ not _ accompanied by morphological change , so that blue field spirals are replaced by red infall spirals ; going further towards the cores all sorts of spirals are gradually replaced by red s0s ( row 1 ) . * at @xmath57 in contrast , blue field spirals are replaced by red infall spheroids as a result of colour change being accompanied by synchronised morphological change . hence , the red fraction among spirals depends strongly on mass and ranges from @xmath59% at low mass ( @xmath60 ) to @xmath61% at high mass ( @xmath62 ) . at @xmath52 it changes strongly from field to the first infall point and stays high throughout infall , while the overall spiral fraction declines . * the contamination of star - forming galaxies in the red sequence ( sf / red , row 4 ) is generally low in the cores , but high in the infall regions at intermediate mass due to dusty red galaxies being common there , and high in the field sample at low mass ( @xmath60 ) , where the cmr cut enters into the contiguous structure otherwise known as blue cloud and leads to 80% of the red sample being star - forming . * again , at @xmath53 all trends are weaker . with respect to red spirals our particular cluster environment appears quantitatively consistent with the cosmic distribution of red spirals at low redshift as studied very recently in the galaxy zoo project : @xcite show that at low density roughly 10% of @xmath63 $ ] spirals are red , which increases to nearly 50% at @xmath64 $ ] ; finally , at high density this number converges to @xmath65% . despite differences in mass determination and density estimator , these numbers match ours surprisingly precisely . another strong signal that could appear spurious , data - specific or possibly a901-specific , is the extreme ( @xmath66% ) sf contamination of the red low - mass field sample ; again this number matches perfectly the recent results of @xcite . hence , we also conclude that although a901/2 has been selected by us to be a convenient laboratory harbouring a range of different environments in a diverse cluster complex that is accessible to single - shot observations , it appears to still be a completely representative environment , from which physical conclusions can be generalised . there are in total exactly 200 galaxies in the four cores of our sample at @xmath67 . of these , the three cores with abell numbers ( 901a , 901b and 902 ) contain between 50 and 70 each , while the sw group has a less rich dense core with only 20 galaxies ( the fact that these galaxy numbers here are integer multiples of 10 is entirely coincidental ) . the sw group also differs from the other cores in its high giant - to - dwarf ratio : 25% of its core galaxies have @xmath40 , compared to only @xmath68% in the abell cores ( see tab . [ cores ] ) . the abell 901a core appears most circular in shape and is further characterized by the lowest fraction of blue galaxies , i.e. @xmath68% vs. @xmath69% for the others , and less significantly the lowest fraction of galaxies with @xmath70 and 24@xmath1-detections , i.e. @xmath71% vs. @xmath72% . it could thus be the most evolved core among the four , although the various possible evolution indicators do not present a clear picture . we consider a regular shape , high e / s0 fraction , low spiral fraction and low blue galaxy fractions indicators for maturity in a cluster core . the other three cores show various properties indicating that they are less evolved , but no clear ranking appears . the abell 901b core shows a particularly high s0 fraction with 44% vs. @xmath71% in the other cores and a correspondingly lower red spiral fraction of 8% vs. @xmath69% for the other cores , so it appears more evolved than a902 or the sw group . as gray et al . ( in preparation ) point out , a901b has the highest x - ray luminosity of all four cores , so ram - pressure stripping by hot gas could well have been responsible for the final suppression of star formation and ensuing wiping - out of disk structure from former red spirals . a902 in particular has the lowest spheroid fraction of all , with 50% vs. @xmath73% for the others , which means that a902 has the properties of a truly less dense region . perhaps some line - of - sight projection has boosted its apparent density ; @xcite show from a weak - lensing analysis , that a902 is the lowest - mass structure among the four with a virial mass of @xmath74 compared to @xmath75 each for a901a and b. further evidence for projection effects is discussed in rhodes et al . ( in preparation ) . it had been established by several authors , that red samples defined by cmr cuts do not only contain passively evolving spheroidal galaxies , but are contaminated with spiral galaxies both in clusters @xcite and in the field @xcite . there is also evidence for a large population of star - forming and at least moderately dusty galaxies in clusters ( wgm05 ) or the field @xcite , who report average fractions of 30 - 40% in red samples , or fractions as a function of mass and environment : we see the most extreme environmental dependence at low mass , where fractions range from @xmath60% in cores to @xmath66% in the field , in excellent agreement with @xcite . while @xcite find red spirals in the field to be mostly edge - on galaxies , we emphasise that in dense environments inclination is not the dominant explanation for their colour : instead lower specific sfrs cause the red colour irrespective of a galaxy s orientation . the key is to look not at contamination fraction in red samples , but at the ratio of red to blue spirals instead : if red spirals were purely due to high inclination , random orientation statistics would suggest in a given mass bin , that the red fraction among spirals ( or star - forming galaxies ) should be independent of environment . in fact , the @xmath76 field sample in wgm05 has a 1:8 ratio of dusty red to blue galaxies , indistinguishable from the red spiral to blue galaxy ratio at @xmath77 reported by @xcite . it is entirely plausible that one in eight spiral galaxies have inclinations high enough to extinguish most of the light from young disk stars due to the increased optical depth of dust projected on the disk . however , the red spiral fraction in our cluster sample exceeds dramatically the small geometric fraction , as it has _ more red spirals than blue galaxies _ at @xmath70 . the majority of these show reduced star formation but do not show enhanced dust extinction of the overall stellar population . their morphologies are also clearly not edge - on , but they have smoothed substructure as expected for a low - sfr system and seen in fig . [ csseq ] . given that these galaxies are the major constituent of the star - forming population , we would like to include them when studying a star - forming sample . as a result , star - forming samples show different colour and star - formation properties than pure blue cloud samples . wgm05 report an average equivalent width of [ oii ] emission in a901 dusty red galaxies of @xmath78 , compared to 17.5 in the blue cloud . [ oii ] line fluxes are an order of magnitude reduced in dusty red compared to blue galaxies . hence , their [ oii ] emission would be undetectable in the cluster studies of @xcite or in the vvds field survey @xcite . their spectra otherwise appear similar to those of truly passive galaxies . it is thus no surprise that spectroscopic surveys have found red spirals _ without optical signatures of star formation_. @xcite point out that the sdss fiber spectra at low redshift target a small nuclear region and fail to detect young stellar populations in many cases , a bias that enhances red - sequence contamination . in order to identify virtually all star - forming galaxies , they suggest to combine nuv data with an h@xmath79-line sensitivity of 2 ( equivalent width ) . finally we note , that @xcite report a fraction of star - forming red galaxies ( their intermediate types ) among all galaxies in their sdss sample ( @xmath80 ) of 20% . surprisingly , it appears to be entirely independent of the galaxy luminosity and its environment ( halo mass , location in a galaxy group ) . a number of works have established two independent observations that are in apparent conflict : \(1 ) the proximity of the virgo cluster has permitted the discovery of a substantial population of spiral galaxies , which have truncated star - forming disks . the idea is that gas is gradually being stripped from galaxies , perhaps via ism - icm interaction , whereby the outer regions of a galaxy are swept first given their lower column densities @xcite , and the truncation radius shrinks with time . at intermediate redshifts there is further evidence that cluster spirals have more centrally - concentrated star formation than field galaxies @xcite . seeing such transition objects abundantly requires a long time scale for the sf - suppressing process . \(2 ) most studies of sfr in star - forming galaxies report an independence of sfr on environment ; they only find that the fraction of star - forming galaxies declines with density , thus requiring a fast transition ( e.g. * ? ? ? the average equivalent width of the [ oii ] line in star - forming galaxies is found by @xcite at @xmath81 $ ] to be @xmath82 in the field , groups or clusters alike ; in @xmath83 clusters @xcite find @xmath84 , and at @xmath85 @xcite find 15 in bright @xmath86 galaxies . why this number should be constant across such a wide range in redshift is unclear , and may coincidentally arise from the mass - dependence of the s - sfr evolution with redshift and the redshift dependence of mass limits in the galaxy samples . our results show that at @xmath4 $ ] the mean s - sfr of star - forming galaxies declines almost by 50% from field to cores , but by much less in lower - mass galaxies . thus , we have found a significant sfr decline with density , as a result of including dusty red galaxies in our star - forming sample and using more reliable sfr estimates . had we restricted ourselves to blue galaxies alone , we would have seen only a 20% decline with very large errors due to the smaller sample , instead of a clear 50% decline . also , our blue sample has [ oii ] properties similar to the literature samples above . the comparison between blue and star - forming samples helps to resolve the contradiction posed between observations of slow transformation in virgo and apparently rapid truncation in more distant clusters : the inclusion of semi - passive spirals in the star - forming sample of cluster galaxies produces a decline of sfr in star - forming galaxies with density , which is consistent with slow quenching in clusters . in contrast , considerations of only blue - cloud galaxies require to invoke nearly instantaneous sf suppression given that their sfrs are apparently constant over density . we note that @xcite report a 20% sfr decline across density in their faintest sample bin at @xmath87 $ ] . their brighter bins show no trends but are affected by an aperture bias , which they argue may hide evolution that may only play out in an extended star - forming disk outside the aperture . we believe that their judgement is correct ; we have a smaller aperture bias in our data , given that our aperture has 4.25 kpc fwhm as opposed to the on average sub - kpc fiber aperture in @xcite . in @xmath88 clusters @xcite reported seeing numerous _ optically passive _ spirals , i.e. spiral galaxies with an absorption - line spectrum characteristic of older stellar populations , and no [ oii ] emission or otherwise apparent star formation . they also found a population of _ k+a galaxies _ characterised by abrupt sf truncation and asked why these quiescent infall populations are so prevalent at @xmath88 , but apparently lacking at low @xmath89 . in the virgo cluster , star formation was still observed and spatially resolved in spirals , and it was later shown that the anemic spirals discovered by @xcite were not the most abundant cluster - specific phenomenon @xcite ; instead , the principal effect of the virgo cluster on infalling spirals seemed to be a slow outside - in truncation of the star - forming disk . in our young @xmath90 cluster environment , we see again a very abundant infall population with similar optical properties as poggianti s _ optically passive _ spirals , except that we have shown them to be on average only _ semi - passive _ : they still form stars at a substantial rate , as shown consistently from either our ir data or from our optical estimates of dust extinction and star formation . their specific sfr , however , is on average lower by 0.6 dex in comparison to regular blue galaxies of the same mass . given an s - sfr scatter in the blue cloud of @xmath91 dex @xcite , these semi - passive spirals are @xmath92-outliers from the regular star - forming sequence in field galaxy samples ( see also the discussion in * ? ? ? altogether , we see spirals in the cluster to span the whole range from normally star - forming to virtually passive . we note that we do not see a star - bursting population in a901 , and wgm05 found no evidence for an abundant post - starburst population either , although that was concluded from stacked spectra , and they could not reliably measure h@xmath93 equivalent widths in individual galaxies . wgm05 showed the optical spectral signatures of dusty red galaxies to be almost indistinguishable from passively evolving ( old red ) galaxies on the basis of high - s / n stacked spectra . individual spectra were too noisy to find the weak [ oii ] lines and were otherwise dominated by the absorption - line spectrum of the older stars . it is now clear that the drop in star formation combined with an increase in the obscuration of star - forming regions , is sufficient to make the optical detection of star - formation signatures a challenge . but although the @xmath94 colours of red spirals and red spheroids are indistinguishable , this is only a coincidence of choosing two bands in which the effects of age and dust are ambiguous . choosing a wider range of colour , it turns out that red spirals are redder than spheroids at the red end and bluer at the blue end of the spectrum , which was the critical differentiation used by wgm05 . in fig . [ ssfrm ] we saw how e.g. the use of a near - uv colour discriminates dusty red from old red galaxies . we thus suggest that the red , or optically passive , spirals in cluster environments discussed by various authors are the same phenomenon as our semi - passive dusty red galaxies . the differentiating criterion for separating them from canonical red - sequence galaxies was the spiral morphology in @xcite and despite similar @xmath94 colour the signature of dust and younger ages in the sed in wgm05 . in some more star - forming and less obscured cases [ oii ] lines could be strong enough to appear above detection limits @xcite ; @xcite also used mir colours to trace star formation in red galaxies . in this paper , the morphology , the spiral - typical dust extinction , the brighter uv luminosity and the substantial ir fluxes all came together to form a consistent picture of semi - passive spirals . finally , we believe we can consider the sf - truncated virgo spirals to be local counterparts to the dusty red galaxies as their properties match at least in global respects : @xcite find the s - sfr in their virgo spirals to be on average @xmath95 reduced compared to the field , with an overall range from 10-fold reduction to slight enhancement ; this echoes the s - sfr properties of our dusty red galaxies . also , the star formation in the virgo spirals is on average more obscured , just as in dusty red galaxies again ; in virgo this is a consequence of the less obscured star formation in the outer disk being much reduced while the more obscured star formation in the central parts remains normal or is even enhanced . their appearance in large numbers suggests that spirals in many clusters undergo only slow quenching of their star formation , and that the time scale of morphological evolution must be longer than that of the spectral evolution . all this applies only in the mass regime of @xmath4 $ ] , where the sfr decline from blue galaxies in the field to dusty red galaxies in the infall region is not accompanied by morphological changes . morphology seems to change only further into the cluster alongside further sf decline that replaces dusty red galaxies by old red ones . at lower masses , in contrast , sfr decline and morphological changes appear more synchronised , and virtually no red spirals are found , except for a near - constant fraction of @xmath96% that may result from high - inclination spirals following random orientation statistics . this situation appears to be similar in the virgo cluster where @xcite find dwarf galaxies ( defined by @xmath97 ) to be either star - forming or quiescent with no intermediate objects being found . already kennicutt ( 1983 ) observed that at fixed hubble type virgo galaxies were redder and less star - forming than their cousins in the field . @xcite confirmed the colour difference for further clusters . however , whether these trends are due to nature ( hidden in the past ) or due to nurture ( change during cluster infall ) could not be identified from snapshot pictures of individual galaxies or clusters , because a snapshot will show the effects of infall and cosmic evolution of the progenitors intertwined . if we can not see the histories of the individual galaxies , we have no means to assess whether the progenitors of present - day cluster galaxies looked like the present - day infall galaxies do . only if this is the case , could we map the density axis onto a time axis and interpret the environmental trends as evolutionary tracks . however , there is strong evidence , e.g. , that higher - redshift clusters have fewer s0s and more spirals @xcite , so for s0s the link is established that a build - up in time corresponds to a build - up in density . higher - redshift cluster cores have similarities with lower - redshift cluster outskirts giving rise to a double downsizing picture where star formation moves progressively to lower - mass objects as you move either forward in cosmic time or higher in density of the environment @xcite . our results for a901/2 suggest , that empirically there are three mass domains with different behaviour along density : * at high mass ( @xmath98 , but certainly at @xmath99 ) only little change is observed with density , as most galaxies are red and spheroidal in all environments . * at intermediate mass ( @xmath100 $ ] ) we see a sequence with density whereby blue spirals in the field are replaced by red spirals in the infall region , which are then replaced by s0s in the core . * at low mass ( @xmath101 ) we see a sequence from blue field spirals and irregulars directly to red cluster spheroids . let us assume that the density axis can be mapped on a time arrow in the evolution of infalling galaxies , and the environmental trends are mostly due to nurture . an average galaxy falling in at 600 km / sec ( the velocity dispersion of a901a / b dusty red galaxies listed , see wgm05 ) needs @xmath102 gyr to fall from the virial radius to the core assuming a plunging orbit . there is thus no need to assume that they have already passed through the core . the results then say that star formation is suppressed at all densities and masses , and already between the field and the outermost infall region in our data , i.e. outside the virial radius of the cluster . this agrees with observations of galaxy pre - processing in infalling groups ( e.g. * ? ? ? * ; * ? ? ? morphological transformation appears in step with the sfr decline in low - mass galaxies , but is delayed for intermediate - mass galaxies . the delay gives rise to spiral galaxies in the infall region that appear red due to a combination of moderate extinction and reduced star formation . they appear broadly less clumpy and have less well - defined spiral arms than their blue cousins in the field , and they are the only progenitor candidates of similar - mass s0 galaxies . the steeper s - sfr decline for massive galaxies demonstrates that they transform slowly and remain visible in transformation for long . in contrast , the near - absence of a trend in mean s - sfr at lower mass requires more abrupt conversions that leave little chance to observe transforming objects , while giving rise to the steep trends in the sf / non - sf composition . we roughly find ( with large errors ) that the red fraction of spirals is increased in the infall region by perhaps 10% over the field value at a mass of @xmath103 and by maybe 50% at @xmath104 . at equal infall velocity this would imply @xmath105 longer quenching time - scales at @xmath105 higher mass or @xmath106 . the time - scale for ram - pressure stripping though is @xmath107 assuming constant surface mass density of disk galaxies @xcite and equal infall velocity . while we do not have the data to be very quantitative here , we argue that at least the quenching time - scale is unlikely to decline with mass . @xcite model the evolution of dwarf galaxies in virgo ( with @xmath108 ) assuming interaction with the intracluster medium , and derive sf quenching time scales of @xmath109 myr . such a rapid sf truncation would be consistent with the absence of a noticeable transition population . also , @xcite see k+a galaxies in the coma cluster only fainter than @xmath110 : the k+a signature requires a sharp truncation in star formation , which argues in favour of a faster shutdown of star formation at lower mass . in higher - mass spirals , gas stripping by hot icm would first remove gas from the outer disks where the gas column densities are lower , and the slow truncation of star - forming disks seen by @xcite would leave morphology largely intact over a slow sf quenching process , and would hence not lead to a k+a signature . it then may remain as puzzling that @xcite found many k+a spirals of normal mass in clusters at @xmath88 . however , the transformation time scale must depend not only on galaxy mass but also on the icm properties . already @xcite found higher s0 fractions with increasing x - ray luminosity of the clusters , although the presence of a residual relation at fixed cluster mass is unclear ( see discussion in * ? ? ? * ) . however , the basic effect has been studied by comparing two @xmath111 clusters with very different icm properties representing a low - mass ( cl0024 ) and high - mass ( ms0451 ) prototype in cluster evolution @xcite . low - mass clusters are irregular with several sub - clumps , perhaps young and still assembling , and have little volume occupied by hot icm ( e.g. virgo and a901/2 ) ; higher - mass clusters are more regular in structure and have a large volume occupied by hot icm ( e.g. coma ) . the galaxies in the two clusters compared by @xcite are on two different tracks of s - sfr - vs .- age as seen from [ oii]-vs.-@xmath112 , showing that ram pressure from the well - developed icm in ms0451 shuts down star formation rapidly even in larger galaxies , while the less developed icm in cl0024 means the transition is a lot slower and may even be driven predominantly by factors other than stripping by the icm , so that many transitioning red spirals with low s - sfr are seen . in line with this picture , we already see variation between our four cluster cores : the most x - ray - luminous core a901b has not only the most prominent icm but also the lowest red spiral and the highest s0 fraction . the discussion whether the observed s0 galaxies have evolved from spiral progenitors has continued for three decades now @xcite . we suggest to tackle the question with a mass - resolved approach , motivated by the strong trends seen at intermediate mass in our cluster . several authors ( e.g * ? ? ? * ) have argued that present - day s0s are too luminous to be made from present - day infalling spirals . in a901/2 we admittedly find s0 galaxies , but no infalling spirals at @xmath46 ; we suggest that these would have formed in the more distant past of this cluster from more massive progenitors that we do not observe in the present snapshot due to downsizing . we have no evidence against this picture , and note the absence of massive star - forming galaxies that could further enrich the massive s0 sample in a901/2 in the near future . in contrast , at @xmath113 $ ] we find a very strong trend of red spirals disappearing towards the core and s0s appearing in comparable numbers ; the s0 concentration towards the cores of a901/2 is markedly different from the flat s0 fraction within a virial radius seen in ensembles of more evolved clusters @xcite . this is consistent with a901/2 being a young clusters with weaker icm , and it suggests that a901/2 has much progenitor material to form more intermediate - mass s0s over the next few gyrs to fill the virial volume more evenly . a further obstacle in the understanding of s0 formation is the difference in the structural properties of s0s compared to their progenitor spirals : s0s are claimed to have larger concentrations or bulges and thicker disks than their progenitors , although recent works claim a smaller degree of difference @xcite . we are not aware of a mass - resolved study of progenitors and s0s that evaluates the effects of gradual gas stripping and continuing centrally concentrated star formation . @xcite studied residual morphological trends in the sdss data after eliminating the primary trend of the dominant colour - density relation and found only a weak trend towards higher concentration with density , and applying only to more luminous galaxies . this would be entirely consistent with a picture where low - mass galaxies experience more abrupt sf shutoff in dense environments , while high - mass galaxies continue forming stars in their inner regions . however , @xcite find no change in concentration with density at @xmath98 , and only a weak trend below . we have investigated the properties of galaxies , especially those of optically passive spirals , in a cluster complex at redshift @xmath0 , which consists of four cores ( a901a , a901b , a902 and sw group ) . we have used the stages dataset published by @xcite , exploiting the restframe near - uv - optical seds , ir 24@xmath1 m data and hst morphologies . the cluster sample is defined on the basis of combo-17 redshifts with an rms precision of @xmath2 km / sec ( at @xmath114 ) . we draw the following conclusions : 1 . optically passive spirals in clusters , dusty red galaxies in a901/2 and virgo spirals with truncated star - formation disks @xcite appear to be basically the same phenomenon . however , these objects are not truly passive galaxies despite their integrated optical appearance from large distances . 2 . they form stars at a substantial rate but reduced in comparison to field spirals ( on average @xmath3 lower sfr at fixed mass in our sample ) . however , their star formation is more obscured and its optical signatures are weak . they appear predominantly in the mass range @xmath4 $ ] where they constitute over half of the star - forming galaxies in the cluster , and thus form an important transition population . we find that the mean s - sfr of star - forming galaxies in the cluster is clearly lower than in the field , in contrast to the s - sfr properties of blue galaxies alone , which appear unchanged . the dusty red spirals are thus a vital ingredient for understanding the overall picture of sf quenching in clusters . 5 . at lower mass the dusty red sample consists mostly of edge - on spirals , which are a small fraction of all spirals with no environmental dependence . there is limited room for a cluster - specific contribution . physically , it seems that star formation quenching is fast in low - mass galaxies and accompanied by morphological change ; hence , no cluster - specific red spiral phenomenon is observed at @xmath5 . at larger masses , sf quenching is a slower process and strong morphological transformation is even more delayed , thus giving rise to abundant red spirals . 7 . the currently observed red spirals are expected to turn into s0s with time when their star formation is terminated by hot icm , although they may end up with different detailed properties from those s0s that are already in place . we thank the anonymous referee for a very careful reading of the paper and very helpful suggestions improving the manuscript . we also thank richard s. ellis , benjamin d. johnson , tommaso treu for comments on this work . support for stages was provided by nasa through go-10395 from stsci operated by aura under nas5 - 26555 . meg and cw were supported by stfc advanced fellowships . ch acknowledges the support of a european commission programme 6th framework marie cure outgoing international fellowship under contract moif - ct- 2006 - 21891 . cyp was supported by the nrc - hia plaskett fellowship , and the stsci institute / giacconi fellowship . efb , ag and kj are greatful for support from the dfg s emmy noether programme of the deutsche forschungsgemeinschaft , ab by dlr grant no . 50 or 0404 , dhm by nasa under ltsa grant nag5 - 13102 , mb and evk by the austrian science foundation fwf under grant p18416 , sfs by the spanish mec grants aya2005 - 09413-c02 - 02 and the pai of the junta de andaluca as research group fqm322 , sj by nasa under ltsa grant nag5 - 13063 , go-10861 , g0 - 11082 , and by nsf under ast-0607748 . balogh , m. l. , navarro , j. f. & morris , s. l. , 2000 , apj , 540 , 113 balogh , m. l. , eke , v. , miller , c. , lewis , i. et al , 2004 , mnras , 348 , 1355 balogh , m. l. , baldry , i. k. , nichol , r. , miller , c. , bower , r. , glazebrook , k. , 2004 , apj , 615 , l101 bamford , s. p.,milvang - jensen , b. & aragn - salamanca , a. , 2007 , mnras , 378 , l6 bamford , s. p. , nichol , r. et al , 2008 , mnras , submitted ( astro - ph/0805.2612 ) barden , m. , rix , h .- w . , somerville , r. s. , et al . , 2005 , apj , 635 , 959 barnes , j. e. , 1992 , apj , 393 , 484 bekki , k. , 1999 , apj , 510 , l15 bell , e. f. , mcintosh , d. h. , barden , m. , wolf , c. et al . , 2004 , apj , 600 , l11 bell , e. f. , zheng , x. z. , papovich , c. , borch , a. , wolf , c. & meisenheimer , k. , 2007 , apj , 663 , 834 benson , a. j. , bower , r. g. , frenk , c. s. , lacey , c. g. , baugh , c. m. , cole , s. , 2003 , apj , 599 , 38 blanton , m. r. , eisenstein , d. , hogg , d. w. , schlegel , d. j. , brinkmann , j. , 2005 , apj , 629 , 143 boselli , a. , boissier , s. , cortese , l. & gavazzi , g. , 2008 , apj , 674 , 742 butcher , h. & oemler , a. , jr . , 1985 , apjs , 57 , 665 bundy , k. , ellis , r. s. , conselice , c. j. , taylor , j. e. et al , 2006 , apj , 651 , 120 burstein , d. , ho , l. c. , huchra , j. p. & macri , l. m. , 2005 , apj , 621 , 246 calzetti , d. , 2001 , pasp , 113 , 1449 calzetti , d. , kinney , a. & storchi - bergmann , t. , 1994 , apj , 429 , 582 cassata , p. , guzzo , l. , frnceschine , a. , scoville , n. , et al . , 2007 , apjs , 172 , 270 cooper , m. c. , newman , j. a. , weiner , b. j. , yan , r. , willmer , c. n. a. et al . , 2008 , mnras , 383 , 1058 croton , d. j. , springel , v. , white , s. d. m. et al , 2006 , mnras , 365 , 11 christlein , d. & zabludoff , a. i. , 2004 , apj , 616 , 192 christlein , d. & zabludoff , a. i. , 2005 , apj , 621 , 201 dressler , a. , 1980 , apj , 236 , 351 dressler , a. , oemler , a. , couch , w. j. , smail , i. et al . 1997 , apj , 490 , 577 elbaz , d. , daddi , e. , leborgne , e. , dickinson , m. et al , 2007 , a&a , 468 , 33 franzetti , p. , scodeggio , m. , garilli , b. , vergani , d. et al . , 2007 , a&a , 465 , 711 gallazzi , a. , bell , e. f. , wolf , c. et al , 2008 , apj , in press giallongo , e. , salimbeni , s. , menci , n. , zamorani , g. et al . 2005 , apj , 622 , 116 gilmour , r. , gray , m. e. , almaini , o. , best , p. , wolf , c. , meisenheimer , k. , papovich , c. & bell , e. f. , 2007 , mnras , 380 , 1467 gomez , p. l. , nichol , r. c. , miller , c. j. et al , 2003 , apj , 584 , 210 gray , m. e. , wolf , c. , barden , m. , peng , c. y. et al , 2008 , mnras , in press ( this issue ) gunn , j. e. & gott , j. r. , 1972 , apj , 176 , 1 guzzo , l. , cassata , p. , finoguenov , a. , massey , r. , scoville , n. z. et al , 2007 , apjs , 172 , 254 haines , c. p. , fargiulo , a. , la barbera , f. , mercurio , a. , merluzzi , p. & busarello , g. , 2007 , mnras , 381 , 7 haines , c. p. , gargiulo , a. & merluzzi , p. , 2008 , mnras , 385 , 1201 heiderman , a. , jogee , s. , marinova , i. , van kampen , e. et al . , 2008 , apj , submitted heymans , c. , gray , m. e. , peng , c. y. , van waerbeke , l. et al . , 2008 , mnras , 385 , 1431 kauffmann , g. , heckman , t. m. , white , s. d. m. et al . , 2003 , mnras , 341 , 54 kauffmann , g. , white , s. d. m. , heckman , t. m. et al . , 2004 , mnras , 353 , 713 koopman , r. a. & kenney , j. d. p. , 1998 , apj , 497 , l75 koopman , r. a. & kenney , j. d. p. , 2004 , apj , 613 , 851 koopman , r. a. & kenney , j. d. p. , 2004 , apj , 613 , 866 kenney , j. d. p. , wong , o. i. et al . , 2008 , submitted ( astro - ph/0803.2532 ) kennicutt , r. c. , 1983 , aj , 88 , 483 kennicutt , r. c. , 1998 , ara&a , 36 , 189 khochfar , s. & ostriker , j. , 2008 , apj , 680 , 54 kodama , t. , smail , i. , nakata , f. , okamura , s. & bower , r. g. , 2001 , apj , 562 , l9 lane , k. p. , gray , m. e. , aragn - salamanca , a. , wolf , c. & meisenheimer , 2007 , mnras , 378 , 716 larson , r. b. , tinsley , b. m. & caldwell , c. n. , 1980 , apj , 237 , 692 laurikainen , e. , salo , h. , buta , r. & knapen , j. h. , 2007 , mnras , 381 , 401 lewis , i. , balogh , m. , de propris , r. , couch , w. et al . , 2002 , mnras , 334 , 673 marinova , i. , jogee , s. , heiderman , a. et al . , 2008 , apj , submitted mehlert , d. , thomas , d. , et al . , 2003 , a&a , 407 , 423 moore , b. , katz , n. , lake , g. , dressler , a. & oemler , a. , 1996 , nature , 379 , 613 moran , s. m. , ellis , r. s. , treu , t. et al , 2007 , apj , 671 , 1503 moss , c. , 2006 , mnras , 373 , 167 moss , c. & whittle , m. , 2000 , mnras , 317 , 667 noeske , k. g. , weiner , b. j. , faber , s. m. , papovich , c. et al . , 2007 , apj , 660 , l43 park , c. , choi , y .- y . , vogeley , m. s. , gott , r. ( iii ) & blanton , m. r. , 2007 , apj , 658 , 898 park , c. , gott , r. & choi , y .- y . , 2008 , apj , 674 , 784 pei , y. c. , 1992 , apj , 395 , 130 peng , c. y. , ho , l. c. , impey , c. d. & rix , h .- w . , 2002 , aj , 124 , 266 poggianti , b. m. , smail , i. , dressler , a. , couch , w. j. et al . , 1999 , apj , 518 , 576 poggianti , b. m. , bridges , t. j. , komiyama , y. , yagi , m. et al . , 2004 , apj , 601 , 197 poggianti , b. m. , desai , v. , finn , r. et al . , 2008 , apj , 684 , 888 postman , m. , franx , m. , cross , n. j. g. , holden , b. et al . , 2005 , apj , 623 , 721 rines , k. & geller , m. j. , 2008 , aj , 135 , 1837 sabatini , s. , davies , j. , scaramella , r. , smith , r. et al . , 2003 , mnras , 341 , 981 salim , s. , rich , r. m. , charlot , s. et al . , 2007 , apjs , 173 , 267 salpeter , e. e. 1955 , apj , 121 , 161 sandage , a. , binggeli , b. & tammann , g. a. , 1985 , aj , 90 , 1759 schawinksi , k. , thomas , d. , et al . , 2007 , mnras , 382 , 1415 schiminovich , d. , wyder , t. k. , martin , d. c. , et al . , 2007 , apjs , 173 , 315 smith , g. p. , treu , t. , ellis , r. s. , moran , s. m. & dressler , a. , 2005 , apj , 620 , 78 tanaka , m. , kodama , t. et al . , 2005 , mnras , 362 , 268 thomas , d. , maraston , c. , bender , r. & mendes de oliveira , c. , 2005 , apj , 621 , 673 trager , s. c. , faber , s. m. & dressler , a. , 2008 , mnras , 386 , 715 tremonti , c. a. , moustakas , j. & diamond - stanic , a. m. , 2007 , apj , 663 , l77 van den bergh , s. , 1976 , apj , 206 , 883 van der wel , a. , 2008 , apj , 675 , l13 verdugo , m. , ziegler , b. l. & gerken , b. , 2008 , a&a , 486 , 9 wilman , d. j. , balogh , m. l. et al . , 2005 , mnras , 358 , 71 wilman , d. j. , pierini , d. , tyler , k. et al . , 2008 , apj , 680 , 1009 weinmann , s. m. , van den bosch , f. c. , yang , x. & mo , h. j. , 2006 , mnras , 366 , 2 weinzirl , t. , jogee , s. , khochfar , s. , burkert , a. & kormendy , j. , 2008 , apj , submitted ( astro - ph/0807.0040 ) wolf , c. , meisenheimer , k. , et al . , 2003 , a&a , 401 , 73 wolf , c. , meisenheimer , k. , kleinheinrich , m. et al . , 2004 , a&a , 421 , 913 wolf , c. , gray , m. , & meisenheimer , 2005 , a&a , 443 , 435 wolf , c. , gray , m. e. , aragn - salamanca , a. , lane , k. p. & meisenheimer , 2007 , mnras , 376 , l1 wolf , c. , hildebrandt , h. , taylor , e. n. & meisenheimer , k. , 2008 , a&a , in press xu , c. k. , sun , y. c. & he , x. t. , 2004 , apj , 603 , l73

we present a complete atlas of the cygnus loop supernova remnant in the light of 3 ( ) , , and 2 ( ) . we include low - resolution ( ) global maps and smaller fields at resolution from observations using the prime focus corrector on the 0.8-m telescope at mcdonald observatory . despite its shell - like appearance , the cygnus loop is not a current example of a sedov - taylor blast wave . rather , the optical emission traces interactions of the supernova blast wave with clumps of gas . the surrounding interstellar medium forms the walls of a cavity through which the blast wave now propagates , including a nearly complete shell in which non - radiative filaments are detected . we identify non - radiative shocks around half the perimeter of the cygnus loop , and they trace a circle of radius ( 19 pc ) in the spherical cavity walls . the cygnus loop blast wave is not breaking out of a dense cloud , but is instead running into confining walls . modification of the shock velocity and gas temperature due to interaction of the blast wave with the surrounding medium introduces errors in estimates of the age of this supernova remnant . the optical emission of radiative shocks arises only where the blast wave encounters inhomogeneities in the ambient medium ; it is not a consequence of gradual evolution to a global radiative phase . distance measurements that rely on this uniform blast wave evolution are uncertain , but the radiative shocks can be used as distance indicators because of the spherical symmetry of the surrounding medium . the interstellar medium dominates not only the appearance of the cygnus loop but also the continued evolution of the blast wave . if this is a typical example of a supernova remnant , then global models of the interstellar medium must account for such significant blast wave deceleration . 3 [ ] 2 [ ] # 1#2 # 1#2#3

astro-ph9805008

supernova remnants greatly determine the large - scale structure of the interstellar medium . the energy of supernova remnants heats and ionizes the interstellar medium ( ism ) , and their blast waves govern mass exchange between various phases of the ism . in doing so , supernova remnants ( snrs ) influence subsequent star formation and the recycling of heavy elements in galaxies . global models of the interstellar medium that include a hot ionized component ( @xcite ; @xcite ) are sensitive to the supernova rate , the persistence of their remnants , and the sizes they attain . a simple calculation of the last of these assumes that the blast wave expands adiabatically in a uniform medium once the blast wave has swept up mass comparable to the mass of the ejecta . during this sedov - taylor phase , the radius of the snr as a function of @xmath5 , the initial energy in units of @xmath6 erg , @xmath7 , the ambient number density in units of @xmath8 , and @xmath9 , time in units of @xmath10 yr , is @xmath11 in a medium where the mean mass per particle is @xmath12 . this phase will last until radiative losses become important . the beginning of this subsequent phase , marked by the initial loss of pressure behind the blast wave , occurs at @xmath13 , when the radius is @xmath14 ( @xcite ) , although the radiating shell is not fully formed yet . we approach these large - scale questions with analysis of complete images of a particular supernova remnant , the cygnus loop , in three optical emission lines . this supernova remnant appears to be a limb - brightened shell at radio ( @xcite ) , infrared ( @xcite ) , optical ( @xcite ) , and x - ray ( @xcite ; @xcite ) energies , which at first glance suggests that it is presently in the transition to the radiative stage . the cygnus loop has the advantages of being nearby , bright , and relatively unobscured by dust . this allows us to examine in detail the evolution of various portions of the shock front and to determine physical parameters , such as shock velocity and local ambient density , as they vary throughout the remnant . despite its appearance , the cygnus loop is not a current example of blast wave propagation in a uniform medium at any stage . instead , its evolution is governed by the inhomogeneous interstellar medium , which we map using the shock as a probe . many of the features we discuss have been noted by others . oort ( 1946 ) first suggested that the cygnus loop is an expanding supernova shell . spectroscopy of radiative shocks in selected locations ( e.g. , @xcite , @xcite , and fesen et al . 1982 ) combined with theoretical models of these shocks ( e.g. , @xcite , @xcite , @xcite , and @xcite ) has been used to derive the physical conditions of the observed shocks . we utilize the radial velocity measures of minkowski ( 1958 ) , kirshner & taylor ( 1976 ) , greidanus & strom ( 1991 ) , and shull & hippelein ( 1991 ) to discern some of the three - dimensional structure that is ambiguous from the data we present . many non - radiative or balmer - dominated shocks in the cygnus loop have been identified ( e.g. , @xcite , @xcite , @xcite , fesen et al . 1982 , @xcite , and @xcite ) . our observations qualitatively match these , and we rely on these works and others ( @xcite ; @xcite ; @xcite ) for quantitative measures of parameters such as shock velocity and preshock density . mccray & snow ( 1979 ) and charles , kahn , & mckee ( 1985 ) have suggested that the cygnus loop is the result of a cavity explosion , and we adapt this global model to interpret the surrounding interstellar medium , as well . this paper is a companion to the soft x - ray survey presently in progress with the _ rosat _ high resolution imager ( @xcite ; @xcite ) . with these two surveys , we examine the cygnus loop as a whole , not restricting our investigation only to those regions that are exceptionally bright or that appear to be particularly interesting . we hope to understand both the global processes and the specific variations that are responsible for the emission we detect . the x - rays probe hot ( temperature @xmath15 k ) gas that shocks with velocities @xmath16 heat . the optical emission is expected from slower shocks ( @xmath17 ) in which the post - shock region cools to temperatures @xmath18 k , yet the most prominent regions at optical wavelengths are also bright in x - rays . mckee & cowie ( 1975 ) suggested that the broad correlation of x - ray and optical emission is the result of a blast wave propagating in an inhomogeneous medium . in this scenario , the shock is significantly decelerated in dense clumps of gas , while portions of it proceed unimpeded through the lower - density intercloud medium . we apply the principles of this basic cloud blast - wave interaction to a range of locations in the cygnus loop . in particular , we refine the cavity model introduced in levenson et al . ( 1997 ) , using these optical data to constrain the current ism in the vicinity of the cygnus loop and to determine how the stellar progenitor modified it in the past . we present the observations in 2 . we describe them in detail , noting individual regions of interest , and we use these data to measure the physical conditions of the blast wave and the ambient medium in particular locations in 3 . the purpose of the detailed examination is to combine the results in a complete map of the surrounding ism . we present this three - dimensional model while providing a coherent explanation of the history that accounts for it in 4 . we predict the future of this snr and relate its fate to more general theories of supernova modification of the interstellar medium in 5 and summarize our conclusions in 6 . we obtained narrow - band images of the cygnus loop in 1995 august and september and 1995 september on the 0.8-m telescope at mcdonald observatory with the prime focus corrector ( pfc ; @xcite ) . the pfc system is a 5-element catadioptric corrector located at the f/3 prime focus of the telescope and projects a flat focal plane onto a loral - fairchild @xmath19 , @xmath20-pixel charge - coupled device . the camera and corrector system has a @xmath21 spot size in the v - band and projects each pixel to @xmath22 on the sky , producing a @xmath23 field of view . the interference filters each have a rectangular transmission profile and full width at half maximum @xmath24 in the converging beam of the pfc . the filter centers in the f/3 beam are 5007 , 6723 , and 6563 to detect line emission from 3 ( @xmath25 ) , 2 ( @xmath26 ) , and + [ ] , which we refer to throughout as . each individual integration of approximately @xmath27 s duration was bias - subtracted , flat - fielded , and scaled by its median sky value before being combined with other images . every region of the cygnus loop was observed at least twice through each filter in order to remove cosmic rays and exclude bad regions of the detector , and most regions were observed three times in . we obtained a total of 62 observations in 3 , 89 in , and 54 and 2 . over the large spatial scale of the cygnus loop ( @xmath28 ) , each observation is a distinct projection of the sky onto the plane of the detector , so images must be remapped to a common projection before being co - added . an astrometric solution for each integration was calculated based on the _ hubble space telescope _ guide star catalog . alignment does not rely on overlapping observations of a particular region but depends instead on the astrometry of individual integrations . the final images are the result of pixel - by - pixel averaging in the combined image plane using nearest - neighbor matching and rejecting bad regions of the detector . the global maps of the cygnus loop with @xmath2 pixels are shown with linear scaling in , 3 , and 2 , and in a false - color combination with logarithmic scaling ( figure [ figwhole ] ) . while the observations at different wavelengths resemble each other broadly , their variation is apparent in the combined image . in this display , blends of and 2 alone are yellow , 3 and 2 together appear as magenta , and and 3 combine to make cyan . only the regions that are displayed as white exhibit strong emission at all three wavelengths . in addition to the global maps , we present 5 smaller fields , each @xmath29 across , at @xmath3 resolution ( figs . [ figsubv][figsubz ] ) . although we first describe the current physical conditions in specific regions of interest , our ultimate goal is to combine the understanding of the particular sites into a complete account of the surrounding ism , its modification by the progenitor star , and the passage of the blast wave through it . the radiative shocks that are responsible for most of the optical emission from the cygnus loop result in two characteristic morphologies : filaments and diffuse emission . figure [ figann ] identifies examples of these features and other structures to which we will refer by name . the morphological distinction is the result of different viewing geometry . as the `` wavy sheet '' model ( hester et al . 1983 ; @xcite ) describes , bright , sharp filaments are the result of long lines of sight through tangencies of the shock front . when the shock front is viewed face - on , the emission appears more diffuse . combined with the x - ray data , these observations provide more specific information on the evolutionary state of the cygnus loop and the distribution of inhomogeneities of the ism around it . the widespread x - ray emission indicates that the global shock velocity is several hundred , so the optical emission comes from regions of denser gas that have decelerated the blast wave . the correlation of the x - rays with optically - bright regions make clear that this snr is not globally in the gradual transition to the radiative phase that is the result of blast wave propagation in a uniform medium . the blast wave has recently and suddenly decelerated where radiative shocks are detected . this rapid , environmentally - imposed evolution is distinct from the simple model prediction for incremental change in a homogeneous medium . these complete data sets demonstrate that many characteristics that have been investigated in small regions of the cygnus loop are widespread , and we adapt the physical explanations of these limited regions in previous work to understand the cygnus loop globally . for example , the bright emission from radiative shocks is found to be stratified at select observed locations ( @xcite ) , and we show this stratification to be typical of all prominent radiative shocks . the 3 extends toward the projected exterior , and the and 2 are located toward the projected interior . the 2 is farthest to the interior and is particularly diffuse . the stratification is a function of the column density that the shock has swept up . the incomplete 3 emission is concentrated behind shocks that have encountered relatively little gas , while where the shock lags because it has swept up a greater column of gas , and then 2 are prominent . the strong 3 emission requires that the shock has progressed through a column density @xmath30 , while the complete recombination zone develops after the swept - up column @xmath31 for @xmath32 ( @xcite ) . the apparent thickness of these zones and the extensive regions of bright emission in a combination of all the observed lines is a consequence of projection effects . the curved surface of the blast wave encounters the cloud at different projected radii , so some lines of sight intersect multiple stages of shock evolution . around most of the perimeter of the cygnus loop , filaments are detected in alone . these balmer - dominated filaments are the result of non - radiative shocks ( @xcite ; @xcite ) . unlike the more common radiative shocks , the emission here does not come from the extended cooling zone behind the shock . instead , it is the result of collisional ionization before excitation in the immediate post - shock region , or the downward cascade toward the ground state following charge exchange . in either case , neutral atoms are required in the post - shock region . because these do not survive long , the emission originates at the shock front . thus , we use the balmer filaments to trace the shock front directly . this emission is intrinsically weak , so it is most obvious when the line of sight through the shock front is long . this is the case at the projected edge of the snr , where the line of sight is tangent to the spherical blast wave . the projected edges of the cygnus loop are the simplest regions to understand because the geometric complications are minimized there . the observed structures may be slightly but not significantly in the foreground or background and still appear near the projected edge . the prominent regions , ngc 6992 at the northeast ( @xmath33 ; fig . [ figsubv ] ) and ngc 6960 at the west ( @xmath34 ; fig . [ figsubw ] ) , are examples of interactions of the blast wave with large clouds ( @xcite ; hester et al . 1994 ; @xcite ; @xcite ) , and the corresponding x - ray enhancement arises from reflected shocks that propagate through the hot , shocked snr interior . the observed length scales of optical and x - ray emission from ngc 6992 and ngc 6960 reveal the sizes of these large clouds . at the 770 pc distance of the cygnus loop , the coherent networks of filaments over @xmath35 imply that the clouds have lengths of about @xmath36 cm . balmer filaments , high - resolution x - ray observations or both constrain the location of clouds along the line of sight . x - ray emission extends beyond both of these bright optical edges , where some portion of the three - dimensional blast wave is not yet impeded by the clouds . thus , the clouds do not extend over the complete edge of the blast wave along the line of sight . in contrast , the southeast knot at @xmath37 ( fesen et al . 1992 ; @xcite ) is an example of a cloud that is clearly extended along the line of sight but is not necessarily large across the plane of the sky . the nearly - circular edge of the x - ray snr and the balmer filaments are distinctly concave at the southeast knot ( fig . [ figsubx]a ) , and there is no x - ray emission to the exterior . thus , this cloud must be at least 11 pc long along the line of sight in order to impede the entire projected edge of the blast wave at this location . again , the presence of the complete optical cooling structure indicates that the shock has progressed through a column @xmath38 here . the breakout to the south , centered on @xmath39 and extending approximately @xmath40 across ( fig . [ figsubz ] ) , is the most significant departure from circularity in the remnant . the greater advance of the blast wave in the breakout compared with the circular loop north of @xmath41 indicates that the ambient density is lower in this southern region . balmer - dominated filaments surround much of the breakout , where the blast wave has now encountered atomic gas . there are sections of incomplete shocks to the east and west ( e.g. , @xmath42 and @xmath43 ) and fully radiative shocks at the east ( @xmath44 ) . emission is absent across the face of this area . in the plane of the sky , the breakout is not very extended . either the blast wave has only recently entered this low - density region , so it has not had enough time to advance significantly , or the extent of the low - density material is limited , so the blast wave no longer continues to progress rapidly . the stratification of emission around the edge of the breakout is apparent on smaller scales than those observed elsewhere in the cygnus loop , and the breakout exhibits smaller regions of common overlap . in the northeast , for comparison , the distinct regions of and 3 are typically @xmath45 wide , and they are observed together over @xmath46 scales , whereas in the breakout , the exclusive and common zones are both typically @xmath47 wide . these characteristics are consequences of the more limited scale of interactions with clouds in the breakout . there is no evidence for indentation or blast wave progress beyond the optical - emitting regions , which could be due to either intrinsically small clouds or early stages of encounter . in either case , at least some dense material surrounds the snr at its southern extremity . the exterior of the southwest quadrant exhibits many filaments that are nearly aligned with one another ( fig . [ figsuby ] ) . some of these are part of the radiative shock structure related to the southern end of ngc 6960 , while farther away from the center , these are non - radiative and incomplete shocks ( fesen et al . 1992 ) . along some of the filaments , the transition from non - radiative to incomplete shock structure is clear where 3 emission arises . the multiple filaments are easily understood as tangencies of a wavy sheet . these filaments are shorter than those at the north and east of the cygnus loop . this indicates that the ism is clumpier here , which would distort the blast wave in many places and account for the number of filaments that we observe . shock diffraction through low - density ( @xmath48 ) regions may be responsible for the large curvature of some of these filaments , as fesen et al . ( 1992 ) suggest , although multiple interfering shock fronts are not necessary to produce the different filaments . they arise from multiple projections of the same shock front . the balmer - dominated filaments around the periphery of the cygnus loop indicate that the fraction of atomic gas surrounding the snr is high . the extreme circularity of these filaments requires that the blast wave remain nearly spherical with little deformation at the projected edge . with the exclusion of the breakout region , we fit them with a circle of radius @xmath49 , which is equivalent to 19 pc at a distance of 770 pc . in figure [ figcirc ] , the model circle is drawn on the soft x - ray image obtained with the _ rosat _ high resolution imager ( @xcite ) . we calculate the correlation of radial variation as a function of angular separation and the corresponding fourier coefficients for the balmer filaments . excluding the breakout , the amplitude of the @xmath50 mode is less than 0.02 , which implies that the surroundings of the cygnus loop are homogeneous on large scales . the individual filaments tend to be very long , typically extending over more than @xmath51 . the continuity and smoothness of the balmer filaments indicate the uniformity of the medium in which these shocks propagate . for constant ram pressure @xmath52 behind the shock , variations in density and velocity are related @xmath53 . in terms of the radius , @xmath54 , and time , @xmath55 , @xmath56 , where @xmath57 is of order unity ; for free expansion @xmath58 , and for an adiabatic supernova remnant in a uniform medium , @xmath59 . thus , @xmath60 . the observed radial variations of a particular filament are @xmath61 . an approximate timescale of these shocks is 1000 yr , and the age of the cygnus loop is about 14,000 yr ( cf . [ secage ] ) , so the overall density variations in the present medium of the balmer - dominated filaments are of order @xmath62 . we expect @xmath63 , so @xmath64 . the balmer - dominated filaments define the current location of the blast wave and mark the presence of neutral material . detailed studies of particular locations at the northeast have used these non - radiative shocks as density probes ( @xcite ; @xcite ; hester et al . 1994 ) and derive densities @xmath65 . the total observed balmer line intensity is sensitive to density , shock velocity , and geometric perspective . all of the filaments around the periphery share the same edge - on geometry . because of the remnant s circularity , it is reasonable to assume that the shock velocity is nearly constant around the extreme projected edge except at the breakout . the filaments that have been studied in detail to determine shock velocity and ambient density are the brightest ones , however , so these regions are most likely in the transition to incomplete radiative emission , and these densities are expected to be higher than elsewhere . thus , we conclude that where typical non - radiative filaments are observed around the periphery , the density is constrained @xmath66 , and most likely @xmath67 . we roughly calculate the surface brightness of a non - radiative shock in the cygnus loop . we examine a portion of the blast wave in the southeast , near @xmath68 . this is not an exceptionally bright filament , and there is no obvious interaction with a large cloud here . the whole filament extends over more than @xmath35 , so the blast wave edge is smooth here . we estimate a surface brightness of @xmath69 from this filament . assuming that this line of sight passes through two folds of the spherical blast wave and the ism does not distort it , the intensity of a single shock surface is @xmath70 . the line intensity , @xmath71 , depends on the energy of the transition , @xmath72 , and is given by @xmath73 where @xmath74 is the preshock neutral density and the ratio of excitation and ionization rates @xmath75 for ( @xcite ) . if we reasonably constrain the shock velocity @xmath76 , these observations limit the density @xmath77 at the location of the non - radiative shock . the lower velocity limit is based on observations of balmer filaments ( @xcite ; @xcite ; hester et al . 1994 ) , which are likely to represent the slowest non - radiative shocks , as noted above , and the upper bound assumes that the forward shock does not heat the x - ray - emitting medium to temperatures greater than @xmath78 . the balmer filaments are related to adjacent x - ray emission : the optical filaments form the exterior boundary of low surface brightness , limb - brightened x - rays ( @xcite ) . this characteristic morphology reveals where the blast wave is decelerated in the cavity walls . we combine the density determined from the balmer observations with x - ray temperature measurements to calculate the unperturbed blast wave velocity and the original density of the cavity , and we compare this with the cavity density based on the observed x - ray surface brightness of the interior of the snr . a bimodal temperature distribution has been detected in the cygnus loop , with peaks at @xmath79 and @xmath80 k in exosat observations with the channel multiplier array ( sensitive over 0.052 kev ) and the medium energy experiment ( sensitive over 150 kev ; @xcite ) . we suggest that the higher temperature component is due to the reflected shock and that previous low single - temperature measures are contaminated by the slow , forward shock in the cavity walls . we model these simply as a reflected shock and ignore non - equilibrium effects , which are probably important . the physical parameters temperature , pressure , and density , are related by the equations of conservation of mass , momentum , and energy flux across shock fronts . assuming the temperature of @xmath81 corresponds to the decelerated shock in the cavity wall and @xmath82 corresponds to the reflected shock , the current blast wave velocity is around @xmath83 , and the original , unperturbed blast wave velocity was around @xmath84 . the reflected shock model also predicts that the original cavity density , @xmath85 , is related to the shell density @xmath86 . combined with the surface brightness measurement of the balmer filaments , this yields an atomic shell density @xmath87 and a cavity density @xmath88 if the shell is entirely neutral . as described below ( [ subsecmodel ] ) , the shell is partially ionized , so the original cavity density was @xmath89 . in addition to neglect of non - equilibrium effects , other sources of error include confusion of the reflected shock and the unperturbed shock in the cavity , uncertainty of the derived x - ray temperature , and uncertainty of the balmer intensity of a single shock surface derived from observation through the spherical edge of the blast wave . the predicted density is consistent with that derived from x - ray observations of the projected interior . in a featureless region of the southeast near @xmath90 , the median hri surface brightness is @xmath91 . we compare this with a model of a raymond - smith plasma at @xmath92 , correcting for absorbing column of @xmath93 . assuming the emitting region is 22 pc deep ( the predicted line of sight through the snr at this radius ) , the observed count rate corresponds to a density @xmath94 . this is the density of the shocked material ; the original cavity density would have been @xmath95 . one unusual feature at the east is a region of very smooth emission adjacent to the bright structure of ngc 6995 at @xmath96 . this area is obvious in and detectable in 2 but absent in 3 . there are no x - rays here ( @xcite ) , so a fast shock wave has not passed through . the x - ray map shows a clear indentation in this region , where dense gas has significantly impeded the blast wave over large scales , similar to the southeast cloud . there is no other direct evidence , such as a balmer filament , that the blast wave is presently located toward the exterior of this region . we propose that this gas is photoionized by a shock precursor , and it has not been shocked . radiation from the shock - heated gas of the interior of the cygnus loop serves as the ionization source . typically , when the shock velocity @xmath97 , much of the subsequent radiation is at uv wavelengths ( @xcite ; @xcite ; @xcite ) , which can ionize the surrounding gas , both ahead of and behind the shock front . the hardness of the ionizing spectrum increases with shock velocity , and the resulting emission from photoionized gas is a function of shock velocity and preshock density ( @xcite ; @xcite ; @xcite ) . the absence of 3 in the ionized precursor indicates that the shock has intermediate velocity ( @xmath98 ) , which is consistent with uv spectroscopy of blair et al . ( 1991 ) who find @xmath99 is typical in this region . the bright emission ahead of the shock front is probably a consequence of relatively high density ; the balmer luminosity of the precursor scales linearly with density in the steady case ( @xcite ; @xcite ; @xcite ) . a second important factor is that the shock is fully radiative , having been decelerated in the dense cloud , so the shock front does not move through the ionized region before the ions have time to radiate ( cf . finally , the column density ahead of the shock must be great for the precursor emission to be observable . the images do allow us to estimate crudely the density of the photoionized material . we assume that the length of the photoionized region along the line of sight is equal to its extent north - south in the plane of the sky and use the observed surface brightness to predict a density @xmath100 . spectral data would allow a more exact determination of the characteristics of the emitting region and quantitative measurements of the photoionizing source . the blast wave history causes the ambient medium to be photoionized here and not elsewhere . the penetrating x - rays of very fast shocks ( @xmath101 ) do not efficiently ionize the surrounding medium ; the optical depth to 200 ev photons @xmath102 . only the softer uv - dominated spectra of slower shocks are effective ionizing sources . in other regions of the cygnus loop , the recent shock velocity has been too great to produce much photoionization , and there has not been enough time for the ionizing field of the post - shock cooling and recombination zone to develop . thus , most of the surrounding medium has remained sufficiently neutral to exhibit balmer - dominated filaments . the exceptional photoionization at the eastern edge is consistent with the cavity model of this snr . the blast wave has not slowed sufficiently elsewhere for most of the shocked gas to be a useful source of ionizing photons . over nearly all of the cygnus loop , the deceleration has been recent , as the blast wave has run into the walls of the cavity . only in this region has the softer spectrum of a slower shock existed long enough to ionize the preshock medium so that its radiative transitions can be detected . photoionized regions have been observed around other snrs , including n132d ( @xcite ) and n49 ( @xcite ) in the large magellanic cloud , for example . one difference between the cygnus loop and the former snr , however , is that while in the cygnus loop the photoionized precursor is the exception , detected only in a limited region , in n132d it is the rule , occurring over the majority of the younger remnant s perimeter . another difference is that the exceptional photoionization ahead of the cygnus loop blast wave is adjacent to the region that contains its brightest soft x - ray knots ( @xcite ) , while the photoionization and x - ray surface brightness variations are not correlated in n132d ( @xcite ) . these differences are likely to be a consequence of the different sources of ionizing photons and the density of the surrounding medium . first , in n132d , the x - ray emitting shock can account for only half of the ionization ( @xcite ) , so another significant source must also contribute to the ionization in this case . the ubiquitous radiative shocks in clouds and o - rich filaments that morse et al . ( 1996 ) propose as this additional source are not preferentially associated with the x - ray emission . second , a large molecular cloud extends over the entire southern hemisphere of this snr ( @xcite ) . while the present blast wave is not in the dense core that is observed in co , the outer regions of the cloud are expected to be over - dense with respect to the typical ambient ism . the higher densities that make the ionizing precursor visible are therefore characteristic of the majority of the perimeter of n132d . in contrast , we suggest that it is the unusual conditions of extreme blast wave deceleration and related high density that result in both the bright x - ray emission and the photoionized precursor in this small area at the eastern edge of the cygnus loop . these conditions make the cygnus loop precursor similar to that of n49 . this middle - aged snr is associated with a small molecular cloud on its eastern limb ( @xcite ) . the blast wave interaction with the cloud results in local optical emission and x - ray enhancement , blast wave deceleration observed in a range of shock velocities ( @xcite ) , and a prominent photoionized precursor on the eastern limb of n49 . the interior of the cygnus loop contains diffuse emission and filaments . both morphologies share some characteristics , such as exclusive 3 emission ( e.g. , @xmath103 in diffuse regions and @xmath104 in filaments ) , and blends of and 2 ( e.g. , @xmath105 and @xmath105 ) . a combination of emission in all three lines is observed only in diffuse emission ( e.g. , @xmath106 ) , and alone is observed only in filaments ( e.g. , @xmath107 ) . projection effects are more severe in the interior , but these optical images and high - resolution spectroscopy ( @xcite ; @xcite ; @xcite ) together help unravel the tangled cygnus loop . in all cases , the optical emission is the result of the interaction of the three - dimensional blast wave surface with dense structures in the ism . the emitting regions are truly at the outside of the snr , but some appear projected to the interior . the brighter filaments tend to have lower absolute radial velocities and narrower velocity dispersions than the fainter diffuse emission ( @xcite ; @xcite ) . this is consistent with the model in which filaments are long lines of sight through the edge of a wavy shock front , and diffuse emission results when the blast wave is observed nearly face - on . the shock velocity is primarily perpendicular to the line of sight in the former case and along the line of sight in the latter case . in some instances , adjacent filamentary and diffuse emission are related , being tangential and normal views of the same portion of the blast wave . elsewhere , spatially separated shock regions are projected onto the same location in the observer s view , causing significant confusion . the extensive diffuse emission at the western interior near @xmath108@xmath109 and @xmath110@xmath111 suggests that a large wall of dense gas is present across the entire face of the snr ( figs . [ figsubw ] and [ figsuby ] ) . the irregularity of the diffuse emission is due to the clumpiness of the cloud . this emission is blueshifted with respect to nearby gas ( @xcite ; @xcite ) so it is on the front face . the bright filaments of the `` carrot '' ( at @xmath112 ) , having positive radial velocities , are on the rear face . the filaments at @xmath113 ( fig . [ figsubw ] ) are distinct on small scales , similar to the edges of the breakout , although here the region of common overlap is absent . the origin of the stratification is the same : the limited physical scale of the interaction with the intervening cloud , either because the cloud is intrinsically small or because the encounter is recent and the blast wave has not progressed far into the cloud . along a particular line of sight , only a single coherent section of the blast wave is detected . unlike the areas where strong emission is observed at all wavelengths , here there are not many projected overlapping regions of various stages of shock evolution . the filaments that are bright in and 2 illustrate an earlier section of the cloud shock than do the 3 filaments . in this area on the western half of the cygnus loop , there is no planar blast wave that actually moves to the east , however . rather , the _ interaction _ with a cloud is progressing from west to east as the blast wave surface expands radially . the intervening cloud must be oriented along the line of sight to some degree , as shown in figure [ figcartc ] . another region of diffuse emission and fainter filaments runs north - south across the center of the remnant , around @xmath114@xmath115@xmath116 ( figs . [ figsubx ] and [ figsuby ] ) . the high - resolution spectra in the vicinity are of the diffuse component alone , which includes very bright central knots . these clumps are blueshifted , so these clouds are on the near face . the westernmost of these filaments , at @xmath117 , is detected only in . the blast wave has run into an extended wall of gas , enabling us to look through a long edge of the shock front . these filaments are broader than most of the other pure balmer filaments , which is expected in this scenario . the broad filaments appear where the blast wave is viewed through a point of inflection , whereas the sharper , narrow ones are viewed at a tangency . for these non - radiative filaments to be observable , the wall of gas must be extremely smooth at the present projected location of the shock front . the wall of gas is likely to be related to the source of the rest of the optical emission at the center of the snr , not the clouds that are illuminated as the `` carrot , '' although this is uncertain . except in the single photoionized section , the optical emission traces dense shocked regions . using some knowledge of shock evolution and geometry , we map the surrounding ism . the basic principles applicable to these observations are : ( 1 ) optical emission due to radiative shocks requires that the blast wave has swept up a significant column density ( @xmath118 ) , which is likely in regions of increased density ( @xmath119 ) ; ( 2 ) sharp filaments are edge - on views of the blast wave surface , and diffuse emission is the result of face - on views of this surface ; ( 3 ) the optical line emission from a radiative shock is progressively dominated by 3 , , then 2 ; and ( 4 ) filaments of alone are the result of non - radiative shocks in material that has a significant neutral fraction . with these data , we construct a three - dimensional model of the interstellar medium around the cygnus loop . over much of its surface , the blast wave is propagating through large clouds . long filaments , detected in around the perimeter and in other lines across the interior , indicate that a spherical shell of atomic gas surrounds the snr . the dense clumps of gas and the atomic shell together form the walls of a cavity in which the supernova occurred . slices through the cygnus loop parallel to the plane of the sky from the near side to the far side ( fig . [ figcartmap ] ) show some of the three - dimensional structure of the ism before the supernova event . the center of the cygnus loop is at latitude @xmath120 . at the rear of the cavity , toward the galactic plane , is a large wall of gas that is detected in neutral hydrogen ( @xcite ) and in the iras sky survey ( @xcite ) . the cloud that causes the bright northeastern emission is also on the far side . clumps at the edge of the breakout and at the northwest are on the near side . some clouds , including the southeast knot and the molecular gas at the west , extend along the line of sight and appear in all slices . the cygnus loop is not the result of a supernova explosion in an arbitrary region of the interstellar medium . rather , the blast wave propagates through a medium that the progenitor star has processed . several characteristics imply a massive progenitor , the sort most capable of modifying the surrounding ism , which charles et al . ( 1985 ) and levenson et al . ( 1997 ) discuss . one strong piece of evidence is the paradoxical morphology of the cygnus loop . at optical and x - ray wavelengths , it appears to be nearly circular , yet it exhibits tremendous structure on smaller scales . the optical data alone could be interpreted as a snr in transition from the adiabatic to the radiative phase , but the presence and correlation of x - rays require that overall the blast wave still has ( or until very recently had ) a high velocity , @xmath121 . a model of blast wave interaction with large clouds accounts for this correlation ( @xcite ; @xcite ; @xcite ) , but such large clouds can not be typical of the pre - supernova interstellar medium . if they were , the snr would not appear so circular , and x - ray emission due to clouds evaporating in the interior would be observed . thus , although large clouds are found around the periphery of the cygnus loop , they were not located within the current blast wave prior to the supernova event . this structure is expected in the vicinity of an o or b star . during its main - sequence lifetime , such a massive star homogenizes the surrounding ism , creating a uniform region at constant density , temperature , and pressure . the star s uv emission removes clumps of denser gas within the h ii region , either destroying the clouds through photoevaporation or relocating them to the exterior by the rocket effect ( @xcite ) . the result is a homogeneous spherical region surrounded by dense clouds . a star of type earlier than b0 will clear a cavity with a radius of about 50 pc ( mckee et al . 1984 ) . the cygnus loop cavity is smaller , so its progenitor must be a later - type star ( @xcite ) , yet one still able to create the photoionized cavity . thus , we suggest that a b0 , @xmath122 star was the progenitor of the cygnus loop . the observer s line of sight through shocked clouds at the projected edge is preferentially longer than through those that appear at the projected center . photoevaporation of clouds progresses radially outward from the central star , so lines of sight tangent to the cavity are biased to be through the longer dimensions of surviving clumps . thus , optical emission on the projected interior is not as bright as on the projected edge of the snr , despite their common cause . the same effect is even more pronounced in the x - ray data . the atomic shell in which the non - radiative filaments appear is also a consequence of the progenitor s evolution . while material is eroded from clouds during the star s main sequence lifetime , increasing the density of the h ii region , the radius of the ionized region decreases as the outer portions recombine , creating the shell ( @xcite ) . the atomic gas of this recombined region together with the cloud remnants on the periphery form the walls of the cavity ( fig . [ figcartmap ] ) . the cygnus loop blast wave has expanded through the uniform cavity and is now hitting the surrounding walls ( @xcite ) . the atomic shell is not actually a sharp density discontinuity . it is more likely a gradual transition from the cavity interior to the maximum shell density . at the end of the progenitor s main sequence lifetime , the pressure in the former h ii region declines as the gas cools and recombines . ( see @xcite for a discussion of this phenomenon applied to n49 . ) in the extreme case of a sudden loss of pressure at the boundary , a rarefaction wave would move back through the dense shell resulting in a density profile of the form @xmath123 ( @xcite ) , where the distance , @xmath124 , from the original shell edge is constrained @xmath125 in terms of sound speed , @xmath126 , and red giant lifetime , @xmath127 . the drop in pressure is not so sudden , however . over @xmath128 for @xmath122 ( @xcite ) , about half the atoms in the former h ii region have recombined . thus , the density profile of the shell edge is likely to be intermediate between a sharp discontinuity and the offset power law above . the cygnus loop has been suggested previously to be the result of an explosion in an incomplete cavity . both falle & garlick ( 1982 ) and shull & hippelein ( 1991 ) , for example , argue for a partial cavity at a density discontinuity , such as the edge of a molecular cloud . in these models , only half of the blast wave runs into the cavity boundary . consequently , the bright radiative emission of ngc 6992 and ngc 6960 and the corresponding x - ray emission are physically separated by great distances along the line of sight : the former are the result of radiative shocks in the cavity wall , and the opposite hemisphere of the blast wave propagates through the low - density medium outside the cloud to cause the latter . in contrast , the model presented here and in levenson et al . ( 1997 ) has a completely defined cavity , the atomic shell and dense clouds together forming its spherical boundary . the blast wave does not break _ out _ of the edge of the molecular cloud into lower - density material ; instead , it slows down when it hits a wall of _ denser _ gas . although ngc 6992 and ngc 6960 and the x - rays that appear to the exterior of these radiative regions are two slightly different portions of the blast wave , they are not significantly separated . furthermore , some of the large clouds are obviously extended along the line of sight , not restricted to the rear boundary ( fig . [ figcartmap ] ) . these optical and previous x - ray data ( @xcite ) demonstrate that the cavity wall is complete . all around the projected edges of the cygnus loop , the balmer filaments and x - ray enhancements mark the smooth component of the cavity wall . _ even the region called the breakout illustrates where the blast wave runs into a boundary of dense gas , not continued expansion in a low - density medium . only the eastern third of the projected interior of the cygnus loop lacks direct evidence for a cavity wall , although the smooth atomic shell that is apparent in balmer - dominated filaments elsewhere would not be detected across the face of the snr . only clumps of gas could be observed , and their distribution depends on the arrangement that existed before the ism was processed by the progenitor . the southeast of the cygnus loop is farthest from the galactic plane , so we expect the original filling factor of clouds to have been lowest there . other snrs , including rcw 86 ( @xcite ) and n132d ( @xcite ) , have been identified as cavity explosions . [ shull ( 1985 ) identifies n49 as a cavity supernova , but vancura et al . ( 1992 ) argue that the dense ambient material associated with the single observed co cloud accounts for the data better . ] in these cases , interaction of the blast wave with the cavity wall enhances the observable properties of the snrs . rcw 86 shares many features of the cygnus loop , including similar size , a limb - brightened ( though very incomplete ) x - ray shell ( vink et al . 1997 ) , and balmer - dominated filaments to the exterior of the x - ray emission ( @xcite ) . n132d is a younger remnant , and the apparent cavity is smaller . the progenitor s wind during its main sequence lifetime may form the cavity boundary , or the scale of photoevaporation may have been limited in this dense environment adjacent to a molecular cloud ( @xcite ) . the bright interior x - ray emission ( @xcite ) does imply that the cavity interior was not uniform . although the shell - like visual morphology of the cygnus loop and its x - ray limb - brightening broadly suggest that this is an example of a supernova remnant in the sedov - taylor or adiabatic expansion phase , in detail this model is inconsistent with the data presented here and in levenson et al . the observed x - ray and optical limb - brightening are a consequence of interaction with the inhomogeneous interstellar medium , not the result of evolution of the blast wave in a uniform medium . assuming that the blast wave has slowed sufficiently during its progress through a uniform medium to produce optical line emission ( i.e. to @xmath17 ) yields an age estimate of about 50,000 yr ( @xcite ; @xcite ) . this is much too great because the age is inversely proportional to the blast wave velocity , and only in propagating through dense clouds has the shock been decelerated to @xmath17 . the balmer filaments mark the location of the unimpeded , spherical blast wave , and the x - ray emission immediately behind these filaments demonstrates that the typical shock velocity is @xmath101 , which is still to great for radiative cooling of the sedov - taylor shell to be significant . this snr has not made the global transition to the radiative evolutionary phase . the x - ray emission alone has been used to determine an age of around 18,000 yr ( @xcite ; @xcite ) , assuming that this is presently an adiabatic blast wave expanding in a uniform medium ; @xmath129 , where @xmath130 and @xmath131 . the x - ray data provide a more accurate measurement of the age , but it is still likely to be an upper limit because these calculations contain uncertainties in the shock temperature and the velocity that is derived from it . only the current unperturbed shock velocity is relevant to the age in this model , but any observation of the snr includes contributions from a variety of temperatures , heated by different stages of the blast wave s history . observing the projected edge of the snr minimizes confusion due to overlapping projections of different temperature components , but the interaction of the blast wave with the inhomogeneous ism greatly alters the shock velocity and the corresponding postshock gas temperature . when the blast wave hits a density discontinuity , as it does over most of the surface of the cygnus loop , the continuing shock is suddenly decelerated and a reflected shock propagates back toward the interior , further heating the dense , hot , shocked material . at a density contrast of 10 , for example , the twice - shocked matter will have a temperature approximately 1.5 times the singly - shocked material at the same radius , which leads to an overestimate of the unperturbed shock velocity by about @xmath132 . both the reflected shock and the forward shock through the dense cloud enhance the x - ray surface brightness and thus bias the mean measured temperature to favor blast - wave cloud interactions rather than the undisturbed blast wave . if the x - ray measurement of two temperature components ( @xcite ) due to the decelerated cloud shock and the reflected shock is correct and the original velocity of the unperturbed blast wave was closer to @xmath84 ( [ subsecedge ] ) , the corresponding age @xmath133 yr . a reliable determination of the age of the cygnus loop is required to determine the initial energy , @xmath134 , accurately ( [ secintro ] ) . adopting the parameters @xmath135 , @xmath136 pc , and @xmath133 yr , and a partially - ionized cavity interior , we find @xmath137 . because the radiative emission shows where the blast wave has run into clumpy material and decelerated , care is required to use it to measure the distance to the cygnus loop . minkowski ( 1958 ) determined the expansion velocity to be @xmath138 , fitting an ellipse to radial velocity measurements as a function of radius . this , combined with hubble s ( 1937 ) observed proper motion of filaments of @xmath139 yields the frequently - cited distance of 770 pc . this calculation assumes that all the optical filaments are part of a common expanding surface . in fact , the optical emission reveals different regions of inhomogeneity in the ism . a second complication is the distinct geometry of the diffuse and filamentary emission on the face of the snr . one is the result of a face - on shock , and the other is an example of a shock viewed edge - on . the latter does not have the radial velocity structure as a function of projected radius that is assumed in the model calculation . a third source of confusion arises because the observed filaments are the cooling zone behind the shock front , so their apparent motion may be the result of further development of the postshock region , not motion of the same clump of gas . the cooling zone may be propagating , without any physical motion of a parcel of gas in the ism . despite these difficulties , however , the radiative shocked gas can be used to measure distance to the cygnus loop or other supernova remnants when the distribution of clumps in the surrounding medium is spherically symmetric . in this case , although the emission comes from distinct regions of the ism , not a single expanding blast wave surface , the different clouds acquire the same velocity when the blast wave passes through them . this approach is not appropriate where the surrounding medium is asymmetric , however . ic 443 , for example , is asymmetrical , and its corresponding distance determinations are highly uncertain ( @xcite ) . the non - radiative filaments can also be used to determine distance , as hester , danielson , & raymond ( 1986 ) have done , finding a distance of about 700 pc . these structures are advantageous because proper motion and spatial velocity can be calculated from the same filament . there is some uncertainty in the shock velocity because the derived value depends on the amount of ion - electron temperature equilibration ( chevalier et al . measuring proper motion of filaments also introduces some uncertainty . all filaments are detected where the observer s line of sight through the shock front is long , which may be the tangent surface of a spherical bubble . if it is instead the deformed surface of the blast wave that appears as a filament , however , the apparent filament will not necessarily persist or be detectable over the timescales required to obtain proper motion measurements . long baselines over which to measure proper motion are desirable , but the oldest observations are sensitive to the brightest filaments . these bright filaments are exactly the locations where the shock is likely slowed in denser gas , so the present velocity differs from the spatial velocity that corresponds to the proper motion . with these caveats in mind , we adopt a distance of 770 pc to the cygnus loop . the cygnus loop blast wave is now in the walls of the cavity , the boundary between the low - density former h ii region and the ambient ism . blast wave propagation in the atomic shell will be a long - lived phenomenon . the smooth material of the shell recombined as clumps were photoevaporated in the interior . we crudely calculate the thickness of the shell assuming a constant number of ions in the h ii region in both the initial stromgren sphere , in which only the less - dense intercloud material was ionized , and in the final stromgren sphere of radius @xmath140 , in which all material , having mean density @xmath141 , is ionized . the typical , unprocessed ism determines the initial conditions ; we assume the intercloud density @xmath142 , and a volume filling factor of 0.03 of clouds of density @xmath143 , so @xmath144 . conserving ion number and neglecting increased recombination as clouds are evaporated , we obtain for the shell thickness @xmath145 so using the observed radius @xmath146 yields @xmath147 . a blast wave of velocity @xmath148 will propagate through the shell for 5000 years . eventually the blast wave will propagate through the denser medium of the cavity walls and clouds everywhere . it will rapidly decelerate , not only in the select locations where it has already encountered dense clumps that protrude into the cavity . some of the surrounding material consists of large dense clumps like the ones observed presently in the very extended regions of radiative shocks . these , too , will light up and will be detectable at optical wavelengths . the southeast cloud may be an example of the earliest stages of interaction with a large cloud , and it may later resemble the northeast region ( @xcite ) . we also expect the entire northern limb , which in the optical regime is now detected primarily in bright balmer filaments , to turn into an extended radiative region . the incomplete shocks along the northern limb are the early stages of interaction with a larger structure that is perhaps directly associated with the bright northeastern region . the entire western face of the cygnus loop will soon be bright when the blast wave has fully encountered the wall of gas that we have noted there . if the long filaments across the center of the snr are the result of another wall of gas , the middle will also light up with optical emission in several hundred years . large clouds of gas will significantly affect blast wave evolution . the shock will no longer be adiabatic when the cooling timescale becomes comparable with the dynamical timescale . in large clouds , where density @xmath143 , radius @xmath149 , and in which the shock velocity @xmath150 , the dynamical time @xmath151 . in these clouds , the cooling time @xmath152 , in terms of the cooling function , @xmath153 , having units @xmath154 . if the covering fraction of clumps in the surrounding medium is sufficiently great , they will force the cygnus loop into the radiative phase , and energy losses will further decelerate the blast wave . the time and radius at which this occurs are distinct from self - similar calculations , such as those of mckee & ostriker ( 1977 ) and shull & draine ( 1987 ) in which the gradual transition to the radiative phase happens as the blast wave propagates through a uniform medium . we use the pressure - driven snowplow model ( @xcite ) to set an upper limit on the final size of the snr . in this model , work of the blast wave pushing into the ambient material and radiative losses of the hot interior are taken into account . in the cygnus loop , the radiative losses of the outer shell also will be significant , slowing the blast wave and making the final radius smaller . during this phase , @xmath155 is constant , where @xmath156 is the pressure inside the snr , @xmath157 is its volume , and @xmath158 is the ratio of specific heats . for @xmath159 , this phase ends when @xmath160 , in terms of the external pressure of the interstellar medium , @xmath161 , and the radius and pressure inside the supernova remnant at the onset of the pressure - driven snowplow stage , @xmath162 and @xmath163 , respectively . at this final time , the pressure inside the supernova remnant is equal to the ambient pressure . for the cygnus loop , we take @xmath164 , @xmath165 , and @xmath166 , which places an upper limit on the final radius @xmath167 . assuming that blast waves propagate through uniform media over great distances yields misleading conclusions about the size of snrs and how they distribute energy through the ism . global models of the interstellar medium that include the hot contribution of snrs ( @xcite ; @xcite ) are sensitive to the supernova rate , the final sizes of supernova remnants , and their persistence as hot bubbles . these factors determine whether the dominant component of the ism is hot and in which smaller regions of warm and cold gas are embedded , or whether the warm and cold components form the substrate in which the distinct hot bubbles of snrs are located . for the former to occur , the onset of the radiative phase of the snr must be postponed until @xmath168 , independent of its prior evolution ( @xcite ) . ( this calculation assumes that the supernova rate is 1 per 30 yr in a disk of radius 15 kpc , yet the conclusion holds for a range of reasonable supernova rates . ) although the cavity of the cygnus loop is small because its progenitor is one of the least massive stars to undergo core collapse , the initial mass function favors these lower mass ( and smaller cavity ) progenitors . furthermore , even more massive stars , where cavities of 50 pc radius are expected , will be forced into the radiative phase and not expand to overlap if the density of the cavity walls is sufficient . the cavity walls are indeed likely to be denser on average than the typical ism because the walls contain much of the material that was cleared out of the cavity interior . the conditions of the ism in general , however , constrain the mass that will be commonly available to form clouds and cavity walls around pre - supernova stars . if the hot , low - density medium is pervasive , cavity walls typically would not be sufficiently massive to induce the radiative phase early , and snrs would then normally attain larger radii before their blast waves decelerate . in any case , the typical environments of supernova progenitors must be considered . if dense , molecular material normally surrounds these stars , the radiative phase will be induced early , even if globally the ism has low density . the cavity structure is significant also in determination of the mass of gas that a supernova remnant heats . the onset of shell formation for a supernova blast wave expanding in a uniform medium occurs at @xmath169 , so the shocked mass is then @xmath170 . if the ambient density varies by three orders of magnitude , the heated mass changes by only a factor of 7 , for example . in contrast , when the supernova environment is a cavity with dense walls , the initial size of the cavity constrains the mass of the hot interior . the cavity interior is rarefied , so the heated cavity mass is significantly lower . in the cygnus loop , assuming a cavity density @xmath171 , the heated mass is @xmath172 , whereas in a uniform medium of this density , the heated mass would be greater than @xmath173 . this optical emission line atlas of the cygnus loop supernova remnant provides the information to render a portrait of the surrounding ism . the ambient medium , which the massive stellar progenitor shaped as it evolved , consists of many large , dense regions . these will significantly affect the subsequent development of the blast wave . the blast wave will decelerate , and it will no longer have sufficient velocity to excite the material through which it propagates to x - ray - emitting temperatures . although this is a study of a particular object , it has far - reaching consequences for the ism as a whole . supernova remnants provide the energy to heat the ism , affect the velocity dispersion of interstellar clouds , and set the stage for future generations of star formation . thus , any global model of the gas in the galaxy critically depends on the evolution of snr blast waves in their inhomogeneous environments , of which this example is typical . while optical emission - line characteristics identify supernova remnants , they may preferentially select those whose evolution the extant ism determines . other snrs need to be examined over a broad range of energies in the same careful way to draw more certain conclusions about their net effect on the interstellar medium . ballet , j. , & rothenflug , r. 1989 , , 218 , 277 banas , k. r. , hughes , j. p. , bronfman , l. , & nyman , l .- . 1997 , , 480 , 607 blair , w. p. , et al . 1991 , , 379 , l33 braun , r. , & strom , r. g. 1986 , , 164 , 208 charles , p. a. , kahn , s. m. , & mckee , c. f. 1985 , , 295 , 456 chevalier , r. a. , kirshner , r. p. , & raymond , j. c. 1980 , , 235 , 186 chevalier , r. a. , & raymond , j. c. 1978 , , 225 , l27 claver , c. f. 1997 , , 24 , 1282 cox , d. p. 1972 , , 178 , 143 cox , d. p. 1986 , , 334 , 252 cox , d. p. , & smith , b. w. 1974 , , 189 , l105 denoyer , l. k. 1975 , , 196 , 479 dopita , m. a. 1977 , , 33 , 437 dopita , m. a. , binette , l. , dodorico , s. , & benvenuti , p. 1984 , , 276 , 653 dopita , m. a. , & sutherland , r. s. 1996 , , 102 , 161 elmegreen , b. g. 1976 , , 205 , 405 falle , s. a. e. g. , & garlick , a. r. 1982 , , 201 , 635 fesen , r. a. , blair , w. p. , & kirshner , r. p. 1982 , , 262 , 171 fesen , r. a. , kwitter , k. b. , & downes , r. a. 1992 , , 104 , 719 graham , j. r. , & aschenbach , b. 1996 , in rosat newsletter # 13 , ed . s. l. snowden , 6 graham , j. r. , levenson , n. a. , hester , j. j. , raymond , j. c. , & petre , r. 1995 , , 444 , 787 green d. a. 1984 , , 209 , 449 greidanus , h. , & strom , r. g. 1991 , , 89 , 15 hanson , g. j. , & winkler , p. f. 1992 , , 24 , 1231 hester , j. j. 1987 , , 314 , 187 hester , j. j. , & cox , d. p. 1986 , , 300 , 675 hester , j. j. , raymond , j. c. , & danielson , g. e. 1986 , , 303 , l17 hester , j. j. , parker , r. a. r. , & dufour , r. j. 1983 , , 273 , 219 hester , j. j. , raymond , j. c. , & blair , w. p. 1994 , , 420 , 721 hubble , e. p. 1937 , carnegie institution yearbook no . 36 , p. 189 keen , n. j. , wilson , w. e. , haslam , c. g. t. , graham , d. a. , & thomasson , p. 1973 , , 28 , 197 kirshner , r. p. , & taylor , k. 1976 , , 208 , l83 ku , w. h .- m . , kahn , s. m. , pisarski , r. & long , k. s. 1984 , , 278 , 615 levenson , n. a. , graham , j. r. , hester , j. j. , & petre , r. 1996 , , 468 , 323 levenson , n. a. , et al . 1997 , , 483 , 304 long , k. s. , blair , w. p. , vancura , o. , bowers , c. w. , davidsen , a. f. , & raymond , j. c. 1992 , , 400 , 214 maeder , a. , & meynet , g. 1989 , , 210 , 155 mccray , r. , & snow , t. p. , jr . 1979 , , 17 , 213 mckee , c. f. , & cowie , l. l. 1975 , , 195 , 715 mckee , c. f. , & ostriker , j. p. 1977 , , 218 , 148 mckee , c. f. , van buren , d. & lazareff , b. 1984 , , 278 , l115 miller , j. s. 1974 , , 189 , 239 minkowski , r. 1958 , rev . mod . phys , 30 , 1048 morse , j. a. et al . 1996 , , 112 , 509 morse , j. a. , winkler , p. f. , & kirshner , r. p. 1995 , , 109 , 2104 morse , j. a. , raymond , j. c. , & wilson , a. s. 1996 , , 108 , 426 oort , j. h. 1946 , , 106 , 159 rappaport , s. , doxsey , r. , solinger , a. , & borken , r. 1974 , , 194 , 329 raymond , j. c. 1979 , , 39 , 1 raymond , j. c. 1991 , , 103 , 781 raymond , j. c. , black , j. h. , dupree , a. h. , hartmann , l. , & wolff , r. s. 1980a , , 238 , 881 raymond , j. c. , blair , w. p. , fesen , r. a. , & gull , t. r. 1983 , , 275 , 636 raymond , j. c. , davis , m. , gull , t. r. , & parker , r. a. r. 1980b , , 238 , l21 raymond , j. c. , hester , j. j. , cox , d. , blair , w. p. , fesen , r. a. , & gull , t. r. , jr . 1988 , , 324 , 869 scoville , n. z. , irvine , w. m. , wannier , p. g. , & predmore , c. r. 1977 , , 216 , 320 shull , j. m. , & draine , b. t. 1987 , in interstellar processes , ed . d. j. hollenbach and h. a. thronson , jr . ( dordrecht : reidel ) , p. 283 shull , j. m. , & mckee , c. f. 1979 , , 227 , 131 shull , p. , jr . , dyson , j. e. , kahn , f. d. , and west , k. a. 1985 , , 212 , 799 shull , p. , jr . , & hippelein , h. 1991 , , 383 714 smith , r. c. 1997 , , 114 , 2664 sutherland , r. s. , bicknell , g. v. , & dopita , m. a. 1993 , , 414 , 510 treffers , r. r. 1981 , , 250 , 213 vancura , o. , blair , w. p. , long , k. s. , & raymond , j. c. 1992 , , 394 , 158 vink , j. , kaastra , j. s. , & bleeker , j. a. m. 1997 , , 328 , 628 zeldovich , ya . b. , & raizer , yu . , physics of shock waves and high - temperature hydrodynamic phenomena , vol . w. d. hayes and r. f. probstein , ( new york : academic press )

the densities in the outer regions of clusters of galaxies are very low , and the collisional timescales are very long . as a result , heavy elements will be under - ionized after they have passed through the accretion shock . we have studied systematically the effects of non - equilibrium ionization for relaxed clusters in thecdm cosmology using one - dimensional hydrodynamic simulations . we found that non - equilibrium ionization effects do not depend on cluster mass but depend strongly on redshift which can be understood by self - similar scaling arguments . the effects are stronger for clusters at lower redshifts . we present x - ray signatures such as surface brightness profiles and emission lines in detail for a massive cluster at low redshift . in general , soft emission ( 0.31.0 kev ) is enhanced significantly by under - ionization , and the enhancement can be nearly an order of magnitude near the shock radius . the most prominent non - equilibrium ionization signature we found is the and line ratio . the ratios for non - equilibrium ionization and collisional ionization equilibrium models are different by more than an order of magnitude at radii beyond half of the shock radius . these non - equilibrium ionization signatures are equally strong for models with different non - adiabatic shock electron heating efficiencies . we have also calculated the detectability of the and lines with the future _ international x - ray observatory _ ( _ ixo _ ) . depending on the line ratio measured , we conclude that an exposure of ksec on a moderate - redshift , massive regular cluster with the x - ray microcalorimeter spectrometer ( xms ) on the _ ixo _ will be sufficient to provide a strong test for the non - equilibrium ionization model .

1012.0039

clusters of galaxies are very sensitive probes to cosmological parameters ( e.g. , * ? ? ? * ; * ? ? ? systematic uncertainties in precision cosmology using galaxy clusters can be minimized by restricting the sample of clusters to the highest degree of dynamical relaxation , and hence studying relaxed clusters is particularly important . in addition , the outer envelopes of clusters have been thought to be less subjected to complicated physics such as active galactic nucleus feedback , and hence these outer regions may provide better cosmological probes . clusters are believed to be formed by the continual merging and accretion of material from the surrounding large - scale structure . based on high resolution @xmath2-body simulations , it has been found that cluster growth can be divided into an early fast accretion phase dominated by major mergers , and a late slow phase dominated by smooth accretion of background materials and many minor mergers @xcite . accretion shocks ( or virial shocks ) are unambiguous predictions of cosmological hydrodynamic simulations . for the most relaxed clusters with roughly spherical morphology in the outer regions , these simulations show that large amounts of material are accreted through filamentary structures in some particular directions , and more spherically symmetrically in other directions . most of the simulations assume the intracluster medium ( icm ) is in collisional equilibrium . however , because of the very low density in the cluster outer regions ( @xmath3 is the radius within which the mean total mass density of the cluster is @xmath4 times the critical density of the universe . the virial radius @xmath5 is defined as a radius within which the cluster is virialized . for the einstein - de sitter universe , @xmath6 , while for the standard @xmath0cdm universe , @xmath7 . ] ) , the coulomb collisional and collisional ionization timescales are comparable to the age of the cluster . it has been pointed out that electrons and ions there may be in non - equipartition @xcite , and also the ionization state may not be in collisional ionization equilibrium @xcite . if these non - equilibrium processes are not properly taken into account , the measured properties may be biased in these regions . studying the outer regions of relaxed clusters not only is valuable in understanding the accretion physics and to test the assumptions concerning plasma physics near the shock regions , but it is also very important to test structure formation theory and to constrain the systematic uncertainties in clusters to be used as precision cosmological probes . before the launch of the _ suzaku _ x - ray observatory , physical properties such as temperature have never been constrained with confidence beyond roughly one - half of the shock radius . most of our understanding of these regions is still based on theoretical models and hydrodynamic simulations . recently , observations by _ suzaku _ have constrained temperatures up to about half of the shock radius for a few clusters for the first time to better than a factor of @xmath8 @xcite . while the uncertainties are still large , there is evidence that the electron pressure in cluster outer regions may be lower than that predicted by numerical simulations assuming collisional equilibrium . observations of secondary cosmic microwave background anisotropies with the south pole telescope ( spt ) and the _ wilkinson microwave anisotropy probe _ ( _ wmap _ ) 7 year data also support these results @xcite . these observational signatures are consistent with electrons and ions in non - equipartition , although it is also possible that the hydrodynamic simulations may simply overestimate the gas pressure . another possibility is that heat conduction outside the cluster may be reducing the gas pressure @xcite . recently , @xcite have shown that cosmological parameters will be biased if non - equilibrium effects ( in particular , non - equipartition ) are not properly taken into account . more observations of the outer regions of clusters are being done or analyzed . in the future , the proposed _ international x - ray observatory _ ( @xmath9 ) will have the sensitivity to constrain cluster properties out to the shock radius . thus , a detailed study of the physics in the outer regions of clusters would be useful . in particular , _ ixo _ will have sufficient spectral resolution to resolve many important x - ray lines , and hence the ionization state of the plasma can be determined . in @xcite , we have studied in detail the x - ray signature of non - equipartition effects in cluster accretion shock regions . in this paper , we extend our study to include non - equilibrium ionization in our calculations . non - equilibrium ionization calculations have been considered in a number of cosmological simulations to study the very low density warm - hot intergalactic medium ( whim ) surrounding galaxy clusters @xcite . at galaxy cluster scales , similar non - equilibrium ionization calculations have focused on merging clusters @xcite . these studies generally agree that in the low density icm and the whim , there are significant deviations from ionization equilibrium , and that the effects on the x - ray emission lines are strong . in this work , we focus on the x - ray signatures of non - equilibrium ionization in the accretion shock regions of relaxed clusters . we also discuss the detectability of non - equilibrium ionization effects with the future _ ixo_. the paper is organized as follows . section [ ion_sec : method ] describes in detail the physical models and techniques to calculate non - equilibrium ion fractions and x - ray observables , which includes the hydrodynamic models we used ( section [ ion_sec : hydro ] ) , and the ionization and spectral calculational techniques ( sections [ ion_sec : ionfraccal][ion_sec : xraycal ] ) . the overall dependence the non - equilibrium ionization effects on the mass and redshift of the cluster is presented in section [ ion_sec : b_vs_mass_z ] . calculated non - equilibrium ionization signatures are presented in section [ ion_sec : signatures ] , which includes descriptions of particular models we used to present the results ( section [ ion_sec : models ] ) , x - ray spectra ( section [ ion_sec : spectra ] ) , surface brightness profiles ( section [ ion_sec : sb ] ) , and the ratio of intensities of and lines ( section [ ion_sec : line ] ) . we discuss the detectability of and lines and non - equilibrium ionization diagnostics with _ ixo _ in section [ ion_sec : detect ] . section [ ion_sec : conclusion ] gives the discussion and conclusions . throughout the paper , we assume a hubble constant of @xmath10 km s@xmath11 mpc@xmath11 with @xmath12 , a total matter density parameter of @xmath13 , a dark energy density parameter of @xmath14 , and a cluster gas fraction of @xmath15 , where @xmath16 is the baryon density parameter for our cluster models in the standard @xmath0cdm cosmology . the clusters have a hydrogen mass fraction @xmath17 for the icm . we calculate the x - ray emission spectrum from the outer regions of cluster of galaxies using one - dimensional hydrodynamic simulations we have developed @xcite . radiative cooling is negligible in cluster outer regions so that it will not affect the dynamics of the plasma . this non - radiative condition allows us to calculate the hydrodynamics and then the ionization structure and x - ray emission separately . the hydrodynamic models simulate the accretion of background material from the surrounding regions onto clusters through accretion shocks in the @xmath0cdm cosmology . the calculated hydrodynamic variables ( e.g. , density and temperature profiles ) in the cluster outer regions are consistent with those calculated by three - dimensional simulations , and a detailed discussion on our hydrodynamic simulations can be found in @xcite . the hydrodynamic simulations were done using eulerian coordinates . in calculating the time - dependent ionization for the hydrodynamic models , we need to follow the lagrangian history of each fluid element . in one - dimensional spherical symmetric systems , it is convenient to transform from the eulerian coordinates to the lagrangian coordinates by using the interior gas mass as the independent variable @xmath18 where @xmath19 is the radius , and @xmath20 is the gas density . to do this , we first determined the values of @xmath21 for each of the grid zones in the hydrodynamical simulations in @xcite at the final redshift of @xmath22 . the values of the density and temperature were also determined for this final time step for each gas element . then , for each earlier time step , the values of gas density and temperature within the shock radius were determined by linear interpolation between the nearest two values of the interior gas mass on the grid at that time . when the values of @xmath21 fall between the discontinuous values of the shocked and preshocked elements , we assume the material to be preshock . this slightly underestimates the ionization timescale parameter defined below , but this effect is small as the grid zones in the simulations are closely spaced . the collisional ionization and recombination rates and the excitation of x - ray emission depend on the electron temperature ( @xmath23 ) . because the accretion shock may primarily heat ions instead of electrons due to the mass difference and the long coulomb collisional timescale between electrons and ions , the electron temperature may not be the same as the ion temperature near the accretion shock regions @xcite . the degree of non - equipartition depends on the non - adiabatic shock electron heating efficiency , which is defined as @xmath24 in @xcite , where @xmath25 and @xmath26 are the changes in electron temperature and average thermodynamic temperature due to non - adiabatic heating at the shock , respectively . while there are no observational constraints on @xmath27 in galaxy cluster accretion shocks , observations in supernova remnants suggest that @xmath28 for shocks with mach numbers similar to those in galaxy cluster accretion shocks @xcite . @xcite have calculated electron temperature profiles for models with a very low electron heating efficiency ( @xmath29 ) , an intermediate electron heating efficiency ( @xmath30 ) , and an equipartition model ( @xmath31 ) . the electron temperature profiles of these models with different values of @xmath27 are used to calculate the ionization fractions and x - ray emission in this paper . the ionization timescale parameter of each fluid element is defined as @xmath32 where @xmath33 is the electron number density , @xmath34 is the time when the fluid element was shocked , and @xmath35 is the time at the observed redshift . the ionization timescale parameters for cluster models with accreted masses of @xmath36 , and @xmath37 are shown in figure [ ion_fig : timescale ] . most ions of astrophysical interests will not achieve ionization equilibrium for the ionization timescale parameters @xmath38 @xmath39 s @xcite . in order to calculate the x - ray emission spectrum for non - equilibrium ionization plasma , the ionization state of each fluid element has to be calculated . we assume h and he are fully ionized . in this paper , we consider only collisional ionization processes , and ignore photoionization . to justify this , we compare the photoionization and collisional ionization rate for the ions we are interested in . the photoionization rate of an ion @xmath40 is given by @xmath41 where @xmath42 is the photoionization cross section of ion @xmath40 at frequency @xmath43 , and @xmath44 is the ionizing flux . near the cluster accretion shock , the dominant photoionization source is the uv background . we approximate the uv ionizing flux at @xmath45 as a power law @xmath46 with @xmath47 erg @xmath48 s@xmath11 hz , @xmath49 hz , and @xmath50 @xcite . for example , for the and ions we are most interested in throughout this paper , adopting the photoionization cross sections given by @xcite gives @xmath51@xmath52 s@xmath11 and @xmath51@xmath53 s@xmath11 . the slowest collisional ionization rate of the oxygen ions is of the order of @xmath54 ) s@xmath11 for typical temperatures ( @xmath55 kev ) and densities in the outer regions of clusters @xcite . this is nearly two orders of magnitude higher than the photoionization rates of and ions . even at @xmath56 when the uv background is roughly 80 times stronger @xcite , the densities in the outer regions of clusters will be higher by a factor of @xmath57 , and hence collisional ionization rates will increase by a similar factor . collisional ionization still dominates over photoionization . for heavier elements , the photoionization rates will be even smaller , and the collisional ionization rates for ions up to those of ni are all higher than @xmath58 ) s@xmath11 @xcite . therefore , we assume that photoionization is not important in our calculations . for ionization dominated by collisional processes at low densities , the ionization states for each of the ions of a given element @xmath59 are governed by @xmath60 f_i \ } \ , , % \end{equation } % \end{widetext}\end{aligned}\ ] ] where @xmath61 is the ionization fraction of the ion @xmath40 with charge @xmath62 , @xmath63 is the number density of that ion , @xmath64 is the total number density of the element @xmath59 , and @xmath65 and @xmath66 are the coefficients of collisional ionization out of and recombination into the ion @xmath40 , respectively . in solving equation ( [ ion_eq : ionization ] ) , we use an eigenfunction technique which is based on the algorithm developed by @xcite , and the method is described in detail in appendix a of @xcite . the eleven heavy elements c , n , o , ne , mg , si , s , ar , ca , fe , and ni are included in our calculations . the eigenvalues and eigenvectors , which are related to the ionization and recombination rates , used to solve equation ( [ ion_eq : ionization ] ) are taken from the latest version of the _ nei _ version 2 model in xspec ( version 12.6.0 ) . the atomic physics used to calculated the eigenvalues and eigenvectors in xspec are discussed in @xcite , and the ionization fractions in the latest version of xspec are calculated using the updated dielectronic recombination rates from @xcite . all of the eleven heavy elements are assumed to be neutral initially when solving equation ( [ ion_eq : ionization ] ) , which is a good approximation as long as the ionization states of the preshock plasma are much lower than that of the postshock plasma ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? we have also tested this by assuming all the ionization states in the preshock regions are in collisional ionization equilibrium initially at different temperatures of @xmath67 and @xmath68 k. we confirmed that the final ionization fractions for and are essentially the same . once the ionization states are calculated by solving equation ( [ ion_eq : ionization ] ) , the x - ray emission spectrum could be calculated by using available plasma emission codes such as the raymond - smith code @xcite , the spex code @xcite , and the apec code @xcite . we chose to use a version of the apec code as implemented in the _ nei _ version 2 model in xspec to calculate the x - ray emission spectrum . the routine uses the astrophysical plasma emission database ( aped ) to calculate the resulting spectrum . thus , the atomic data we used to calculate the ionization fractions and spectrum are mutually consistent , both coming from the same _ nei _ model version 2.0 in xspec . the publicly available line list _ apec_nei_v11 _ was used throughout the paper . the inner - shell processes are missing in this _ nei _ version 2 . we have compared our results to an updated line list which includes the inner - shell processes but is not yet publicly released ( k. borkowski , private communications ) , and we found that the difference is less than 10% which is smaller than the @xmath69 uncertainties in the atomic physics of the x - ray plasma code . we assume the abundance to be a typical value for galaxy clusters , which is 0.3 of solar for all models , and use the solar abundance tables of @xcite . we calculated the x - ray emissivity in photons per unit time per unit volume per unit energy , @xmath70 , for each fluid element in our hydrodynamic models at redshift zero , which can be expressed in terms of an emissivity function , @xmath71 , by @xcite @xmath72 where @xmath73 is the proton density . the emissivity function @xmath71 depends on the ionization fractions and the electron temperature , but is independent of the gas density . the projected spectrum in the rest frame is given by integrating the x - ray emissivity along the line of sight @xcite @xmath74 where @xmath75 is the distance along the line of sight . the broadband rest frame surface brightness in energy per unit time per unit area is then given by @xmath76 where the integral is across the energy band of interest . we consider how the degree on non - equilibrium ionization of a cluster depends on its mass and redshift . we first consider how the cluster parameters which affect the ionization depend on mass and redshift . in order to compare equivalent locations in the different clusters , we estimate the ionization parameters at a given over - density radius @xmath77 . we consider a simple self - similar scaling argument for galaxy clusters . by definition , the average total density within the radius @xmath77 is @xmath78 where @xmath79 is the critical density at redshift @xmath80 , and @xmath81 is the hubble constant at redshift @xmath80 . the quantity @xmath82 , so that @xmath83 \nonumber \\ & \approx & \omega_m ( 1 + z)^3 + 1 - \omega_m \ , .\end{aligned}\ ] ] the final expression in equation ( [ ion_eq : ez ] ) follows from the assumption that the universe is flat ( @xmath84 ) and the fact that the radiation density parameter is small ( @xmath85 ) in the present - day universe . the gas density at @xmath77 is roughly @xmath86 the inflow timescale is roughly @xmath87^{-1 } \ , , \ ] ] where @xmath88 is the hubble time at @xmath22 . thus , the ionization timescale parameter @xmath89 varies as @xmath90 thus , the value of @xmath89 at a fixed characteristic radius should be nearly independent of the cluster mass but will increase significantly with redshift . figure [ ion_fig : timescale ] shows that the variation of @xmath89 with radius is fairly self - similar for clusters with differing masses . when these curves are scaled to a fixed cluster characteristic radius , they are very nearly identical . the ionization state of the gas depends on @xmath89 and the collisional ionization and recombination rates . for an under - ionized plasma , the collisional ionization rates are more important , so that the ionization state should depend mainly on @xmath91 . for an under - ionized plasma where the electron temperature is greater than the ionization potential of the relevant ions , @xmath92 varies slowly , @xmath93 . we have confirmed that this dependence fits the temperature dependence of the ionization rate of over the interesting temperature range ( @xmath94 15 kev ) . below , we show that the ratio of to lines is the best diagnostic of departures from ionization equilibrium ( section [ ion_sec : line ] ) . we assume that the electron temperature increases in proportion to the mean temperature @xmath95 at the radius @xmath77 . the mass within @xmath77 is @xmath96 , so that the radius @xmath77 is given by @xmath97^{-2/3 } \ , .\ ] ] the condition of hydrostatic equilibrium or the shock jump condition at the accretion shock imply that @xmath98 . thus , the gas temperature varies as @xmath99^{2/3 } \ , .\ ] ] this implies that @xmath100^{1/3 } \ , .\ ] ] combining equations ( [ ion_eq : tauscale ] ) and ( [ ion_eq : cscale ] ) gives @xmath101^{4/3 } \ , .\ ] ] equations ( [ ion_eq : tauscale ] ) and ( [ ion_eq : ctau ] ) suggest that the ionization state should depend only very weakly on the cluster mass , but should depend strongly on cluster redshift . the increase of @xmath89 and @xmath102 with @xmath80 implies that collisional ionization will be faster at high redshifts , and hence non - equilibrium ionization will be most important in low redshift clusters . later , we will show that the ratio of to lines is the best diagnostic of departures from ionization equilibrium ( section [ ion_sec : line ] ) . figure [ ion_fig : ctau_vs_m ] shows the dimensionless ionization parameter @xmath91 for as a function of the scaled radius ( @xmath103 ) for clusters with different masses at @xmath45 . here , @xmath104 is the radius of the cluster accretion shock . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . all the curves for the same @xmath27 but with different masses nearly overlap each others . this confirms that the non - equilibrium ionization effect at the same characteristic radius is nearly independent of mass . figure [ ion_fig : ctau_vs_z ] shows the dimensionless ionization parameter as a function of the scaled radius for cluster models at different redshifts . the cluster model with a total accreted mass @xmath105 at @xmath45 is used . models with @xmath106 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . in contrast to the dependence in mass , we find that there is significant evolution with redshift . the non - equilibrium ionization effect should be strongest for low - redshift clusters . the preceding arguments assumed that departures from ionization equilibrium could be assessed through the variation of @xmath91 . to test this explicitly and to determine if the preceding arguments apply to spectral diagnostics for non - equilibrium ionization , we study the effects of non - equilibrium ionization on and ion fractions using numerical simulations . we calculate the non - equilibrium ionization bias , @xmath107 , of the ionization fraction ratios @xmath108@xmath109@xmath110 for the non - equilibrium ionization ( nei ) and collisional ionization equilibrium ( cie ) models with the same electron heating efficiency @xmath27 . the non - equilibrium ionization bias is defined as @xmath111_{\rm nei } } { \left [ f({\rm o~viii})/f({\rm o~vii } ) \right]_{\rm cie } } \,.\ ] ] figure [ ion_fig : ofracbias_vs_m ] shows the non - equilibrium ionization bias as a function of the scaled radius for clusters with different masses . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . the nearly self - similar curves justify the semi - analytical argument that non - equilibrium ionization effect at the same characteristic radius is nearly independent of mass . note that the non - equilibrium ionization effect on the @xmath108@xmath109@xmath110 ratio is only significant for the outer 10% of the shock radius . however , the projected and line emission will be significantly affected even at a radius as small as one - fourth of the shock radius . this is because the projected emission at the inner radius can be dominated by the line emission from the under - ionized outer shell ( section [ ion_sec : line ] below ) . figure [ ion_fig : ofracbias_vs_z ] shows the non - equilibrium ionization bias as a function of the scaled radius for cluster models at different redshifts . the cluster model with a total accreted mass @xmath105 at @xmath45 is used . models with @xmath112 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . the non - equilibrium ionization effect is larger for clusters at lower redshifts which agrees with the semi - analytical argument given above . note that at a redshift higher than 1 , only a thin shell with a width of less than 5% of the shock radius is in non - equilibrium ionization compared to the wider shell with a width of @xmath113 of the shock radius for zero redshift clusters . massive clusters at low redshifts are ideal candidates to study the non - equilibrium ionization effects in the outer regions since the departures from ionization equilibrium are larger at low redshift and nearly independent of cluster mass ( section [ ion_sec : b_vs_mass_z ] ) . clusters with higher masses are more luminous in x - rays , and hence the spectral signatures are easier to detect . in the following , we present x - ray spectra for the hydrodynamic cluster model with an accreted mass of @xmath114 at @xmath45 calculated in @xcite . this model represents a typical massive cluster in the present - day universe . the shock radius is @xmath115 mpc for this model . the virial radius is @xmath116 mpc , and the total mass within @xmath117 is @xmath118 . another commonly used radius and mass are @xmath119 mpc and @xmath120 . we calculate spectra for three different values of the shock electron heating efficiency @xmath29 , 0.5 , and 1 . the last case corresponds to electron - ion equipartition . we also calculate spectra both for non - equilibrium ionization ( nei ) models and for collisional ionization equilibrium ( cie ) models for comparison . for the nei models , the results for the model with the small shock electron heating efficiency @xmath29 will be discussed extensively throughout the paper , as this model maximizes the departures from equilibrium in the outer regions of clusters . for the cie models , we present a non - equipartition model with @xmath29 ( cie non - eq ) and an equipartition model with @xmath31 ( cie eq ) . the projected rest frame spectra for several different models at two projected radii are shown in figure [ ion_fig : spec ] . in the upper panels , we show the nei model with a very small shock heating efficiency ( @xmath29 ) , and electrons and ions are in non - equipartition . the cie non - eq model is shown in the lower panels . the left panels show spectra at a radius of @xmath121 mpc , while the right panels show spectra at a radius of @xmath122 mpc . each spectrum is binned with a bin size of @xmath123 . at @xmath121 mpc , the overall spectra are dominated by the free - free continuum emission over a wide range of energy for both the nei and the cie non - eq models . the continuum spectra for both models are nearly identical because of the dominant free - free emission with the same electron temperature . the line emission is also very similar for both models because at this radius , the ionization timescale parameter is rather large ( @xmath124 @xmath39 s ) . the most notable difference is the line intensity of the triplet lines near @xmath125 ev . this line intensity for the nei model is much higher than that of the cie non - eq model . there is almost no line emission for the cie non - eq model . the strong at @xmath121 mpc for the nei model is mainly due to the projection of emission from the under - ionized ions in the outer regions with much shorter ionization timescale parameters . the differences for other strong emission lines are much smaller between the two models . there is slightly more soft line emission below about 1 kev for the nei model . at @xmath122 mpc , there are significant differences in the spectra between the nei and the cie non - eq models . for the cie non - eq model , the continuum emission still dominates the overall spectrum ; for the nei model , the soft emission is dominated by lines . one of the most obvious differences in the line emission between the two models is again for the triplet . the line intensity for the nei model is much higher than that of the cie non - eq model . the triplet is weak for the cie non - eq model . by inspecting a number of line ratios at different radii , we found that the ratio of the and line intensities can be used as a diagnostic for the degree of ionization equilibrium , and this will be discussed in section [ ion_sec : line ] below . the rest frame radial surface brightness profiles integrated over various energy bands for the outer regions of clusters are shown in figure [ ion_fig : sb ] . the nei models with @xmath29 are shown in thick lines in both the upper left and upper right panels . the cie non - eq and the cie eq models are shown as thin lines on the upper left and upper right panels , respectively . the ratios of @xmath126 and @xmath127 are shown below the corresponding panels . comparing the nei and the cie non - eq models tells us the effects of non - equilibrium ionization alone , while comparing the nei and the cie eq model tells us the total effects of both non - equilibrium ionization and non - equipartition . from the left panels of figure [ ion_fig : sb ] , we can see that non - equilibrium ionization significantly enhances the soft ( 0.31.0 kev ) emission in the outer regions . for the nei model , the soft emission has been increased by more than 20% at around 3 mpc , and up to nearly an order of magnitude around the shock radius compared to the cie non - eq model . the increase of the soft emission is due to the line emission by the under - ionized ions . for the cie non - eq model , the surface brightness profiles in all energy bands shown in figure [ ion_fig : sb ] decrease rapidly out to the shock radius . for the nei model , the decrease in surface brightness in the soft band as a function of radius slows down near @xmath128 mpc , and the soft band surface brightness actually increases with radius from @xmath13.7 mpc out to nearly the shock radius , where the surface brightness drops rapidly . within about 2.3 mpc , non - equilibrium ionization effect is less than 5% in the soft band . the non - equilibrium ionization effect on the medium ( 1.02.0 kev ) band is not as dramatic as the soft band , and the maximum increase is about 70% near 3.8 mpc . the non - equilibrium ionization effect on the hard ( 2.010.0 kev ) band is less than 5% in most regions , and the effect is to lower the hard emission near the shock radius . the soft emission dominates the overall x - ray band ( 0.310.0 kev ) outside of @xmath13 mpc , and the overall x - ray emission decreases more slowly than for the cie non - eq model beyond that radius . near the shock radius , the surface brightness in the overall x - ray band for the nei model is about a fact of 6 higher than that for the cie non - eq model . the right panels of figure [ ion_fig : sb ] show that in addition to the non - equilibrium ionization effect , non - equipartition will increase the soft emission by a significant factor near the shock radius . this occurs because in the cie non - eq and nei models , the electron temperatures in the outer regions are lower than for the cie eq model , and this also leads to more soft x - ray line emission . the surface brightness profile in the overall x - ray band for the nei model is a factor of 5 higher than that for the cie eq model near the shock radius . a detailed discussion on the difference between the cie non - eq and cie eq models can be found in @xcite . the most prominent non - equilibrium ionization signature in the x - ray lines are for the line ratio of and . figure [ ion_fig : spec_line ] shows the spectra for models at @xmath121 mpc . the spectra are shown in the 0.50.7 kev range which covers the rest frame energies of the and lines . the upper panel shows the nei model with a very low electron heating efficiency @xmath129 . the middle panel shows the nei model with an intermediate @xmath130 . the lower panel shows the cie non - eq model with @xmath129 . each spectrum is binned with a bin size of @xmath131 kev . in the upper panel of figure [ ion_fig : spec_line ] , the most prominent lines are the he - like triplets at 561.0 , 569.6 , and 574.0 ev , the h - like doublets at 653.5 and 653.7 kev , and the he - like line at 665.6 ev . here , we focus on the line ratio between the he - like triplets and the h - like doublets as a diagnostic for non - equilibrium ionization . these two lines have been used to search for the whim as well as to study the non - equilibrium ionization of the whim @xcite . we find that the he - like and the h - like lines also show strong signatures of non - equilibrium ionization in the outer regions of clusters . the upper and lower panels of figure [ ion_fig : spec_line ] show that non - equilibrium ionization strongly enhances the he - like triplets compared to the cie non - eq model . the h - like doublets for the nei model are only slightly stronger than for the cie non - eq model . this suggests that the ratio of the and lines can be used as a diagnostic for non - equilibrium ionization . the and lines are similarly strong for the nei models with electron heating efficiencies @xmath130 and @xmath132 , and we suggest that the and line ratio is a good diagnostic for the non - equilibrium ionization for a wide range of electron heating efficiencies . figure [ ion_fig : line_flux ] shows the surface brightness for the and lines for the nei model with @xmath29 . the surface brightness of the lines are calculated by subtracting the continuum surface brightness from the total surface brightness within narrow energy ranges of 556.0579.0 ev and 648.5658.7 ev for the triplets and the doublets , respectively . the energy bands were chosen to cover the and the lines with a spectral resolution of 10 ev which is the expected value for the outer arrays of the x - ray microcalorimeter spectrometer ( xms ) on the _ ixo_. the continuum surface brightness were calculated by fitting the spectrum to a power law model in the energy range of 0.50.7 kev , excluding the lines . this simulates the techniques likely to be used to analyze real observations . we also show the continuum surface brightness within the narrow energy bands used to extract the line surface brightness in figure [ ion_fig : line_flux ] . the surface brightness of the triplets is nearly constant from @xmath133 mpc to @xmath128 mpc , and then rises gradually . note that a flat surface brightness at inner radii and rising surface brightness at larger radii is the signature of a shell of emission at large radii seen in projection . that is , most of the emission is actually at large radii where the ionization time scale is short . beyond about 3.9 mpc , the surface brightness rises rapidly to a peak value , and then drops . the continuum in the 556.0579.0 ev energy band drops from a radius of 1 mpc out to the shock radius . beyond @xmath134 mpc , the line emission dominates over the continuum in the 556.0579.0 ev energy band . for the doublets , the surface brightness drops from a radius of 1 mpc out to @xmath128 mpc , and then rises to a peak value at @xmath135 mpc . the surface brightness then drops beyond @xmath135 mpc . the line emission dominates over the continuum in the 648.5658.7 ev energy band beyond @xmath8 mpc . figure [ ion_fig : lineratio ] shows the line ratios of and , @xmath136@xmath137@xmath110 , for different models . to compare the effect of non - equilibrium ionization alone , we can compare the line ratios between the nei model ( solid line ) and the cie non - eq model ( dashed line ) , while both models assume an electron heating efficiency @xmath29 . both models use the same non - equipartition electron temperature to calculate the spectra , but one of them assumes equilibrium ionization while the other one does not . at @xmath133 mpc , the line ratio for the nei model is more than a factor of 2 lower than that of the cie non - eq model . the difference increases as the radius increases , and the differences are over an order of magnitude for radii beyond @xmath8 mpc . for the cie eq model ( dotted line ) , the line ratio is very similar to that of the cie non - eq model , except near the shock regions where the line ratio for the the cie eq model is a factor of a few higher than the cie non - eq model . both the line ratios of the cie eq and the cie non - eq models are above 10 in most regions between 14 mpc . nei models with electron heating efficiencies @xmath138 and @xmath139 are also shown in figure [ ion_fig : lineratio ] . the effect of increasing the electron heating efficiency is to raise the electron temperature at the shock , and hence increase the ionization rates . from figure [ ion_fig : lineratio ] , we can see that increasing the electron heating efficiency only affects the line ratio by less than a factor of two in most regions shown compared to the nei model with @xmath140 . the line ratios for all the nei models with different electron heating efficiencies are less than 10 for radii beyond about 1.3 mpc . in summary , the line ratios in the outer regions for all the nei models we calculated are significantly smaller than those for the cie models , and the differences are larger than an order of magnitude for most regions beyond @xmath8 mpc . such large differences can be used to distinguish between nei and cei models in real observations . in this section , we estimate whether the and lines in cluster outer regions and the non - equilibrium ionization signatures can be detected with _ ixo_. we consider the nei model with @xmath29 for a cluster with an accreted mass of @xmath114 at low redshift . the choice is justified by the fact that the non - equilibrium ionization effect does not depend strongly on mass and the effect is larger at lower redshift ( section [ ion_sec : b_vs_mass_z ] ) , and that the surface brightness for massive and low redshift clusters is higher . the x - ray microcalorimeter spectrometer ( xms ) planned for the _ ixo _ can potentially detect the and lines in cluster outer regions . the wide field imager ( wfi ) does not have enough spectral resolution ( @xmath141 ev ) at around 0.6 kev , and the triplets can not be resolved from the 0.5 kev nitrogen line . the x - ray grating spectrometer covers the interesting energy range , but the collecting area is too small to detect the weak and lines . therefore , we only consider the xms in our estimations . the xms has inner ( core ) and outer microcalorimeter arrays with expected spectral resolutions of 2.5 and 10 ev , respectively . we consider two cases when observing with the xms . the first case ( xmsc ) is that only the inner core array is used , and the second case ( xmsf ) is that the full array ( both inner and outer arrays ) is used . for simplicity , when considering the xmsf , we assume the spectral resolution to be the same as the outer array throughout the full array . this does nt strongly affect the detectability of the lines , since the bands used to determine the fluxes in the lines are set by the line width in the outer array , and the line width in the inner core does nt affect the line flux . the expected effective areas ( @xmath142 ) for the xms is about 10000 @xmath143 at around 0.6 kev . the relevant instrument parameters for the two cases we considered are listed in table [ ion_tab : ixo ] . the major background for the _ ixo _ observations are non - x - ray background ( nxb ) , the soft emission from the local galactic background ( gxb ) , and these are included in our signal - to - noise ratio calculations . the cluster continuum emission will be much weaker than the line emission in the band widths we are interested in , but we also include the cluster continuum emission in our calculations . with the spatial resolution of 5 and the very long exposure time needed for the line detection , point source contaminations will be negligible , and hence it is not included in our calculations . in fact , the gxb we used has included a component from unresolved agns , and this may overestimate the total background . the total count rates of the backgrounds we used for the two instrument setups are listed in table [ ion_tab : bg ] , and are discussed below . we use the gxb simulated by @xcite which included two thermal components to represent the local hot bubble emission and the transabsorption emission @xcite , and one continuum emission component to represent unresolved agn background . the parameters used for their background model are based on @xcite . the most prominent emission around 0.6 kev is the line emission from nitrogen and oxygen ions ( figure 6 in * ? ? ? with the very high xms spectral resolutions , the and lines from clusters beyond @xmath144 should be separated from the strong line emission in the gxb ( see table [ ion_tab : min ] below ) . we adopt the continuum intensity value of @xmath145 photons @xmath48 s@xmath11 sr@xmath11 kev@xmath11 at around 0.6 kev , which is used in @xcite . the total count rate for the whole field of view ( @xmath146 ) which covers the energy range of the lines is given by @xmath147 , where @xmath148 is the bandwidth covered the lines . for the doublets , since the line separation is much smaller than the spectral resolutions of all the instruments we considered , @xmath148 is simply the spectral resolution of the corresponding instrument . for the triplets , the three lines should be well separated by the xms core array . hence , @xmath148 of is then three times the spectral resolution of the xms core array . however , for the xms full array , the triplets can not be resolved . therefore , @xmath149 is then given by the maximum separation of the triplets ( 13 ev ) plus the corresponding spectral resolutions . for a future mission like the _ ixo _ , the nxb is rather uncertain . we use the count rate of @xmath150 photons s@xmath11 arcmin@xmath151 kev@xmath11 at 0.6 kev for the xms estimated by the _ team @xcite . the total count rate for the whole field of view which covers the energy range of the lines is given by @xmath152 . to address the effects of the uncertainties , we also multiply the nxb by factors of 0.5 and 2 in our calculations ( section [ ion_sec : sn ] below ) . the cluster continuum emission within the very narrow energy bands we are interested in are much weaker than the line emission in cluster outer regions ( figure [ ion_fig : line_flux ] ) and the gxb . to detect the and lines , it is also important to observe regions where the cluster continuum emission is weak compared to the line emission . therefore , the cluster continuum emission should not be important when estimating the signal - to - noise ratio . nevertheless , we have included the cluster continuum emission in our calculations . to be conservative , we simply take the cluster continuum emission at @xmath153 mpc where the cluster continuum emission across 648.5658.7 ev ( energy range where the triplets are covered by the xms full array ) equal to the emission . the continuum count rates @xmath154 are listed in table [ ion_tab : bg ] . in order for the lines to be separated from the local gxb , clusters to be observed should be at high enough redshifts . since the triplets spread across 13 ev , the best targets should have redshifts such that the lines can be shifted by at least 13 ev plus the spectral resolution . the targets should also not to have too high redshift such that the triplets will not overlap with the 0.5 kev nitrogen line . table [ ion_tab : min ] lists the minimum and maximum redshifts for clusters to be observed by the xms core and the xms full arrays which meet these requirements . to calculate the signal of the and lines for clusters for both the xmsc and xmcf cases , we assume the nominal massive cluster model to be at a redshift of 0.05 . the redshift is chosen such that it is slightly higher than @xmath155 for the xms full array in table [ ion_tab : min ] . it is also important that there actually be clusters which are at or within the selected redshift which are fairly regular in shape in their outer regions and with masses which are comparable to the cluster model we calculated ( e.g. , abell 85 , abell 1795 ) . to calculate the total count rates of the lines from a cluster at the assumed redshift as observed by _ ixo _ ( @xmath156 ) , we first convolve the rest frame surface brightness of the lines ( figure [ ion_fig : line_flux ] ) with a top hat function with a width of 0.114 ( 0.286 ) mpc which corresponds to the physical distance at @xmath157 covered by the xmsc ( xmsf ) field of view , and then multiply the convolved rest frame surface brightness by a factor of @xmath158 to give the count rate @xmath156 . in calculating the signal - to - noise ratios , count rates at a radius where the surface brightness of the weaker line is maximum in the outer regions are used as the optimum model count rates . the optimum count rates are also listed in table [ ion_tab : bg ] . we calculate the signal - to - noise ratios for the optimum models ( 1 nxb , 1 gxb ) which use the count rates listed in table [ ion_tab : bg ] . the signal for each line is given by @xmath159 , where @xmath160 is the exposure time . the noise for each line is given by @xmath161 the signal - to - noise ratio of the line ratio is given by @xmath162^{-1/2 } \ , , % \end{equation}\end{aligned}\ ] ] where @xmath163 and @xmath164 are the signal - to - noise ratios of the and lines , respectively . to address the effects of the nxb uncertainties , we also calculate the signal - to - noise ratios for models with the nxb multiplied by a factor of 2 ( 2 nxb , 1 gxb ) and a factor of 0.5 ( 0.5 nxb , 1 gxb ) . the gxb varies with sky position . to address the uncertainties in the gxb , we vary the gxb by a factor of 0.8 ( 1 nxb , 0.8 gxb ) , 1.2 ( 1 nxb , 1.2 gxb ) , and 1.5 ( 1 nxb , 1.5 gxb ) . in real observations , the line signals may not be optimum . we consider models with the count rates of the lines to be a factor of 0.75 ( 0.75 @xmath156 ) , 0.50 ( 0.50 @xmath156 ) , and 0.25 ( 0.25 @xmath156 ) of the optimum models . the signal - to - noise ratios as a function of exposure time for the different models are shown in figure [ ion_fig : sn1 ] . to ensure that there are enough counts for the lines , we only plot the signal - to - noise ratios if the total count for each line @xmath165 is larger than 30 . for the xmsc instrument setup with the optimum model ( 1 nxb , 1 gxb ) , in order to get a signal - to - noise ratio of 3 for the triplets , about 220 ksec is needed . with a deeper observation of 600 ksec , the signal - to - noise ratio can exceed 5 . for the doublets , a shorter exposure time of @xmath166 ksec is need to have @xmath167 with a signal - to - noise ratio of 3.2 . about 450 ksec is needed to get a signal - to - noise ratio of 5 , and it is possible to get a signal - to - noise ratio up to 7 with a very deep observation of 900 ksec . if no single cluster is observed for such a long exposure , the very deep exposure can be achieved by stacking many observations of outer regions of many clusters . in this case , the line emission measured is average over many clusters . for a given exposure time , increasing the nxb by a factor of 2 or increasing the gxb by a factor of 1.5 only decrease the signal - to - noise ratios by about 10% for both the or lines . as expected , decreasing the line signals significantly decreases the signal - to - noise ratios . if the line signals are smaller than half of the optimum values , it will be impossible to detect the ( ) line with an exposure time less than @xmath168 ( 0.5 ) msec . limited by the detection , an exposure time of about 220 ksec is sufficient to have a 2.3-@xmath169 determination for the line ratio for our optimum model . a 3-@xmath169 measurement of the line ratio will require an exposure time of about 380 ksec . if the electron temperature can be measured from observations such as x - ray spectroscopy or hardness ratios , the measured line ratio can be compared to the cie ratio inferred by the electron temperature , and this will provide a direct test to the ionization states of the plasma without ambiguity . in the outermost regions of a cluster , the surface brightest may be too low and hence reliable temperature measurement may not be feasible . however , as shown in figure [ ion_fig : lineratio ] , beyond a radius of @xmath170 mpc where the electron temperatures are always higher than about 0.5 kev for either the non - equipartition model with @xmath29 or the equipartition model , the cie line ratios are always higher than @xmath171 . in contrast , the nei line ratios are always lower than @xmath8 for the electron temperature range of interest in realistic clusters for all the models we considered . hence , a 3-@xmath169 measurement of the line ratio @xmath172 ( @xmath173@xmath137@xmath174 \lesssim$ ] 2/3 at 1-@xmath169 ) is a strong evidence that the ions are in non - equilibrium ionization due to the huge difference between the nei and cie line ratios for the temperature range of realistic clusters . if the line ratio is measured to be even lower ( e.g. , @xmath175 ) , only a 2 or 3-@xmath169 will be sufficient to rule out the cie model . on the other hand , if the line ratio is measured to be as high as 4 in the outermost regions , a 6-@xmath169 ( @xmath173@xmath137@xmath110 ] = 2/3 at 1-@xmath169 ) will be necessary to rule out the nei low line ratio of @xmath8 at a 3-@xmath169 level . even if the line ratio is measured within about @xmath177 mpc where the line ratio for the nei model can be as high as @xmath171 , the surface brightness at this relative small radius is high enough so that electron temperature can be measured by the _ ixo _ or even _ suzaku _ ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) with confidence . to study whether the _ ixo _ xms can measure the electron temperature at this radius , we have carried out simulations using xspec with our model spectrum taken at radius @xmath8 mpc the detector response and background files were taken from the latest _ ixo _ simulator simx . both the nxb and the gxb have been included in the background file . we assumed an absorption model _ wabs _ with @xmath178 @xmath48 . we found that the _ ixo _ xms core detector can collect about 2400 photon counts ( 0.23.0 kev ) for a 150 ksec observation . when fitting the continuum of the spectrum , we assumed an absorbed thermal model ( _ wabs*apec _ ) and ignored the strong lines . using the _ wabs*nei _ model , the temperature determined is within the uncertainty of the _ wabs*apec _ model and does not affect the results significantly . we found that the electron temperature can be constrained to @xmath179 kev . the cie line ratios within the determined temperature range are always higher than 13 , and hence a 2 to 3-@xmath169 measurement for the low line ratio will also be enough to test the non - equilibrium ionization model . in fact , the _ ixo _ wfi should constrain the electron temperature better because of its lower particle background @xcite . for the xmsf instrument setup , with the optimum model ( 1 nxb , 1 gxb ) , about @xmath180 ksec is needed to get a signal - to - noise ratio of 3 for the triplets ( doublets ) . a longer exposure of @xmath181 ksec is needed to get a signal - to - noise ratio of 5 for the ( ) lines , and a very deep exposure of @xmath182 msec is needed for a 8 ( 10)-@xmath169 detection . increasing the nxb by a factor of 2 or increasing the gxb by a factor of 1.5 decreases the signal - to - noise ratios by less than @xmath183 for both the and lines . similarly to the xmsc instrument setup , the signal - to - noise ratios decrease as the line signals decrease . for the optimal model , a shorter exposure time of about 130 ksec ( also limited by the detection ) compared to the xmsc setup will give the same 2.3-@xmath169 detection for the line ratio . an exposure time of about 230 ksec will be needed for a 3-@xmath169 measurement of the line ratio . overall , the xmsf instrument setup can achieve higher signal - to - noise ratios for a given exposure time compared to the xmsc setup for the optimum models . this is mainly due to the larger field of view of the detector which can collect more photons with the same exposure time . the xmsf is slightly more subject to the continuum background uncertainties ( cluster continuum emission , nxb , and gxb ) due to the poorer spectral resolution . for both the xmsc and xmsf setups , detecting the and lines and testing non - equilibrium ionization in cluster outer regions are promising . studying the physics in the outer regions of clusters is very important to understand how clusters are formed , how the the intracluster gas is heated , as well as to constrain the formation of large - scale structure . because of the very low density in cluster outer regions , the collisional timescales are very long and comparable to the cluster age . electrons and ions passed through the accretion shocks may not have enough time to reach equipartition and the ions may be under - ionized . in a previous paper @xcite , we have studied the non - equipartition effects on clusters using one - dimensional hydrodynamic simulations . in this paper , we systematically studied non - equilibrium ionization effects on clusters and the x - rays signatures using the same set of simulations we have developed @xcite . by using semi - analytic arguments together with numerical simulations , we have shown that the non - equilibrium ionization effect is nearly independent of cluster mass but depends strongly on redshift . in particular , non - equilibrium ionization effects are stronger for low - redshift clusters . therefore , the brighter massive clusters at low - redshifts are good candidates for studying the non - equilibrium ionization effects . we systematically studied non - equilibrium ionization signatures in x - rays for a massive cluster with @xmath184 . we first calculated the ionization fractions for 11 elements heavier than he following the electron temperature and density evolutions of each fluid element . we then calculated the x - ray emissivity of each fluid element and the resulting projected spectra for the cluster . since the electron temperature profiles depend on electron heating efficiency @xmath27 , we have considered three different possibilities which represent a very low heating efficiency ( @xmath29 ) , an intermediate heating efficiency ( @xmath30 ) , and equipartition , @xmath31 . we also considered models which assume equilibrium ionization for comparison . at a radius ( e.g. , 2 mpc ) where the ionization timescale is long , the overall spectra for the nei and cie non - eq models are very similar . this is because of the dominant free - free emission , and both models assume the same electron temperature . however , in the outer regions , e.g , at @xmath185 mpc which is between @xmath5 and @xmath186 , the soft emission in the nei model is dominated by line emission , where the cie non - eq spectrum is still dominated by the continuum free - free emission . by analyzing the surface brightness profiles , we found that soft emission ( 0.31.0 kev ) for the nei model can be enhanced by more than 20% at around 3 mpc , and up to nearly an order of magnitude near the shock radius compared to the cie non - eq model . the soft emission enhancement is mainly due to the line emission from under - ionized ions . the non - equilibrium ionization effects on the medium ( 1.02.0 kev ) and hard ( 2.010.0 kev ) band emissions are smaller . the overall x - ray band ( 0.310.0 kev ) emission is dominated by the soft emission , and the total x - ray emission for the nei model decrease much slower than that of the cie non - eq model . thus , if cluster outer regions are in non - equilibrium ionization , the shock region will be much more luminous compared to the cie non - eq model . by inspecting a number of spectra , we found that the most prominent non - equilibrium ionization signature in line emission is the line ratio of the he - like triplets and the h - like doublets , @xmath136@xmath137@xmath110 . the line ratios for the cie models are higher than 10 for most regions between @xmath187 14 mpc , while the line ratios are smaller than 10 for the nei models . the differences in the line ratios between the nei and cie models increase with radius , and the differences are more than an order of magnitude for radii beyond @xmath188 mpc . these results are insensitive to the degree of non - equipartition or electron heating efficiency @xmath27 . we suggest that the line ratios can be used to distinguish between the nei and cie models . the electron temperature profile can be determined from fits to the continuum spectra of the outer regions of clusters , allowing the cie line ratios to be determined . comparison to the observed ratios should show the effects of non - equilibrium ionization . note that a line ratio of @xmath136@xmath189@xmath190 in the outer region of a massive clusters is a clear signal of nei . we have also studied the detectability of the and lines around cluster accretion shock regions with _ ixo _ , as well as the test for non - equilibrium ionization using the line ratio . for our optimum model , we found that with the xms core array , an exposure time of 220 ksec is need to have a 3.0-@xmath169 detection of the lines and about 180 ksec is need to have @xmath191 counts for a 3.2-@xmath169 detection of the lines . the uncertainties in nxb and gxb will not affect the results significantly . for the xms full array while we assume the spectral resolution to be the same as the outer array throughout the detector , we found that the signal - to - noise ratios for our optimum model are higher for the same exposure time as the xms core array . in particular , only about 130 ( 100 ) ksec is needed to detect the ( ) line . the xms full array is only slightly more subject to nxb and gxb uncertainties due to the poorer spectral resolution . to test the non - equilibrium ionization model without ambiguity requires measurements of both the electron temperature ( or hardness ratio ) and the line ratio so that the measured line ratio can be compared to the cie line ratio inferred by the electron temperature . we have shown that this can be done within @xmath172 mpc of a cluster by the _ ixo _ with sufficient confidence . beyond @xmath170 mpc where the surface brightness may be too low and measuring the electron temperature may be difficult , we have shown that if the line ratio is measured to be as low as @xmath8 at 3-@xmath169 ( @xmath192@xmath137@xmath174 \sim 2/3 $ ] at 1-@xmath169 ) in the outermost regions , this will rule out the cie model at a 3-@xmath169 level since the cie line ratios are always higher than 4 for realistic cluster temperatures . a 3 or 4-@xmath169 measurement of such a low line ratio is sufficient to provide a strong test of the non - equilibrium ionization . if the line ratio is measured to be even lower ( e.g. , @xmath175 ) , only a 2 or 3-@xmath169 will be sufficient to rule out the cie model . on the other hand , if the line ratio is measured to be as high as 4 , a 6-@xmath169 measurement ( @xmath173@xmath137@xmath110 ] = 2/3 at 1-@xmath169 ) will be necessary to rule out the nei low line ratio of @xmath8 at a 3-@xmath169 level . we found that an observation with about 130 ( 220 ) ksec with the xms full ( core ) array is enough to measure the line ratio at 2.3-@xmath169 . for a 3-@xmath169 measurement of the line ratio , about 230 ( 380 ) ksec will be needed for the xms full ( core ) array , and this will provide a strong test for non - equilibrium ionization . in summary , detecting the and lines around the cluster accretion shock regions and testing non - equilibrium ionization in cluster outer regions with _ ixo _ are promising . it is expected that the and lines from whim will also be strong . because of the high spectral resolution of the xms , emissions from different redshifts should be easily separated . only the emission from whim immediately surrounding the target cluster will be potentially confused with the emission from cluster outer regions . to observe the and lines and study nei effects in cluster accretion shock regions , it will be best to avoid observing directions along the filaments where it is believed that denser preheated whims and subclusters are preferentially accreted onto more massive and relaxed galaxy clusters . what do we learn about clusters from the ionization state of the outer gas ? since collisional ionization and recombination rates involve straightforward atomic physics , the processes are not in question and the rates are reasonably well - known . unlike shock electron heating or rates for transport processes like thermal conduction , the basic physics is not uncertain and magnetic fields do not affect the results in a significant way . what we mainly learn about is the pre - shock physical state of materials which are being accreted by the cluster . if most of the whim is ionized beyond , then the effects described in this paper will be greatly reduced . if most of the material currently being accreted by clusters comes in through filaments which have a higher ionization , then nei effects will be diminished significantly . if most of the gas being added to clusters at present comes in through mergers with groups which deposit most of the gas in the inner regions of clusters , the gas will achieve cie quickly . from the theoretical point of view , with the increasing number of observations of galaxy cluster outer regions ( @xmath193 ) and the potential to extend observations out to the shock radius with _ ixo _ in the future , it is necessary to perform more detailed simulations than ours . it is also interesting to extend our work to study the connections between the shocked icm and the more diffuse whim surrounding clusters . three - dimensional simulations will be essential to understand the effects of mergers or filament accretion on the degree of ionizations in different regions of clusters . this will allow us to characterize the variation of non - equilibrium signatures in the clusters ; such calculations are essential to compare observational signatures with our understanding of the cluster physics near the accretion shocks . cosmological simulations have been performed recently to study the nei signatures @xcite . these studies have shown that both non - equipartition and non - equilibrium ionization effects are important in cluster outer regions ; although they focus more on the lower density and lower temperature whim . high resolution simulations were also performed for studying nei effects in clusters , but these are limited to binary mergers with idealized initial conditions and focus on the denser merger shocks @xcite . re - simulating representative clusters and the surrounding whim from cosmological simulations with higher resolutions and including realistic physics ( e.g. , cooling , conduction , turbulent pressure , magnetic pressure , and relativistic support by cosmic rays ) will be necessary to provide realistic model images and spectra . the different observational signatures and connections between the icm and the more diffuse whim can also be addressed self - consistently by these simulations . we thank daniel wang and todd tripp for helpful discussions . support for this work was provided by nasa through _ chandra _ grants go7 - 8129x , go7 - 8081a , go8 - 9083x , go9 - 0135x , and go9 - 0148x , _ xmm - newton _ grants nnx08az34 g , and nnx08aw83 g , and _ suzaku _ grants nnx08az99 g , nnx09ah25 g , and nnx09ah74 g . we thank the referee for helpful comments . smith , r. k. , bautz , m. w. , bookbinder , j. , garcia , m. r. , guainazzi m. , & kilbourne , c. a. 2010 , spie , 7732 , 773246 ( http://ixo.gsfc.nasa.gov/resources/published_papers/docs/ spie_2010/ixo - bkgnd - spie.pdf ) ( equation [ ion_eq : iontime ] ) versus radius @xmath19 scaled to the cluster shock radius @xmath104 for cluster models with total accreted masses of @xmath196 ( dashed line ) , 1.53 ( solid line ) , and 3.06 ( dotted line ) @xmath197 @xmath198 at a redshift of @xmath45 . the shock radii for the three clusters from small to high mass are @xmath199 , and 5.41 mpc , respectively . all three lines lie almost on top of one another . , width=453 ] for versus the scaled radius ( @xmath103 ) for clusters with total accreted masses of @xmath200 ( dashed line ) , 1.53 ( solid line ) , and 3.06 ( dotted line ) @xmath197 @xmath198 at a redshift of @xmath45 . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . , width=453 ] for versus the scaled radius ( @xmath103 ) for clusters at different redshifts . the cluster model with a total accreted mass @xmath201 at @xmath45 is used . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . , width=453 ] ) for clusters with total accreted masses of @xmath200 ( dashed line ) , 1.53 ( solid line ) , and 3.06 ( dotted line ) @xmath197 @xmath198 at a redshift of @xmath45 . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . , width=453 ] ) for clusters at different redshifts . the cluster model with a total accreted mass @xmath201 at @xmath45 is used . models with @xmath29 ( non - equipartition ) and @xmath31 ( equipartition ) are shown in thick and thin lines , respectively . , width=453 ] at @xmath45 are shown at two projected radii ( left : 2 mpc ; right : 3.5 mpc ) . upper panels : the nei model with @xmath140 and non - equipartition of electrons and ions . lower panels : projected rest frame spectra for the cie model with the same @xmath27 and non - equipartition . , width=453 ] ( thick lines ) and the cie non - eq model ( thin lines ) . upper right panel : rest frame projected surface brightness profiles for the nei model with @xmath140 ( thick lines ) and the cie eq model ( thin lines ) . lower left panel : ratios of the surface brightness profiles @xmath126 . lower right panel : ratios of the surface brightness profiles @xmath127 . , width=453 ] . middle panel : projected rest frame spectra for the nei model with @xmath202 . lower panel : projected rest frame spectra for the cie non - eq model with @xmath29 . all spectra are for the @xmath184 at @xmath45 , and are for a projected radius of @xmath121 mpc . the spectra are binned with @xmath203 ev . , width=453 ] @xmath137@xmath110 for the nei models with @xmath29 ( solid ) , 0.5 ( dash - dotted ) , and 1.0 ( dash - dot - dotted ) . line ratios for the cie non - eq ( dashed ) and cie eq ( dotted ) models are also shown . , width=453 ]

let be a complex finite - dimensional semisimple lie algebra and be any-subalgebra of . in this paper we prove an earlier conjecture by penkov and zuckerman claiming that the first derived zuckerman functor provides an equivalence between a truncation of a thick parabolic category for and a truncation of the category of admissiblemodules . this latter truncated category consists of admissiblemodules with sufficiently large minimal-type . we construct an explicit functor inverse to the zuckerman functor in this setting . as a corollary we obtain an estimate for the global injective dimension of the inductive completion of the truncated category of admissiblemodules . * key words : * generalized harish - chandra module , zuckerman functor , thick category . mathematics subject classificiation 2010 : primary 17b10 , 17b55

1604.04672

let @xmath0 be a complex finite - dimensional semisimple lie algebra and @xmath5 be a reductive in @xmath0 subalgebra . admissible @xmath6-module _ is a @xmath0-module on which @xmath1 acts semisimply , locally finitely , and with finite multiplicities . the study of the category of admissible @xmath7-modules is a main objective of the theory of generalized harish - chandra modules , see @xcite . in the case of a general reductive in @xmath0 subalgebra @xmath1 , a central result of the existing theory of generalized harish - chandra modules is the classification of simple admissible @xmath6-modules with generic minimal @xmath1-type @xcite . other notable results for a general @xmath1 are established in @xcite , @xcite , @xcite , and @xcite . there are three special cases for @xmath1 in which more detailed information on admissible @xmath6-modules is available . first of all , this is the case when @xmath1 is a symmetric subalgebra of @xmath0 , i.e. , @xmath1 coincides with the fixed points of an involution on @xmath0 . this case , the theory of harish - chandra modules , is in the origin of the studies of generalized harish - chandra modules . there is an extensive literature on harish - chandra modules , see for instance @xcite , @xcite , and references therein . ( in particular , some remarks on the history of harish - chandra modules can be found in @xcite . ) another case which has drawn considerable attention is the case when @xmath8 , see for instance @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , and references therein . in both these cases , a classification of simple admissible @xmath6-modules is available and there has been progress in the study of the category of admissible @xmath6-modules . a third natural choice for @xmath1 is to let @xmath1 be isomorphic to @xmath2 . this case `` interpolates '' between the above two cases and is a natural experimentation ground when aiming at the case of a general @xmath1 . for @xmath9 , there is no classification of simple admissible @xmath6-modules for a general @xmath0 and an arbitrary @xmath2-subalgebra @xmath10 ; however , for @xmath11 the partial classification of @xcite can be carried out under much less severe restrictions on the minimal @xmath1-type : the details are explained in @xcite and @xcite . since the @xmath1-types are parametrized here simply by nonnegative integers , one can talk about a truncated category of admissible @xmath6-modules : it consists of finite - length admissible modules whose minimal @xmath1-type is larger or equal a bound @xmath12 depending on the pair @xmath6 . the simple objects of this truncated category have been classified in @xcite ( see also @xcite ) . the purpose of this paper is to describe the above truncated category of admissible @xmath6-modules for @xmath13 by proving that it is equivalent to an explicit full subcategory of a thick parabolic category @xmath3 for @xmath0 . in fact , the objects of the truncated category of @xmath6-modules are constructed by simply applying the zuckerman ( first derived ) functor @xmath14 to a subcategory of a thick parabolic category @xmath3 . it was conjectured in @xcite that the functor @xmath14 yields an equivalence of these categories , and here we prove this conjecture . we construct a left adjoint to @xmath14 defined on all finitely generated admissible @xmath6-modules , and then show that , when restricted to the truncated category of admissible @xmath7-modules , this functor is an inverse to the appropriately restricted functor @xmath14 . the history of @xmath15-modules goes back to the 1940 s : a classical example here is the lorentz pair @xmath16 studied by harish - chandra @xcite , gelfand - minlos - shapiro @xcite , and others . explaining how exactly the theorem proved in this paper fits in the 70-year history of the topic is a task so complex that we do not really attempt to tackle it nonetheless , we would like to mention that in this subject many equivalences of categories have been established ; some relate algebraic categories of @xmath0-modules to geometric categories of sheaves , others relate algebraic categories of @xmath0-modules to other algebraic categories of @xmath0-modules . the equivalence we establish is clearly of the second kind and could be seen as an analogue of bernstein - gelfand s equivalence of a certain subcategory of harish - chandra bimodules ( or @xmath17-modules ) with category @xmath3 . an extension of the geometric techniques introduced by beilinson and bernstein from the theory of harish - chandra modules to generalized harish - chandra modules is not straightforward ( see some results in this direction in @xcite , @xcite , and @xcite ) , and fitting the main result of the present paper into a geometric contex is an open problem . we show , however , that the algebraic methods from the 1970 s ( where , in addition to the third author s contribution , we would like to mention to important contributions by enright - varadarajan and enright ) , together with the more recent ideas of @xcite , @xcite , and @xcite ( which are building up on vogan s work ) , are well suited to yield concrete results about the structure of categories of generalized harish - chandra modules . the paper is structured as follows . we state the main result in section [ sec : main ] . in particular , we introduce the functor @xmath18 which will then be shown to be inverse to the functor @xmath14 . in section [ sec : prelim ] we present some results which deal mostly with the structure of a semi - thick parabolic category @xmath19 we work with . section [ sec : conjugacy ] contains the proof of the adjointness of @xmath14 and @xmath18 . the proof of the fact that @xmath14 and @xmath18 are mutually inverse equivalences of categories is carried out in steps throughout sections [ sec : bijection],[sec : exactness ] and [ sec : endofproof ] . in section [ sec : ex ] we show that for some blocks of the semi - thick parabolic category @xmath19 , the truncation condition is vacuous , which then helps establish a stronger equivalence of categories for certain central characters . finally , in section [ sec : appl ] we provide an application of our equivalence of categories by proving an estimate for the global dimension of the truncated category of admissible @xmath7-modules via a correponding estimate for the truncated semi - thick parabolic category @xmath19 . * acknowledgements . * the first and second authors acknowledge the hospitality of the mittag - leffler institute in djursholm where a significant part of this work was written up . the first author thanks the dfg for partial support through priority program spp 1388 and grant pe 980/6 - 1 . the second author acknowledges partial support from nsf through grant number dms 1303301 , and both the second and third authors acknowledge the hospitality of jacobs university bremen . the ground field is @xmath20 . by @xmath0 , we will denote a fixed finite - dimensional semisimple lie algebra . we fix also an @xmath2-subalgebra @xmath21 by @xmath22 we denote the orthogonal ( with respect to the killing form ) complement of @xmath23 in @xmath24 . the classification of all possible subalgebras @xmath1 up to conjugacy is equivalent to describing all nilpotent orbits in @xmath0 , and goes back to malcev and dynkin ( see @xcite and the references therein ) . by a _ @xmath1-type _ we mean a simple finite - dimensional @xmath2-module . a simple finite - dimensional @xmath2-module with highest weight @xmath25 is denoted by @xmath26 . by @xmath27 ( respectively , @xmath28 ) , we denote the socle ( respectively , the top ) of a @xmath24-module @xmath29 of finite length . @xmath27 is the maximal semisimple submodule of @xmath29 and @xmath28 is the maximal semisimple quotient of @xmath29 . by @xmath30 $ ] we denote the multiplicity as a subquotient of a simple module @xmath31 in a module @xmath32 . @xmath33 stands for the restriction of a module @xmath29 to a subalgebra @xmath34 , and @xmath35 stands for the direct sum of @xmath36 copies of @xmath29 . the main result of this paper states that certain categories of @xmath0-modules are equivalent via explicit mutually inverse functors . here we define these categories and functors . recall that @xmath0 is a finite - dimensional semisimple lie algebra and @xmath1 is an arbitrary @xmath2-subalgebra of @xmath0 . fix a standard basis @xmath37\big\}$ ] of @xmath1 and note that @xmath38 is a semisimple element of @xmath0 . let @xmath39 be the toral subalgebra of @xmath0 spanned by @xmath38 . define the parabolic subalgebra @xmath40 of @xmath0 by setting @xmath41 by @xmath42 we denote the opposite parabolic subalgebra @xmath43 where @xmath44 is the centralizer of @xmath38 in @xmath45 . we also set @xmath46 let @xmath47 be the category of finite - length @xmath24-modules which are @xmath42-locally finite , @xmath48-semisimple , and @xmath48-integral ( i.e. , @xmath38 acts with integer eigenvalues ) . informally , @xmath47 is a `` semi - thick '' ( `` thick in all directions except the @xmath48-direction '' ) parabolic category @xmath3 . by @xmath49 for @xmath50 , we denote the @xmath51-truncated category @xmath47 , i.e. , the full subcategory of @xmath47 consisting of objects all @xmath48-weights @xmath52 of which satisfy @xmath53 . we also assign an integer @xmath12 to the pair @xmath7 : we set @xmath54 , where @xmath55 ( respectively , @xmath56 ) is the maximum ( resp . , submaximum ) weight of @xmath48 in @xmath57 . here and below , we identify @xmath48-weights with integers via the correspondence @xmath58 . denote by @xmath59 the category of admissible @xmath7-modules of finite length , i.e. , the category of @xmath24-modules @xmath29 of finite length on which @xmath23 acts locally finitely and such that @xmath60 for any @xmath23-type @xmath61 . by @xmath62 for @xmath50 , we denote the full subcategory of @xmath59 consisting of @xmath24-modules @xmath29 such that @xmath63 implies @xmath64 . we now describe two functors : @xmath65 and @xmath66 . @xmath65 is the functor of @xmath23-finite vectors in a @xmath67-module . that is , if @xmath29 is a @xmath67-module then @xmath68 and @xmath69 is a @xmath24-submodule of @xmath29 . it is well known ( and easy to see ) that @xmath65 is a left - exact functor . in what follows we set @xmath70 and denote the right derived functors @xmath71 by @xmath72 . ( @xmath72 is known as the _ @xmath73-th zuckerman functor_. ) by definition , @xmath72 is a functor from @xmath74 to @xmath75 . it is proved in @xcite that the restriction of @xmath72 to @xmath76 is a well - defined functor from @xmath76 to @xmath62 . we denote this functor also by @xmath72 . we now define a functor @xmath77 where @xmath78 stands for the category of finitely generated @xmath67-modules . throughout the rest of the paper , @xmath79 will denote a fixed central character . if @xmath29 is a @xmath24-module , then @xmath80 stands for the vectors in @xmath29 on which @xmath81 acts locally nilpotently for any @xmath82 . by @xmath83 we denote a variable positive integer . let @xmath84 be the subcategory of @xmath85 consisting of modules @xmath29 with @xmath86 and such that @xmath87 acts via jordan blocks of size at most @xmath83 . we note that @xmath84 is a finite - length category which has an injective cogenerator @xmath88 . this fact is proved in lemma [ lem : lemma1 ] below . we set @xmath89 for @xmath90 . lemma [ lem : lemma3 ] below claims that @xmath91 , which shows that @xmath92 is the `` largest quotient '' of @xmath93 lying in @xmath84 . next , we notice that there is a canonical surjective homomorphism @xmath94 i.e. , that @xmath95 is an inverse system of @xmath42-locally finite @xmath67-modules . we set @xmath96 it is easy to see that @xmath97 is a right - exact functor from @xmath98 to @xmath99 , and we denote by @xmath100 its left derived functors , that is @xmath101 for @xmath90 . let @xmath102 and @xmath103 be the respective subcategories of @xmath62 and @xmath76 consisting of @xmath24-modules @xmath29 with @xmath86 . corollary [ cor : cor1_of_prop2 ] below states that in fact @xmath100 is a well - defined functor from @xmath104 to @xmath105 . as @xmath23 and @xmath48 are fixed , in what follows we set @xmath106 , @xmath107 , and @xmath108 . by the same letters we also denote the restrictions of these functors to the category @xmath109 . the main result of this paper is the following . [ main ] for @xmath110 , let @xmath111 . then , for any @xmath112 , the functors @xmath113 and @xmath114 are mutually inverse equivalences of categories . let us note that for the lorentz pair @xmath115 a description of the category @xmath59 was given by i. gelfand and v. ponomarev in @xcite already in 1967 . we start with two general results . [ prop : spec ] a. for any @xmath7-module @xmath93 there exists a singly graded spectral sequence converging to @xmath116 such that its @xmath117-term has the form @xmath118 b. if @xmath110 for @xmath119 , then the @xmath120-homology @xmath121 is finite dimensional . a. the statement follows from proposition 1 in @xcite if one sets @xmath73 to be the total degree @xmath122 in the notation of @xcite . the statement holds more generally for any @xmath24-module @xmath93 but the assumption that @xmath93 is a @xmath7-module is sufficient for us . b. this follows from the more general statement of proposition 3.5 in @xcite . let @xmath123 be an admissible @xmath67-module where @xmath124 is the @xmath48-weight space in @xmath29 of weight @xmath125 : by definition , @xmath126 for @xmath127 . set @xmath128 . then @xmath129 is a well - defined admissible @xmath67-module . similarly let @xmath130 be an admissible @xmath7-module . here @xmath131 stands for the @xmath26-isotypic component in @xmath93 . then @xmath132 is a well - defined admissible @xmath7-module . moreover , @xmath133 and @xmath134 are well - defined contravariant functors ( in fact antiequivalences ) on the respective categories of admissible @xmath67-modules and @xmath7-modules . in what follows we will use the composition of the functors @xmath133 and @xmath134 with the twist by the cartan involution of @xmath24 . the so obtained new functors are denoted respectively by @xmath135 and @xmath136 . these functors preserve the respective @xmath48- and @xmath23-characters of the modules . a duality theorem proved in @xcite implies [ prop : duality ] for any admissible @xmath67-module @xmath29 there is a natural isomorphism of admissible @xmath7-modules @xmath137 in what follows @xmath138 stands for a finite - dimensional simple @xmath139-module on which @xmath38 acts via a natural number @xmath140 . often , we consider @xmath138 as a @xmath42-module by setting @xmath141 . in this case , we set also @xmath142 and we let @xmath143 be the unique simple quotient of @xmath144 . then @xmath145 , and @xmath146 is an indecomposable object of @xmath47 with @xmath147 . the following proposition is a summary of preliminary results concerning the specific categories we study in this paper . [ prop : zuck ] let @xmath148 . then a. @xmath149 is an exact functor which maps a simple object to a simple object , and induces a bijection on the isomorphism classes of simple objects in @xmath85 and in @xmath62 , respectively . moreover , the simple @xmath7-module @xmath150 has minimal @xmath23-type @xmath151 . b. @xmath150 and @xmath152 are isomorphic simple @xmath7-modules with minimal @xmath23-type @xmath151 , and the isotypic components of the minimal @xmath23-types of @xmath153 and @xmath150 are isomorphic . part a ) follows directly from the results of @xcite , see corollary 6.4 and section 9 . part b ) is a consequence of the above mentioned results and the fact that the functor @xmath14 commutes with @xmath154 according to proposition [ prop : duality ] . since @xmath14 preserves central character , proposition [ prop : zuck ] , a ) implies in particular the existence of a bijection between the isomorphism classes of simple objects in @xmath103 and @xmath109 for @xmath112 . without the condition @xmath112 , no such bijection exists in general . for instance , if @xmath7 is the lorentz pair and @xmath155 is the central character of a finite - dimensional @xmath24-module of the form @xmath156 for a simple finite - dimensional @xmath2-module @xmath157 , then @xmath158 has @xmath159 pairwise nonisomorphic simple objects while @xmath160 has two nonisomorphic simple objects . the rest of the section is devoted to results on @xmath42-locally finite modules . note that @xmath139 is a reductive subalgebra : we denote by @xmath161 the derived subalgebra of @xmath139 , and by @xmath162 the center of @xmath139 . we choose a borel subalgebra @xmath163 of @xmath24 such that @xmath164 . let @xmath87 be a cartan subalgebra of @xmath163 containing @xmath38 . then @xmath165 . let @xmath166 be the category of locally finite @xmath139-modules semisimple over @xmath48 with integral @xmath38-weights . by @xmath167 we denote the subcategory of @xmath166 consisiting of modules on which @xmath162 acts via jordan blocks of size less than or equal to @xmath83 . clearly @xmath168 note that @xmath169 is also a well - defined functor on the category @xmath167 ( but not on @xmath170 ) . [ small ] let @xmath171 be a simple finite - dimensional @xmath161-module and @xmath172 be a @xmath48-integral weight . define @xmath138 as @xmath173 where @xmath174 is a one - dimensional @xmath162 module with weight @xmath175 . let @xmath176 denote the ideal in @xmath177 generated by @xmath178 and @xmath179 for all @xmath180 . then a. every simple object in @xmath166 is isomorphic to @xmath138 for some choice of @xmath171 and @xmath175 . furthermore , @xmath181 . b. @xmath182 is a projective cover of @xmath138 and @xmath183 is an injective hull of @xmath138 in @xmath167 . c. @xmath184 is an injective hull of @xmath138 in @xmath166 . [ lem : lem_inj_hull ] \(a ) is obvious . to show ( b ) , note that @xmath171 is projective in the category of locally finite @xmath161-modules , and that @xmath185 is the maximal quotient of the induced module @xmath186 lying in in @xmath167 . then ( c ) is clearly a corollary of ( b ) . the following generalizes basic results in @xcite . for any @xmath187 and any @xmath119 , the abelian category @xmath84 has a unique , up to isomorphism , minimal injective cogenerator @xmath88 . moreover , the @xmath48-weight spaces of @xmath88 are finite dimensional . [ lem : lemma1 ] we denote by @xmath188 the category of @xmath189-modules whose restrictions to @xmath139 belong to @xmath167 , and by @xmath190 the subcategory consisting of modules whose @xmath48-weights are bounded from below by @xmath191 . let @xmath138 be as in lemma [ lem : lem_inj_hull ] . endow @xmath138 with a @xmath189-module structure by letting @xmath120 act trivially on @xmath138 , and consider the @xmath189-module @xmath192 since @xmath193 preserves injectivity and the functor of @xmath48-weight vectors @xmath194 is right adjoint to the inclusion of the category of @xmath189-modules semisimple over @xmath48 into the category of all @xmath189-modules , and hence also preserves injectivity , the @xmath189-module ( [ eq : gammapromodule ] ) is an injective hull of @xmath138 in @xmath188 . consequently , the truncated submodule @xmath195 of ( [ eq : gammapromodule ] ) , spanned by all @xmath48-weight spaces with weights greater or equal than @xmath191 , is an injective hull of @xmath138 in @xmath190 . set @xmath196^\theta.\ ] ] then , by a similar argument , @xmath197 is injective in @xmath84 and we have an embedding of @xmath24-modules @xmath198 induced by the embedding of @xmath189-modules @xmath199 . moreover , it is easy to check that @xmath197 has finite - dimensional @xmath48-weight spaces . note that , up to isomorphisms , @xmath84 has finitely many simple objects @xmath200 . each of them has a unique , up to isomorphism , injective hull @xmath201 which is a submodule of @xmath202 . then @xmath88 is the direct sum @xmath203 . [ cor : cor2 - 0 ] let @xmath204 . then @xmath205 is equivalent to the category of finite - dimensional @xmath206-modules . let @xmath207 be the category of inductive limits of objects from @xmath208 . [ cor : cor2 ] for any @xmath51 , the category @xmath207 has a unique , up to isomorphism , minimal injective cogenerator @xmath209 . moreover , @xmath210 . in particular , the category @xmath207 has enough injectives . in fact , if @xmath211 is any injective object in @xmath207 , then @xmath211 is a direct limit , @xmath212 , for injective objects @xmath213 . recall the definition of @xmath214 from lemma [ small ] , and let @xmath215 [ injectivecoverma ] let @xmath216 be the category of locally finite @xmath42-modules such that their restrictions to @xmath139 lie in @xmath166 . then @xmath217 is an injective hull of @xmath138 in @xmath216 . moreover , @xmath218 for some finite - dimensional irreducible @xmath139-modules @xmath219 . by the poincare birkhoff witt theorem we have an isomorphism @xmath220 since @xmath221 is an injective module in @xmath170 , @xmath222 is an injective module in @xmath223 . hence @xmath224 is an injective module in @xmath223 . let @xmath225 be the socle of @xmath224 as a module over @xmath139 . since @xmath226 is locally @xmath139-finite , it is an essential extension of @xmath225 . we conclude that @xmath226 is an injective hull of @xmath227 for some finite - dimensional irreducible @xmath139-modules @xmath219 . we now show that @xmath228 . let @xmath229 be the direct sum of the @xmath139-modules @xmath230 . then @xmath229 is an essential extension of @xmath225 . because @xmath231 is left noetherian , @xmath229 is injective . consequently , @xmath229 is an injective hull of its socle . but @xmath232 . since @xmath226 is also an injective hull of @xmath225 , there is an isomorphism @xmath233 . we say that an object @xmath234 admits a _ co - verma filtration _ if there exists a finite filtration @xmath235 whose successive quotients @xmath236 are isomorphic to @xmath237 for simple @xmath139-modules @xmath238 . [ covermafilt ] let @xmath29 be an object of @xmath239 . then @xmath29 admits a co - verma filtration if and only if @xmath240 is injective in @xmath216 with socle of finite length . if @xmath29 admits a co - verma filtration , then @xmath240 is a direct sum of modules of the form @xmath241 , and by lemma [ injectivecoverma ] @xmath240 is injective in @xmath216 with @xmath242-socle of finite length . to prove the opposite assertion , choose a simple @xmath242-submodule @xmath243 with minimal @xmath140 . the existence of @xmath138 follows from the fact that the socle of @xmath240 has finite length . let @xmath244 be the multiplicity of @xmath138 in @xmath245 . then we have a surjective morphism @xmath246 of @xmath189-modules ( @xmath247 being the identity map ) which induces a morphism @xmath248 of @xmath24-modules by frobenius reciprocity . since @xmath240 is injective in @xmath223 , @xmath240 is an injective hull of its socle , i.e. , @xmath249 by lemma [ injectivecoverma ] . moreover , @xmath250 if @xmath251 . therefore @xmath252 and @xmath253 is an isomorphism of @xmath242-modules since it is induced by the identity map @xmath254 . set @xmath255 . then @xmath256 is isomorphic to @xmath257 , and hence @xmath258 satisfies all conditions of the lemma . so we can finish the proof by induction on the length of the socle of @xmath240 . [ technical ] let @xmath259 for some @xmath260 . suppose that @xmath261 admits a co - verma filtration . then @xmath29 and @xmath262 also admit co - verma filtrations . a direct summand of an injective module is injective , so the statement follows from lemma [ covermafilt ] . [ lem : lemma2 ] let @xmath263 be an injective hull of @xmath143 in @xmath207 . then @xmath264 admits a co - verma filtration with successive quotients isomorphic to @xmath265 for @xmath266 . let @xmath267^\theta$ ] . the @xmath189-module @xmath268 has a finite filtration with successive quotients @xmath269 with @xmath270 . moreover , the quotient @xmath271 has a filtration with successive quotients @xmath269 for @xmath266 . therefore @xmath272 admits a co - verma filtration with successive quotients @xmath265 with @xmath266 . similarly as in the proof of lemma [ lem : lemma1 ] , @xmath263 is a direct summand of @xmath273 . therefore , by lemma [ technical ] , @xmath263 has a filtration as desired . @xmath209 admits a co - verma filtration . in this section , @xmath51 is an arbitrary nonnegative integer . for any @xmath274 , @xmath275 is a right - exact functor from @xmath98 into @xmath84 ( in particular , @xmath276 has finite length for @xmath90 ) . [ lem : lemma3 ] fix @xmath277 . then @xmath278 is finite dimensional . this follows from the fact that the @xmath48-weight spaces of @xmath279 are finite dimensional . as a consequence , @xmath280 is isomorphic to a submodule of a finite direct sum of copies of @xmath88 . since @xmath88 has finite length , @xmath276 also has finite length , and is an object of @xmath205 . the fact that @xmath275 is right - exact follows from the observation that @xmath275 is left adjoint to the inclusion functor @xmath281 , i.e. , @xmath282 for any @xmath90 and @xmath283 . indeed , a left adjoint to a left - exact functor is right - exact . for any @xmath284 and @xmath285 , we have @xmath286 where @xmath287 is the @xmath288-th left derived functor of @xmath275 . [ lem : inverselimitbeta ] let @xmath289 be a projective resolution of @xmath93 in the category @xmath290 . by definition , @xmath291 is the homology of the complex @xmath292 , and @xmath293 is the homology of the complex @xmath294 . moreover , @xmath295 by lemma [ lem : lemma3 ] , for every @xmath83 , @xmath288 and @xmath296 , @xmath297 has finite length as a @xmath24-module . so , lemma [ lem : inverselimitbeta ] follows from the mittag - leffler principle . if @xmath298 is an inverse system of objects from @xmath299 for variable @xmath300 , we set @xmath301 for any @xmath24-module @xmath29 . let @xmath211 be injective in @xmath302 . then , for any finitely generated @xmath67-module @xmath93 and for any @xmath285 , @xmath303 [ prop : prop1 ] let @xmath289 be as in the proof of lemma [ lem : inverselimitbeta ] . then @xmath304 since @xmath305 by the mittag - leffler principle , we have @xmath306 next , the injectivity of @xmath211 in @xmath302 implies @xmath307 consequently , @xmath308 and since homology commutes with direct limits , @xmath309 recalling that @xmath310 , we notice that @xmath311 furthermore , since @xmath289 is finitely generated , @xmath312 therefore , ( [ eq : eqhomdirectlimits1 ] ) , ( [ eq : eqhomdirectlimits2 ] ) , and ( [ eq : eqhomdirectlimits3 ] ) yield @xmath313 since @xmath314 , we obtain @xmath315 as desired . recall that @xmath209 is an injective cogenerator of the category @xmath207 . for any @xmath110 and any @xmath316 , @xmath317 is finite dimensional . [ prop : prop2 ] by lemma [ lem : lemma2 ] , it suffices to show that @xmath318 for any @xmath138 with @xmath319 . by the shapiro lemma , @xmath320 since by the injectivity of @xmath221 as a @xmath139-module we have @xmath321 , we conclude that @xmath322 now the statement follows from the finite - dimensionality of @xmath323 , see proposition [ prop : spec ] , b ) . for @xmath110 and any @xmath324 , we have @xmath325 . [ cor : cor1_of_prop2 ] by lemma [ lem : lemma3 ] , @xmath326 is finite dimensional for any @xmath83 . by propositions [ prop : prop1 ] and [ prop : prop2 ] , @xmath327 is finite dimensional . by the definition of the direct limit functor , we have , for sufficiently large @xmath328 , @xmath329 under the @xmath24-module map @xmath330 . since @xmath209 is an injective cogenerator for the category @xmath331 , we conclude that the map @xmath332 is an isomorphism for sufficiently large @xmath328 . by lemma [ lem : inverselimitbeta ] and the definition of the inverse limit functor , we conclude that @xmath333 for large enough @xmath328 . since @xmath334 for only finitely many @xmath335 , @xmath336 has finite length , or equivalently @xmath337 . for any @xmath110 and any injective object @xmath338 , @xmath339 [ cor : cor2_of_prop2 ] this follows from the isomorphism @xmath340 for large enough @xmath328 . for any @xmath110 and @xmath341 , @xmath342 so @xmath343 and @xmath344 are adjoint functors . [ prop : prop3 ] it clearly suffices to prove ( [ eq : prop3 ] ) for @xmath345 and @xmath346 . by proposition 7.9 in @xcite , @xmath347 for any injective object @xmath348 . by proposition [ prop : prop1 ] and corollary [ cor : cor2_of_prop2 ] , @xmath349 consider a part of an injective resolution of @xmath29 in @xmath350 , @xmath351 . since @xmath14 is an exact functor ( proposition [ prop : zuck ] ) , the following sequence is also exact : @xmath352 . next , applying @xmath353 , we obtain an exact sequence @xmath354 by ( [ eq : prop3_1 ] ) and ( [ eq : prop3_2 ] ) , we have a diagram @xmath355{diagram1.eps}\ ] ] which is commutative as the identifications @xmath356 are functorial in @xmath211 . since @xmath357 is left - exact , we conclude that @xmath358 finally , @xmath359 , and we are done . @xmath343 is a right - exact functor . [ corr : corr11 ] this is a direct consequence of the fact that @xmath18 is a left adjoint functor . as stated in proposition [ prop : zuck ] , for @xmath148 the functor @xmath14 induces a bijection between the sets of isomorphism classes of simple objects in the categories @xmath76 and @xmath62 . the main result of this section is : for @xmath112 , the functor @xmath18 induces a bijection of sets of isomorphism classes of simple objects of @xmath62 and @xmath76 , inverse to the bijection induced by @xmath14 . [ prop : prop11 ] in the rest of the paper we assume that @xmath112 . we set @xmath360 for @xmath361 . then @xmath362 is a simple object of @xmath62 , and all simple objects in @xmath62 are of this form for appropriate simple @xmath139-modules . let @xmath345 have the property that the isotypic component of the minimal @xmath23-type of @xmath93 is isomorphic to the isotypic component of the minimal @xmath23-type of @xmath363 . then @xmath364 = \left\ { \begin{array}{ll } 0 & \text{for } |f|<|e|\ , , \\ \leq 1 & \text{for } |f|=|e|\ , . \end{array } \right.\ ] ] [ lem : lem12 ] observe first that the @xmath48-weights of @xmath365 are less than or equal to @xmath366 . therefore the @xmath48-weights of @xmath367 are less than or equal to @xmath368 . this shows that @xmath369 for @xmath370 . indeed , the inequalities @xmath371 and @xmath372 yield @xmath373 , which contradicts our assumption that @xmath374 . next , note that kostant s theorem applied to @xmath23 gives @xmath375 therefore the spectral sequence ( [ spectral ] ) of proposition [ prop : spec ] implies @xmath376 hence , @xmath377 furthermore , for any @xmath378 @xmath379 by ( [ eq : eq23 ] ) . this yields @xmath380 consequently , since the injective hull @xmath381 of @xmath382 in @xmath383 admits a co - verma filtration with successive quotients isomorphic to @xmath265 for @xmath384 , and @xmath217 enters @xmath263 with multiplicity @xmath385 , we have @xmath386 finally , @xmath387\,,\ ] ] and the lemma is proved . set @xmath388 under the assumption that @xmath389 . then @xmath390=\left[\mathrm{b}_1x(e):l(f)\right]=\left\ { \begin{array}{ll } 0 & \text{for } |f|<|e|\ , , \\ 1 & \text{for } f\simeq e\ , . \end{array } \right.\ ] ] [ cor : cor13 ] for @xmath391 , the statement follows directly from lemma [ lem : lem12 ] as the isotypic components of the minimal @xmath23-types of @xmath392 and @xmath362 are isomorphic by proposition [ prop : zuck ] , b ) . if @xmath393 , then @xmath394 so the identity homomorphism @xmath395 provides a nonzero homomorphism @xmath396 . since @xmath14 is exact and @xmath18 is right - exact , this homomorphism is in fact a composition of surjections @xmath397 , in particular , @xmath398\geq 1 $ ] and @xmath399\geq 1 $ ] . on the other hand , @xmath398\leq 1 $ ] by lemma [ lem : lem12 ] ; hence , @xmath400=\left[\mathrm{b}_1x(e):l(e)\right]=1\,.\ ] ] @xmath401 . [ cor : cor14 ] by the adjointness of @xmath18 and @xmath14 , we have a canonical nonzero homomorphism @xmath402 induced by the identity homomorphism @xmath403 . note that @xmath404 is isomorphic to @xmath362 by proposition [ prop : zuck ] , b ) . next , by the adjointness of @xmath18 and @xmath14 , @xmath405 and as @xmath406 only for @xmath393 , we see that @xmath407 since @xmath408 , @xmath409 induces an isomorphism @xmath410 consequently , @xmath409 is surjective . let @xmath411 . since the exact sequence @xmath412 does not split , the assumption @xmath413 would imply @xmath414\neq 0 $ ] for some @xmath391 , i.e. , a contradiction with corollary [ cor : cor13 ] . hence , @xmath415 and @xmath409 is an isomorphism . @xmath416 . [ cor : cor15 ] @xmath18 is right - exact , hence the surjective homomorphism @xmath417 yields a surjective homomorphism @xmath418 . by applying @xmath14 we obtain a surjective homomorphism @xmath419 , hence @xmath420 . @xmath421 . [ cor : cor16 ] by corollary [ cor : cor13 ] , @xmath422 . we have an isomorphism @xmath423 hence , we have a nonzero ( and therefore injective ) homomorphism @xmath424 corresponding to the identity homomorphism @xmath425 . once again , corollary [ cor : cor13 ] implies @xmath399=1 $ ] . since @xmath14 is exact and is a bijection on isomorphism classes of simple modules , we have @xmath426=1\ , . \label{eq : cor16_0}\ ] ] by corollary [ cor : cor15 ] , there is a surjective homomorphism @xmath427 equation ( [ eq : cor16_0 ] ) now implies that @xmath428 and @xmath429 . thus , @xmath362 is a direct summand of @xmath430 . but corollary [ cor : cor15 ] shows that @xmath430 is indecomposable . we conclude that @xmath431 as before , @xmath14 is exact and is a bijection on isomorphism classes of simple modules . so , @xmath432 is a simple @xmath24-module , and @xmath433 . the goal of this section is to prove the following . [ prop : exact ] @xmath343 is an exact functor . [ prop : prop17 ] our main effort will go into proving @xmath435 for @xmath436 . [ lem : lem18 ] note that it suffices to show that @xmath437 for all @xmath438 . indeed , the implication @xmath439 follows from the following three facts : 1 . @xmath440 ; 2 . @xmath441 has a co - verma filtration with factors isomorphic to @xmath442 for @xmath443 ( lemma [ lem : lemma2 ] ) ; 3 . @xmath444 , see ( [ eq : eq23 ] ) . to prove that @xmath437 for all @xmath438 , we give another construction of the functor @xmath445 . denote by @xmath446 the enveloping algebra @xmath447 localized by the multiplicative set @xmath448 . the localized algebra @xmath449 is a subalgebra of @xmath446 . for any @xmath24-module ( resp . , @xmath23-module ) @xmath29 , set @xmath450 [ lem : localization ] if @xmath29 is a @xmath24-module on which @xmath451 acts injectively , we have an isomorphism of @xmath23-modules @xmath452 there is an embedding @xmath453 . by frobenius reciprocity @xmath454 induces a morphism @xmath455 . since @xmath456 , the morphism @xmath457 is surjective . let us show that @xmath457 is also injective . since @xmath451 acts injectively on @xmath29 , we see that @xmath458 is an essential extension of @xmath29 . therefore , the fact that @xmath459 suffices to conclude that @xmath457 is injective . suppose that a @xmath24-module @xmath29 is free over @xmath460 $ ] and locally finite over @xmath461 $ ] . then we have an embedding @xmath462 where @xmath463 is the functor of @xmath464-finite vectors . set @xmath465 cf . @xcite . since @xmath466 satisfies the above assumptions , we have constructed a new functor @xmath467 [ character ] if @xmath466 , @xmath119 , then for some @xmath468 @xmath469 and @xmath470 where @xmath471 for an integral @xmath48-weight @xmath52 . the first assertion is obvious and the second follows from lemma [ lem : localization ] and the straightforward check that @xmath472 [ cor : restriction ] if @xmath466 , @xmath119 , then @xmath473 [ cor : fexact ] the functor @xmath474 is exact . [ prop : isomf ] the functors @xmath474 and @xmath149 are isomorphic . let us start with the construction of a morphism of functors @xmath475 . let @xmath466 . then the exact sequence @xmath476 does not split over @xmath23 , and therefore does not split over @xmath24 . set @xmath477 and let @xmath478 be the koszul resolution of @xmath29 as introduced in lemma 2.2 of @xcite . the complex @xmath479 is functorial with respect to @xmath29 and yields an injective resolution of @xmath29 in the category of @xmath67-modules . hence , we have a commutative diagram @xmath480 for some morphisms @xmath481 and @xmath482 , unique up to homotopy . we recall from @xcite that @xmath483 . by construction , @xmath484 and @xmath485 . by applying @xmath486 to the above diagram , we obtain a new commutative diagram @xmath487 the morphism @xmath488 induces a unique morphism @xmath489 , by the definition of @xmath14 . since our diagram is functorial in @xmath29 , we obtain a morphism of functors @xmath490 . it remains to show that @xmath491 is an isomorphism for all @xmath341 . since both functors @xmath492 and @xmath14 are exact , it is sufficient to check this for simple @xmath29 . suppose that @xmath491 is nonzero . then , we have a surjective morphism @xmath489 , since @xmath493 is also simple . but then , by corollary [ cor : restriction ] , @xmath491 is an isomorphism . on the other hand , assume @xmath494 . by a diagram chase , we obtain a nonzero morphism @xmath495 such that @xmath496 . moreover , @xmath497 . because @xmath498 is an essential extension of @xmath29 , @xmath481 is an injection , and hence @xmath499 . on the other hand , @xmath500 a contradiction . hence , @xmath501 . [ cor : verma ] there is an isomorphism of @xmath24-modules @xmath502 we are now ready to give a proof of lemma [ lem : lem18 ] . set @xmath503 from the explicit form of @xmath504 as a @xmath23-module it is easy to check that @xmath505 by the spectral sequence of proposition [ prop : spec ] this implies @xmath506 the exact sequence @xmath507 where @xmath508 , yields @xmath509 . it is easy to check that @xmath510 , hence the input into @xmath511 in the spectral sequence ( [ spectral ] ) comes from @xmath512 the maximum possible @xmath48-weight of @xmath513 is @xmath514 , hence the maximum possible @xmath48-weight of ( [ eq : vhc ] ) is @xmath515 . however , for any @xmath516 such that @xmath517 , we have @xmath518 as @xmath519 . we obtain @xmath520 . next , we note that corollary [ cor : verma ] shows the existence of an exact sequence @xmath521 as @xmath522 . since @xmath144 is free as an @xmath120-module , @xmath523 . together with @xmath520 , this yields @xmath524 as @xmath144 is free as an @xmath120-module . the proof of lemma [ lem : lem18 ] is complete . to prove proposition [ prop : exact ] , it now suffices to establish the following . @xmath525 for any @xmath436 . [ lem : lem22 ] we will prove the statement by induction on the bruhat height on the @xmath163-lowest weight of @xmath143 ( @xmath526 is the borel subalgebra we fixed in section [ sec : prelim ] ) . if @xmath175 is @xmath163-dominant ( i.e. @xmath527antidominant ) , @xmath528 and we are done . for an arbitrary @xmath175 , we consider the exact sequences @xmath529 and @xmath530 the long exact sequence corresponding to ( [ eq : ses2 ] ) is the top row of the following commutative diagram @xmath531{diagram2.eps}\ ] ] the vertical isomorphisms are explained as follows : @xmath532 @xmath533 @xmath534 and @xmath535 the exactness of the bottom row of the diagram now yields @xmath525 , and lemma [ lem : lem22 ] is proved . proposition [ prop : prop17 ] now follows . the results of sections [ sec : conjugacy]-[sec : exactness ] imply that , under the assumption @xmath148 , the functors @xmath113 and @xmath536 are exact functors between finite - length abelian categories which induce mutually inverse bijections on isomorphism classes of simple objects . the isomorphisms @xmath537 and @xmath538 for @xmath110 and @xmath539 induce morphisms of functors @xmath540 and @xmath541 in addition , it follows from section [ sec : bijection ] that for any simple @xmath539 the morphism @xmath542 is an isomorphism . we show now by induction on the length of objects that @xmath543 is an isomorphism for any @xmath539 . consider the morphism @xmath543 , and let @xmath544 be an exact sequence of modules in @xmath76 with @xmath545 and @xmath546 . then , since both functors @xmath14 and @xmath18 are exact , we have an exact sequence @xmath547 in @xmath62 . since the lengths of @xmath548 and @xmath549 are less than the length of @xmath29 , we can assume that we have a commutative diagram @xmath550{diagram3.eps}\ ] ] where the left and the right vertical arrows are isomorphisms . by the short five lemma the middle vertical arrow @xmath551 is an isomorphism too . the case of @xmath552 is analogous . it is interesting to see when the functor @xmath553 establishes an equivalence of the category @xmath554 with the entire category @xmath555 . this is equivalent to the question : for which central characters @xmath155 does the equality @xmath556 hold ? consider in more detail the case when @xmath24 is simple and @xmath23 is a principal subalgebra of @xmath24 . here @xmath38 is a regular element of @xmath24 and @xmath557 is a borel subalgebra . let the simple roots of @xmath163 be @xmath558 , and @xmath559 be the highest root . then @xmath560 for some positive integers @xmath561 . moreover , @xmath562 . we would like to find central characters @xmath155 such that @xmath563 . for a weight @xmath564 denote by @xmath565 the central character of the verma module @xmath566 . the equality @xmath567 holds if and only if @xmath568 and @xmath569 belong to the same orbit of the weyl group , where @xmath570 is the half - sum of roots of @xmath163 . this orbit contains a unique @xmath163-antidominant weight @xmath568 . then @xmath571 for any other @xmath572 on the weyl group orbit . therefore we need to find @xmath573 such that @xmath568 is antidominant and @xmath574 . let @xmath575 denote the simple coroots . then @xmath576 for some positive integers @xmath577 . we set @xmath578 . since @xmath579 for all @xmath580 , the condition that @xmath568 is antidominant can be written in the form @xmath581 the equality @xmath582 shows that @xmath583 . hence , the condition @xmath584 is equivalent to @xmath585 let @xmath586 denote the set of weights satisfying conditions ( [ antidominant ] ) and ( [ notrun ] ) . clearly @xmath586 is not empty as soon as @xmath587 the latter inequality can be rewritten as @xmath588 for example , let @xmath589 . then @xmath590 , hence @xmath591 and @xmath592 . therefore ( [ notrun1 - 1 ] ) holds for @xmath593 . let @xmath24 be a simple lie algebra not equal to @xmath2 or @xmath594 . then @xmath586 is not empty . if in addition @xmath595 , then @xmath586 is infinite . @xmath586 is infinite as soon as the inequality ( [ notrun ] ) is strict . we can further rewrite ( [ notrun1 - 1 ] ) as @xmath596 if @xmath597 , the inequality [ notrun2 ] is strict as in this case the sum of positive non - highest @xmath48-weights in the @xmath23 submodule generated by the highest root vector is greater than the highest @xmath48-weight . therefore , the statement holds for all @xmath24 of rank greater than @xmath598 and for @xmath599 . for @xmath600 we have @xmath601 , @xmath602 , and hence @xmath586 consists of one element : @xmath603 . for @xmath604 we have @xmath605 . note that the set of integral weights lying in @xmath586 is always finite since @xmath586 is compact . moreover , the cardinality of this finite set grows with rank . using translation functors we can strengthen theorem [ main ] for certain central characters . let us call a central character @xmath155 @xmath23_-adapted _ if @xmath563 . a central character @xmath606 is _ weakly _ @xmath23_-adapted _ if there exists a @xmath23-adapted character @xmath155 and a translation functor @xmath607 establishing an equivalence between the categories of @xmath24-modules admitting respective generalized central characters @xmath155 and @xmath606 . recall that , if @xmath608 for some @xmath609 such that @xmath610 is antidominant and @xmath611 for some @xmath612 , then @xmath613 must be integral and the stabilizers of @xmath568 and @xmath610 in the weyl group of @xmath24 must be the same @xcite . assume that @xmath606 is weakly @xmath23-adapted . then a. @xmath614 is simple for any simple module @xmath615 . b. let @xmath616 denote the full subcategory of @xmath617 consisiting of modules whose simple constituents are of form @xmath614 for simple modules @xmath615 . then the functor @xmath618 is an equivalence of categories inverse to @xmath14 . both assertions follow from the following commutative diagram of functors @xmath619{diagram4.eps}\ ] ] where @xmath620 are appropriate translation functors . the commutativity of the diagram is a consequence of theorem [ main ] and of the fact that the zuckerman functor commutes with translation functors . this latter fact is essentially a reformulation of proposition 2.6 and corollary 2.8 in @xcite . in this section @xmath23 is an arbitrary @xmath2-subalgebra of @xmath24 and @xmath112 . by @xmath621 ( respectively , @xmath622 ) we denote the category of inductive limits of objects from @xmath62 ( respectively , @xmath623 ) . theorem [ main ] implies the following [ lem : gd1 ] for every simple @xmath139-module @xmath138 such that @xmath632 , the module @xmath217 has an injective resolution in @xmath633 of length not greater than @xmath634 . hence @xmath635 for any @xmath636 and any @xmath637 . recall the category @xmath638 from section [ sec : prelim ] . in this category @xmath214 has an injective resolution with terms @xmath639 then @xmath640 provides an injective resolution for @xmath217 in @xmath633 of length at most @xmath641 . this is a standard fact about parabolic category @xmath19 . indeed let @xmath655 . then @xmath656 . consider the koszul complex @xmath657 as the complex of polynomial differential forms on the open orbit of @xmath658 on @xmath659 . it gives a resolution of the trivial module by modules which admit finite filtrations with succesive quotients isomorphic to @xmath654 . to obtain a similar resolution for @xmath143 , we tensor the above resolution with @xmath143 and project on the subcategory of modules with the central character of @xmath143 . 20 g. benkart , d. britten , f. lemire , modules with bounded weight multplicities for simple lie algebras , math . z. * 225 * ( 1997 ) , 333353 . j. bernstein , s. gelfand , tensor products of finite and infinite - dimensional representations of semisimple lie algebras , compositio math . * 41 * ( 1980 ) , 245285 . j. bernstein , i. m. gelfand , s. gelfand , a category of @xmath24-modules , funktional anal . i prilozhen * 5 * , no . 2 ( 1971 ) , 18 ; english translation , functional anal . and appl . * 10 * ( 1976 ) , 8792 . d. j. britten , f. w. lemire , a classification of simple lie modules having a @xmath385-dimensional weight space , trans . * 299 * ( 1987 ) , 683697 . e. b. dynkin , semisimple subalgebras of semisimple lie algebras , mat . ( n.s . ) * 30(72)*:2 ( 1952 ) , 349462 . t. j. enright , on the fundamental series of a real semisimple lie algebra , annals of mathematics * 110 * ( 1979 ) , 182 . t. j. enright , n. r. wallach , notes on homological algebra and representations of lie algebras , duke math . j. * 47 * ( 1980 ) , 115 . v. futorny , the weight representations of semisimple finite dimensional lie algebras , ph.d . thesis , kiev university , 1987 . s. fernando , lie algebra modules with finite dimensional weight spaces i , transactions of ams * 322 * ( 1990 ) , 757781 . i. m. gelfand , r. a. minlos , and z. ya . shapiro . , representations of the rotation and lorentz groups and their applications , translated from the russian edition ( moscow , 1958 ) by g. cummins and t. boddington , pergamon , london ; macmillan , new york , 1963 . i. gelfand , v. ponomarev , the category of harish - chandra modules over the lie algebra of the lorentz group , soviet . doklady * 8*(5 ) ( 1967 ) , 1065 - 1068 . d. grantcharov , v. serganova , category of @xmath666-modules with bounded weight multiplicities , mosc . * 6 * ( 2006 ) , 119134 . d. grantcharov , v. serganova , cuspidal representations of @xmath667 , adv . math . * 224 * ( 2010 ) , 15171547 . harish - chandra , infinite irreducible representations of the lorentz group , proceedings of the royal society of london , series a , mathematical and physical sciences , vol . * 189 * , issue 1018 ( 1947 ) , 372401 . a. w. knapp , d. a. vogan , jr . , cohomological induction and unitary representations , princeton university press , princeton , nj , 1995 . o. mathieu , classification of irreducible weight modules , annales de linstitut fourier ( grenoble ) * 50 * ( 2000 ) , 537592 . a. petukhov , bounded reductive subalgebras of @xmath668 , transformation groups * 16 * , 2011 , 11731182 . i. penkov , v. serganova , on bounded generalized harish - chandra modules , annales de linstitut fourier * 62 * ( 2012 ) , 477496 . i. penkov , v. serganova , g. zuckerman , on the existence of @xmath7-modules of finite type , duke math . j. * 125 * ( 2004 ) , 329349 . i. penkov , g. zuckerman , generalized harish - chandra modules with generic minimal @xmath23-type , asian journal of mathematics * 8 * ( 2004 ) , 795812 . i. penkov , g. zuckerman , a construction of generalized harish - chandra modules with arbitrary minimal @xmath23-type , canadian mathematical bulletin * 50 * ( 2007 ) , 603609 . i. penkov , g. zuckerman , on the structure of the fundamental series of generalized harish - chandra modules , asian journal of mathematics * 16 * ( 2012 ) , 489514 . i. penkov , g. zuckerman , algebraic methods in the theory of generalized harish - chandra modules , developments and retrospectives in lie theory : algebraic methods , developments in mathematics , vol . * 38 * , springer verlag , 2014 , 331350 . d. vogan , representations of real reductive lie groups , progress in mathematics , vol . * 15 * , birkhauser , boston , 1981 . g. zuckerman , generalized harish - chandra modules , in highlights of lie algebraic methods , progress in mathematics , vol . * 295 * , birkhauser , 2012 , 123143 .

the very large array ( vla ) has been used in c - configuration to map an area at 1.4 ghz with sensitivities of 0.305 , 0.325 , 0.380 and 0.450 mjy beam over four equal subareas . radio properties are presented for 62 detected sources . deep optical imaging to gunn mag using the hale 5-m telescope covering is reported for a subset of 43 sources . this optical follow - up is much deeper than that of existing larger area radio surveys of similar radio sensitivity . archival , and-band photometry from the two - micron all sky survey ( 2mass ) is also presented . using a robust likelihood - ratio technique , we optically identified 26 radio sources with probabilities , nine are uncertain / ambiguous detections , and eight are empty fields . comparisons with a stellar synthesis model that includes radio emission and dust reddening suggests that the near - infrared to optical emission in a small , bright sub - sample is reddened by ` optically thin ' dust with absorption mag , regardless of morphological type . this is consistent with other , more direct determinations of absorption . the radio to optical(near - infrared ) flux ratios of early - type galaxies require significant contamination in the radio by an active galactic nucleus ( agn ) , consistent with the current paradigm . using our simple modeling approach , we also discuss a potential diagnostic for selecting ultraluminous infrared galaxies ( uligs ) to from microjansky radio surveys .

astro-ph0002490

the advent of deep radio surveys reaching flux densities well below 1 mjy ( mitchell & condon 1985 ; windhorst et al . 1985 ; oort 1987 ; windhorst et al . 1993 ; hopkins et al . 1998 ; richards 2000 ) revealed a new population of faint sources more numerous than the agn - powered radio galaxies dominating the strong - source population . this corresponds to a steepening of the differential source counts over non - evolving predictions at levels @xmath11 mjy . the faint counts suggest that significant evolution has occured over the redshift range spanned by the observed population . photometric and spectroscopic studies ( thuan et al . 1984 ; windhorst et al . 1985 ; thuan & condon 1987 ; benn et al . 1993 ) suggest that the faint excess at 1.4 ghz is composed predominately of star - forming galaxies similar to the nearby starburst population dominating the _ infrared astronomical satellite _ ( _ iras _ ) 60-@xmath12 m counts ( benn et al . indeed , this is supported by the strong correlation between radio ( 1.4 ghz ) and far - infrared ( 60-@xmath12 m ) flux densities of disk galaxies ( helou , soifer & rowan - robinson 1985 ) , implying a significant proportion of starburst galaxies at faint radio flux densities . the overall observed source - count distribution from faint ( @xmath12jy ) to bright ( @xmath13mjy ) flux densities can not be explained by starbursts alone . evolutionary models of radio source counts need to invoke two separate populations ( eg . danese et al . 1987 ; rowan - robinson et al . 1993 ; hopkins et al . 1998 ) . condon ( 1989 ) describes these populations as starbursts and monsters , each powered by different mechanisms : ` starbursts ' deriving their radio emission from supernova remnants and hii regions and ` monsters ' from compact nuclear related activity ( eg . active galactic nuclei ; agns ) . the proportion of agn is much greater at higher flux densities @xmath13 mjy ( kron , koo & windhorst 1985 ; gruppioni et al . 1998 ) , where a majority are associated with classical radio galaxies exhibiting extended ( fri and frii - type ) morphologies ( fanaroff & riley 1974 ) . the optical counterparts to sources at bright radio flux densities @xmath14 mjy is comprised mostly of ellipticals while at sub - mjy to @xmath12jy levels , the optical counterparts are identified as blue galaxies exhibiting peculiar ( compact , interacting and irregular ) morphologies ( kron , koo & windhorst 1985 ; gruppioni et al . studies of faint radio sources , namely their stellar population , how they evolve with redshift , and how they relate to local normal galaxies is progressing rapidly , however , much remains to be learned from the faintest ( @xmath12jy ) radio populations at redshifts @xmath15 . radio surveys are insensitive to the effects of absorption by dust which is known to bias surveys severely at optical / uv wavelengths . this is particularly important for derivations of the cosmic history of star formation and its relation to hierarchical models of galaxy formation . optical / uv studies have shown that there is an increase in both the space density of star - forming , morphologically disturbed galaxies ( eg . richards et al . 1999 ) and also , the global star formation rate with redshift to @xmath16 ( eg . madau et al . similar evidence is emerging from studies of sub - mm sources ( blain et al . 1999 ) and amongst the faint radio population at @xmath15 - the redshift range probed by the deepest surveys ( eg . cram 1998 ; haarsma et al . 1999 ) . there has been much speculation as to whether global star - formation rates ( sfrs ) derived from radio observations exceed those determined from optical / uv studies . cram et al . ( 1998 ) note that systematic discrepancies may exist between the various star formation indicators , which are not well understood ( see also schaerer 1999 ) . it is encouraging to see however that haarsma et al . ( 2000 ) find global sfrs derived from 1.4ghz observations of the hubble deep field to exceed optically detetermined values by a factor of a few out to @xmath17 . indeed , an analysis of balmer decrements and optical near - infrared colors in star forming galaxies by georgakakis et al . ( 1999 ) from the phoenix deep survey ( hopkins et al . 1998 ) , finds evidence for visual extinctions from one to a few magnitudes . currently , about 20% of existing micro - jansky radio samples remain unidentified to @xmath18 mag in hubble deep field images ( eg . richards et al . a majority of these could represent a significant population of dust - enshrouded starbursts and/or agn at high redshift . these results are in support of efforts to further understand the dust properties of star - forming galaxies . the primary aim of this paper is to introduce a new complete sample of radio sources selected at 1.4 ghz , uniformly selected over the flux range @xmath19 mjy ( 5@xmath20 ) from an area covering @xmath21 . although much larger area surveys to deeper radio sensitivities have been carried out ( eg . hopkins et al . 1998 ) , the present study reports the results of more sensitive optical observations . archival near - infrared data for a subset of the sample are also presented . the near - infrared data is from the ongoing 2mass project , and represents a unique aspect of this study in the identification of radio - selected starbursts . although we currently lack valuable spectroscopic information , we combine radio near - infrared optical flux ratios , radio maps , and optical images to explore the properties of the entire sample . our deep optical identifications provide the opportunity to asses the importance of dust in star - forming systems via the observed radio to optical and near - infrared to optical colors . simple stellar synthesis models that include radio emission and reddening are used to constrain possible amounts of absorption . this paper is organised as follows . in 2 , we discuss the radio observations and data reduction , present the radio catalog , and compare our results with data available from ( shallower ) all sky radio surveys . the optical photometric observations , their reduction and astrometric calibrations are discussed in 3 . our method for radio - optical identification , the archival near - infrared data , radio - optical image overlays , and our optical / near - infrared catalog are presented in 4 . a study of the radio , near - infrared , and optical colors and constraints on synthesis models incorporating dust is presented in 5 . an application of our color - color analysis to select high - redshift ultraluminous infrared galaxies from deep radio surveys is discussed in 6 . all results and future prospects are summarised in 7 . observations were made with the vla c - configuration at 1.4 ghz on 1998 december 19 . this configuration yields a good compromise for resolution and surface - brigntness sensitivity . the @xmath1 confusion limit for this configuration is only @xmath22jy / beam at 1.4 ghz ( mitchell & condon 1985 ) since the synthesised beam size ( full width at half power ; fwhp ) is @xmath23 arcsec . the resulting radio positions have rms errors @xmath24 arcsec , except for extended sources with multiple components , sufficient for making optical identifications . at 1.4 ghz the fwhp of the vla primary beam is 31 arcmin and approximately corresponds to the diameter of our final imaged field . this relatively large coverage avoids field - to - field variations in source counts induced by high - redshift clustering . although surveys at higher frequencies ( eg . 8 ghz ) can reach lower flux densities than at 1.4 ghz , most radio sources have spectral indices @xmath25 ( @xmath26 ) so that the population being sampled is similar . to provide uniform sensitivity over the full area of our field , we observed seven positions arranged in a filled hexagonal pattern with a separation of @xmath27 between pointing centers . the resulting root mean square ( rms ) map noise is thus nearly constant ( cf . condon et al . our field was centered on ra(2000)=@xmath28 , dec.(2000)=@xmath29 . this field was chosen for its relatively low foreground galactic - cirrus emission likely to affect optical / near - infrared identifications and also for the absence of bright radio galaxies . the integration time on each pointing was @xmath24 hour . this allowed us to reach an rms noise of @xmath30jy in regions free from bright contaminating sources ( see [ radnoise ] for more details ) . our observations were made in spectral line mode with four intermediate frequencies ( ifs ) , each divided into 7 spectral channels of width 3.125 mhz . the advantage of this mode is to minimize bandwidth smearing ( i.e. chromatic aberration ) which reduces the point - source sensitivity away from the pointing center and causes appreciable image distortion . additionally , the spectral line mode is less prone to narrowband interference noise spikes . with continuum mode however , we would have had a little over twice the bandwidth and a factor @xmath31 lower noise . the data were analysed with the nrao aips reduction package . we observed 3c 48 to calibrate the visibility amplitudes , using @xmath32 and @xmath33 jy at 1.365 and 1.435 ghz , respectively . the calibration was applied using the split task . the calibrated data from each pointing were edited and imaged separately , cleaned , and restored with 15 arcsec fwhp gaussian beams . the seven separate images were weighted and combined as described in condon et al . ( 1998 ) to produce a final @xmath34 map with nearly uniform sensitivity and corrected for primary - beam attenuation . the resulting rms noise of our final map after correcting for primary - beam attenuation is not uniform over the entire field , but increases by up to 30% near a strong ( @xmath35 mjy ) source near the edge of our field . despite this variation in sensitivity , we were able to divide our @xmath36 field into four equal ( @xmath37 ) regions within which the rms noise varies by no more than a few percent . these constant - noise regions simplify the application of an automated source - finding algorithm over a single continuous region ( see below ) . the lowest and highest rms noise amongst these regions was @xmath38jy / beam and @xmath39jy / beam respectively . see table [ tbl1 ] for the region definitions . the rms noise of each region was estimated from the gaussian core of the amplitude distribution of the pixel values as produced by the aips task imean . in figure [ pixhist ] , we show the distribution in pixel values of our entire @xmath40 field . the rms deviation in peak flux density derived from a fit to this histogram is @xmath41jy / beam . figure [ contfull ] shows a contour map of our entire radio field . each constant - noise region in table [ tbl1 ] was used for the source extractions . within each region , we searched for radio sources down to a peak flux density @xmath42 times the rms value of the region . the sources were extracted by the aips task sad which uses gaussian fits to estimate the fluxes , positions , and angular sizes of the selected sources . for faint sources however , unconstrained gaussian fits may be unreliable ( see condon 1997 ) . for this reason , we adopted the following method for the source extraction : first , we ran the task sad with a @xmath43 detection threshold to obtain an initial list of candidates , we then derived the peak flux densities of the faint sources ( with @xmath44 ) using the maxfit task . this task uses a second order interpolation algorithm and is known to be more accurate . only sources with a maxfit peak flux density @xmath45 were retained . for these faint sources , the total flux density was estimated using the task imean , which integrates the ( median background subtracted ) pixel values in a specific rectangle . the rectangle was chosen to encompass as much of the source as possible . for all other parameters ( sizes , positions and position angles ) we retained the values obtained from the initial gaussian fits . only two sources had irregular morphologies showing multiple components . for these , the total ( background subtracted ) flux was determined using the task tvstat which allows an integration over a specific irregular area defined to encompass as much of the source as possible . the numbers of sources detected in each constant - noise region are summarised in table [ tbl1 ] . a total of 62 sources were detected with flux densities @xmath45 over an area of @xmath46 . within poisson uncertainties , this number is consistent with source counts from surveys by ciliegi et al . ( 1999 ) and hopkins et al . ( 1998 ) . to our limiting ( mean ) sensitivity of 0.35mjy and within @xmath46 , they report a source count of typically @xmath47 . this confirms the reliability of our radio source detections and flux density estimates . table 2 shows the full radio catalog which reports ( in column order ) : the source name ; ra and dec.(j2000 ) ; errors in ra and dec . ; the peak flux density @xmath48 ; error in @xmath48 ; the total flux density @xmath49 ; error in @xmath49 ; the full width half maximum ( fwhm ) of the major and minor axes @xmath50 and @xmath51 ( determined from gaussian fits ) ; the position angle @xmath52 of the major axis ( in degrees ) ; and the rms errors associated with @xmath50 , @xmath51 , and @xmath52 respectively . the different components of multiple sources are labelled `` c1 '' and `` c2 '' . in figure [ fhist ] , we show the distribution of peak flux densities and the total to peak flux ratio as a function of peak flux for all sources . sources with ratios @xmath53 in figure [ fhist]b are primarily due to uncertainties on measured fluxes as estimated from the two dimensional gaussian fits . uncertainties in peak fluxes are typically @xmath54 , while total flux estimates are more uncertain due to a sensitive dependence on the gaussian fitting procedure . errors in total fluxes are typically @xmath55 . contour maps of radio sources with available optical data are shown in figure [ optradim ] . the @xmath36 region that we observed with the vla was also covered by the nrao vla sky survey ( condon et al . 1998 , nvss ) . the nvss covers the sky north of @xmath56 at 1.4 ghz with @xmath57 resolution and limiting flux density of @xmath58 mjy ( @xmath1 ) . to this limit , 17 of our sources were found to be in common with the nvss public catalog . one source in our catalog however is a double component source and is unresolved by the nvss . a comparison of flux densities derived in this study with those from this catalog is shown in figure [ nvss ] . it is evident that our derived flux densities are on average slightly lower than those from the nvss . this was also reported in a larger comparison study by ciliegi et al . ( 1999 ) using a similar observational set - up and can be explained by the difference in resolutions used in the two surveys : @xmath59 vs. @xmath57 . high - resolution surveys tend to miss flux from low - surface - brightness emission . although the effect is only marginal for bright sources , it may become important for attempts to detect faint , low - surface - brightness objects at high redshift . optical ccd photometry of our radio field was carried out on the 5-m hale telescope at palomar observatory during the nights of 27 - 29 august 1997 . these fields were initially selected for subsequent deep mid - ir imaging with the _ wide field infrared explorer _ mission ( _ wire _ ) , but the mission failed to perform to expectations . the carnegie observatories spectroscopic multislit and imaging camera ( cosmic ; kells et al . 1998 ) mounted at the @xmath60 prime focus with a tek 2k ccd was used to image nine @xmath61 fields in the gunn @xmath62 ( 6550 ) filter . each optical field comprised of three optical pointings offset by @xmath63 , each with integrations of 600 sec giving a total 1800 sec per field . this resulted in a limiting magnitude of @xmath3 mag ( @xmath1 ) . the seeing was typically @xmath64 ( fwhm ) . the optical fields do not cover our entire @xmath36 radio field . the nine slightly overlapping optical fields correspond to an areal coverage @xmath65 , or about 70% of the radio map . the ccd data were reduced with standard tasks in the iraf package . frames were first bias - subtracted , then flat - fielded using dome flats . bad pixels and columns were removed by interpolating between adjacent pixels and lastly , the individual dithered frames were median combined to remove cosmic ray hits . calibration was performed using standard stars from thuan & gunn ( 1976 ) . these were used to correct for atmospheric extinction from varying airmass and provide the instrumental zero point . photometric uncertainties , estimated using these standards , are no more than @xmath66 mag . sources in the reduced optical frames were extracted using the daophot package in iraf ( stetson 1987 ) . this package has the benefit of performing photometry in crowded fields , which is the case in most regions of our optical fields . it performed the following steps : first , sources were extracted above a given threshold , given as a multiple of the total ccd noise ( sky and read noise , @xmath67 ) . we adopted threshold values of 4.5 - 5@xmath67 , just high enough to avoid large numbers of spurious detections . second , simple aperture photometry was performed on these identified sources . this required a specification of the aperture radius which is likely to contain most of the light of our target source , and width of a surrounding annulus to estimate and subtract the sky background . we adopted a radius of six arcsec and annulus width of four arcsec . next , a point spread function ( psf ) was determined in each of our nine fields . this involved an interative technique to remove contamination from neighbouring sources in crowded fields near our psf candidates . simultaneous psf fitting on all sources was then performed to identify sources which were previously hidden in crowded regions . finally , the magnitudes determined from psf fits were aperture corrected to a common aperture size as used on our standard stars . aperture corrections were typically 0.22 mag . a final visual inspection removed any spurious detections . a total of @xmath68 - 390 sources were extracted from each @xmath61 field to a limit of @xmath3 mag . previous optical surveys find typically 380 - 520 sources within this area to this limit , and the variation is primarily due to clustering . such fluctuations are found to be significant on such relatively small scales ( eg . metcalfe et al . 1991 ) . the astrometry on the optical images was based on 10 - 12 apm catlog stars in each field ( maddox et al . 1990 ) . the ccmap and cctran tasks in the iraf immcoords package were used to compute plate solutions relating pixel positions to astrometric coordinates . astrometric coordinates for all sources on the frames were then determined . by comparing the positions of several sources common to the apm catalog and our fields , we found that our rms position uncertanties are typically @xmath69 . we assigned optical identifications and estimate their reliability using a robust likelihood ratio ( @xmath70 ) analysis . this general method has frequently been used to assess identification probabilities for radio and infrared sources ( eg . de ruiter , willis & arp 1977 ; prestage & peacock 1983 ; sutherland & saunders 1992 ; lonsdale et al . the method , which computes the probability that a suggested identification is the ` true ' optical counterpart , is outlined as follows : for each optical candidate @xmath71 in the search area of some radio source @xmath72 , we calculated the value of the dimensionless difference in radio and optical positions : @xmath73^{1/2 } , \label{rij}\ ] ] where the @xmath74 s and @xmath75 s represent right ascensions and declinations respectively , and @xmath20 s standard deviations . we chose a moderately large search radius of @xmath76 to allow for the maximal position uncertainties : @xmath77 and @xmath78 ( assuming @xmath79 , where @xmath80 ) . such a radius is also small enough to avoid large numbers of chance associations . given @xmath81 , we must now distinguish between two mutually exclusive possibilities : ( 1 ) the candidate is a confusing background object that happens to lie at distance @xmath81 from the radio source or ( 2 ) the candidate is the ` true ' identification that appears at distance @xmath81 owing solely to radio and optical position uncertainties . we assume that the radio and optical positions would coincide if these uncertainties were zero . this assumption however is not valid when the centroid of an extended radio source is used , and is further discussed below . to distinguish between these cases , we compute the likelihood ratio @xmath82 , defined as : @xmath83}{2\pi n(<m_{i } ) \left[(\sigma_{\alpha_{i}}^{2 } + \sigma_{\alpha_{j}}^{2 } ) ( \sigma_{\delta_{i}}^{2 } + \sigma_{\delta_{j}}^{2})\right]^{1/2 } } , \label{lr}\ ] ] where @xmath84 is the _ local _ surface density of objects brighter than candidate @xmath71 . the likelihood ratio @xmath82 is simply the ratio of the probability of an identification ( the rayleigh distribution : @xmath85 ) , to that of a chance association at @xmath62 ( @xmath86 ) . @xmath82 therefore represents a ` relative weight ' for each match @xmath87 , and our aim is to find an optimum cutoff @xmath88 above which a source is taken to be a reliable and likely candidate . its advantage over alternative methods ( purely based on finding the lowest random chance match , eg . downes et al . 1986 ) is that it allows for a possible distant candidate to still be the ` true ' identification even when there is still a high chance of it being a spurious background source . it is important to note that our form for @xmath82 ( eq . [ lr ] ) slightly differs from that used by earlier studies ( eg . lonsdale et al . 1998 ) in that it does nt contain the multiplicative `` @xmath89 '' factor in the numerator . this factor represents the apriori probability that a `` true '' optical counterpart brighter than the flux limit exists amongst the identifications . for our purposes , we will treat @xmath82 as simply a a relative weight measure for each radio - optical match , just for the purposes of assigning an optimal cutoff for reliable identification ( see below ) . we are not concerned with its absolute value , which is required when computing formal probability measures from @xmath70 . for simplicity , we have therefore set @xmath90 in this work . the optical surface density as a function of magnitude to be used in computing @xmath70 was determined from the total number of objects visible in our optical frames . the variation in surface density in the vicinity of each radio source caused by possible clustering effects was found to be small : no more than 5% on @xmath91 scales . the distribution of @xmath70 values for all possible radio source - candidate matches is shown by the shaded histogram in figure [ lrhisto ] . following lonsdale et al . ( 1998 ) , we generate a truly random background population with respect to the radio sources by offseting the radio source positions by @xmath92 . @xmath70 values for each radio source were then re - computed and their distribution is given by the thick - lined histogram in figure [ lrhisto ] . a comparison of the number of associations for ( true ) radio source positions with the number of associations found for random ( offset ) positions will enable us to determine a critical value @xmath88 for reliable identification . from these distributions , we compute the reliability as a function of @xmath70 : @xmath93 where @xmath94 and @xmath95 are the number of true and random associations respectively . the reliability computed in this way also represents an approximate measure of the identification probability for a candidate with given @xmath70 . figure [ relfig ] illustrates the reliability as a function of @xmath70 . above @xmath96 , the reliabilities are @xmath97 because few random associations exceed this value of @xmath70 ( fig . [ lrhisto ] ) . as a good working measure we therefore assume a cutoff @xmath98 above which a source is taken to be a likely candidate . the determination of reliabilities via the @xmath70 method is insensitive to variations in @xmath84 across the field or uncertainties in its derivation , and also the assumption of gaussian error ellipses in the radio and optical positions . such uncertainties are `` normalised out '' when one computes the ratio of random to true number of associations within a search radius when estimating the reliability ( eq . [ reliability ] ) . lonsdale et al . ( 1998 ) have shown that the absolute value of @xmath70 itself depends on the characteristics of the source population being identified ( eg . stars versus galaxies ) . different populations ( assuming they could be classified a - priori using some diagnostic ) map into different underlying surface densities at the ` identifying ' wavelength , implying that distributions in @xmath70 ( eq . [ lr ] ) will also be different . for a robust determination of the reliability in such situations , see lonsdale et al . ( 1998 ) . there are two complications to consider in the above method . the first is when one attempts to identify extended ( or resolved ) radio sources with this method . for all radio sources , we have used the positions of centroids derived from two - dimensional gaussian fits in computing the @xmath70 for optical candidates . for unresolved sources with ( gaussian fitted ) sizes @xmath99 ( the synthesised fwhp beam width ) , the source is likely to have a compact central component and the optical position is expected to lie close to its quoted radio centroid . for an extended ( resolved ) source however , the radio and optical positions may differ considerably since errors in the radio centroids are only @xmath100 . in such cases , if @xmath101 the identification may still be valid , since its low @xmath70 value could purely be due to a real large positional offset . the second complication is when a radio source has more than one optical candidate within its search radius with @xmath102 . this occured in about 20% of cases and was primarily due to contaminating stars . we assess these ambiguities and increase the robustness of our identifications by visually examining all optical candidates according to the following criteria : 1 . if candidates have @xmath101 for a radio source with @xmath103 , then identification is classified as _ uncertain_. 2 . if candidates have _ very low _ reliability , @xmath104 ( for _ unresolved _ radio sources ) , or there are no objects in the search radius , then radio source is classified as _ empty field_. 3 . if @xmath105 , i.e. where reliability is _ moderately `` low '' _ , then identification is also uncertain . if more than one optical candidate exists with @xmath102 , then only source(s ) with extended ( galaxy - like ) optical profile is taken as the identification . point sources associated with quasar nuclei are not considered in our identification scheme due to their relatively low surface density compared to galaxies ( @xmath106 ) in sub - millijansky radio samples . 5 . for unique , @xmath102 candidates , its optical profile is also checked for confirmation . of the 62 radio sources , 43 lie within our optically imaged @xmath65 field . we found optical identifications for 26 to @xmath3 mag with reliabilities @xmath107 . four sources have identifications classified as uncertain owing to a moderately low identification reliability of @xmath108 ( and @xmath109 ) . five more are uncertain because they have extended radio structure and large possible positional offsets between optical and radio centroid positions . eight radio sources lie in `` definite '' optical empty fields with no candidates brighter than @xmath3 mag . other optical follow - up studies found similar results . georgakakis et al . ( 1999 ) identified @xmath110 of sources to @xmath111 mag from the phoenix deep survey ( @xmath112 mjy ) ( hopkins et al . deeper indentifications of sources as faint as @xmath113jy from hubble deep field images revealed a 80% success rate to @xmath114 mag ( richards et al . 1999 ) . ignoring the uncertain identifications in our study ( from criteria 1 and 3 above and which are excluded from our analysis ) , we find that @xmath11518% of our sources are unidentified to @xmath116 . accounting for differences in bandpasses and sensitivity , this is broadly consistent with the above studies . figure [ rhisto ] shows the distribution of apparent magnitude @xmath62 for all reliable ( robust ) and uncertain identifications in our sample . near - infrared data in the @xmath5(1.25@xmath12 m ) , @xmath6(1.65@xmath12 m ) and @xmath117(2.17@xmath12 m ) photometric bandpasses were obtained from the two micron all sky survey ( _ 2mass _ ) project database . for multi - band detection of point sources , this survey is currently scanning the sky to sensitivities 16.5 , 16.0 and 15.5 mag at signal - to - noise ratios @xmath118 , @xmath119 and @xmath118 respectively in @xmath5 , @xmath6 and @xmath117 . the data relevant to this study are not yet released in the public catalogs , and was retrieved from the ` internal working database ' at _ ipac_. photometry in this database was determined using custom psf - fitting software and algorithms are described in cutri et al . ( 2000 ) . since such data have not been subjected to the rigorous quality assurance as that in the public release catalogs , we have examined individual images for quality and any possible systematic uncertainties in the photometry . to maximise the possible number of detections , we searched for near - infrared counterparts to each radio source with a conservatively low signal - to - noise ratio threshold of @xmath120 in each band . in cases where a source was detected in only one or two of the three bandpasses : @xmath5 , @xmath6 and @xmath117 , we note its `` band - filled '' 95% confidence upper - limit in the undetected band . in other words , the _ 2mass _ catalog also reports upper limits to measurements in an undetected band by placing an aperture over the position inferred from detections in other bands . the image pixel scale of the _ 2mass _ detectors is @xmath121 and the positional uncertainties are @xmath122 . owing to the relatively shallow flux limits of the _ 2mass _ survey , the background source surface density is low enough that chance associations with radio positions are very unlikely . we searched the _ 2mass _ database for near - infrared counterparts to our 43 radio sources that have _ available optical _ information from our deep optical survey and examined their images for quality . we found 7 reliable matches with 6 detections in @xmath5 , 7 in @xmath6 , and 4 in @xmath117 at the @xmath123 level . upper limits were available for the remainding `` band - filled '' values . results of our optical identification analysis and available near - infrared data are shown in table 3 . in column order , this table reports : the radio source name ( see table 2 ) ; ra and dec.(j2000 ) of the optical counterpart of the radio source ; optical - radio position separation ( @xmath124 ) in arcsec ; logarithm of the liklihood ratio ( @xmath70 ) ; reliability of the optical identification ( see eq.[[reliability ] ] ) ; apparent @xmath62-band magnitude ; @xmath5 , @xmath6 and @xmath117 magnitudes with errors or @xmath125 upper limits ; @xmath126 color ; @xmath127 color , and last , the optical morphology if the optical counterpart is visually resolved with size @xmath128 . optical morphologies were determined from light profile fitting of individual sources and comparison with @xmath129-law ( elliptical - like ) and exponential ( disk - like ) profiles . in cases where a disturbed or interacting morphology is apparent , then it is designated to have an irregular ( labelled as ` irr ' ) morphology . sources unresolved in the optical with typically @xmath130 are designated as ` unknown ' and labelled as ` ? ' in table 2 . 8. contd ... fig . 8. contd ... in figure [ optradim ] we show the optical image radio map overlays for the 43 radio sources with available optical information . a visual inspection of the optical images of resolved counterparts shows a diverse morphological mix , consistent with previous studies . about 40% of our optical identifications have elliptical / disk - like morphology , while @xmath131 can be identified as exhibiting peculiar ( either interacting or disturbed ) morphologies . it is important to note that these ` disturbed ' sources are based on visual inspection alone and their morphology could still be uncertain until future spectroscopy confirms their nature . the elliptical / disk hosts also tend to be associated on average with sources of relatively brighter radio flux density ( @xmath132 mjy ) than the irregular class . this is consistent with previous radio - optical identification studies which find an increasing fraction of irregular - type galaxies at @xmath133 mjy ( eg . kron , koo , & windhorst 1985 , hammer et al . 1995 , gruppioni et al . 1998 ) and a significant number of elliptical galaxies hosting the brighter extended radio galaxies and agn ( condon 1989 ) . a further observation is the unique radio structure exhibited by our eight optical empty field sources with @xmath134 mag . these are represented by maps in fig . [ optradim ] labelled by the letter `` e '' . all show compact and symmetric ( presumably unresolved ) structures and could represent either of the following : distant ( possibly dusty ) agn where with our radio sensitivity , we could have detected a nominal fr - i galaxy to @xmath135 , or , nearby , compact dusty starbursts at @xmath136 as constrained by typical starburst luminosities : @xmath137 . the second explanation for the nature of the empty fields is more plausible , given that the majority of submillijansky radio sources are associated with star - forming galaxies and less than 5% are usually identified with bright fr - is at @xmath138 ( kron , koo & windhorst 1985 ) . this section presents an analysis of flux ratios between the radio , near - infrared , and optical bands to explore possible contributions from agn and starbursts to the observed radio emission as well as the importance of absorption by dust . because we lack spectroscopic information , our analysis treats the sub - mjy sources as one homogeneous population and uses a simple stellar synthesis model to interpret its properties quantitatively . we can predict the radio to optical(near - ir ) flux ratios and @xmath139 colors for a range of galaxy types using the stellar population synthesis code of bruzual & charlot ( 1993 ) ( hereafter bc ) . on its own however , the bc model does not directly predict the amount of radio emission expected from a star - forming galaxy , which we need for the determination of flux ratios involving the radio band . we do this by relating the star - formation rates derived from empirical calibrations involving the uv and radio bands as follows . the 1.4 ghz radio emission from star - forming systems is believed to be primarily synchrotron emission from cosmic rays accelerated in supernova remnants plus a small ( @xmath131 ) thermal contribution from hii regions ( condon & yin 1990 , condon 1992 ) . thus to a good approximation , the radio luminosity is taken to be proportional to the formation rate of stars with @xmath140 : @xmath141 ( condon 1992 ) . these same massive stars will also contribute significantly to the uv continuum emission in the range @xmath142-@xmath143 . in particular , there have been many different calibrations of the sfr from the uv - flux . for a salpeter initial mass function ( imf ) from @xmath144 to 100@xmath145 , the calibration of madau et al . ( 1998 ) ( which assumes _ no dust _ correction ) yields @xmath146 we have modified the initial relation of madau et al . ( 1998 ) to include the factor @xmath147 , which represents the fraction of stellar masses contributing to the observed sfr , @xmath148 where @xmath149 is the imf . for @xmath150 , we have @xmath151 . assuming a salpeter imf ( @xmath152 ) , we find that for stars with @xmath140 , @xmath153 . with this fraction , and equating the two sfr calibrations ( eqs [ [ sfrrad ] ] and [ [ sfruv ] ] ) , we find that the luminosity densities at 1.4 ghz ( @xmath154 ) and @xmath155 ( @xmath156 ) are very nearly equal . we therefore assume that the rest frame 1.4 ghz flux density is directly given by the flux density at @xmath155 as specified by the synthesized model spectrum . in general terms , the observed radio flux ( in the same units as the synthesised uv spectrum ) can be written : @xmath157 where @xmath158 is the radio spectral index ( condon 1992 ) and @xmath159 is the rest frame ( unreddened ) uv spectral flux . we must emphasise that this relative radio flux is only that associated with the star - formation process . possible additional contributions , such as contaminating agn , are not considered in this model . we calculated flux ratios involving the radio , near - infrared and optical bands using evolutionary synthesis models for ellipticals ( e / so ) , early ( sab / sbc ) and late ( scd / sdm ) type spirals , and `` very blue '' starbursts ( sb ) . these are meant to represent the possible contributions to the sub - mjy radio sources , and each class is defined by a characteristic star formation rate as a function of time . as supported by local observations ( eg . gavazzi & scodeggio 1996 ) , we assumed that e / so and sab / sbc galaxies have an exponentially decaying sfr of the form @xmath160 , where @xmath161 is the e - folding time . values of @xmath162 and @xmath163 gyr were adopted for the e / so and sab / sbc galaxies respectively . for late - type spirals ( scd / sdm ) and young starbursts ( sb ) , we assumed constant sfrs with different ages . all models used to generate the spectral energy distributions ( seds ) are summarised in table [ tbl4 ] . the models assume @xmath164 , @xmath165 and a galaxy formation redshift @xmath166 , which corresponds to an age of 12.7 gyr at @xmath167 . to explore the effects of dust on our flux ratios and colors , each model sed was reddened in the source rest frame with an extinction curve @xmath168 characteristic of the small magellanic cloud ( smc ) , which appears to be a good approximation for the ism of nearby galaxies ( calzetti et al . this approximation is most accurate for the reddest wavelengths of starburst galaxy spectra ( @xmath169-@xmath170 m ) , although breaks down severely at @xmath171(calzetti et al . 1999 ) . we have used the analytical fit for @xmath172 as derived by pei ( 1992 ) for the smc . for simplicity , we assumed that the dust is distributed in a homogeneous foreground screen at the source redshift . in figure [ rvsr ] we plot the radio to optical flux ratio , @xmath173 , defined as @xmath174 where @xmath175 and @xmath62 are the radio flux and optical @xmath62-band magnitude respectively . the distribution seen in observed values of @xmath173 is larger at the faintest optical magnitudes @xmath176 . there are few galaxies however with @xmath177 , and nonetheless , the scatter at @xmath176 is consistent with that found at @xmath178 in a larger follow - up study of sub - mjy radio sources by georgakakis et al . ( 1999 ) . figure [ rr ] shows @xmath173 as a function of @xmath126 color for all sources with available optical and near - infrared data . our reason for using @xmath126 color is that the @xmath6 band yielded more `` definite '' detections than the other near - infrared bandpasses . although the numbers are still relatively small , this facilitates the best comparison with the synthesis models . the predictions for four galaxy types ( see table [ tbl4 ] and 5.1 ) are shown for no dust reddening ( thin curves ) and a rest - frame extinction @xmath179 mag . the morphological mix of data shows a relatively large dispersion in @xmath126 color that is more consistent with the range predicted by the models that _ include dust_. this suggests that on average , the optical to near - infrared continua of most sources in fig . [ rr ] are reddened by a uniform ( possibly `` optically thin '' ) dust component with @xmath9 mag absorption . this measure is consistent with spectroscopic studies of nearby starbursts by calzetti et al . ( 1996 ) , meurer et al . ( 1997 ) and photometric modelling by nakata et al . ( 1999 ) . we must emphasise that our models only account for radio emission produced from star - formation processes . the sources labelled as elliptical ( or early type ) in fig . [ rr ] are not expected to lie on any of the star - formation derived locii . an agn most likely dominates their radio emission . we include them here merely as a comparision , and their relationship to normal starbursts is discussed further below . the sources in fig . [ rr ] appear more - or - less consistent with the dusty `` 0.1 gyr starburst '' model . this could in principle apply to the two sources with spiral / disk - like morphology ( labelled `` s '' ) , but is unconventional for the five elliptical morphologies ( see below ) . a comparison with radio to near - infrared flux ratios further constrains the underlying properties of these sources . figure [ rk ] shows the radio to near - infrared flux ratio @xmath180 , defined analagous to eq.[[ro ] ] , as a function of @xmath126 color . the near - infrared emission is dominated by old stars and is less affected by dust than the optical . the radio to near - infrared flux ratio should therefore be relatively insensitive to dust . given the simplicity of our models , the two disk - like sources may not necessarily represent `` 0.1 gyr starbursts '' as indicated in fig . they could also belong to the sab / sbc or scd / sdm classes . for this to be true , the following additional components may play an important role in more ` realistic ' models : ` optically - thick ' dust that _ completely obscures _ both the @xmath6 and @xmath62 emission without causing appreciable reddening in @xmath126 color , or , contamination by at least an order of magnitude times more radio emission from a central agn than that produced purely by supernovae . the second possibility is favored by radio observations of a number of luminous infrared galaxies by norris et al . ( 1988 ) , where some showed evidence for significant radio emission from compact seyfert - like nuclei . the large discrepancy between the four sources with elliptical morphologies ( labelled `` e '' ) and predictions from the early - type e / so models suggests the importance of a significant agn contribution to the radio emission . appreciable amounts of optically - thick dust suppressing the optical and near - infrared light ( except for extinction by diffuse , optically - thin dust ) is not favored by observations ( eg . goudfrooij & de jong 1995 ) . most , if not all of these ellipticals are likely to be radio - powered by agn . at the limiting sensitivity of our radio survey ( @xmath181 mjy ) , a nominal fr - i galaxy ( with @xmath182 ) could be detected to @xmath183 , and indeed the spread in @xmath126 color for the ellipticals in fig . [ rk ] is consistent with the e / so ( optically - thin dust ) model to this redshift . a comparison between the model @xmath180 values with actual observed values implies that such agn will contribute a factor @xmath184 times more radio emission than that produced by any underlying star formation activity in these systems . it is important to note that the ratio of agn - to - stellar powered radio activity has a huge spread for the elliptical population in general , and that the factor @xmath184 only illustrates a property specific to the ellipticals in our radio sample . to summarise , our use of a simple synthesis model that includes radio emission and dust reddening to analyse the properties of sub - mjy radio sources has shown the following : first , the presence of dust with extinctions @xmath185 mag and possibly greater , consistent with previous more direct determinations , and second , that the level of radio emission from non - stellar processes such as agn could be easily inferred and constrained . this will be particularly important for starbursts hosting seyfert nuclei where a comparison with more sophisticated dust models may be required to infer the relative contributions . since the emission ( and dust absorption ) properties from normal galaxy populations are reasonably well known , a color - color diagram such as fig . [ rr ] could provide a potential diagnostic for selecting ultraluminous infrared galaxies ( uligs ) to high redshift . the relatively low sensitivity of the infrared - astronomical satellite ( iras ) has primarily confined ulig selection to the local universe ( sanders & mirabel 1996 ) , although there is some speculation that recently discovered faint `` scuba '' sources at sub - millimetre wavelengths could represent their high - redshift counterparts ( eg . blain et al . approximately 80% of local uligs are believed to be powered by starbursts and the remainder show evidence for an agn contribution ( genzel et al . 1998 ; lutz et al . far - infrared observations have shown that dust and molecular gas in local uligs is concentrated in compact regions @xmath186kpc ( okumura et al . 1991 , bryant 1996 ) and that a large fraction of the optical / uv emission is hidden by optically - thick dust ( sanders et al . a study of their properties and importance to galaxy evolution therefore requires observations at wavelengths virtually immune to dust absorption . radio frequencies provide an excellent window of opportunity . figure [ rkother ] illustrates the predicted locus in color - color space using the synthetic seds of three local far - ir selected systems : arp 220 ( @xmath187 ) - a ulig undergoing a powerful starburst as seen via high resolution radio observations by smith et al . ( 1998 ) ; m82 ( @xmath188 ) - a system undergoing a weak - to - moderate starburst , and mrk 273 ( @xmath189 ) - a ulig whose bolometric emission is believed to be dominated by a hidden central agn from the presence of strong seyfert-2 lines and moderately strong hard x - ray ( 2 - 10 kev ) emission ( turner et al . 1997 ) . we have used the synthetic seds generated by devriendt et al . ( 1999 ) to model the starburst emission . these authors used a self - consistent modelling approach to predict the stellar optical / uv / near - ir emission , its reprocessing into the mid - ir to sub - mm by dust , and the nonthermal stellar - powered radio emission based on the empirical radio to far - ir luminosity correlation . due to its strong agn - dominated nature , the starburst synthetic sed predicted by devriendt et al . for mrk 273 differs appreciably from that observed in the _ radio_. for this source , we therefore used the devriendt et al . sed at wavelengths @xmath190 mm and extrapolated into the radio using its actual _ observed _ radio to1 mm spectral slope and fluxes ( obtained from the nasa / ipac extragalactic database ) . figure [ rkother ] shows that a galaxy characteristic of the ( low - ir luminous ) m82 system will occupy a region similar to that occupied by normal galaxies in this study ( and also their predicted synthetic colors in fig . luminous systems classified as uligs however ( arp 220 and mrk 273 ) , will tend have higher radio to optical flux ratios which could be easily selected . this can be explained by the well - observed correlation between far - infrared luminosity and far - ir ( @xmath191m)to optical spectral slope ( soifer et al . consequently , the most ir - luminous systems with the largest far - ir to optical ratios are also likely to have a high level of radio - emission due to its strong correlation with ir luminosity . this will lead to a larger than average radio to optical flux ratio for uligs in general as shown in fig . [ rkother ] . the existence of systems with either larger rest - frame optical / uv extinction or excess agn contribution to the radio than the uligs considered here will be shifted further upwards on this plot . diagnostics to distinguish between agn and starburst dominated uligs using radio - to - optical color alone will not be trivial and is left to a future study . the three uligs in fig . [ rkother ] represent a range of known uligs and their location on this plot simply serves as a diagnostic to pre - select ulig candidates for further study . a system like arp 220 ( with @xmath192 ) could be observed to redshift @xmath193 , @xmath194 if initially selected from a radio survey limited to @xmath195jy . thus , to limiting sensitivities reached by existing 1.4 ghz surveys , such a method may not probe the highest redshifts . nonetheless , as shown in fig . [ rkother ] , such systems could still be well separated from normal galaxies to this redshift . assuming an arp 220-like sed and moderate luminosity evolution ( @xmath196 ) , the surface density of uligs to @xmath193 is expected to be of order @xmath197 at @xmath198jy , or about @xmath199 of the integral count to this sensitivity ( richards 2000 ) . they should therefore exist in significant numbers in deep large - area radio surveys . we have used the vla radio telescope to image a contiguous @xmath36 area to a ( mean ) limiting ( @xmath1 ) sensitivity of @xmath200 mjy . from a total of 62 detections , the results of optical and near - infrared photometry are reported for 43 sources . our optical photometry is more sensitive than previous optical follow - up studies of radio surveys of similar depth . our main findings are : \(1 ) we have used a robust , likelihood - ratio method for determining optical identifications and their reliability . this method is seldom used in identification studies and is insensitive to assumptions concerning fluctuations in background source density and gaussian error distributions . we assigned optical candidates to 26 radio sources with reliability @xmath8 . nine radio sources are uncertain and/or ambiguous , and eight are empty fields . near - infrared photometry from the _ 2mass _ database was reported for 7 sources . \(2 ) the eight optical empty field sources all display compact and symmetric radio morphologies and most probably represent compact starbursts at @xmath136 strongly obscured by dust . they may require at least 4 magnitudes optical extinction to account for their large radio to optical flux ratio compared to the identified population . our conclusion for them being ` compact starbursts ' is very tentative as it is purely based on starburst versus agn number statistics expected from sub - mjy radio surveys . further deep infrared / optical imaging and spectroscopy will be necessary . \(3 ) consistent with previous studies , our deep ( @xmath3 ) optical imaging shows that the optical appearence can be divided into two classes according to radio flux - density : elliptical - like morphologies for @xmath132 mjy , and peculiar or disturbed for @xmath201 mjy . \(4 ) using a stellar synthesis model which includes radio emission and dust reddening , we find that the near - infrared to optical emission in a small , bright sub - sample is reddened by ` optically thin ' dust with @xmath9 mag , regardless of morphological type . this appears consistent with other more direct determinations . consistent with previous studies , the radio emission from early - type systems seems to be powered by agn rather than star - formation to account for their anomalously large radio to optical(near - infrared ) ratios . despite the lack of spectroscopic information , our study of a homogeneous population of faint radio sources has stressed the importance of dust on studies of intrinsic galaxy properties and their evolution at optical wavelengths . a future goal would be to obtain spectra , or multi - color optical / near - infrared photometry to better explore these sources and the validity of the simple stellar synthesis models presented in this paper . the ever improving resolution ( and sensitivity ) capabilities of optical / near - ir detectors over those feasible at ( the longest ) radio wavelengths requires robust identification techniques to better ascertain their properties . likelihood - ratios provide one such technique . the present study complements other deep optical studies of faint radio sources to constitute a statistically significant sample for inferring their nature and importance to galaxy evolution . fjm thanks glenn morrison and joann olinger for valuable assistance with the data reduction and rosalie ewald for assistance with radio / optical image overlays . we thank the staff at palomar observatory for technical assistance during the observing run . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by the national aeronautics and space administration and the national science foundation . the national radio astronomy observatory is operated by associated universities , inc . , under cooperative agreement with the national science foundation . this research has made use of the nasa / ipac extragalactic database ( ned ) which is operated by the jet propulsion laboratory , caltech , under contract with the national aeronautics and space administration . fjm acknowledges support from a jpl / nasa postdoctoral fellowship grant . benn , c. r. , rowan - robinson , m. , mcmahon , r. g. , broadhurst , t. j. , & lawrence , a. , 1993 , , 263 , 98 blain , a. w. , smail , i. , ivison , r. j. , & kneib , i. -p . , 1999 , , 302 , 632 bruzual , a. g. , & charlot , s. , 1993 , , 405 , 538 bryant , p. , 1996 , in _ co observations of luminous infrared galaxies _ , phd thesis , california institute of technology calzetti , d. kinney , a. l. , & storchi - bergmann , t. , 1994 , , 429 , 582 calzetti , d. kinney , a. l. , & storchi - bergmann , t. , 1996 , , 458 , 132 calzetti , d. , armus , l. , bohlin , r. c. , kinney , a. l. , koornneef , j. , storchi - bergmann , t. , 2000 , , 533 , 682 ciliegi , p. , mcmahon , r. g. , miley , g. , gruppioni , c. , rowan - robinson , m. et al . , 1999 , , 302 , 222 condon , j. j. , 1989 , , 338 , 13 condon , j. j. , & yin , q. f. , 1990 , , 357 , 97 condon , j. j. , 1992 , , 30 , 575 condon , j. j. , 1997 , , 109 , 166 condon , j. j. , cotton , w. d. , greison , e. w. , yin , q. f. , perley , r. a. , taylor , g. b. , & broderick , j. j. , 1998 , , 115 , 1693 cram , l. e. , 1998 , , 506 , l85 cram , l. e. , hopkins , a. m. , mobasher , b. , & rowan - robinson , m. , 1998 , , 507 , 155 cutri et al . , 2000 , explanatory supplement to the 2mass second incremental data release , http://www.ipac.caltech.edu/2mass/releases/second/doc/explsup.html danese , l. , de zotti , g. , franceschini , a. , & toffolatti , l. , 1987 , , 318 , l15 de ruiter , h. r. , willis , a. g. , & arp , h. c. , 1977 , , 28 , 211 devriendt , j. e. , guiderdoni , b. , & sadat , r. , 1999 , , 350 , 381 downes , a. j. b. , peacock , j. a. , savage , a. , carrie , d. r. , 1986 , , 218 , 31 fanaroff , b. l. , & riley , j. m. , 1974 , , 167 , 31 gavazzi , g. , & scodeggio , m. , 1996 , , 312 , l29 georgakakis , a. , mobasher , b. , cram , l. , hopkins , a. , & lidman , c. , 1999 , , 306 , 708 genzel , r. , lutz , d. , sturm , e. , egami , e. , kunze , d. , moorwood , a. f. m. , rigopoulou , d. , spoon , h. w. w. , sternberg , a. , tacconi - garman , l. e. , tacconi , l. , and thatte , n. , 1998 , , 498 , 579 goudfrooij , p. , & de jong , t. , 1995 , , 298 , 784 gruppioni , c. , mignoli , m. , & zamorani , g. , 1998 , , 304 , 199 haarsma , d. b. , partridge , r. b. , waddington , i. , & windhorst , r. a. , 1999 , 19th texas symposium on relativistic astrophysics and cosmology , paris 1998 , ed . j. paul , t. montmerle , & e. aubourg ( cea sarclay ) , 600 haarsma , d. b. , partridge , r. b. , windhorst , r. a. , & richards , e. a. , 2000 , astro - ph/0007315 , to appear in hammer , f. , crampton , d. , lilly , s. j. , le fvre , o. , & kenet , t. , 1995 , , 276 , 1085 helou , g. , soifer , b. t. , & rowan - robinson , m. , 1985 , , 298 , l7 hopkins , a. m. , mobasher , b. , cram , l. , & rowan - robinson , m. , 1998 , , 296 , 839 kells , w. , dressler , a. , sivaramakrishnan , a. , carr , d. , koch , e. , epps , h. , hilyard , d. , & pardeilhan , g. , 1998 , , 110 , 1487 kron , r. g. , koo , d. c. , & windhorst , r. a. , 1985 , , 146 , 38 lonsdale , c. j. , conrow , t. , evans , t. , fullmer , l. , moshir , m. , chester , t. , yentis , d. , wolstencroft , r. , macgillivray , h. , & egret , d. , 1998 , new horizons from multi - wavelength sky surveys , proc . iau colloquium 179 , johns hopkins university , ed . b. mclean , d. golombek , j. hayes , & h. payne , kluwer academic publishers , p. 450 lutz , d. , spoon , h. w. w. , rigopoulou , d. , moorwood , a. f. m. , genzel , r. , 1998 , , 505 , l103 madau , p. , ferguson , h. c. , dickinson , m. e. , giavalisco , m. , steidel , c. c. , & fruchter a. , 1996 , , 283 , 1388 madau , p. , pozetti , l. , dickinson , m. , 1998 , , 498 , 106 maddox , s. j. , sutherland , w. j. , efstathiou , g. , loveday , j. , 1990 , , 243 , 692 metcalfe , n. , shanks , t. , fong , r. , jones , l. r. , 1991 , , 249 , 498 meurer , g. , heckman , t. m. , lehnert , m. d. , leitherer , c. , & lowenthal , j. , 1997 , , 114 , 54 mitchell , k. j. & condon , j. j. , 1985 , , 90 , 1957 mobasher , b. , cram , l. , georgakakis , a. , & hopkins , a. , 1999 , , 308 , 45 nakata , f. , shimasaku , k. , doi , m. , kashikawa , n. , kawasaki , w. et al . , 1999 , , 309 , l25 norris , r. p. , allen , d. a. , & roche , p. f. , 1988 , , 234 , 773 okumura , s. k. , kawabe , r. , ishiguro , m. , kasuga , t. , morita , k. i. , ishizuki , s. , 1991 , in _ dynamics of galaxies and their molecular cloud distributions _ f. combes , f. casoli , p. 425 , dordrecht : reidel oort , m. j. a. , 1987 , s , 71 , 221 pei , y. c. , 1992 , , 395 , 130 prestage , r. m. , & peacock , j. a. , 1983 , , 204 , 355 richards , e. a. , fomalont , e. b. , kellermann , k. i. , windhorst , r. a. , partridge , r. b. , cowie , l. l. , barger , a. j. , 1999 , , 526 , l73 richards , e. a. , 2000 , astro - ph/9908313 , to appear in rowan - robinson , m. , benn , c. r. , lawrence , a. , mcmahon , r. g. , & broadhurst , t. j. , 1993 , , 263 , 123 sanders , d. b. , soifer , b. t. , elias , j. h. , madore , b. f. , matthews , k. , neugebauer , g. , scoville , n. z. , 1988 , , 325 , 74 sanders , d. b. , & mirabel , i. f. , 1996 , , 34 , 749 schaerer , d. , 1999 , astro - ph/9906014 , to appear in building the galaxies : from the primordial universe to the present , xixth moriond astrophysics meeting , f. hammer et al . ( eds . ) , editions frontieres ( gif - sur - yvette ) scoville , n. z. , sargent , a. i. , sanders , d. b. , & soifer , b. t. , 1991 , , 366 , l5 serjeant , s. , gruppioni , c. , & oliver , s. , 1998 , preprint astro - ph/9808259 smith , h. e. , lonsdale , c. j. , lonsdale , c. j. , & diamond , p. j. , 1998 , , 493 , 17 soifer , b. t. , sanders , d. b. , madore , b. f. , neugebauer , g. , danielson , g. e. , elias , j. h. , lonsdale , c j. , & rice , w. l. , 1987 , , 320 , 238 stetson , p. b. , 1987 , , 99 , 191 sutherland , w. , & saunders , w. , 1992 , , 259 , 413 thuan , x. t. , & gunn , j. e. , 1976 , , 88 , 543 thuan , x. t. , & condon , j. j. , 1987 , , 322 , l9 . thuan , x. t. , windhorst , r. a. , puschell , j. j. , isaacman , b. r. , & owen , f. n. , 1984 , , 285 , 515 turner , t. j. , george , i. m. , nandra , k. , mushotzky , r. f. , 1997 , , 113 , 23 windhorst , r. a. , miley , k. g. , owen , f. n. , kron , r. g. , & david , c. k. , 1985 , , 289 , 494 windhorst , r. a. , fomalont , e. b. , partridge , r. b. , lowenthal , j. d. , 1993 , , 405 , 498

over 3 billion astronomical objects have been detected in the more than 22 million orthogonal transfer ccd images obtained as part of the pan - starrs1 survey . over 85 billion instances of those objects have been automatically detected and characterized by the pan - starrs image processing pipeline photometry software , . this fast , automatic , and reliable software was developed for the pan - starrs project , but is easily adaptable to images from other telescopes . we describe the analysis of the astronomical objects by in general as well as for the specific case of the 3rd processing version used for the first public release of the pan - starrs survey data .

1612.05244

this is the fourth in a series of seven papers describing the pan - starrs1 surveys , the data reduction techiques and the resulting data products . this paper ( paper iv ) describes the details of the source detection and photometry , including point - spread - function and extended source fitting models , and the techniques for `` forced '' photometry measurements . * paper i ) provides an overview of the pan - starrs system , the design and execution of the surveys , the resulting image and catalog data products , a discussion of the overall data quality and basic characteristics , and a brief summary of important results . * paper ii ) describes how the various data processing stages are organised and implemented in the imaging processing pipeline ( ipp ) , including details of the the processing database which is a critical element in the ipp infrastructure . * paper iii ) describes the details of the pixel processing algorithms , including detrending , warping , and adding ( to create stacked images ) and subtracting ( to create difference images ) and resulting image products and their properties . * paper v ) describes the final calibration process , and the resulting photometric and astrometric quality . * paper vi ) describes the details of the resulting catalog data and its organization in the pan - starrs database . * paper vii ) describes the medium deep survey in detail , including the unique issues and data products specific to that survey . the medium deep survey is not part of data release 1 . ( dr1 ) the pan - starrs1 filters and photometric system have already been described in detail in @xcite . the photometric and astrometric precision goals for the pan - starrs1 surveys were quite stringent : photmetric accuracy of 10 millimagnitudes , relative astrometric accuracy of 10 milliarcseconds and absolute astrometric accuracy of 100 milliarcseconds with respect to the icrs reference stars . an additional constraint on the pan - starrs analysis system comes from the high data rate . ps1 produces typically @xmath1 exposures per night , corresponding to @xmath2 billion pixels of imaging data . the images range from high galactic latitudes to the galactic bulge , so large numbers of measurable stars can be expected in much of the data . the combination of the high precision goals of the astrometric and photometric measurements and the high data rate ( and a finite computing budget ) mean that the process of detecting , classifying , and measuring the astronomical objects in the image data stream in a timely fashion are a significant challenge . in order to achieve these ambitious goals , the object detection , classification , and measurement process must be both precise and efficient . not only is it necessary to make a careful measurement of the flux of individual objects , it is also critical to characterize the image point - spread - function , and its variations across the field and from image to image . since comparisons between images must be reliable , the measurements must be stable for both photometry and astrometry . a variety of astronomical software packages perform the basic object detection , measurement , and classification tasks needed by the pan - starrs ipp . each of these programs have their own advantages and disadvantages . below we discuss some of the most widely used of these other packages , highlighting the features of the programs which are particularly desirable , and noting aspects of the programs which are problematic for the ipp . * dophot : analytical fitted model with aperture corrections . pro : well - tested , stable code . con : limited range of models , algorithm converges slowly to a psf model , limited tests of psf validity , inflexible code base , fortran @xcite . * daophot : pixel - map psf model with analytical component . pro : well - tested , high - quality photometry . con : difficult to use in an automated fashion , does it handle 2d variations well ? @xcite . * sextractor : pure aperture measurement with rudimentary object subtraction . pro : fast , widely used , easy to automate . con : poor object separation in crowded regions , psf - modeling was only in beta , not widely used at the time @xcite . * galfit : detailed galaxy modeling . not a multi - object psf analysis tool . con : does not provide a psf model , not easily automated . very detailed results in very slow processing . only a galaxy analysis program @xcite . * sdss phot : con : tightly integrated into the sdss software environment @xcite . when the ipp development was starting , the existing photometry packages either did not meet the level of accuracy required or were required too much human intervention to be considered for the needs of ps1 . in the case of the sdss photo tool , the software was judged to be too tightly integrated to the architecture of sdss to be easily re - integrated into the pan - starrs pipelin . a new photometry analysis package was developed using lessons learned from the existing photometry systems . in the process , the object analysis software was written using the data analysis c - code library written for the ipp , . components of the photometry code were integrated into the ipp s mid - level astronomy data analysis toolkit called . the result software , psphot , can be used either as a stand - alone c program , or as a set of library functions which may be integrated into other programs the main version of psphot is a stand - alone program which is run on a single image or a group of related images representing the data read from a camera in a single exposure . the images are expected to have already been detrended so that pixel values are linearly related to the flux . the gain may be specific by the configuration system , or a variance image may be supplied . a mask may also be supplied to mark good , bad , and suspect pixels . several variants of psphot have also been used in the ps1 pv3 analysis . the version called psphotstack accepts a set of images , each representing the same patch of sky in a different filter , nominally the full @xmath3 filter set for the analysis of the ps1 pv3 stack images , though where insufficient data were available in a given filter , a subset of these filters was processed as a group . as discussed in detail below , the psphotstack analysis includes the capability of measuring forced psf photometry in some filter images based on the position of sources detected in the other filters . it also include an option to convolve the set of images to a single , common psf size across the filters for the purpose of fixed aperture photometry . a second version of psphot used in the pv3 analysis is called psphotfullforce . in this version , a set of image all representing the same pixels are processed together , with the positions of sources to be analysed loaded from a supplied file . in this version the analysis , sources are not discovered only the supplied sources are considered . psf models are determined for each exposure and the forced psf photometry is measured for all sources . a subset of sources may also be used to measure forced galaxy shape parameters . as described below , a grid of galaxy models are fitted based on the supplied guess model . psphot has a number of important requirements that it must meet , and a number of design goals which we believe will help to make usable in a wide range of circumstances . the critical requirements of the pan - starrs ipp which drive the requirements for psphot : * * 10 millimagnitude photometric accuracy*. for psphot , this implies that the measured photometry of stellar objects must be substantially better than this 10 mmag since the photometry error per image is combined with an error in the flat - field calibration and an error in measuring the atmospheric effects . we have set a goal for psphot of 3mmag photometric consistency for bright stars between pairs of images obtained in photometric conditions at the same pointing , ie to remove sensitivity to flat - field errors . this goal splits the difference between the three main contributors and still allows some leeway . this requirement must be met for well - sampled images and images with only modest undersampling . * * 10 milliarcsecond astrometric accuracy*. relative astrometric calibration depends on the consistency of the individual measurements . the measurements from psphot must be sufficiently representative of the true object position to enable astrometric calibration at the 10mas level . the error in the individual measurements will be folded together with the errors introduced by the optical system , the effects of seeing , and by the available reference catalogs . we have set a goal for psphot of 5mas consistency between the true source postion and the measured position given reasonable psf variations under simulations . this level must be reached for images with 250 mas pixels , implying psphot must introduce measurement errors less than 1/50th of a pixel . the choice of 32 bit floating point data values for the source centroids places a numerical limit of 1e-7 on the accuracy of a pixel relative to the size of a chip ( since a single data value is used for x or y ) . for the @xmath4 gpc chips , this yields a limit of about 0.25 milliarcsecond . the design goals for psphot are chosen to make the program flexible , general , and able to meet the unknown usages cases future projects may require : * * flexible psf model * different image sources require different ways of representing the psf . ideally , both analytical and pixel - based versions should be possible . * * psf spatial variation * most images result in some spatial psf variations at a certain level . the psf representation should naturally incorporate 2-d variations . * * flexible non - psf models * psphot must be able to represent psf - like objects as well as non - psf sources ( e.g. , galaxies ) . it must be easy to add new object models as interesting representations of sources are invented . * * clean code base * psphot should incorporate a high - degree of abstraction and encapsulation so that changes to the code structure can be performed without pulling the code apart and starting from scratch . * * psf validity tests * psphot should include the ability to choose different types of psf models for diffent situations , or to provide the user with methods for assessing the different psf models . * * careful systematic corrections * psphot must carefully measure and correct for the photometric and astrometric trends introduced by using analytical psf models . * * user configurable * psphot should allow users to change the options easily and to allow different approaches to the analysis . 1 . * image preparation * load data , characterize the image background , load or construct variance and mask images . 2 . * initial object detection * smooth , find peaks , measure basic properties . * psf determination * select psf candidates , perform model fits , build psf model from fits , select best psf model class . * bright object analysis * fit objects with psfs , determine psf validity , subtract psf - like objects , fit non - psf model(s ) , select best model class , subtract model . n sources * detect low - level sources , measure properties ( aperture or psf ) 6 . * extended source analysis * detailed measurements relevant to galaxies and/or other extended ( non - psf ) sources . 7 . * aperture corrections * measure the curve - of - growth , spatial aperture variations , and background - error corrections . 8 . * output * write out objects in selected format , write out difference image , variance image , etc , as selected . psphot is highly configurable . users may choose via the configuration system which of the above analyses are performed . this is useful for testing , but also allows for specialized use cases . for example , the psf model may already be available from external information , in which case the psf modeling stage can be skipped . the first step is to prepare the image for detection of the astronomical objects . we need three separate images : the measured flux ( signal image ) , the corresponding variance image , and a mask defining which pixels are valid and which should be ignored . the signal and variance images are represented internally as 32-bit floating point values . the variance and mask images may either be provided by the user , or they may be automatically generated from the input image , based on configuration - defined values for the image gain , read - noise , saturation , and so forth . for the function - call form of the program , the flux image is provided in the api , and references to the mask and variance are provided in the configuration information . as in the stand - alone c - program , the variance and mask may be constructed automatically by psphot . the mask is represented as 16-bit integer image in which a value of 0 represents a valid pixel . each of the 16 bits define different reasons a pixel should be ignored . this allows us to optionally respect or ignore the mask depending on the circumstance . for example , in some cases , we ignore saturated pixels completely while in other circumstances , it may be useful to know the flux value of the saturated pixel . in addition , the mask pixels are used to define the pixels available during a model fit , and which should be ignored for that specific fit . the initial mask , if not supplied by the user , is constructed by default from the image by applying three rules : 1 ) pixels which are above a specified saturation level are marked as saturated ( configuration keyword : ) . 2 ) pixels which are below a user - defined value are considered unresponsive and masked as dead . 3 ) pixels which lie outside of a user - defined coordinate window are considered non - data pixels ( eg , overscan ) and are marked as invalid . the valid window is defined by the configuration variables , , , . psphot ( and other ipp ) functions understand two types of masked pixels : `` bad '' and `` suspect '' . bad pixels are those which should not be used in any operations , while suspect pixels are those for which the reported signal may be contaminated or biased , but may be useable in some contexts . for example , a pixel with poor charge transfer efficiency is likely to be too untrustworthy to use in any circumstance , while a pixel in which persistence ghosts have been subtracted might be useful for detection or even analysis of brighter sources . table [ tab : mask_values ] lists the 16 bit values used for ps1 mask images , along with their description ( see * ? ? ? * for additional information ) . the variance image , if not supplied is constructed by default from the flux image using the configuration supplied values of and to calculate the appropriate poisson statistics for each pixel . in this case , the image is assumed to represent the readout from a single detector , with well - defined gain and read noise characteristics . this assumption is not always valid . for example , if the input flux image is the result of an image stack with a variable number of input measurements per pixel ( due to masking and dithering ) , the variance can not be calculated from the signal image alone . it is necessary in such a case to supply a variance image which accurately represents the variance as a function of position in the image . some image processing steps introduce cross - correlation between pixel fluxes . an obvious case is smoothing , but geometric transformations which redistibute fractional flux between neighboring pixels also introduces cross - correlations . in the noise model , it is necessary to track the impact of the cross correlations on the per - pixel variance . in the general case , this would require a complete covariance image , consisting of the set of cross - correlated pixels for each image pixel . since a typical smoothing or warping operation may introduce correlation between 25 - 100 neighboring pixels , the size of such a covariance image is prohibitive . before sources are detected in the image , a model of the background is subtracted . the image is divided into a grid of background points with a spacing of 400 pixels . superpixels of size @xmath5 pixels are used to measure the local background for each background grid point , thus over - sampling the background spatial variations by a factor of 2 . in the interest of speed , 10,000 randomly selected _ unmasked _ pixels in these regions are sampled to determine the background . bilinear interpolation is used to generate a full - resolution image from the grid of background points , and this image is then subtracted from the science image . the background image and the background standard deviation image are kept in memory from which the values of and are calculated for each object in the output catalog . see also the discussion in @xcite . the objects are initially detected by finding the location of local peaks in the image . the flux and variance images are smoothed with a small circularly symmetric kernel using a two - pass 1d gaussian . the smoothed flux and variance images are combined to generate a significance image in signal - to - noise units , including correction for the covariance , if known . at this stage , the goal is only to detect the brighter sources , above a user defined s / n limit ( configuration keyword : ) . a maximum of are found at this stage . the detection efficiency for the brighter sources is not strongly dependent on the form of this smoothing function . the local peaks in the smoothed image are found by first detecting local peaks in each row . for each peak , the neighboring pixels are then examined and the peak is accepted or rejected depending on a set of simple rules . first , any peak which is greater than all 8 neighboring pixels is kept . any peak which is lower than any of the 8 neighboring pixels is rejected . any peak which has the same value as any of the other 8 pixels is kept if the pixel @xmath6 and @xmath7 coordinates are greater than or equal to the other equal value pixels . this simple rule set means that a flat - topped region will maintain peaks at the maximum @xmath6 and @xmath7 corners of the region . we use the 9 pixels which include the source peak to fit for the position and position errors . we model the peak of the sources as a 2d quadratic polynomial , and use a very simple bi - quadratic fit to these pixels . we use the following function to describe the peak by approximating the error per pixel as the error on just the peak , and pulling that term out of the above equation , and recognizing that the values x , y in the 3x3 grid centered on the peak pixel have values of only 0 or 1 , we can greatly simplify the chi - square equation to a square matrix equation with the following values : @xmath10 the peaks detected in the image may correspond to real sources , but they may also correspond to noise fluctuations , especially in the wings of bright stars . psphot attempts to identify peaks which may be formally significant , but are not locally significant . it first generates a set of `` footprints '' , contiguous collections of pixels in the smoothed significance image above the detection threshold . these regions are grown by a small amount to avoid errors on rough edges an image of the footprints is convolved with a disk of radius 3 pixels . peaks are assigned to the footprints in which they are contained ( note by definition all peaks must be located in a footprint ) . for any peak which is not the brightest peak in that footprint it is possible to reach the brightest peak by following the highest valued pixels between the two peaks . the lowest pixel along this path is the _ key col _ for this peak ( as used in topographic descriptions of a mountain ) . if the key col for a given peak is less than ( 4.0 ) sigmas below the peak of interest , the peak is considered to be _ locally insignificant _ and removed from the list of possible detections ( see figure [ fig : peaks ] ) . in the vicinity of a saturated star , the rule is somewhat more agressive as the flat - topped or structured saturated top of a bright star may appear as multiple peaks with highly significant cols between them . however , this is an artifact of the proximity to saturation . in this regime , we require the col to also be a fixed fraction ( 5% ) of the saturation below the peak to avoid being marked as locally insignificant . example of the biases encountered when measuring the second moments . a simulated image was generated using the ps1 psf profile . each panel corresponds to a different value of @xmath15 , as marked . the solid red line is the true fwhm of the psf used to generate the stars . the blue solid line is the fwhm of the window function ( @xmath16 ) . the gray dots are the fwhm derived from the measured second moments for stars in the image . the dotted blue line is the target ( 65% of the window function ) . in this example , we would choose @xmath15 between 0.5 and 0.8 arcseconds so the dotted blue line would match the bright end of the gray dots . ] once a collection of peaks has been identified , a number of basic properties of the objects related to the first and second moments are measured . below , the second moments are used to select candidate stellar sources to be used in modeling the psf . in order to measure the moments , it is necessary to define an appropriate aperture in which the moments are measured . we also apply a `` window function '' , down - weighting the pixels by a gaussian of size @xmath17 which is chosen to be large compared to the psf size , @xmath18 . this window function reduces the noise of the measurement of the first and second moments by suppressing the noisy pixels at high radial distance as well as by reducing the contaminating effects of neighboring stars . the choice of the window function @xmath17 and the aperture is an iterative process : for a given value of @xmath17 , the psf stars will have a measured value of @xmath18 which is modified by the effect of the window function . in addition , depending on the size of the window function compared to the true psf size , the measured value of the psf size , @xmath18 , will be biased high or low depending on the signal - to - noise of the object . these effects are illustrated in figure [ fig : moments.window ] using simulated data . an image was generated with a psf model matching the radial profile of the ps1 psf model with a fwhm of 1.4 arcseconds . as the window function @xmath17 is increased , the measured fwhm for the bright simulated stars rises to meet the truth value . for small values of @xmath17 , fainter stars are biased to low measured values of the fwhm . for large values of @xmath17 , the faint stars are biased to higher values and the scatter increases . in a real image , we do not know the true value of the psf size . if we simply choose a very large window function and rely on bright stars , our estimate of the psf size will be quite noisy . compounding this problem are the two additional facts that ( 1 ) we do not know which are the real stars ( as opposed to bright galaxies or possible image artifacts ) and ( 2 ) the brighter stars are themselves subject to additional biases due to saturation and other non - linear effects ( c.f . , `` the brighter - fatter '' effect , ref ) . to make a robust choice for the window function @xmath19 , we choose a value such that the measured value of @xmath18 is 65% of @xmath15 . the resulting second moment values are biased somewhat low ( 75% of the truth value for the ps1 psf profile ) , but are relatively unbiased as a function of brightness . to choose the value of @xmath17 , we try values of ( 1 , 2 , 3 , 4.5 , 6 , 9 , 12 , 18 ) pixels @xmath20 ( 0.26 , 0.51 , 0.77 , 1.15 , 1.54 , 2.3 , 3.1 , 4.6 ) arcseconds . for each of these values , we then select candidate psf stars based on the distribution of the measured @xmath21 values . for each test value of @xmath15 , we determine the ratio @xmath22 , i.e. , the ratio of the window size to the observed psf size . we interpolate to find a value of @xmath17 for which @xmath23 is expected to be 0.65 . we call this value the . we use an aperture with a radius of = 4@xmath24 to select the pixels for the measurement . where @xmath26 is the flux in a pixel ; @xmath27 is the local sky value for that pixel ; @xmath28 is the value of the window function for the pixel ; @xmath29 is the window - weighted sum of the source flux , used to re - normalize the moments ; @xmath30 is the radius of a pixel , @xmath31 ; the sum is performed over all pixels in the aperture . for the centroid calculation ( @xmath32 ) , the peak coordinate ( see [ sec : peaks ] ) is used to define the aperture and the window function ; for higher order moments , the centroid is used to center the window function . if the measured centroid coordinates ( @xmath33 ) differs from the peak coordinates be a large amount ( ) , then the peak is identified as being of poor quality and is rejected . in both of these cases , it is likely that the ` peak ' was identified in a region of flat flux distribution or many saturated or edge pixels . in addition to the moments above , a preliminary kron radius and flux are also calculated at this stage . in this analysis , the 1st and half - radial moments are calculated : @xmath34 note that the window function is not applied in the calculation of these moments . the kron radius @xcite is defined the be 2.5@xmath24 the first radial moment . the kron flux is the sum of ( sky - subtracted ) pixel fluxes within the kron radius . we also calculate the flux in two related annular apertures : the kron inner flux is the sum of pixel values for the annulus @xmath35 , while the kron outer flux is the sum of pixel values for @xmath36 . the first radial moment is limited at the low and high ends by @xmath37 where @xmath38 is the first radial moment of the psf stars , or 0.75@xmath24 if that can not be determined . @xmath39 is set to , the size of the moments aperture . psphot uses an analytical model to represent the shape and flux of an object . an important concept within the psphot code is the distinction between a model which describes an object on an image and a model with describes the point - spread - function ( psf ) across an image . any object in an image may be represented by some analytical model , for example , a 2-d elliptical gaussian : @xmath40 the object model will have a variety of model parameters , in this case the centroid coordinates ( @xmath41 ) , the elliptical shape parameters ( @xmath42 ) , the model normalization ( @xmath43 ) and the local value of the background ( @xmath44 ) . a specific object will have a particular set of values for these different parameters . the point - spread - function ( psf ) of an image describes the shape of all unresolved objects in the image . in a typical image , the shape of point sources is not well described by a single function . instead , the shape will vary as a function of position in the image . the psf model therefore must describe the parameter variation as a function of the position of the object on the image . note that the object model consists of a certain number of parameters which are defined by the psf model , and another set of parameters which are independent from object to object . for the case of the elliptical gaussian model , the psf parameters would be the shape terms ( @xmath45 ) while the independent parameters would be the centroid , normalization and local sky values ( @xmath46 ) . thus these parameters are each a function of the object centroid coordinates : @xmath47 psphot represents the variation in the psf parameters as a function of position in the image in two possible ways , specified by the configuration . the first option is to use a 2-d polynomial which is fitted to the measured parameter values across the image . the second option is to use a grid of values which are measured for objects within a subregion of the image . in the latter case , the value at a specific coordinate in the image is determined by interpolation between the nearest grid points . the order of the polynomial or the sampling size of the grid is dynamically determined depending on the number of available of psf stars . in the case of the pv3 analysis , the grid of values was used , with a maximum of @xmath48 samples per gpc1 chip image . for the earlier pv2 analysis , the maximum grid sampling was @xmath49 per gpc1 chip image . for the pv1 analysis , the polynomial representation was used , with up to 3rd order terms . the higher order representation was used for pv3 in order to follow some of the observed psf variations in the images psphot uses a single structure to represent the object model and another structure to represent the psf model . the object model structure consists of the collection of measured object model parameters , carried as a vector ( ) along with an equal - length vector with the parameter errors . the structure also includes an integer giving the identifier of the model used in the particular case , as well as model fit statistics such as the chi - square of the fit and the magnitude representation of the ratio between the model flux and an aperture flux ( see below for more details on this value ) . the psphot representation of the psf consists of an array of polynomials , each representing the variation in the object model psf parameters ( of ) . the psf model structure also includes the same integer used to identify which model corresponds to particular instance of the psf . at the moment , the number of psf parameters is a fixed number ( 4 ) fewer than the number of parameters of the corresponding object model . for example , the elliptical gaussian model uses 7 parameters to represent the object and 3 for the psf model . psphot is written so that the object detection , measurement , and classification code does not depend on the specific form of the available object model functions . access to the characteristics of the models is provided through a simple function abstraction method . throughout psphot , there are many places where it is necessary for the code to refer to an aspect of the object or psf model . often , these quantities are needed deep within other parts of the code . for example , when attempting to fit the pixel flux values for an object , it is necessary to generate a guess for the model parameters . or , in order to limit the domain of the fit , it is necessary to determine an isophotal radius for a model . in order to avoid having the code depend on the specific form of a model , the function calls needed in these types of circumstances are abstracted , and a method is provided to return the necessary function to the higher - level software . for example , each model type has its own function to define an initial guess for the model , or a function to determine the radius for a given flux level . these are then registered as part of the model function code . another function is then used to return the appropriate function for a specific model type . for example , the will return the function for a given model type . this mechanism makes it very easy to add new model functions into the psphot code base . to add a new model function , the programmer simply defines a new model name ( a string ) , the set of all necessary model lookup functions , and places the reference to the model code at the appropriate location in the psmodelinit.c routine . when a new model is provided to psphot , it is not necessary to specify the intended use of the object model function ( ie , psf - like object , galaxy , comet , etc ) . any model can be used for the psf model , or to describe the flux distributions of the non - psf objects . the code currently uses a fixed translation between the object model parameters and the psf model parameters . it also defines a specific order for the 4 independent parameters . the first stage of determining the psf model for an image is to identify a collection of objects in the image which are _ likely _ to be psf - like . psphot uses the object moments to make the initial guess at a collection of psf - like objects . at this point , the program has measured the second order moments for all objects identified by their peaks , as well as an approximate signal - to - noise ratio . all objects with a s / n ratio greater than a user - defined parameter ( = 20.0 ) are selected by psphot , though objects which have more than a certain number of saturated pixels are excluded at this stage . psphot then examines the 2-d plane of @xmath50 in search of a concentrated clump of objects ( see figure [ fig : moment.class ] ) . to do this , it constructs an artificial image with pixels representing the value of @xmath50 , using a user - defined scale for the size of a pixel in this artificial image ( note that the units of the @xmath50 plane are the size of the second - moment in pixels in the original image ) . a typical value for the bin size is approximately 0.1 image pixels . the binned @xmath50 plane is then examined to find a peak which has a significance greater than xxx . unless the image is extremely sparse , such a peak will be well - defined and should represent the objects which are all very similar in shape . other objects in the image will tend to land in very different locations , failing to produce a single peak . to avoid detecting a peak from the unresolved cosmic rays , objects which have second - moments very close to 0 are ignored . the only danger is if the psf is very small and too many of these objects are rejected as cosmic rays . once a peak has been detected in this plane , the centroid and second moments of this peak are measured . all objects which land within xxx @xmath51 of this centroid are selected as likely psf - like objects in the image . illustration of psf star selection using the fwhm derived from the second moments in @xmath52 and @xmath53 directions . the dominant clump is located in this diagram . galaxies tend to have a range of sizes and thus spread out above the stars . cosmic rays also have a range of sizes , with one dimension smaller than the psf . the red circle represents the psf star candidates . ] all candidate psf objects are then fitted with the selected object model , allowing all of the parameters ( psf and independent ) to vary in the fit . psphot uses the levenberg - marquardt minimization technique for the non - linear fitting . non - linear fitting can be very computationally intensive , particularly for if the starting parameters are far from the minimization values . psphot uses the first and second moments to make a good guess for the centroid and shape parameters for the psf models . any objects which fail to converge in the fit are flagged as invalid . for the resulting collection of object model parameters , the psf - dependent parameters of the models are all fitted as a function of position to a 2-d polynomial . the order of this polynomial is a user - defined parameter . the fitting process for these polynomials is iterative , and rejects the @xmath54 outliers in each of three passes . this fitting technique results in a robust measurement of the variation of the psf model parameters as a function of position without being excessively biased by individual objects which fail drastically . objects whose model parameters are rejected by this iterative fitting technique are also marked as invalid and ignored in the later psf model fitting stages . all of the psf - candidate objects are then re - fitted using the psf model to specify the dependent model parameter values for each object . for example , in the case of the elliptical gaussian model , the shape parameters ( @xmath55 ) for each object are set by the coordinates of the object centroid and fixed ( not allowed to vary ) in the fitting procedure . the resulting fitted models are then used to determine a metric which tests the quality of the psf model for this particular image . the metric used by psphot to assess the psf model is the scatter in the differences between the aperture and fit magnitudes for the psf objects . the difference between the aperture and fit magnitudes ( _ apresid _ ) is a critical parameter for any psf modeling software which uses an analytical model to represent the flux distribution of the objects in an image . an approximate correction is measured here , with a more detailed correction measured after all object analysis is performed . the psf model with the best consistency of the aperture correction is judged to be the best model . before the detailed analysis of the objects is performed , it is convenient to subtract off all of the sources , at least as well as possible at this stage . we make the assumption that all sources are psf - like . if the centroid of the source has been determined , we use this value for its position ; otherwise , we use the interpolated position of the peak . a single linear fit is used to simultaneously measure all source fluxes . since the local sky has been subtracted , this measurement assumes the local sky is zero . we can write a single @xmath56 equation for this ) ^2\ ] ] where @xmath57 is image flux for each pixel , @xmath58 $ ] is the psf model realized at the position of source @xmath59 , and @xmath60 is the normalization for the source . minimizing this equation with respect to each of the @xmath60 values results in a matrix equation : @xmath61 where @xmath62 is the vector of @xmath60 values , the elements of @xmath63 consist of the dot products of the unit - flux psf for source @xmath59 and source @xmath64 , and @xmath65 is the dot product of the unit - flux psf for source @xmath64 with the pixels corresponding to source @xmath64 . the dot products are calculated only using pixels within the source apertures . since most sources have no overlap with most other sources , this matrix equation results in a very sparse , mostly diagonal square matrix . the dimension is the number of sources , likely to be 1000s or 10,000s . direct inversion of the matrix would be computationally very slow . however , an interative solution quickly yields a result with sufficient accuracy . in the iterative solution , a guess at the solution @xmath66 is made assuming @xmath63 is purely diagonal ; the guess is multiplied by @xmath63 , and the result compared with the observed vector @xmath65 . the difference is used to modify the initial guess . this proces is repeated several times to achieve a good convergence . once a psf model has been selected for an image , psphot attempts to fit all of the detected objects , above a user - defined signal - to - noise ratio with the psf model . for these fits , the dependent parameters are fixed by the psf model and only the 4 independent object model parameters are allowed to vary in the fit . psphot again uses levenberg - marquardt minimization for the non - linear fitting . the objects are fitted in their s / n order , starting with the brightest and working down to the user - specified limit . once a solution has been achieved for an object , psphot attempts to judge the quality of the psf model as a representation of the object shape . to do this , it calculates the next step of the minimization _ allowing the shape parameters to vary_. this step , essentially the gauss - newton minimization distance from the current local minimum , should be very small if the object is well represented by the psf , but large if the psf is not a good representation of the object flux . the model quality is judged by the change in the two shape parameters which represent the 2d size of the object . for the case of the elliptical gaussian , these two parameters are @xmath67 and @xmath68 . for a generic model , the shape parameters may be defined differently , but there should always be two parameters which scale the object size in two dimensions . currently , psphot requires the two relevant shape parameters to be the first two dependent parameters in the list of model parameters ( ie , parameters 4 & 5 ) . the expected distribution of the variation of the two shape parameters will be a function of the signal - to - noise of the object in question and the value of the shape parameter itself . the expected standard deviation on the shape parameter is , eg , @xmath69 . if the object is well - represented by the psf , then the shape parameter values should be close to their minimization value . we can thus ask , for each object , given the measured amplitude of the gauss - newton step , how many standard deviations from the expected value ( of 0.0 ) is this particular value ? objects for which the variation in the shape parameters is a large positive number of standard deviations are likely to be better represented by a larger flux distribution than the psf ( eg , a galaxy or comet , etc ) . objects for which the variation in the shape parameters is a large negative number of standard deviations are likely to be better represented by a smaller flux distribution than the psf ( ie , a cosmic ray or other defect ) . a user - defined number of standard deviations is used to select these two cases , and to flag the object as a likely galaxy ( really meaning extended ) or as a likely defect . at this stage of the analysis , psphot uses two additional indicators to identify good and poor psf fits . the first of these is the signal - to - noise ratio . it is possible for the peak finding algorithm to identify peaks in locations which are not actually a normal peak . some of these cases are in the edges of saturated , bleeding columns from bright stars , in the nearly flat halos of very bright stars , and so on . in these cases , a local peak exists , with a lower nearby sky region . however , the fitted psf model can not converge on the peak because it is very poorly defined ( perhaps only existing in the smoothed image ) . the fit can either fail to converge or it can converge on a fit with very low or negative peak flux / flux normalization . psphot will flag any non - convergent psf fit and any object with psf s / n ratio lower than a user - defined cutoff . it is also useful to identify very poor fits by setting a maximum chi - square cutoff for objects . as the objects are fitted to the psf model , those which survive the exclusion stage are subtracted from the image . the subtraction process modifies the image pixels ( removing the fitted flux , though not the locally fitted background ) but does not modify the mask or the variance images . the signal - to - noise ratio in the image after subtraction represents the significance of the remaining flux . if the subtractions are sufficiently accurate models of the psf flux distribution , the remaining flux should be below 1 @xmath51 significance . in practice the cores of bright stars are poorly represented and may have larger residual significance . sources which are blended with other sources are fitted together as a set of psfs . a single multi - object fit is performed on all blended peaks . the resulting fits are evaluated independently and any which are determined to be psfs are subtracted from the image . sources which are judged to be non - psf - like are confronted with two possible alternative choices . first , the object is fitted with a double - source model . in this pass , the assumption is made that there are two neighboring sources , but the peaks are blended together , or otherwise not distinguished . the initial guess for the two peaks is made by splitting the flux of the single source in half and locating the two starting peaks at + /- 2 pixels from the original peak along the direction of the semi - major axis of the sources , as measured from the second moments . in order for the two - source model to be accepted , both sources must be judged as a valid psf source . otherwise , the double - psf model is rejected and the source is fitted with the available non - psf model or models . after the psf model has been fitted to all sources , and the kron flux has been measured for all sources , psphot uses these two measurements , along with some additional pixel - level analysis , to determine the size class of the object . if the object is large compared to a psf , it is considered to be _ extended _ and will be fitted with a galaxy model ( or possibly another type of extended source model in special cases ) . if the object is small compared to a psf , it is considered to be a _ cosmic ray _ and masked . extended sources are identified as those for which the kron magnitude is significantly brighter than the psf magnitude when compared to a psf star . the value @xmath70 , the difference between the psf and kron magnitudes , is calculated for each object . the median of @xmath71 is calculated for the psf stars . this median is subtracted from @xmath71 for each star . the result is divided by the quadrature error of the psf and kron magnitudes and called . if is larger than ( 3.0 ) , the object is considered to be extended . cosmic rays are identified by a combination of the kron magnitude and the second - moment width of the object in the minor axis direction . the second - moment in the minor axis direction is calculated from @xmath72 as follows : @xmath73 if @xmath74 pixels@xmath75 and the instrumental kron magnitude is @xmath76 , then the object is identified as a cosmic ray and the associated pixels are masked . once every object ( above the s / n cutoff ) has been confronted with the psf model , the objects which are thought to be galaxies ( extended ) can now be fit with appropriate models for the galaxies ( or other likely extended shapes ) . again , the fitting stage starts with the brightest sources ( as judged by the rough s / n measured from the moments aperture ) and working to a user defined s / n limit . psphot will use the user - selected galaxy model to attempt the galaxy model fits . in the configuration system , the keyword is set to the model of interest . all suspected extended objects are fitted with the model , allowing all of the parameters to float . the initial parameter guesses are critical here to achieving convergence on the model fits in a reasonable time . the moments and the pixel flux distribution are used to make the initial parameter guess . many of the object parameters can be accurately guessed from the first and second moments . the power - law slope can be guessed by measuring the isophotal level at two elliptical radii and comparing the ratio to that expected . for each of the galaxy models ( in fact for all object models ) , a function is defined which examines the fit results and determines if the fit can be consider as a success or a failure . the exact criteria for this decision will depend on the details of the model , and so this level of abstraction is needed . for example , in some case , the range of valid values for each of the parameters must be considered in the fit assessment . in other cases , we may choose to use only the parameter errors and the fit chi - square value . all galaxy model fits which are successful are then subtracted from the image as is done for the successful psf model fits . of course , the background flux is retained , with the result that only the object is subtracted from the image . again , the variance image is ( currently ) not modified . after a first pass through the image , in which the brighter sources above a high threshold level have been detected , measured , and subtracted , psphot optionally begins a second pass at the image . in this stage , the new peaks are detected on the image with the bright objects subtracted . in this pass , the peak detection process uses the variance image to test the validity of the individual peaks . all peaks with a significance greater than a user - defined minimum threshold are accepted as objects of potential interest . the objects which are measured in this faint - object stage are clearly low significance detections . the pv3 threshold for the bright object analysis is a signal - to - noise of 20 . the lower limit cutoff for the faint object analysis in pv3 is a signal - to - noise of 5.0 . objects detected in the faint object stage are fitted with the psf model using the linear , ensemble fitting process . the important concept here is that an analytical model will always fail to describe the flux of the objects at some level . in the end , all astronomical photometry is in some sense a relative measurement between two images . whether the goal is calibration of a science image taken at one location to a standard star image at another location , or the goal is simply the repetitive photometry of the same star at the same location in the image , it is always necessary to compare the photometry between two images . if this measurement is to be consistent , then the measurement must represent the flux of the stars in the same way regardless of the conditions under which the images were taken , at least within some range of normal image conditions . so , for example , two images with different image quality , or with different tracking and focus errors , will have different psf models . since an analytical model will always fail to represent the flux of the star at some level , the measured flux of the same object in the two images will be different ( even assuming all other atmospheric and instrumental effects have been corrected ) . the amplitude of the error will by determined by how inconsistently the models represent the actual object flux . for example , if the first image psf model flux is consistently 10% too low and the second is 5% too high , then the comparison between the two images will be in error by 15% . aperture photometry avoids these problems , by trading for other difficulties . in aperture photometry , if a large enough aperture is chosen , the amount of flux which is lost will be a small fraction of the total object flux . even more importantly , as the image conditions change , the amount lost will change by an even smaller fraction , at least for a large aperture . this can be seen by the fact that the dominant variations in the image quality are in the focus , tracking and seeing . all of these errors initially affect the cores of the stellar images , rather than the wide wings . the wide wings are largely dominated by scattering in the optics and scattering in the atmosphere . the amplitude and distribution of these two scattering functions do not change significantly or quickly for a single telescope and site . the difficulty for aperture photometry is the need to make an accurate measurement of the local background for each object . as the aperture grows , errors in the measurement of the sky flux start to become dominant . if the aperture is too small , then variation in the image quality are dominant . the brighter is the object , the smaller is the error introduced by the large size of the aperture . however , the number of very bright stars is limited in any image , and of course the brighter stars are more likely to suffer from non - linearity or saturation . psphot measures the aperture correction ( _ apresid _ ) for every psf candidate object and applies this correction to the psf model photometry . psphot allows a collection of psf model functions to be tried on all psf candidate objects . for each model test , the above corrected apresid scatter is measured . the psf model function with the smallest value for the apresid scatter is then used by psphot as the best psf model for this image . the number of models to be tested is specified by the configuration keyword . the configuration variables , , through specify the names of the models which should be tested . where @xmath82 ( @xmath83 ) . the pseudo - gaussian is a taylor expansion of the gaussian and is used by dophot @xcite . the latter profiles are similar to the moffat profile form @xcite , with small differences . for the ps1 gpc1 analysis , we used the model , which we found by experimentation to match well to the observed profiles generated by ps1 . figure [ fig : radial.profiles ] shows example radial profiles for moderately bright stars in fairly good ( 0.9 arcsec ) and poor ( 2.2 arcsec ) seeing . using a fixed power - law exponent results in somewhat faster profile fitting compared to the variable power - law exponent model . radial profiles of stellar images from ps1 . these two profiles illustrate the radial trend of the ps1 psfs for a star with fwhm 0.9 arcsec ( red ) and 2.2 arcsec ( blue ) . the black line shows the psf model with radial trend of the form @xmath84 . ] galaxies with regular profiles , such as elliptical galaxies and regular spiral galaxies , may be described as primarily a radial surface brightness profile , with additional structure acting as a perturbation on that profile . for many galaxies , the azimuthal shape at a given isophotal level may be described as an elliptical contour . to first order , a galaxy may be well decribed with a single elliptical contour and radial profile . in order to facilitate the petrosian photometry analysis below , psphot generates a radial profile for each suspected galaxy . this analysis starts by generating a radial profile in 24 azimuthal segments . near the center of the galaxy , the profile is defined for radial coordinates in steps of 1 pixel , with the closest pixel values interpolated to that radial position . further from the center , profile is defined using the median of the pixels landing in an annular segment of size @xmath85 , rounded up to the nearest integer pixel value . the median of all pixels within a rectangular approximation to the radial wedge is used . the resulting 24 radial profiles are subject to contamination from neighboring sources or to nan values from masked pixels . to clean the profiles , pairs of radial profiles from opposite sides of the source are compared . any masked values are replaced by the corresponding value in the other profile . the minimum of both profiles is the kept for both profiles . the result of this analysis is a set of profiles of the form @xmath86 . in this case , @xmath26 is effectively the surface brightness for each radius in instrumental counts per pixel . the surface brightness profiles are then used to define the radial contour at a specific isophotal level . this contour will be used to rescale the radial profiles into a single set of profiles normalized by the elliptical contour . this contour is defined by determining the median radius for profile bins with surface brightness in the range @xmath87 to @xmath88 . the result of this analysis is a value for the radius as a function of the angle for a well - defined surface brightness regime . we then determine the elliptical shape parameters for this elliptical contour : @xmath89 . this ellipse is then used to redefine a single radial profile normalized by the elliptical contour : @xmath90 the surface brightness values are sampled at a number of radial annuli , with the radii defined in the configuration ( & ) . for each source , the resulting surface brightness profile is saved in the output cmf - file as an n - element value in the fits table ( ) . the flux at each radial position and the fill - factor ( fraction of pixels used to the total possible ) as also saved as equal - length vectors in the fits table ( and ) . the values of the radial bins are saved in the cmf header ( , ) . @xcite defined an adaptive aperture based on a ratio of surface brightnesses . the motivation is to define an aperture which can be determined for galaxies without significant biases as a function of distance . since surface brightness in a resolved object is conserved , using a ratio of surface brightness to define a spatial scale results in a spatial scale which is constant regardless of galaxy distance . to measure the petrosian radius and flux , we start by defining a series of radial apertures with power - law spacing : @xmath91 . we calculate the surface brightness for the annulus from @xmath92 by calculating the median of the values in the range @xmath93 to @xmath94 and dividing the the effective area of the annulus corresponding to @xmath95 . for any annulus @xmath59 spanning the radii @xmath96 to @xmath97 , the petrosian ratio for that annulus is defined as the ratio of the surface brightness in the annulus to the average surface brigthness within @xmath98 . the petrosian radius is defined to be @xmath99 for the annulus for which the petrosian ratio = 0.2 , i.e. , the point on the galaxy radial profile at which the surface brightness is 20% of the average surface brightness at that point . we determine the petrosian radius for the galaxy by quadratic interpolation between the last two of the fixed annuli with petrosian ratio @xmath100 and the first annulus with petrosian ratio @xmath101 . in general , the petrosian ratio for a galaxy with a regular morphology ( spiral or elliptical ) is falling monotonically , so this interpolation is unambiguous . however , irregular galaxy morphologies , noise , and/or significant masking can cause the petrosian ratio to have rises as well as drops . we track the petrosian ratio until the value is no longer significant ( @xmath102 ) . if the petrosian ratio drops below 0.2 for more than one radius , we choose the largest such radius . once the petrosian radius has been determined , we can now measure the petrosian flux : this is defined to be the total flux within an aperture corresponding to 2 @xmath24 the petrosian radius . using the petrosian flux , we can calculate two other interesting radii : @xmath103 and @xmath104 , the radii inside which 50% and 90% of the total petrosian flux is contained . we attempt to measure the radial profile of sources in order to find the radius at which the flux of the object is matches the sky . in this analysis , a series of up to 25 radial bins with power - law spacing are defined and the flux of the object in each annulus is measured . the `` sky radius '' is defined to be the radius at which the ( robust median ) flux in the annulus is within 1 @xmath51 of the local sky level . if this limit is not reached before the slope of the flux from one annulus to the next is less than a user - defined limit , then the annulus at which the slope reaches this limit is used to define the sky radius . these values are saved in the output smf / cmf files , but not sent to the psps . the sky radius value is used below in the calculation of the kron magnitude . preliminary kron radius and flux values @xcite are calculated soon after sources are detected ( section [ sec : moments ] ) . however , these preliminary values are not accurate due to the window - functions applied . after sources have been characterized and the psf model is well - determined , the kron parameters are re - calculated more carefully . in this version of the calculation , the image is first smoothed by gaussian kernel with @xmath105 pixels , corresponding to a fwhm of 1.0 in the ps1 stack images . next , the kron radius is determined in an iterative process : the first radial moment is measured using the pixels in an aperture 6@xmath24 the first radial moment from the previous iteration . on the first iteration , the sky radius is used in place of the first radial moment . by default , 2 iterations are performed . the kron radius is defined the be 2.5@xmath24 the first radial moment . the kron flux is the sum of pixel fluxes within the kron radius . we also calculate the flux in two related annular apertures : the kron inner flux is the sum of pixel values for the annulus @xmath35 , while the kron outer flux is the sum of pixel values for @xmath106 . two details in the calculation above should be noted . first , for faint sources , noise in the measurement of the 1st radial moment may result in an excessively small aperture for the successive calculations . the window used for the calculations is constrained to be at least the size of the aperture based on the psf stars ( section [ sec : moments ] ) . at the other extreme , noise may make the radius grow excessively , resulting in an unrealistically low effective surface brightness . the aperture is constrained to be less than a maximum value defined such that the minimum surface brightness is 1/2@xmath107 the effective surface brightness of a source detected at the @xmath108 limit . second , the measurement of the 1st radial moment includes a filter to reduce contamination from outlier pixels . pairs of pixels on opposites sides of the central pixel are considered together . the geometric mean of the two fluxes is used to replace the flux values . if the object has 180 symmetry , this operation has no impact . however , if one of the two pixels is unusually high , the value will be surpressed by the matched pixel on the other side . this trick has the effect of reducing the impact of pixels which include flux from near neighbors . in the galaxy model fittting stage , sources which meet certain criteria are fitted with analytical models for galaxies . three traditional analytical galaxy models are implemented in and used in the pv3 analysis : where @xmath112 is a normalized radial term : @xmath113 . the terms ( @xmath114 , @xmath115 , @xmath116 ) describe the elliptical contour and the profile scale in all three models and the coordinates @xmath117 & @xmath118 are determined relative to the centroids ( @xmath119 ) . including the normalization ( @xmath120 ) and a local sky value , the exponential and devaucouleur profiles have 7 free parameters and the sersic profile has the additional free parameter of the sersic index @xmath121 . in this stage , the galaxy model is convolved with an approximation to our best guess for the psf model at the location of the galaxy . for the pv3 analysis , all sources detected in the bright source analysis step ( @xmath122 ) were fitted with all three galaxy models , unless ( a ) the morphological test identified the source as a likely cosmic ray ( section [ sec : source.size ] ) or ( b ) the peak of the psf profile was above the saturation limit for the chip ( see the discussion in * ? ? ? * regarding the masking of saturated pixels ) . sources in the denser portions of the galactic plane and bulge were not included in the analysis . this restriction limited the total time spent on the galaxy modeling analysis at the expense of galaxy photometry in the plane ( though kron photometry is available for those objects ) . the galactic plane region was defined by @xmath123 where @xmath124 . for the pv3 analysis , @xmath12520 , @xmath12615 , @xmath12750 . before the non - linear fitting may be performed , it is necessary to determine initial values for the parameters to be fitted . for each of the three model types , the position determined from the psf fitting analysis is used as the initial centroid @xmath128 . a guess for the terms ( @xmath114 , @xmath115 , @xmath116 ) is generated based on the second moments . the guess does not attempt to use the psf model to adjust the ( @xmath114 , @xmath115 , @xmath116 ) values ; it was found that such a guess tended to be too small and resulted in more iterations rather than fewer . the 1st radial moment ( see [ sec : moments ] ) is used to estimate the effective radius of the model based on the results of graham & driver ( 2005 , table 1 ) . they quantive the relationships between the first radial moment used to calculated a kron magnitude and the effective radius for different sersic index values , @xmath121 . since the exponential and devaucouleur models are equivalent to sersic models with @xmath121 = 1 and 4 , respectively , this work can be used to generate the initial effective radius values for all 3 model types . once the effective radius is chosen , the second moments are used to define the aspect ratio and position angle of the elliptical contour . the kron flux is used to generate a guess for the normalization , applying an appropriate scale factor based on the ( @xmath114 , @xmath115 , @xmath116 ) values , generated by integrating normalized sersic models and determining the relationship between the central intensity and the integrated flux as a function of the sersic index . the psf - convolved galaxy model fitting analysys uses the levenberg - marquardt minimization method to determine the best fit . in this process , the @xmath56 value to be minimized is : @xmath129 ^ 2\ ] ] where @xmath130 represents the pixel values in the image ( within some aperture ) and @xmath131 represents the unconvolved galaxy model , a function of a number of parameters @xmath132 , which is then convolved with the psf model . to determine the minimization , we need the gradient and laplacian of @xmath56 with respect to the model parameters , @xmath134 : @xmath135 where we have approximated the laplacian with the hessian matrix , @xmath136 by dropping the second - derivatives ( which are assumed to be a small perturbation ) . since @xmath137 and since the order of the derivative and convolution may be exchanged , we can write these in terms of the convolved image of the model and the convolved images of the derivatives of the model @xmath138 with respect to the model parameters , @xmath134 : @xmath139 the gradient vector and hessian matrix are used in the levenberg - marquardt minimization analysis using the standard techinique of determining a step from the current set of model parameters to a new set by solving the matrix equation : @xmath140 where @xmath141 is zero for @xmath142 and for @xmath143 set to be large when the last iteration produced a large change in the parameters compared to the local - linear expectation and small when the last change was small . the iteration ends when the change in the parameters is small and/or the change in the @xmath56 value is small . in the analysis , convolved galaxy fit , the galaxy model image and the model derivative images must be convolved with the psf at each iteration step . to save computation time , this convolution is performed using a circularly symmetric approximation of the psf model , with the psf model scale size set to the average of the major and minor axis direction scale size of the full psf model , with the same radial profile term as the psf model . the convolution is performed directly using the circular symmetry to reduce the number of multiplications performed : all points in the 2d circularly symmetric psf model which have the same radial pixel coordinate can be evaluated in the convolution by summing up the corresponding pixels in the ( galaxy model ) image to be convolved before multiplying by the psf model profile at that radial coordinate . this approximation reduces the number of multiplications by a factor of near 8 for larger radii . for the small size of the psf model used to convolve the galaxy model images , it was found that this direct convolution was faster than using an fft - based convolution . for the exponential and devaucouleur fits , all parameters are fitted in the non - linear minimization stage . for the sersic model , we do not fit the index within the levenberg - marquardt analysis . instead , we start with a coarse grid search over a range of possible index values , ( @xmath144 ) and a range of possible values for @xmath145 based on the value of @xmath146 , the first radial moment . for a given value of the sersic index , the @xmath145 is related to the 1st radial moment by the scale factor specificy by graham & driver . we use the observed value of the 1st radial moment and try @xmath145 values of a factor of ( 0.8 , 0.9 , 1.0 , 1.12 , 1.25 ) times the value predicted by the graham and driver equation . for each of these steps , the aspect ratio and position angle are held constant and the normalization is determined to minimize the @xmath56 . we next perform 3 levenberg - marquardt minimization fits allowing the shape parameters ( @xmath114 , @xmath115 , @xmath116 ) and the normalization to be fitted , holding the centroid ( @xmath33 ) , sersic index @xmath121 , and sky constant . in these fits , the index @xmath121 is set to the minimum value previously calculated as well as values halfway to the next , and previous , values in the grid above . e.g. , if the minimum fitted index value is 3.0 , then the lmm fits are performed using @xmath121 = 2.5 , 3.0 , 3.5 . the resulting @xmath56 values are then used to perform quadratid interpolation to find the index @xmath121 which produces the locally minium @xmath56 value . finally , this best - fit index value is held constant while levenberg - marquardt minimization is used to find the best fit values of all other parameters . for some science goals , a well - measured color of a galaxy is more important than an accurate total magnitude . in the case of ps1 , the image quality variations for stacks of different filters presents a serious challenge for the determination of precise colors . psphot determines a set of psf - matched radial aperture flux measurements in order to minimize the impact of the stack image quality variations . in psphotstack , the stack analysis version of psphot , the 5 filter images are processed together . after the psf models have been fitted and a best set of galaxy models have been determined , three sets of radial apertures are measured . in the first set , the fluxes in the radial apertures are measured using the raw stack images . the centers of the apertures for each object across the 5 filters are fixed so that the pixels represent the equivalent portions of the same galaxy for all 5 filters . in this analysis , the best model for each object is subtracted from the image pixels for all objects excluding the object in consideration . the best model is determined based on the minimum @xmath56 value for the model fits . in addition to the raw radial apertures , the stack images are each convolved with a circular gaussian with @xmath51 chosen to yield an image with a typical fwhm of 6 . the full set of radial apertures are again measured on these convolved images . again , the best object models are subtracted from the image for objects not being measured . this subtraction includes the convolution to smooth the model to the effective fwhm of the convolved image . the entire procedure is then repeated with a target fwhm of 8 . the pan - starrs1 surveys ( ps1 ) have been made possible through contributions of the institute for astronomy , the university of hawaii , the pan - starrs project office , the max - planck society and its participating institutes , the max planck institute for astronomy , heidelberg and the max planck institute for extraterrestrial physics , garching , the johns hopkins university , durham university , the university of edinburgh , queen s university belfast , the harvard - smithsonian center for astrophysics , the las cumbres observatory global telescope network incorporated , the national central university of taiwan , the space telescope science institute , the national aeronautics and space administration under grant no . nnx08ar22 g issued through the planetary science division of the nasa science mission directorate , the national science foundation under grant no . ast-1238877 , the university of maryland , and eotvos lorand university ( elte ) and the los alamos national laboratory .

recently , we developed a method for calculating the _ lifetime _ of a particle inside a magnetic trap with respect to spin flips , as a first step in our efforts to understand the quantum - mechanics of magnetic traps . the 1d toy model that was used in this study was physically _ unrealistic _ because the magnetic field was _ not _ curl - free . here , we study , both classically and quantum - mechanically , the problem of a neutral particle with spin , mass and magnetic moment , moving in 3d in an inhomogeneous magnetic field corresponding to traps of the ioffe - pritchard , ` clover - leaf ' and ` baseball ' type . defining by , and the precessional , the axial and the lateral vibrational frequencies , respectively , of the particle in the adiabatic potential , we find classically the region in the- plane where the particle is trapped . quantum - mechanically , we study the problem of a spin - one particle in the same field . treating and as small parameters for the perturbation from the adiabatic hamiltonian , we derive a closed - form expression for the transition rate of the particle from its trapped ground - state . in the extreme cases the expression for reduces to{c}% 4\pi\omega_{r}\exp\left [ -\dfrac{2\omega_{p}}{\omega_{r}}\right ] \text { ; for } \omega_{p}\gg\omega_{r}\gg\omega_{z}\\ 8\sqrt{2\pi}\sqrt{\omega_{p}\omega_{i}}\exp\left [ -\dfrac{2\omega_{p}}% { \omega_{i}}\right ] \text { ; for } \omega_{p}\gg\omega_{r}=\omega_{z}% \equiv\omega_{i}\\ \sqrt{\dfrac{\pi}{2}}\omega_{r}\left ( \dfrac{\omega_{z}}{\omega_{p}}\right ) ^{3/2}\exp\left [ -\dfrac{2\omega_{p}}{\omega_{z}}\right ] \text { ; for } % \omega_{p}\gg\omega_{z}\gg\omega_{r}% \end{array } \right . .\ ] ]

quant-ph9907044

recently there has been rapid progress in techniques for trapping samples of neutral atoms at elevated densities and extremely low temperatures . the development of magnetic and optical traps for atoms has proceeded in parallel in recent years , in order to attain higher densities and lower temperatures @xcite . we should note here that traps for neutral particles have been around much longer than their realizations for neutral atoms might suggest , and the seminal papers for neutral particles trapping as applied to neutrons and plasmas date from the sixties and seventies . many of these papers are referenced by the authors of refs.@xcite . in this paper we concentrate on the study of _ magnetic _ traps . such traps exploit the interaction of the magnetic moment of the atom with the inhomogeneous magnetic field to provide spatial confinement . microscopic particles are not the only candidates for magnetic traps . in fact , a vivid demonstration of trapping large scale objects is the hovering magnetic top @xcite . this ingenious magnetic device , which hovers in mid - air for about 2 minutes , has been studied in the past few years by several authors @xcite . the physical mechanism underlying the operation of magnetic traps is the adiabatic principle . the common way to describe their operation is in terms of _ classical _ mechanics : as the particle is released into the trap , its magnetic moment points antiparallel to the direction of the magnetic field . inside the trap , the particle experiences translation oscillations with vibrational frequencies @xmath13 which are small compared to its precession frequency @xmath14 . under this condition the spin of the particle may be considered as experiencing a _ slowly _ rotating magnetic field . thus , the spin precesses around the _ local _ direction of the magnetic field @xmath15 ( adiabatic approximation ) and , on the average , its magnetic moment @xmath16 points _ antiparallel _ to the local magnetic field lines . hence , the magnetic energy , which is normally given by @xmath17 , is now given ( for small precession angle ) by @xmath18 . thus , the overall effective potential seen by the particle is @xmath19 in the adiabatic approximation , the spin degree of freedom is rigidly coupled to the translational degrees of freedom , and is already incorporated in eq.([energy ] ) such that the particle may be considered as having only translational degrees of freedom . when the strength of the magnetic field possesses a _ _ minimum _ _ , the effective potential becomes attractive near that minimum , and the whole apparatus acts as a trap . as mentioned above , the adiabatic approximation holds whenever @xmath20 . as @xmath14 is inversely proportional to the spin , this inequality can be satisfied provided that the spin of the particle is sufficiently small . if , on the other hand , the spin of the particle is too large , it can not respond fast enough to the changes of the direction of the magnetic field . in this limit @xmath21 , the spin has to be considered as fixed in space and , according to earnshaw s theorem @xcite , becomes unstable against _ translations_. note also that @xmath14 is proportional to the field @xmath22 . to prevent @xmath14 of becoming too small , resulting in spin - flips ( majorana transitions ) , most magnetostatic traps include a _ bias _ field , so that the effective potential @xmath23 possesses a _ nonvanishing _ minimum . the discussion of magnetic traps in the literature is , almost entirely , done in terms of _ classical _ mechanics . in microscopic systems , however , quantum effects become important , giving rise to a finite lifetime of the particle within the trap . this requires a quantum - mechanical treatment @xcite . an even more interesting issue is the understanding of how the classical and the quantum descriptions of a _ given _ system are related . it is important to note here that there are _ two _ mechanisms by which the particle can escape from the trap : the first one is the usual tunneling of the particle from the trap , without a change of its spin state , to regions where the magnetic field decreases to zero . the time scale for this process can be evaluated by standard methods . the second way the particle can escape from the trap is by flipping its spin state ( majorana transitions ) . this process , which is different from the first one because there is no potential barrier , is the subject of this paper . as a first step in our efforts to understand the quantum - mechanics of magnetic traps , we recently developed a method for calculating the _ lifetime _ of a particle inside a magnetic trap with respect to such a spin flip process @xcite . the toy model that was used in this study consisted of a particle with spin , having only a single translational degree of freedom , in the presence of a 1d inhomogeneous magnetic field . we found that the trapped state of the particle decays with a lifetime given by @xmath24 where @xmath25 , and where the result is valid for @xmath26 . though the field that was used in this model did trap the particle , it was not _ realistic _ in the sense that it was not curl - free . our next step was to study the case of a particle with spin , having _ two _ translational degree of freedom , in the presence of a physically more realistic trapping field that , in contradistinction to the toy model , _ is _ curl - free @xcite . this model is reminiscent of a ioffe - pritchard trap @xcite , but without the axial translational degree of freedom . here we found that the lifetime is given by @xmath27 which is similar to the result found in the 1d case . in the present paper we describe an analysis of a ioffe - pritchard type trap which _ includes _ the axial translational degree of freedom . we neglect the effect of interactions between the particles in the trap , and we analyze the dynamics of a _ single _ particle inside the trap . unlike in our previous papers , where we studied the case of a spin @xmath28 particle , we treat here the case of a spin @xmath29 particle , both as an example to show the validity of our approach for higher spins , and also because it is more relevant in view of the recent development in bose - einstein condensation experiments . the structure of this paper is as follows : in sec.([def ] ) we start by defining the system we study , together with useful parameters that will be used throughout this paper . next , we carry out a _ classical _ analysis of the problem in sec.([class ] ) . here , we find two stationary solutions for the particle inside the trap . one of them corresponds to a state whose spin is _ parallel _ to the direction of the magnetic field whereas the other one corresponds to a state whose spin is _ antiparallel _ to that direction . when considering the dynamical stability of these solutions , we find that only the _ antiparallel _ stationary solution is stable , as expected from the discussion in sec.([desc ] ) above . in sec.([quant ] ) we reconsider the problem from a _ quantum - mechanical _ point of view for a spin-_one _ particle . here , we find states that refer to _ antiparallel _ ( @xmath30 , where @xmath31 is the magnetic quantum number ) and _ orthogonal _ ( @xmath32 ) orientations of the spin , the first of these being bounded while the second one is unbounded . we argue that the third possible situation , in which the spin is _ parallel _ ( @xmath33 ) to the direction of the field , has negligible coupling to the bound state , and therefore can be neglected . we show that the _ antiparallel _ and _ orthogonal _ states are _ coupled _ due to the inhomogeneity of the field , and we calculate the transition rate from the bound state to the unbounded state . finally , in sec.([dis ] ) we compare the results of the classical analysis with those of the quantum analysis and comment on their implications for practical magnetic traps . we consider a particle of mass @xmath1 , magnetic moment @xmath2 and intrinsic spin @xmath0 ( aligned with @xmath2 ) moving in an inhomogeneous magnetic field @xmath15 corresponding to traps of the ioffe - pritchard , ` clover - leaf ' and ` baseball ' type @xcite , and given by @xmath34 \mathbf{\hat{z}% } \label{d0}\\ & + \left ( b^{\prime}-\dfrac{1}{2}b^{\prime\prime}z\right ) x\mathbf{\hat { x}+}\left ( -b^{\prime}-\dfrac{1}{2}b^{\prime\prime}z\right ) y\mathbf{\hat { y}}\text{.}\nonumber\end{aligned}\ ] ] this field possesses a nonzero minimum of amplitude at the origin , which is the essential part of the trap . the hamiltonian for this system is@xmath35 where @xmath36 is the momentum of the particle . we define @xmath3 as the precessional frequency of the particle when it is at the origin @xmath37 . since at that point the magnetic field is @xmath38 we find that @xmath39 next , we define @xmath4 and @xmath5 as the small - amplitude axial and lateral vibrational frequencies of the particle when it is placed with _ spin into the adiabatic potential given by @xmath40 from which we get @xmath41 } { mb_{0}}}\text{.}\nonumber\end{aligned}\ ] ] in what follows we assume that @xmath42 such that @xmath5 is real . we also define the ratios , @xmath43 s^{2}}{\mu mb_{0}^{3}}}\nonumber\end{aligned}\ ] ] these parameters will be our ` measure of adiabaticity ' . it is clear that as @xmath44 and @xmath45 become smaller and smaller , the adiabatic approximation becomes more and more accurate . note that when the bias field @xmath46 vanishes , both @xmath44 and @xmath45 become infinite , and the adiabatic approximation fails . we will show below that , under this condition , the system becomes _ unstable _ against spin flips , which is in agreement with our discussion at the beginning . this shows that the introduction of the bias field @xmath46 , is _ essential _ to the operation of the trap with regard to spin - flips . we denote by @xmath47 a unit vector in the direction of the spin ( and the magnetic moment ) . thus , the equations of motion for the center of mass of the particle are @xmath48 and the evolution of its spin is determined by @xmath49 the two equilibrium solutions to eqs.([c1 ] ) and ( [ c2 ] ) are @xmath50 with@xmath51 representing a motionless particle at the origin with its magnetic moment ( and spin ) pointing _ antiparallel _ ( @xmath52 ) to the direction of the field at that point and a similar solution but with the magnetic moment pointing _ parallel _ to the direction of the field ( @xmath53 ) . to check the stability of these solutions we now add first - order perturbations . we set @xmath54 ( note that , to first order , the perturbation @xmath55 is _ orthogonal _ to the vector @xmath47 for the stationary solution @xmath56 , since @xmath47 is a unit vector ) , substitute these into eqs.([c1 ] ) and ( [ c2 ] ) , and retain only first - order terms . we find that the resulting equations for @xmath57 , @xmath58 , @xmath59 , @xmath60 and @xmath61 are @xmath62 the motion of the @xmath63-coordinate is decoupled from the others . if @xmath64 , it is stable only when the upper sign is taken , corresponding to a spin _ antiparallel _ to the direction of the field . it can be shown that when @xmath65 then , even if the system is stable under axial vibrations ( by choosing the lower sign ) , it can not be stable as a whole . we therefore disregard the lower sign , and the equation for the @xmath63-coordinate for the rest of the derivation . the normal modes of the reduced system transform as the irreducible representations of the symmetry group . the 4-dimensional linear space spanned by the deviations @xmath66 from the stationary state carries the irreducible representations @xmath67 with characters @xmath68 and @xmath69 with characters @xmath70 , and may thus be decomposed into the two 2-dimensional invariant subspaces transforming as @xmath67 and @xmath69 , respectively . these subspaces are spanned by the circular position coordinates and precessional spin coordinates@xmath71 thus , the normal modes consist of a circular motion in the @xmath72-plane coupled to a precession of the spin vector in the _ opposite _ sense . indeed , after introducing the @xmath73-coordinates into eqs.([c5 ] ) , this set of four equations decomposes into one pair of equations for @xmath74 and another pair for @xmath75 . we now look for oscillatory ( stable ) solutions of these equations and set @xmath76 this yields the algebraic equations @xmath77{cc}% \dfrac{1}{2}\mu b^{\prime\prime}+m\omega^{2 } & \mu b^{\prime}\\ i\mu b^{\prime } & i\left ( \omega s+\mu b_{0}\right ) \end{array } \right ) \cdot\left ( \begin{array } [ c]{l}% \rho_{+,0}\\ \epsilon_{-,0}% \end{array } \right ) = \left ( \begin{array } [ c]{l}% 0\\ 0 \end{array } \right ) , \label{c8.1}%\ ] ] @xmath78@xmath79{cc}% \dfrac{1}{2}\mu b^{\prime\prime}+m\omega^{2 } & \mu b^{\prime}\\ -i\mu b^{\prime } & i\left ( \omega s-\mu b_{0}\right ) \end{array } \right ) \cdot\left ( \begin{array } [ c]{l}% \rho_{-,0}\\ \epsilon_{+,0}% \end{array } \right ) = \left ( \begin{array } [ c]{l}% 0\\ 0 \end{array } \right ) . \label{c8.2}%\ ] ] these equations have non - trivial solutions whenever the determinant of either of the two matrices vanishes . this yields the secular equations @xmath80 which determine the eigenfrequencies @xmath81 of the various modes . since the reduced system has three degrees of freedom , we expect to have three normal modes . indeed , when @xmath81 is a solution of the first equation , then @xmath82 is a solution of the second equation . we define the mode frequencies in eq.([c11.1 ] ) to be positive ( or , in the case of complex @xmath81 , to have positive real part ) ; the negative @xmath81-values are needed to construct real solutions . then , the @xmath67-modes describe vibrational motions turning counter - clockwise coupled to spin precessions turning clockwise , i.e. , opposite to the natural spin precession , and the @xmath69-modes describe vibrational motions turning clockwise coupled to spin precessions turning counter - clockwise , i.e. , in the same sense as the natural spin precession . stability requires that all three solutions of , say , eq.([c11.2 ] ) be _ real_. we note that at the edge of the stability region ( and when @xmath83 ) , two out of the three roots of eq.([c11.2 ] ) for @xmath81 become identical . in this case , the third order polynomial eq.([c11.2 ] ) takes the form @xmath84 @xmath85 , which satisfy @xmath86 . the edge of the stability region is then found by simultaneously solving the equations @xmath87 and @xmath88 . the result is given in the form of the parametric curve in the @xmath89-plane@xmath90{c}% k_{r}^{2}\left ( t\right ) = 2t^{3}-t^{2}\\ k_{z}^{2}\left ( t\right ) = 4t-6t^{2}% \end{array } \right\ } \text { ; with } \dfrac{1}{2}<t<\dfrac{2}{3}%\ ] ] which is shown in fig.([fig1 ] ) . note that by eliminating @xmath91 from the second equation and substituting it in the first gives @xmath92 _ explicitly _ in terms of @xmath93 . in this section we consider the problem of a neutral particle with spin _ one _ ( @xmath94 ) in a 3d inhomogeneous magnetic field from a quantum - mechanical point of view . unlike the classical analysis , in which the derivation was valid for any value of the adiabaticity parameters @xmath45 and @xmath44 , we concentrate here on the behavior of the system when @xmath45 and @xmath44 are _ small_. note also that , quantum - mechanically , the magnetic moment @xmath2 and the spin @xmath0 of a particle are related by@xmath95 where @xmath96 is the gyromagnetic ratio of the particle . setting @xmath97 and @xmath94 in eqs.([d3 ] ) gives@xmath98 \hbar}{\gamma mb_{0}^{3}}}.\end{aligned}\ ] ] now , it is convenient to transform to cylindrical coordinates @xmath99 by setting @xmath100 , @xmath101 . we denote by @xmath102 the _ amplitude _ of @xmath15 , by @xmath103 its direction with respect to the @xmath63-axis and by @xmath104 the angle between the projection of @xmath15 onto the @xmath105-plane and the @xmath106-axis . thus , eq.([d0 ] ) is rewritten as @xmath107 . \label{h1}%\ ] ] the approximate expressions for @xmath102 , @xmath103 and @xmath104 near the origin are given by @xmath108 \right ) \simeq -\phi+\mathcal{o}\left ( z\sin\left ( 2\phi\right ) \right ) .\nonumber\end{aligned}\ ] ] thus approximately , @xmath102 and @xmath103 depend only on @xmath109 , whereas @xmath104 depends only linearly on @xmath110 . the time - independent schrdinger equation for this system is @xmath111 \phi(r,\phi , z)=e\phi(r,\phi , z ) \label{h5}%\ ] ] where @xmath112 , @xmath113 and @xmath114 are the spin one matrices , given by @xmath115{ccc}% \hat{s}_{x}=\dfrac{1}{\sqrt{2}}\left ( \begin{array } [ c]{lll}% 0 & 1 & 0\\ 1 & 0 & 1\\ 0 & 1 & 0 \end{array } \right ) & \hat{s}_{y}=\dfrac{1}{\sqrt{2}}\left ( \begin{array } [ c]{lll}% 0 & -i & 0\\ i & 0 & -i\\ 0 & i & 0 \end{array } \right ) & \hat{s}_{z}=\left ( \begin{array } [ c]{lll}% 1 & 0 & 0\\ 0 & 0 & 0\\ 0 & 0 & -1 \end{array } \right ) , \end{array}\ ] ] @xmath116 is the eigenenergy , and @xmath117 is the three - components spinor @xmath118{c}% \phi_{+}(r,\phi , z)\\ \phi_{0}(r,\phi , z)\\ \phi_{-}(r,\phi , z ) \end{array } \right ) . \label{h5.1}%\ ] ] in matrix form eq.([h5 ] ) becomes @xmath119{c}% \phi_{+}(r,\phi , z)\\ \phi_{0}(r,\phi , z)\\ \phi_{-}(r,\phi , z ) \end{array } \right ) = e\left ( \begin{array } [ c]{c}% \phi_{+}(r,\phi , z)\\ \phi_{0}(r,\phi , z)\\ \phi_{-}(r,\phi , z ) \end{array } \right ) , \label{h6.0}%\ ] ] where @xmath120 and @xmath121 , given by @xmath122{lll}% \cos\theta & \dfrac{1}{\sqrt{2}}\sin\theta e^{-i\varphi } & 0\\ \dfrac{1}{\sqrt{2}}\sin\theta e^{i\varphi } & 0 & \dfrac{1}{\sqrt{2}}\sin\theta e^{-i\varphi}\\ 0 & \dfrac{1}{\sqrt{2}}\sin\theta e^{i\varphi } & -\cos\theta \end{array } \right ) , \nonumber\end{aligned}\ ] ] are the kinetic part and the magnetic part of the hamiltonian @xmath123 , respectively . in order to diagonalize the magnetic part of the hamiltonian , we make a local _ passive _ transformation of coordinates on the wavefunction such that the spinor is expressed in a new coordinate system whose @xmath124 axis coincides with the direction of the magnetic field at the point @xmath125 . we denote by @xmath126 the required transformation and set @xmath127 . thus , @xmath128 represent _ the same _ direction of the spin as before the transformation but using the _ new _ coordinate system . the hamiltonian in this newly defined system is given by @xmath129 . we represent the rotation matrix @xmath130 in terms of the three euler angles : first , we perform a rotation through an angle @xmath104 around the @xmath131 axis . second , we make a rotation through an angle @xmath103 around the _ new _ position of the @xmath132 axis . at the end of this process the new @xmath131 axis coincide with the direction of the magnetic field . now the value of the last euler angle , which is a rotation around the new @xmath131 axis , has no effect on this axis . for simplicity we choose this angle to be @xmath133 . thus , the representation of the complete transformation for spin - one particle is given by @xcite @xmath134 \exp\left [ i\varphi\hat{s}% _ { z}\right ] , \ ] ] while its inverse is given by @xmath135 \exp\left [ -i\theta\hat { s}_{y}\right ] .\ ] ] it is easily verified that the transformation indeed diagonalizes the magnetic part of the hamiltonian as @xmath136 as for the kinetic part we show at the appendix that @xmath137{c}% -i\nabla^{2}\varphi\left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}_{x}\right ) -\left| \nabla\varphi\right| ^{2}\left ( \cos\theta\hat{s}_{z}-\sin \theta\hat{s}_{x}\right ) ^{2}\\ -2i\left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}_{x}\right ) \nabla \varphi\cdot\left ( -i\mathbf{\nabla}\theta\hat{s}_{y}+\mathbf{\nabla}\right ) \\ -i\nabla^{2}\theta\hat{s}_{y}-\left| \mathbf{\nabla}\theta\right| ^{2}% \hat{s}_{y}^{2}-2i\hat{s}_{y}\mathbf{\nabla}\theta\cdot\nabla+\nabla^{2}% \end{array } \right ] . \label{trans}%\ ] ] since we are interested in the behavior near the origin , we substitute the approximate expressions eqs.([h1.1 ] ) into eq.([trans ] ) , replace @xmath138 by @xmath29 and @xmath139 by @xmath133 ( since @xmath103 changes very slowly as compared to the extent over which @xmath140 changes significantly ) , and neglect the terms that are proportional to @xmath141 and @xmath142 ( being of higher order with respect to @xmath143 ) . this gives@xmath144 \text{.}%\ ] ] thus , the hamiltonian of the system in the rotated frame may be written approximately as @xmath145 where @xmath146 -\mu b\hat{s}_{z}\label{h7.01}\\ h_{int } & = i\dfrac{\hbar^{2}b^{\prime}}{mb_{0}}\hat{s}_{y}\dfrac{\partial } { \partial r}.\nonumber\end{aligned}\ ] ] the first part of the hamiltonian @xmath147 is diagonal with respect to the spin degrees of freedom . it contains the kinetic part @xmath148 , a term whose form is @xmath149 which is identified as the adiabatic effective potential , and the terms @xmath150 which appear due to the rotation . the second part of the hamiltonian @xmath151 contains only non - diagonal components . generally , @xmath151 should contain terms which couple a spin state @xmath31 to the two nearest spin states @xmath152 and to the two next - to - nearest spin states @xmath153 ( see the appendix ) . in the limit where @xmath44 and @xmath45 are small , we see that the coupling of the state with spin projection value @xmath31 to the states @xmath153 is _ negligible _ compared to its coupling to the @xmath152 states . we proceed to find the eigenstates of @xmath147 . since @xmath147 is diagonal , the three spin states of the wavefunction are decoupled . we seek a solution for the spin - down ( @xmath30 ) state@xmath154{c}% 0\\ 0\\ \psi_{-}(r,\phi , z ) \end{array } \right ) \text { ; } e = e_{- } , \label{h8.01}%\ ] ] and for the @xmath32 state @xmath155{c}% 0\\ \psi_{0}(r,\phi , z)\\ 0 \end{array } \right ) \text { ; } e = e_{0}. \label{h8.02}%\ ] ] we do not consider the _ spin - up _ ( @xmath33 ) state , since its coupling to the trapped spin - down state is negligible , as explained above . the equation for the non - vanishing component of the spin - down state reads @xmath156 + \mu b\right\ } \psi_{-}=e_{-}\psi_{- } , \label{h8.1}%\ ] ] whereas the equation for the non - vanishing component of the spin - zero state is @xmath157 the solutions of these equations is outlined in the next two subsections . eq.([h8.1 ] ) represents a particle in a cylindrically symmetric _ attractive _ 3d potential . if the extent of the wave function is small enough we can expand @xmath102 in eq.([h1.1 ] ) to second order in @xmath109 and @xmath63 as given by eq.([h1.1 ] ) , and apply the well - known solution of the harmonic oscillator @xcite in 3d . under this approximation , eq.([h8.1 ] ) becomes@xmath90{c}% -\dfrac{\hbar^{2}}{2m}\left ( \dfrac{\partial^{2}}{\partial z^{2}}% + \dfrac{\partial^{2}}{\partial r^{2}}+\dfrac{1}{r}\dfrac{\partial}{\partial r}-\dfrac{1}{r^{2}}\left ( i\dfrac{\partial}{\partial\phi}+1\right ) ^{2}\right ) \\ + \left [ \dfrac{m\omega_{z}^{2}}{2}z^{2}+\dfrac{m\omega_{r}^{2}}{2}% r^{2}\right ] \end{array } \right\ } \psi_{-}=\left ( e_{-}-\mu b_{0}\right ) \psi_{-}.\label{down0.1}%\ ] ] the @xmath63-coordinate decouples and we assume that it is in the ground - state . we thus seek a solution whose form is@xmath158 , \label{down0.2}%\ ] ] with @xmath159 an integer . the equation satisfied by @xmath160 is then@xmath161 + \dfrac { m\omega_{r}^{2}r^{2}}{2}f=\left ( e_{-}-\mu b_{0}-\dfrac{1}{2}\hbar\omega _ { z}\right ) f.\label{down0.3}%\ ] ] this is an eigenvalue problem for @xmath162 . the smallest eigenvalue for this equation is obtained by setting@xmath163 for which the eigenfunction @xmath162 is@xmath164 .\ ] ] thus , under the harmonic - oscillator approximation , the normalized down - part of the spin - down state is given by @xmath165 \exp\left [ -\dfrac{m\omega_{z}}{2\hbar}% z^{2}\right ] .\label{down1}%\ ] ] note that the extent of this wave function over which it changes appreciably is given by @xmath166 whereas the extent over which @xmath140 changes significantly ( see eq.([h1.1 ] ) ) is @xmath167 thus , the ratio between these two length scales is @xmath168 we therefore conclude that when @xmath44 and @xmath45 are small enough , the harmonic approximation is justified . the wave function @xmath169 , given by eq.([down1 ] ) , then represents the lowest possible _ bound _ state for this system . this state corresponds to a _ trapped _ particle . the energy of this state is @xmath170 while its full spinor representation is @xmath154{c}% 0\\ 0\\ \sqrt{\dfrac{m\omega_{r}}{\pi\hbar}}\text { } \left ( \dfrac{m\omega_{z}}% { \pi\hbar}\right ) ^{1/4}e^{i\phi}\exp\left [ -\dfrac{m\omega_{r}}{2\hbar } r^{2}\right ] \exp\left [ -\dfrac{m\omega_{z}}{2\hbar}z^{2}\right ] \end{array } \right ) . \label{down6}%\ ] ] eq.([h8.2 ] ) describes a free particle . it corresponds to an unbounded state representing an _ untrapped _ particle . in this case there is a continuum of states , each with its own energy . as we are interested in non - radiative decay , we focus on finding a solution with an energy which is _ equal _ to the energy found for the trapped state , that is @xmath171 we seek a solution in the form@xmath172 , \ ] ] where @xmath173 is an integer . substituting this , together with eq.([up0 ] ) into eq.([h8.2 ] ) gives@xmath174 g=0,\ ] ] where@xmath175 the non - singular solution for @xmath176 is@xmath177 where @xmath178 is the bessel function of the first kind of order @xmath173 . for what follows , it is convenient to introduce an angle @xmath96 such that@xmath179 with @xmath180 . we note that @xmath151 does not operate on the @xmath110 coordinate . hence , in order to have a non - vanishing matrix element between the zero - state and the down - state , they must have the _ same _ @xmath110-dependence . thus , @xmath181 , and as a result , the state with angle @xmath96 is given by @xmath182 \text{. } \label{up0.2}%\ ] ] with@xmath183{c}% 0\\ c_{\gamma}j_{1}\left ( k_{0}r\sin\gamma\right ) \exp\left [ i\left ( \phi+k_{0}z\cos\gamma\right ) \right ] \\ 0 \end{array } \right ) , \text { } \label{up0.3}%\ ] ] where @xmath184 is the normalization constant which is chosen to be real , and depends on @xmath96 . to evaluate @xmath184 we temporarily introduce boundary conditions under which the wavefunction @xmath185 vanishes at @xmath186 , and satisfies periodic boundary conditions along @xmath63 with period @xmath187 . thus , normalization of @xmath185 gives @xmath188^{2}=1,\ ] ] such that@xmath189 where we have used @xcite @xmath190^{2}rdr=\frac{1}{2}r^{2}[j_{2}(kr)]^{2}.\ ] ] in the asymptotic region @xmath191 , the function @xmath192 takes the values @xmath193 at the zeros of @xmath194 . thus,@xmath195 and hence@xmath196 we calculate the transition rate from the bound state given by eq.([down6 ] ) to the unbounded state eq.([up0.3 ] ) , according to fermi s golden rule @xcite . thus , the infinitesimal decay time from the trapped state to the untrapped state defined by @xmath96 is given by @xmath197 where @xmath198 is the density @xmath199 of states @xmath200 with an angle between @xmath96 and @xmath201 and energy between @xmath202 and @xmath203 , and @xmath204 is the matrix element of @xmath151 between the bound state and the unbounded state defined by @xmath96 and @xmath202 . to find @xmath205 we note that the final state is defined by the two quantized @xmath206-vectors @xmath207 and @xmath208 . the possible @xmath209 values are equally - spaced with lattice constant @xmath210 . since the bessel function @xmath211 is very close to its asymptotic behavior at large arguments @xmath212 the @xmath213 are also very much equally - spaced ( even when @xmath213 is small , it is still a good approximation ) with lattice constant @xmath214 . thus , in the @xmath215-space , the allowed @xmath206 vectors form a regular lattice , and the number of states @xmath216 in the volume element @xmath217 is given by@xmath218 with @xmath219 , this gives the density of states@xmath220 this , together with eq.([cg2 ] ) yields@xmath221 evaluation of @xmath204 gives@xmath222 \left ( -\dfrac{m\omega_{r}% } { \hbar}\right ) c_{\gamma}\\ \times & % tcimacro{\dint \limits_{0}^{\infty}}% % beginexpansion { \displaystyle\int\limits_{0}^{\infty } } % endexpansion drr^{2}j_{1}\left ( k_{0}r\sin\gamma\right ) \exp\left [ -\dfrac{m\omega_{r}% } { 2\hbar}r^{2}\right ] \int_{-\infty}^{\infty}dz\exp\left [ -\dfrac { m\omega_{z}}{2\hbar}z^{2}-ik_{0}z\cos\gamma\right ] .\end{aligned}\ ] ] and hence@xmath223 .\nonumber\end{aligned}\ ] ] substituting eqs.([hi2 ] ) and ( [ c2rho ] ) into eq.([t1 ] ) and integrating over @xmath96 from @xmath133 to @xmath224 gives @xmath225 with@xmath226 , \label{i}%\ ] ] where we have substituted our previous definitions for @xmath3 , @xmath5 and @xmath4 . the integral @xmath227 may be expressed in terms of the simpler integral@xmath228 \label{i0}\\ & = 2\exp\left ( -a\right ) \int_{0}^{1}\exp\left [ \left ( a - b\right ) t^{2}\right ] dt\nonumber\end{aligned}\ ] ] by@xmath229 in the isotropic case where @xmath230 , the integral in eq.([1/tt ] ) can be evaluated analytically with the result that@xmath231 .\ ] ] in the extreme cases @xmath232 and @xmath233 we obtain from the asymptotic behavior of the error function of real and imaginary argument @xcite@xmath234{cc}% \sqrt{\dfrac{\pi}{b}}\exp\left ( -a\right ) & \text { ; } b\gg a\\ \dfrac{1}{a}\exp\left ( -b\right ) & \text { ; } b\ll a \end{array } \right . , \label{i0app}%\ ] ] and hence@xmath235{cc}% \sqrt{\dfrac{\pi}{b}}\exp\left ( -a\right ) & \text { ; } b\gg a\\ \dfrac{1}{a^{2}}\exp\left ( -b\right ) & \text { ; } b\ll a \end{array } \right . .\label{i2}%\ ] ] substituting eqs.([i0app ] ) and ( [ i2 ] ) into eq.([1/tt ] ) gives@xmath12{c}% 4\pi\omega_{r}\exp\left [ -\dfrac{2\omega_{p}}{\omega_{r}}\right ] \text { ; for } \omega_{p}\gg\omega_{r}\gg\omega_{z}\\ 8\sqrt{2\pi}\sqrt{\omega_{p}\omega_{i}}\exp\left [ -\dfrac{2\omega_{p}}% { \omega_{i}}\right ] \text { ; for } \omega_{p}\gg\omega_{r}=\omega_{z}% \equiv\omega_{i}\\ \sqrt{\dfrac{\pi}{2}}\omega_{r}\left ( \dfrac{\omega_{z}}{\omega_{p}}\right ) ^{3/2}\exp\left [ -\dfrac{2\omega_{p}}{\omega_{z}}\right ] \text { ; for } % \omega_{p}\gg\omega_{z}\gg\omega_{r}% \end{array } \right . , \label{final}%\ ] ] with the conclusion that the transition rate is dominated by the _ largest _ of the two frequencies @xmath4 and @xmath5 . the problem we have studied has three important time scales : the shortest time scale is @xmath236 , which is the time required for _ one _ precession of the spin around the axis of the local magnetic field . the intermediate time scale is given by @xmath237 , which are the times required to complete one cycle of the center of mass around the center of the trap in the lateral and axial directions , respectively . these two time scales appear both in the classical and the quantum - mechanical analysis . the longest time scale ( provided that @xmath45 and @xmath44 are small ) @xmath238 , which is not present in the classical problem , is the time it takes for the particle to escape from the trap . whereas the classical analysis yields an upper bound for @xmath44 and @xmath45 for trapping to occur , no such sharp bound exists in the quantum - mechanical analysis . nevertheless it is interesting to compare the classical bound with the values of @xmath44 and @xmath45 for which the exponent in the expression for the quantum - mechanical lifetime becomes equal to @xmath29 : according to fig.([fig1 ] ) , we find that @xmath239 when @xmath240 , and @xmath241 when @xmath242 . from eq.([final ] ) on the other hand , we conclude that @xmath243 when @xmath244 , and @xmath245 when @xmath242 . thus , the quantum - mechanical condition for trapping to occur is roughly the same as the classical condition . these results however , should be taken with caution since our quantum - mechanical analysis is valid only for small values of @xmath45 and @xmath44 . though our derivation was for the case of a spin - one particle , it is clear that it can be extended to particles with higher spin , and also to _ half - integer _ spin particles . in view of the results obtained by our recent study of spin half particles in 1d field @xcite and 2d field @xcite , we believe that the expression for the lifetime in these cases is similar to the result which is obtained in the present paper . as an example , we apply our results to the case of a spin @xmath29 atom that is trapped in a field with @xmath246 oe and @xmath247 cm . these parameters correspond to typical traps used in bose - einstein condensation experiments @xcite . the results , being correct to within an order of magnitude , are outlined in table [ tab1 ] . we note that in both cases the values of @xmath44 and @xmath45 are much smaller than @xmath29 . also , the calculated lifetime of the particle in the trap is extremely large , suggesting that the particle is tightly trapped in this field . in this study we have been interested in the _ ground - state _ trapped state . in the case of a particle with spin @xmath28 or spin @xmath29 , this is the only _ one _ trapped spin state . however , when particles with higher spin are considered there are more than one trapped states . a natural question in connection with these is what is the lifetime of these trapped states . another interesting issue is the lifetime of an _ excited _ state in a given trapped spin state . our preliminary results show that some of these excited states may have a _ short _ lifetime , being _ algebraically _ dependent on @xmath248 and @xmath249 rather than _ exponentially _ dependent . this question is still under study . 99 a. l. migdall , j. v. prodan , w. d. phillips , t. h. bergeman and h. j. metcalf , `` first observation of magnetically trapped neutral atoms '' , _ phys . _ , * 54 * ( 24 ) , 2596 - 2599 ( 1985 ) . t. bergeman , g. erez and h. j. metcalf , `` magnetostatic trapping fields for neutral atoms '' , _ phys . a. , _ * 35 * ( 4 ) , 1535 - 1546 ( 1987 ) . v. s. bagnato , g. p. lafyatis , a. g. martin , e. l. raab , r. n. ahmad - bitar and d. e. pritchard , `` continous stopping and trapping of neutral atoms '' , _ phys . _ , * 58 * ( 21 ) , 2194 - 2197 ( 1987 ) . w. petrich , m. h. anderson , j. r. ensher and e. a. cornell , `` stable , tightly confining magnetic trap for evaporative cooling of neutral atoms '' , _ phys . _ , * 74 * ( 17 ) , 3352 - 3355 ( 1995 ) . m. o. mewes , m. r. andrews , n. j. van - druten , d. m. kurn , d. s. durfee , w. ketterle , `` bose - einstein condensation in a tightly confining dc magnetic trap '' , _ phys . rev . lett . _ , * 77*(3 ) , 416 - 419 ( 1996 ) . the levitron is available from ` fascinations ' , 18964 des moines way south , seattle , wa 98148 . the u - cas is available from masudaya international inc . , 6 - 4 , kuramae , 2-chome , taito - ku , tokyo , 111 japan . r. harrigan , u.s . patent number : 4,382,245 , date of patent : may 3 , 1983 . hones et al . patent number : 5,404,062 , date of patent : apr . 4 , 1995 . r. edge , `` levitation using only permanent magnets '' , _ phys . teach . _ * 33 * , 252 - 253 ( 1995 ) and `` corrections to the levitation paper '' , _ ibid_. * 34 * , 329 ( 1996 ) . m. v. berry , _ proc . lond . _ a * 452 , * 1207 - 1220 ( 1996 ) . s. gov and s. shtrikman , _ _ proc . of the 19__@xmath250 _ ieee conv . in israel _ , 184 - 187 ( 1996 ) . m. d. simon , l. o. heflinger and s. l. ridgway , _ am . j. phys . _ * 65 * ( 4 ) , 286 - 292 ( 1997 ) . s. gov , s. shtrikman and h. thomas , `` on the dynamical stability of the hovering magnetic top '' , _ physica d _ * 126 * , 214 - 224 ( 1999 ) . s. gov , s. shtrikman and h. thomas , `` on the spinning motion of the hovering magnetic top '' , _ physica d _ * 126 * , 225 - 235 ( 1999 ) . s. earnshaw , _ trans . cambridge philos . soc . _ * 7 * , 97 - 112 ( 1842 ) . the quantized motion of atoms in a quadrupole magnetic trap has been studied _ numerically _ by t. h. bergeman , p. mcnicholl , j. kycia , h. metcalf and n. l. balazs , `` quantized motion of atoms in a quadrupole magnetostatic trap '' , _ j. opt . soc . am . b _ , * 6 * ( 11 ) , 2249 ( 1989 ) . here , we use an _ analytic _ method to find the lifetime of the particle in such a trap . s. gov , s. shtrikman and h. thomas , _ los - alamos e - print archive _ , http://xxx.lanl.gov/physics/9808007 , _ am . j. phys . _ in press . s. gov , s. shtrikman and h. thomas , submitted to _ am . _ d. e. pritchard , _ phys . lett . _ * 51 * , 1336 ( 1983 ) . j. d. weinstein and k. g. libbrecht , _ phys . a. , _ * 52 * ( 5 ) , 4004 - 4008 ( 1995 ) . l. d. landau and e. m. lifshitz , quantum mechanics ( pergamon press ) , 3@xmath251 ed . 213 - 214 . ` quantum mechanics ' by e. merzbacher , _ john wiley & sons . _ , 2@xmath252 ed . , ch . 5 , sec . 3 , 57 - 61 this integral may be found in ` table of integrals , series , and products ' by i. s. gradshteyn and i. m. ryzhik , _ academic press _ , 5@xmath253 ed . , 6.521.1 , pp . 697 . note that this integral make explicit use of the fact that @xmath254 . ` quantum mechanics ' by e. merzbacher , _ john wiley & sons . _ , 2@xmath252 ed . , 18 , sec . 8 , 475 - 481 . ` handbook of mathematical functions ' by m. abramowitz and i. a. stegun , _ dover publications _ , 9@xmath253 ed . , 7.1.23 , pp . m. h. anderson , j. r. ensher , m. r. matthews , c. e. wieman and e. a. cornell , `` observation of bose - einstein condensation in a dilute atomic vapor '' , _ science _ * 269 * , 198 ( 1995 ) . k. b. davis , m. o. mewes , m. r. andrews , n. j. van druten , d. s. durfee , d. m. kurm and w. ketterle , `` bose - einstein condensation in a gas of sodium atoms '' , _ phys . rev . lett . _ * 75 * , 3969 ( 1995 ) . c. c. bradley , c. a. sackett , j. j. tollett and r. g. hulet , `` evidence of bose - einstein condensation in an atomic gas with attractive interactions '' , _ phys . _ * 75 * , 1687 ( 1995 ) ; _ ibid _ * 79 * , 1170 ( 1997 ) . e. a. cornell and c. e. wiemann , `` the bose - einstein condensate '' , _ sci . * 278 * , 26 - 31 ( 1998 ) . the transformation of @xmath255 is given by @xmath256 q\exp\left [ -i\theta\hat{s}_{y}\right ] , \label{a1}%\ ] ] where@xmath257 \nabla^{2}\exp\left [ -i\varphi\hat{s}_{z}\right ] . \label{a2}%\ ] ] evaluating @xmath258 first gives@xmath259 a\right ) & = \mathbf{\nabla}\cdot\mathbf{\nabla}\left ( \exp\left [ -i\varphi\hat{s}% _ { z}\right ] a\right ) \\ & = \mathbf{\nabla}\cdot\left [ \mathbf{\nabla}\left ( \exp\left [ -i\varphi\hat{s}_{z}\right ] \right ) a+\exp\left [ -i\varphi\hat{s}% _ { z}\right ] \mathbf{\nabla}a\right ] \\ & = \mathbf{\nabla}\cdot\left [ -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}\varphi\hat{s}_{z}a+\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}a\right ] \\ & = \mathbf{\nabla}\cdot\left [ -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}\varphi\hat{s}_{z}a+\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}a\right]\end{aligned}\ ] ] but@xmath260 \mathbf{\nabla}\varphi\hat{s}_{z}a\right ] \\ & = \mathbf{\nabla}\varphi\cdot\mathbf{\nabla}\left [ -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \hat{s}_{z}a\right ] -i\exp\left [ -i\varphi \hat{s}_{z}\right ] \mathbf{\nabla}^{2}\varphi\hat{s}_{z}a\\ & = \mathbf{\nabla}\varphi\cdot\left [ -i\exp\left [ -i\varphi\hat{s}% _ { z}\right ] \hat{s}_{z}\mathbf{\nabla}a-\exp\left [ -i\varphi\hat{s}% _ { z}\right ] \mathbf{\nabla}\varphi\hat{s}_{z}^{2}a\right ] -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}\varphi\hat{s}_{z}a\end{aligned}\ ] ] and@xmath261 \mathbf{\nabla}a\right ] \\ & = -i\mathbf{\nabla}\varphi\exp\left [ -i\varphi\hat{s}_{z}\right ] \hat { s}_{z}\cdot\mathbf{\nabla}a+\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}a\end{aligned}\ ] ] hence@xmath262 a\right ) \\ & = \mathbf{\nabla}\varphi\cdot\left [ -i\exp\left [ -i\varphi\hat{s}% _ { z}\right ] \hat{s}_{z}\mathbf{\nabla}a-\exp\left [ -i\varphi\hat{s}% _ { z}\right ] \mathbf{\nabla}\varphi\hat{s}_{z}^{2}a\right ] -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}\varphi\hat{s}_{z}a\\ & -i\mathbf{\nabla}\varphi\exp\left [ -i\varphi\hat{s}_{z}\right ] \hat { s}_{z}\cdot\mathbf{\nabla}a+\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}a\\ & = -2i\mathbf{\nabla}\varphi\exp\left [ -i\varphi\hat{s}_{z}\right ] \hat { s}_{z}\cdot\mathbf{\nabla}a-\exp\left [ -i\varphi\hat{s}_{z}\right ] \left| \mathbf{\nabla}\varphi\right| ^{2}\hat{s}_{z}^{2}a\\ & -i\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}\varphi \hat{s}_{z}a+\exp\left [ -i\varphi\hat{s}_{z}\right ] \mathbf{\nabla}^{2}a\end{aligned}\ ] ] thus@xmath263 \nabla^{2}\left ( \exp\left [ -i\varphi\hat{s}_{z}\right ] a\right ) \\ & = -2i\hat{s}_{z}\mathbf{\nabla}\varphi\cdot\mathbf{\nabla}a-\left| \mathbf{\nabla}\varphi\right| ^{2}\hat{s}_{z}^{2}a - i\mathbf{\nabla}% ^{2}\varphi\hat{s}_{z}a+\mathbf{\nabla}^{2}a\end{aligned}\ ] ] or in an operatorial form@xmath264 \nabla^{2}\exp\left [ -i\varphi \hat{s}_{z}\right ] = -2i\hat{s}_{z}\mathbf{\nabla}\varphi\cdot\mathbf{\nabla } -\left| \mathbf{\nabla}\varphi\right| ^{2}\hat{s}_{z}^{2}-i\mathbf{\nabla } ^{2}\varphi\hat{s}_{z}+\mathbf{\nabla}^{2}. \label{a3}%\ ] ] substituting in eq.([a1 ] ) each of the four terms in eq.([a3 ] ) we find@xmath265 \nabla^{2}\exp\left [ -i\theta\hat { s}_{y}\right ] = -i\nabla^{2}\theta\hat{s}_{y}-\left| \mathbf{\nabla}% \theta\right| ^{2}\hat{s}_{y}^{2}-2i\hat{s}_{y}\mathbf{\nabla}\theta \cdot\nabla+\nabla^{2}\text { , } \label{a3.1}%\]]@xmath265 \left ( -i\mathbf{\nabla}^{2}% \varphi\right ) \hat{s}_{z}\exp\left [ -i\theta\hat{s}_{y}\right ] = -i\mathbf{\nabla}^{2}\varphi\left ( \cos\theta\hat{s}_{z}-\sin\theta\hat { s}_{x}\right ) \text { , } \label{a3.2}%\]]@xmath266 \hat{s}_{z}^{2}\exp\left [ -i\theta \hat{s}_{y}\right ] & = \left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}% _ { x}\right ) ^{2}\label{a3.3}\\ & = \cos^{2}\theta\hat{s}_{z}^{2}-\sin\theta\cos\theta\left ( \hat{s}_{z}% \hat{s}_{x}+\hat{s}_{x}\hat{s}_{z}\right ) + \sin^{2}\theta\hat{s}_{x}% ^{2}\text{,}\nonumber\end{aligned}\]]@xmath266 \hat{s}_{z}\mathbf{\nabla}\exp\left [ -i\theta\hat{s}_{y}\right ] & = \exp\left [ i\theta\hat{s}_{y}\right ] \hat{s}_{z}\exp\left [ -i\theta\hat{s}_{y}\right ] \left ( -i\mathbf{\nabla } \theta\hat{s}_{y}+\mathbf{\nabla}\right ) \label{a3.4}\\ & = \left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}_{x}\right ) \left ( -i\mathbf{\nabla}\theta\hat{s}_{y}+\mathbf{\nabla}\right ) \nonumber\\ & = -i\cos\theta\mathbf{\nabla}\theta\hat{s}_{z}\hat{s}_{y}+i\sin \theta\mathbf{\nabla}\theta\hat{s}_{x}\hat{s}_{y}+\cos\theta\hat{s}% _ { z}\mathbf{\nabla}-\sin\theta\hat{s}_{x}\mathbf{\nabla,}\nonumber\end{aligned}\ ] ] substituting eqs.([a3.1 ] ) to ( [ a3.4 ] ) into eq.([a1 ] ) gives@xmath267{c}% -i\nabla^{2}\varphi\left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}_{x}\right ) -\left| \nabla\varphi\right| ^{2}\left ( \cos\theta\hat{s}_{z}-\sin \theta\hat{s}_{x}\right ) ^{2}\\ -2i\left ( \cos\theta\hat{s}_{z}-\sin\theta\hat{s}_{x}\right ) \nabla \varphi\cdot\left ( -i\mathbf{\nabla}\theta\hat{s}_{y}+\mathbf{\nabla}\right ) \\ -i\nabla^{2}\theta\hat{s}_{y}-\left| \mathbf{\nabla}\theta\right| ^{2}% \hat{s}_{y}^{2}-2i\hat{s}_{y}\mathbf{\nabla}\theta\cdot\nabla+\nabla^{2}% \end{array } \right ] .\ ] ] note that the transformed @xmath255 is composed of terms containing @xmath268 with @xmath269 or @xmath270 . this is a consequence of the fact that the original operator @xmath255 is a second order differential operator . thus , a spin state @xmath271 for which @xmath272 is coupled , in first order , only to the states @xmath273 and @xmath274 . [ c]|l|l| & spin @xmath29 atom + @xmath1 gr & @xmath275 + @xmath2 emu & @xmath276 + @xmath44 , @xmath45 & @xmath277 + @xmath278 sec & @xmath279 + @xmath280 , @xmath281 sec & @xmath282 + @xmath238 sec & @xmath283 +

robust generation of gamma - ray bursts ( grbs ) implies the formation of outflows with very low baryon loads and highly relativistic velocities , but more baryon - rich , slower outflows are also likely to occur in most grb central engine scenarios , either as `` circum - jet winds '' or `` failed grbs '' . here we study the possibility of nucleosynthesis within such baryon - rich outflows by conducting detailed reaction network calculations in the framework of the basic fireball model . it is shown that high baryon load fireballs attaining mildly relativistic velocities can synthesize appreciable quantities of heavy neutron capture elements with masses up to the platinum peak and beyond . small but interesting amounts of light elements such as deuterium and boron can also be produced . depending on the neutron excess and baryon load , the combination of high entropy , rapid initial expansion and gradual expansion at later times can cause the reaction flow to reach the fission regime , and its path can be intermediate between those of the- and-processes ( ``-process '' ) . the nucleosynthetic signature of these outflows may be observable in the companion stars of black hole binary systems and in the most metal - poor stars , potentially offering an important probe of the inner conditions of the grb source . contribution to the solar abundances for some heavy elements may also be possible . the prospects for further developments in various directions are discussed .

astro-ph0301392

although many fundamental aspects of the nature of gamma - ray bursts ( grbs ) remain mysterious , an essential requirement for any type of central engine to successfully generate a grb is the formation of an ultrarelativistic outflow with bulk lorentz factor @xmath3 100 , so that the gamma - ray compactness of the emitting region can be sufficiently small . this in turn implies the realization of a very low baryon load within the outflow , typically of isotropic equivalent mass @xmath4 . regardless of its detailed properties , the conditions at the base of such an outflow should generically be radiation - dominated with very high entropy , high temperature ( @xmath5 mev ) and very high optical depth ( see reviews by piran 1999 , mszros 2002 ) . the baryonic matter in the outflow is also expected to be enriched in neutrons , as the grb engine is likely to involve some kind of collapsed stellar mass object , possibly with intense thermal neutrino emission ( fuller , pruet & abazajian 2000 , pruet , woosley & hoffman 2003 , beloborodov 2003b ) . matter expanding from mev - temperature , neutron - rich conditions is known to be conducive to fusion of nuclei from initially free protons and neutrons . it is thus of interest to study the effects and implications of nucleosynthesis inside grb outflows , which was the subject of several recent papers ( lemoine 2002 , pruet , guiles & fuller 2002 , beloborodov 2003b ) . production of nuclei turns out to be quite limited in the very high entropy ( @xmath6 per baryon ) and extremely rapid expansion ( @xmath7 msec ) characteristic of grb outflows , resulting only in the synthesis of some d and @xmath8he . the low total mass ejected per event and the low intrinsic event rate expected for grbs imply that these products themselves are probably difficult to observe . besides the baryon - poor , ultrarelativistic outflow triggering the observable grb emission , most grb progenitors should also give rise to associated outflows with much higher baryon - loading and lower velocities under different circumstances . mounting evidence points to the high @xmath9 grb outflow being narrowly collimated into jets ( rhoads 2001 , frail et al . 2001 , panaitescu & kumar 2002 ) . it is then highly probable that an outer , sheath - like wind of slower , baryon - rich material surrounds and coexists with the inner , fast jet . generically speaking , whatever physical mechanism that initiates the grb jet ( pair production via neutrino annihilation , magnetohydrodynamic ( mhd ) stresses , blandford - znajek process , etc . ) must act on regions where the baryon content is tolerably low , presumably nearest to the jet axis ( e.g. mszros & rees 1992 , 1997 , levinson & eichler 1993 ) . however , this mechanism is also likely to affect more baryon - contaminated areas peripheral to the jet , inducing a baryon - rich , circum - jet flow as a natural byproduct . for models requiring jet penetration through envelopes of massive stars ( woosley 1993 , paczyski 1998 , nakamura 1998 , wheeler et al . 2000 , wheeler , meier & wilson 2002 ) , similar flows can additionally occur after jet formation through entrainment and mixing with the ambient stellar material ( aloy et al . 2000 , 2002 , macfadyen , woosley & heger 2001 , zhang , woosley & macfadyen 2003 ) . alternatively , if an accretion disk around a black hole is formed , as envisioned in many current scenarios ( woosley 1993 , popham , woosley & fryer 1999 , fryer , woosley & hartmann 1999 , narayan , piran & kumar 2001 ) , the outer parts of the disk may drive a baryon - rich wind by mechanisms different from the grb jet ( e.g. mszros & rees 1997 , narayan et al . 2001 , janka & ruffert 2002 ) . this is anticipated in models invoking rotating core collapse of massive stars ( e.g. woosley 1993 , macfadyen & woosley 1999 , vietri & stella 1998 ) , as well as compact binary mergers ( e.g. ruffert et al . 1997 , janka et al . 1999 , rosswog & ramirez - ruiz 2002 ) . such winds may possibly convey total energies comparable to or even larger than the grb jet itself ( levinson & eichler 2000 , macfadyen 2003 ) . the presence of such circum - jet winds or flows may be consistent with interpretations of the statistics of afterglow light curve breaks ( rossi , lazzati & rees 2002 , zhang & mszros 2002 ) . the recently recognized class of x - ray flashes ( heise et al . 2001 , kippen et al . 2002 ) may also be triggered by such kinds of flows ( zhang et al . 2003 ) . baryon - rich outflows may also occur without concomitant grbs in various progenitor scenarios , when conditions are such that the baryon - loading process acts more thoroughly , or the mechanism for grb jet production and/or propagation operates less efficiently ( e.g. woosley 1993 , narayan et al . 2001 , janka & ruffert 2002 , wheeler et al . 2000 , 2002 , woosley , zhang & heger 2002b , macfadyen 2003 ) . such `` failed grbs '' may have event rates higher than successful grbs ( dermer 2000 , woosley et al . 2002b , huang , dai & lu 2002 ) . depending on the identity of the actual grb engine , perhaps some identification of failed grbs can be made with observed `` hypernovae '' ( i.e. explosions similar to ordinary supernovae but with higher energies ; iwamoto et al . 1998 , nomoto et al . 2002 , wheeler et al . 2002 , woosley et al . 2002b ) , x - ray flashes ( e.g. dermer , chiang & mitman 2000 ) , or with some hitherto undiscovered type of transient ( e.g. li & paczyski 1998 , macfadyen et al . they may also comprise transitory phenomena preceding or ensuing a phase of genuine grb emission , possibly appearing as x - ray precursors ( murakami et al . 1991 ) or x - ray tails ( yoshida et al . 1989 ) observed in some grbs . either as circum - jet winds or failed grbs , these baryon - rich outflows ( hereafter bros ) should be typified by lower entropy , slower expansion and higher ejecta mass compared to grb jets , which are all more propitious from a nucleosynthesis viewpoint . besides the production of light elements , a particularly interesting issue is whether appreciable heavy element synthesis by neutron capture reactions can take place in such environments . elements heavier than fe in the solar system are believed to arise from two distinct types of neutron capture processes , the @xmath0-process , for which the timescale of neutron capture is much faster than @xmath10-decay , and the @xmath1-process , for which the relation is vice - versa ( see reviews by cowan , thielemann & truran 1991 , meyer 1994 , qian 2003 ) . the @xmath1-process proceeds with a nuclear reaction flow close to the line of @xmath10-stability , and is favored in situations with relatively low neutron abundances . its main astrophysical site is thought to be asymptotic giant branch stars , a very different environment compared to grb - bros . in contrast , exposure to high neutron fluxes is necessary for the @xmath0-process , in which the reaction flow occurs far into the neutron - rich region close to the neutron dripline . currently , the most widely discussed @xmath0-process site is the neutrino - driven winds of core - collapse supernovae ( e.g. woosley et al . 1994 , takahashi , witti & janka 1994 , terasawa et al . 2002 , terasawa 2002 , woosley , heger & weaver 2002a ) . efficient operation of the @xmath0-process in such winds requires high entropy ( @xmath11 per baryon ) , rapid expansion ( @xmath12 msec ) and high neutron excess ( @xmath13 0.43 in terms of the electron fraction ) . grb - bros are intriguing in this regard , as their physical properties should be similar in a number of ways . the typical conditions in relativistic bros can be even more extreme , but depending on the baryon load and other parameters , a successful @xmath0-process may also be possible . while this has been the subject of speculation by various authors in different contexts ( levinson & eichler 1993 , ruffert et al . 1997 , macfadyen et al . 2001 , lemoine 2002 , pruet et al . 2003 ) , quantitative nucleosynthesis calculations in the relevant physical conditions have not been performed . ( see 5 for a discussion relating to previous studies of the @xmath0-process in jet - driven supernova explosions . ) pruet et al . ( 2003 ; see also beloborodov 2003b ) recently evaluated the degree of neutronization expected in grb central engines based on black hole accretion disk models , and discussed the implications for nucleosynthesis in winds emanating from such disks , but no reaction network calculations were carried out . this work is an exploratory investigation of nucleosynthesis in bros associated with grbs and their observational consequences . we utilize the simplified dynamical framework of the basic fireball model , accounting for the crucial coasting phase ( 2.1 ) . a key feature is the incorporation of large and detailed nuclear reaction networks including light , neutron - rich nuclei ( 2.2 ) . the treatment in this paper is limited to the relativistic regime , which may apply to bros in compact binary mergers or supranovae ( but probably not to those occurring inside massive stars ) . although this is a first step in a potentially broader problem , we find interesting results with regard to both light and heavy element production ( 3 ) . some important observational implications of these nucleosynthetic products are pointed out , particularly for understanding the nature of the ill - understood grb central engine ( 4 ) . we conclude with a discussion of various interesting possibilities for future studies ( 5 ) . we start with the simplest possible dynamical model : a spherical , adiabatic , thermally - driven flow that is freely expanding and in steady state ( paczyski 1986 , 1990 ) . the key parameters in this wind fireball model are the luminosity @xmath14 , the initial radius @xmath15 , and the dimensionless entropy @xmath16 , where @xmath17 is the baryon - loading rate . the initial bulk velocity is taken to be subrelativistic , @xmath18 . assuming that the available energy is entirely dissipated into thermal form at the base of the outflow , the fireball plasma is radiation - dominated if @xmath19 , and its initial temperature is @xmath20 where @xmath21 is the radiation constant and @xmath22 the degrees of freedom including photons and electron - positron pairs . its optical depth is very high due to the presence of the pairs . the entropy per baryon @xmath1 is conserved in the flow and related to @xmath23 and @xmath24 as @xmath25 , where @xmath26 is the nucleon mass ( fuller et al . the fireball starts to expand adiabatically and first undergoes an acceleration phase , where the bulk lorentz factor @xmath9 grows with radius as @xmath27 while the comoving temperature @xmath28 falls as @xmath29 . this continues until the internal energy is fully converted to the bulk kinetic energy of the loaded baryons at the saturation radius @xmath30 . the fireball then enters a coasting phase with constant velocity @xmath31 , where @xmath32 . these simple scalings derive from conservation of energy and rest mass along with adiabaticity and are strictly valid for @xmath33 . lorentz transformations to the comoving frame give the profiles ( trajectories ) for the temperature @xmath28 and baryon density @xmath34 with respect to time @xmath35 : for the acceleration phase ( @xmath36 ) , @xmath37 for the coasting phase ( @xmath38 ) , @xmath39 here @xmath40 is the initial light crossing time or dynamical time , @xmath41 is the comoving saturation time , and @xmath42 is the initial baryon density . note that @xmath43 and @xmath44 asymptotically for @xmath45 . the trajectories above are in fact slightly modified by the effect of pair annihilation at @xmath46 mev , which is included in our calculations . we see that the acceleration phase is characterized by an exponential expansion with e - folding time equal to the dynamical time at the outflow base ( fuller et al . 2000 ) , whereas the coasting phase corresponds to a power - law expansion . a high @xmath23 ( low baryon load ) fireball suffers from a long period of exponential drop in @xmath28 and @xmath47 in addition to a very high entropy , both severely hindering nucleosynthesis . the ensuing coasting phase is then irrelevant and can be neglected for nucleosynthesis in the high @xmath23 outflows of successful grbs ( lemoine 2002 , pruet et al . 2002 , beloborodov 2003b ) . in contrast , a low @xmath23 ( high baryon load ) would allow the fireball to reach the power - law regime in a short time , where the more gradual decrease in @xmath28 and @xmath47 along with a lower entropy can greatly facilitate nucleosynthesis . we note that the forms of these trajectories are not specific to grb - related flows , but are generic descriptions for adiabatic , radiation - dominated expansion reaching a terminal velocity . eqs.[eqn : trajacc ] and [ eqn : trajcoa ] can also approximate the trajectories for neutrino - driven winds in core collapse supernovae ( e.g. qian & woosley 1986 , thompson , burrows & meyer 2001 ) . the main difference here is in the range of values explored for the physical parameters , such as the entropy , density and expansion timescale ( see also meyer 2002 ) . the description above is also applicable for an outflow confined to a conical channel so long as the opening half - angle @xmath48 , in which case @xmath49 and @xmath50 should be interpreted as isotropic equivalent quantities . flows which are strongly collimated , e.g. by external pressure gradients , would possess somewhat different scalings of physical variables with radius ( lemoine 2002 , beloborodov 2003b ) . note that when compared to the narrow grb jets , the bros may inherently have a much wider geometry ( levinson & eichler 2000 ) , so that the present formulation may be useful even for modest lorentz factor flows . as is apparent from the above arguments and also verified by actual calculations below , the lower the value of @xmath23 , the more advantageous it is for nucleosynthesis . this is in terms of both the reaction flow to heavier nuclei and the total product mass . however , the above dynamical description must be modified if @xmath51 , i.e. the baryon load is sufficiently large that the fireball becomes nonrelativistic . here we will limit ourselves to the relativistic scalings while imposing a lower limit of @xmath52 ( i.e. a mildly relativistic fireball ) , which corresponds to total baryon mass @xmath53 for total outflow energy @xmath54 , in isotropic equivalent terms . this is on two accounts . first , the physical regime of completely non - relativistic flows would be rather distinct from the high @xmath23 grb fireballs , and their actual properties , if they occur , are more uncertain . second , for relatively baryon - free progenitor models such as neutron star mergers ( ruffert et al . 1997 , janka & ruffert 2002 ) or possibly supranovae ( vietri & stella 1998 ) , @xmath55 may be a reasonable upper bound to the potential amount of baryon contamination ( e.g. woosley 1993 ) . as the assumption of free expansion should also suit these scenarios , the simple formulation employed here may be an adequate approximation for bros in such models . ( however , bros of the `` disk wind '' type may have @xmath24 , @xmath1 and @xmath56 different from the simple fireball prescription ; 5 ) . cases of much higher baryon loading can happen in other models like collapsars , where the outflows form and propagate deep inside massive stars , and @xmath57 or more may be realized ( macfadyen & woosley 1999 , macfadyen 2003 , wheeler et al . 2000 , 2002 ) , but this will be considered elsewhere . other effects that can be important for realistic grb - bro dynamics are touched upon in 5 . a critical quantity influencing the nuclear composition of baryonic matter is the neutron excess , commonly described in terms of the electron fraction @xmath58 , i.e. the net number of electrons per baryon . the value of @xmath59 at the base of the outflow should be predictable if reliable physical models of the central engine can be constructed , but this is currently a difficult task . the material in the collapsing cores of massive stars typically have @xmath60 , whereas material near the surfaces of cold neutron stars have @xmath61 . however , these can be significantly altered by capture of electrons ( positrons ) as well as electron - type neutrinos ( antineutrinos ) on protons ( neutrons ) during the course of the central engine evolution . pruet et al . ( 2003 ) and beloborodov ( 2003b ) have recently evaluated the @xmath59 values that may be realized in black hole accretion disk models appropriate for grb central engines ( popham et al . pruet et al . give numbers ranging from @xmath62 to @xmath63 at the innermost edge of the disk , depending on parameters such as the viscosity , accretion rate , mass and angular momentum of the central black hole . models invoking fireball formation in the envelopes of supermassive stars can instead be highly proton - rich , @xmath64 ( fuller & shi 1998 ) . in view of these uncertainties , we will treat @xmath59 as an additional free parameter in this work . except when @xmath65 mev , electron / positron captures within the fireball will not be fast enough to change @xmath59 from its initial value ( fuller et al . 2000 , beloborodov 2003b ) . for @xmath66 and @xmath67 typical of grb outflows , the baryons should initially be in nuclear statistical equilibrium ( nse ) consisting predominantly of free protons and neutrons , so the assembly of nuclei always begins from the lightest elements . a key feature of the nuclear reaction network codes implemented here is the inclusion of a large number of reactions involving light , very neutron - rich nuclei ( terasawa et al . 2001 , orito & iwamoto 2003 , in preparation ) . besides the well - known paths of @xmath68(@xmath69,@xmath70)@xmath71c and @xmath68(@xmath68n,@xmath70)@xmath72be(@xmath68,n)@xmath71c bridging the mass gaps at @xmath73 and @xmath74 , new reaction channels allowed via these light neutron - rich nuclides have been shown by terasawa et al . ( 2001 ) to critically influence the synthesis of even the heaviest nuclei , especially for expansion with short timescales and/or high neutron excesses . this is indeed the case for our conditions here . most previous network codes including only a limited number of such reactions would give erroneous results for the present problem . in total , our code used for heavy element synthesis calculations includes over 3000 nuclides , mainly on the neutron - rich side of stability . on the other hand , we only deal with a limited number of heavy nuclei on the proton - rich side , which , if fully included , may possibly affect the reaction flow when @xmath59 is near 0.5 . we also choose to neglect fission , the detailed physics of which is currently very uncertain ( e.g. goriely & arnould 2001 , cameron 2001 ) , but which may in reality have important consequences for some parameter regimes studied here . these microphysical aspects will be accounted for in subsequent studies ( 5 ) . we do include the effects of the increase in temperature and entropy from electron - positron pair annihilation occuring at @xmath46 mev ( e.g. meyer & brown 1997 ) . in many grb models , strong neutrino irradiation from the central source can also affect the nuclear composition ( e.g. beloborodov 2003b ) , but this will not be treated here as it is highly model - dependent ( see 5 ) . we remark that for all cases under consideration here the neutrons remain kinematically well - coupled to the rest of the fireball plasma throughout the duration of nucleosynthesis ( c.f . derishev , kocharovsky & kocharovsky 1999 , fuller et al . the main parameters in the present problem are @xmath49 , @xmath75 , @xmath23 and @xmath59 . grb observations put reasonable constraints on @xmath49 and @xmath75 , so we will keep these to our fiducial numbers of 10@xmath76 erg s@xmath77 and 10@xmath78 cm respectively ( hence @xmath79 ; see however 5 ) . we thus concentrate on the effect of variations in @xmath23 and @xmath59 , for which a wide range of values can be realized . the trajectories are very different for the limiting cases of high and low @xmath23 . for @xmath80 typical of successful grbs , @xmath81 , and the acceleration phase lasts for @xmath82 msec during which @xmath28 and @xmath34 fall extremely rapidly by 2 and 6 orders of magnitude , respectively . in comparison , for @xmath83 , @xmath84 , and the coasting phase is entered at @xmath85 msec when @xmath28 and @xmath34 are lower by just factors of 2 and 8 , respectively . much of the reactions can then proceed during the power - law expansion , especially neutron - captures . since the production of the more common @xmath68 and fe - group elements at intermediate @xmath86 should be dominated by normal stars and supernovae in the universe , the signatures of grb - bro nucleosynthesis are to be sought in other elements which are not readily synthesized in ordinary stellar environments ( see however macfadyen 2003 ) . we therefore focus our interest here on the light element region of d , li , be and b , and on the neutron capture element region much heavier than fe . for calculating light element production , a relatively small network limited to nuclei with mass number @xmath87 suffices , as the reaction flow to higher @xmath86 is not very large . here we utilize a code developed for big bang nucleosynthesis ( bbn ) , which includes the relevant weak interaction processes and has been extended and updated with reactions for 40 light nuclei in total to study inhomogeneous bbn ( orito & iwamoto 2003 , in preparation ) . the relative ease of these computations allows us to survey large regions of parameter space . we first discuss the yields of 4 and d for a wide range of @xmath23 . these elements can be produced appreciably even in the high @xmath23 fireballs of successful grbs ( lemoine 2002 , pruet et al . 2002 , beloborodov 2003b ) , and we extend the regime of consideration down to @xmath83 . like bbn , 4 is the dominant final species besides remaining free nucleons for all cases here , as the majority of the available protons ( neutrons ) can become locked up in @xmath68 particles for neutron - rich ( proton - rich ) conditions . the left panel of fig . [ fig : light ] shows the final abundance of d as contours of constant mass fraction in the @xmath23-@xmath59 parameter plane . for @xmath88 , the reactions mainly occur during the exponentially expanding acceleration phase . d production depends on @xmath89 such that its final abundance is smaller for lower @xmath23 due to larger reaction flows to @xmath90 , similar to the case of bbn . the production rate is also proportional to @xmath91 , with a maximum at @xmath92 . contrastingly , for @xmath93 , the nucleosynthesis proceeds mainly during the coasting phase with power - law expansion , where neutron capture reactions can be active until very late times , @xmath94 s ( see 3.2 ) . here d synthesis can occur via a novel process : when @xmath60 , free neutrons remain abundant in the outflow until they start decaying into protons , which subsequently can undergo the p(n,@xmath70)d reaction during the latest stages of the expansion . in this regime , d production does not depend strongly on @xmath59 , as long as @xmath60 . this is an interesting d production mechanism that has not been discussed in previous contexts . notable amounts of nuclei with @xmath95 4 such as li , be and b are created only when @xmath93 . we display the results for 11 in the right panel of fig . [ fig : light ] for the range of @xmath96 . under neutron - rich conditions ( @xmath97 ) , 11 is produced by the reaction chains @xmath98 in proton - rich environments ( @xmath99 ) , the main path for @xmath100 production is @xmath101 . other light elements such as 6 , 7 and 9 can also be generated at levels of @xmath102 . we remark that under certain conditions , light element production may be considerably enhanced through nonthermal reactions such as spallation and/or photodisintegration , but these effects are outside of our present scope ( see 5 ) . the synthesis of heavy elements through neutron capture is the focal point of this work . such processes obviously encompass a vastly larger number of nuclides compared to light element production , so full network calculations are much more cumbersome . parameter studies with the small , light element network discussed above indicate that the flow to heavy nuclei ( @xmath9516 ) becomes considerable only for @xmath103 . we thus concentrate here only on the most favorable case of @xmath83 , and choose selected values for @xmath59 ( @xmath104 ) . using the full network described in terasawa et al . ( 2001 ) , the calculations below have been followed until all reactions have frozen out , typically occurring at @xmath105 s. the @xmath83 outflow is characterized by very high entropy ( @xmath106 ) , low density ( @xmath107 ) , and very short initial expansion timescale ( @xmath108 ms ) , which can be compared with the values typical of sn neutrino - driven winds ( 1 ) . these conditions lead to heavy element synthesis which is markedly different from the usual @xmath0-process in a number of respects . displayed in figure [ fig : heavy ] are the final abundances by number as a function of mass number , for the cases of @xmath109 0.1 , 0.3 , 0.4 , 0.48 and 0.498 . the abundance patterns are clearly discrepant from the solar distribution , shown here with arbitrary normalization as the uppermost curve . when @xmath110 0.4 , implying copious supplies of neutrons , considerable reaction flows can occur to high mass numbers , ending up in large fractions of the heavier nuclei . however , we also see that the abundances are at levels of @xmath111 even at the peaks , which is quite low compared to the sn wind @xmath0-process . this is due to the combination of small @xmath56 , which limits the initial , charged particle reaction flow at small @xmath86 , and the low density , which hinders the subsequent neutron capture flow to higher @xmath86 . note that the apparent abundance peak at @xmath112 250 is an artifact of our neglect of fission and results from artificial pileup at the limit of our reaction network . in situations where this peak becomes dominant , the realistic effects of fission ( and possible consequent fission cycling ) may significantly affect the final abundance pattern ( 5 ) . for @xmath109 0.48 and 0.498 , the flows to high mass nuclei are drastically reduced due to the low neutron excess , and only small amounts of elements result at @xmath113 and @xmath114 , respectively . our abundance distributions for @xmath110 0.4 show three peaks around @xmath11287 , 135 and 205 located in between the solar @xmath1- and @xmath0-process peaks . this is a clear manifestation of nuclei with neutron magic numbers forming on a reaction flow path intermediate between those of the @xmath1- and @xmath0-processes . figure [ fig : path ] shows a snapshot of the calculated reaction flow in the @xmath115-@xmath116 plane for the case of @xmath109 0.3 , at @xmath117 4.2 s when the density has fallen to 1.2 @xmath118 g @xmath119 . unlike the @xmath0-process in sn winds , the neutron capture rates here are too low to effectively establish ( n , @xmath70)-(@xmath70 , n ) equilibrium due to the low density , even when the neutrons constitute a major fraction of the baryons . instead , as @xmath34 and @xmath28 decrease steadily in the coasting phase , @xmath10-decays begin to compete with the neutron capture reactions , leading to a flow path that gradually evolves in time from near the dripline ( @xmath0-process - like ) to that near the @xmath10-stability line ( @xmath1-process - like ) . we also note that the differences in the abundances between the peak and interpeak elements are larger than that of the solar pattern . this is due to the fact that nuclei in the moderately neutron - rich region have longer @xmath10-decay lifetimes than compared to those near the dripline . therefore , reaction flows at the magic nuclei are halted for a longer time , resulting in larger peak - to - interpeak contrasts . a final point is that the low density here keeps the neutrons from ever being exhausted by neutron capture , in contrast to the @xmath0-process in sne . instead the neutrons remain abundant after the capture reactions freeze out until they freely decay into protons . neutron capture nucleosynthesis in environments with low neutron abundances , where the flow path is determined by the balance of neutron captures and @xmath10-decays rather than ( n,@xmath70)-(@xmath70,n ) equilibrium , has been discussed previously by blake & schramm ( 1976 ) and blake et al . ( 1981 ) and christened the `` @xmath2-process '' . our physical conditions are quite different from those considered by the above authors , in that we have a very large neutron abundance but also a low total density . however , the basic physical processes are similar . bros associated with grbs can therefore be natural astrophysical sites for operation of the @xmath2-process . the calculations presented above are based on a number of simplifications , so they should be regarded as exemplary and tentative expectations from nucleosynthesis in actual grb - bros . in particular , they may be fair representations for at least some types of bros in neutron star mergers or supranovae , but probably not for bros triggered inside massive stars ( 2 ) . various realistic physical effects can modify these results in important ways ( 5 ) . nevertheless , some qualitative trends seen here should remain generally valid . below we discuss the observational implications of grb - bro nucleosynthesis , taking at face value the results of our @xmath83 , mildly relativistic fireball as a fiducial case . these considerations should also serve as a reference point for more detailed investigations in the future . since it was found that grb - bros can be interesting sites of neutron capture nucleosynthesis , the first important comparison we can make is between the heavy element yields from a single grb - bro event and that expected from a normal core collapse sn . for neutrino - driven wind models of sne , the typical abundance by number of @xmath0-process peak elements in the ejecta is @xmath120 at @xmath121 , and the total ejecta mass should be @xmath122 ( e.g. meyer 1994 , qian 2003 ) . to be compared are the bro final abundances for the realistic range of @xmath123 , where the peak abundance is @xmath124 ( fig.[fig : heavy ] ) , and the total ejecta mass can be of order @xmath125-@xmath126 . , @xmath127 implies a total outflow energy @xmath128 . this would be consistent with @xmath129 if , for example , the grb duration is @xmath130 and the bro is mildly collimated into a fraction @xmath131 of the sky . although this bro energy is larger than the inferred intrinsic ( i.e. beaming - corrected ) total energies for grb jets , @xmath132 ( frail et al . 2001 ) , such values may not be so unreasonable from certain physical viewpoints ( e.g. levinson & eichler 2000 ) , and is also concordant with the observed energetics of `` hypernovae '' , which may possibly correspond to failed grbs . ] this gives a total heavy element yield per grb - bro event comparable to or somewhat less than sne , albeit with large uncertainties . as a first guess , the contribution of grb - bros to the solar / galactic abundances then is a matter of event rates compared to sne , if sne are in fact the main sources of the solar @xmath0-process elements . the intrinsic , beaming - corrected event rates for successful grbs are estimated to be @xmath133-@xmath134 ( e.g. frail et al . 2001 , woosley et al . 2002b ) , which is @xmath135 0.1 - 1 % of the sn rate @xmath136 ( although the rates can be less for neutron star mergers ; e.g. fryer et al . if a bro with the above parameters is associated with every grb as a circum - jet wind , then the nominal allowance of grb - bros to the solar abundances is @xmath135 0.1 - 1 % . if failed grbs with similar properties occur at a higher rate , the contribution will be even greater . however , we must remember that the abundance patterns arising from our fiducial outflow are different from solar , particularly the location of the peaks . in order to avoid overproduction of the elements between the solar @xmath0- and @xmath1-peaks , the true contribution may be even less . at the very least , such grb - bros may be the origin of select rare elements not easily produced under any other conditions ( blake et al . alternatively , some realistic effects discussed in 5 may alter the nucleosynthesis in grb - bros so as to make them potential sources of the solar @xmath0-process elements . even if grb - bros hold a minor share of the solar abundances , some observable environments may directly manifest the local abundance pattern of the bro ejecta . detailed spectroscopic observations of metal - poor stars with metallicities down to [ fe / h ] @xmath135 -3 have so far shown abundance patterns very similar to that of the solar @xmath0-process for elements with 56 @xmath137 70 , at least for the so - called @xmath0-process enhanced stars ( truran et al . 2002 , qian 2003 and references therein ) . this fact , together with the observed large scatter in the abundances of @xmath0-process elements as well as iron - group elements ( mcwilliam 1997 ) , indicate that chemical evolution in the early galaxy proceeded very inhomogeneously , and the abundance patterns in metal - poor stars are set mainly by a small number of ( or even individual ) sn explosions of the previous stellar generation ( audouze & silk 1995 , ryan , norris & beers 1996 , shigeyama & tsujimoto 1998 ) . if grbs are associated with the most massive range of progenitor stars , these explosions may have dominated in the earliest stages of the galaxy , so that their nucleosynthesis products can be discerned in metal - poor stars with the lowest metallicities . @xmath138 reflect explosive nucleosynthesis products of energetic `` hypernovae '' ( nakamura et al . 2001 , umeda & nomoto 2002 , nomoto et al . 2002 , maeda & nomoto 2003 ) . ] we can estimate the total mass of gas swept up by a grb - bro and with which the bro ejecta mixes to be @xmath139-@xmath140 ( ryan et al . 1996 , shigeyama & tsujimoto 1998 ) . the dilution factor per event is then @xmath141 , and the expected heavy element abundance is @xmath142 or roughly 2.5 dex below solar , an observable level . it is intriguing to see if future observations of extremely metal - poor stars ( e.g. christlieb et al . 2002 ) reveal a heavy element abundance pattern distinct from that of the solar @xmath0-process . another interesting locale is the companion star in a black hole binary system , where the explosive event that formed the black hole may have left traces of the explosion ejecta on the surface of the companion . ( this argument applies to massive star core collapse models like collapsars and supranovae , but excludes neutron star merger scenarios . ) indeed , observations of the companion stars of gro j1655 - 40 ( nova sco ; israelian et al . 1999 ) and other black hole binaries ( orosz et al . 2001 , israelian et al . , in preparation ) have uncovered anomalous @xmath68-element enhancements indicating contamination by supernova ejecta , possibly of the energetic `` hypernova '' variety ( israelian et al . 1999 , podsiadlowski et al . 2002 ) . very crudely , the fraction of the bro ejecta captured by the companion may be @xmath143 , considering a typical solid angle subtended by the star , provided that it lies within the ( presumably wide ) collimation cone of the bro . the mass of the mixing zone between the ejecta and the companion surface could be in the range @xmath144 ( israelian et al . 1999 , qian 2000 ) , so that the dilution factor here is @xmath145 , although these values are very uncertain . this leads to @xmath146 , which is a staggeringly high overabundance with respect to solar @xmath147 , and should be clearly detectable . high resolution spectroscopic observations of such stars are warranted to search for neutron - capture elements and study their detailed abundance patterns . in discerning light element products from bros , the only hope is offered by the black hole binary companions , as their contribution to the solar abundances and even very metal - poor stars can be shown to be miniscule . taking the @xmath83 case , we will be most optimistic and assume the least dilution @xmath148 , and also select @xmath59 where the maximum final abundance is achieved for each element . for d , @xmath78li , @xmath149li , @xmath72be and @xmath150b , respectively , these are @xmath151 0.1 , 0.7 , 0.5 , 0.4 and 0.1 , at which the abundance by number on the companion surface would be @xmath152 and @xmath153 . this must be compared to the solar abundances for the respective elements , @xmath154 and @xmath155 . overabundance is expected to be high only for d , and is marginal for @xmath78li and @xmath150b . more dilution or deviations from the optimal @xmath59 , as well as stellar depletion effects that can be severe for d , would render them difficult to observe , if at all . however , light element production may become more prolific for bros which are nonrelativistic and/or collimated , or when circumstances allow for significant spallation ( 5 ) . either in extremely metal - poor stars or in black hole binary companions , observations of the pure ejecta abundance pattern may provide us with valuable insight into the origin of grbs . it is evident in figs.[fig : light ] and [ fig : heavy ] that the final abundance patterns of both the light and heavy elements are quite sensitive to @xmath59 , which in turn implies that observations can significantly constrain its value . since @xmath59 is an important quantity characterizing the nature and conditions of the central engine , the nucleosynthetic products can shed light on the innermost regions of grb sources , which are otherwise very difficult to probe . the salient points of this work are summarized . we have explored the possibility of nucleosynthesis in baryon - rich outflows that are expected to accompany grbs . detailed reaction network calculations were performed in the context of the fireball model with varying baryon load . it was found that mildly relativistic outflows can synthesize appreciable quantities of heavy neutron capture elements with masses up to the actinides via the `` @xmath2-process '' , as well as interesting amounts of light elements such as deuterium . their characteristic abundance patterns may be discernible in black hole binary companion stars or in extremely metal - poor stars , possibly offering important information on the grb central engine conditions . some contribution to the solar heavy element abundances may also exist . our intention here was to carry out a pilot study , focusing on the most general and model - independent aspects of nucleosynthesis in bros related with grbs . in reality , various different types of grb - bros can arise ( 1 ) , and the present calculations may be fairly indicative for some kinds of bros , but not for others . for accretion disk winds that originate in regions and by mechanisms distinct from the grb jet , the fireball model may not be so suitable for describing their basic properties such as the temperature , entropy and expansion timescale ( 2 ) . likewise , the characteristics of bros occurring in failed grbs are dependent on the particular manner in which grb jet fails . detailed studies of nucleosynthesis in realistic grb - bro conditions require a good understanding of both the dynamics and microphysics of the true grb engine . since we are still far from this , one approach for undertaking more in - depth investigations is to work in the framework of specific models for the grb progenitor and outflow formation . alternatively , we can also make progress by considering in steps various physical aspects that can be generically important for a range of possible grb scenarios . below we give brief discussions of some effects that have not been accounted for in this work , but which may have significant consequences for the outcome of grb - bro nucleosynthesis . these considerations point us toward a variety of subsequent studies . in our reaction network calculations , fission was not included for the sake of simplicity . however , we saw that for a plausible range of @xmath59 the reaction flow can extend well into the actinide region where fission should be unavoidable . since the free neutrons in our bros always stay abundant until very late times ( 3.2 ) , a highly probable consequence is fission cycling , whereby fission products act as seed nuclei for successive cycles of reaction flows into the fission regime . this should affect the resulting heavy element abundances in interesting ways and is currently under investigation ( terasawa et al . , in preparation ) . another facet not included in our reaction network is heavy nuclei in the proton - rich region , which should be important for situations when @xmath63 and may potentially give rise to @xmath156-process nuclei ( iwamoto et al . , in preparation ; see also meyer 2002 ) . our dynamical framework here was limited to the basic spherical fireball model . although the bros may not be as strongly collimated as the grb jets , some degree of collimation ( e.g. through external gas pressure gradients or magnetic stresses ) is naturally expected , which modify in the density and temperature trajectories from eqs.[eqn : trajacc ] and [ eqn : trajcoa ] . while the details are model - dependent , collimation should generally tend to prolong the phase of initial expansion when the charged particle reactions mainly proceed ( lemoine 2002 , beloborodov 2003b ) . this would permit the initial reaction flow to reach larger @xmath86 , resulting in increased heavy element production for cases with high neutron excess , as well as larger light element yields . magnetic fields can also crucially modify the flow acceleration in addition to collimation , but assessing their effect on nucleosynthesis requires detailed dynamical modeling . for reasons discussed in 2 , we had also restricted the dimensionless entropy to @xmath157 , meaning that the outflows are at least mildly relativistic and the total baryon masses ( isotropic equivalent ) are @xmath158 . however , much larger baryon loading can conceivably happen in grb models that activate outflows inside massive stars ( macfadyen & woosley 1999 , macfadyen 2003 , wheeler et al . 2002 ) , leading to bros in the nonrelativistic regime . these should be even more efficient nucleosynthetic cauldrons , as they involve a lower entropy , longer initial expansion time , and larger total ejecta mass . in fact , the effects of nonrelativistic baryon loading ( and to a lesser extent , collimation ) may allow the realization of physical conditions that are ideal for reproducing the solar @xmath0-process abundances ( pruet et al . this interesting prospect of grb - bros as potential @xmath0-process sites is worth exploring in some detail . ( see below for a discussion of previous work on the @xmath0-process in magneto - rotational jet - powered supernovae . ) another significant but model - dependent process is neutrino irradiation of the bro from a central source . strong thermal neutrino emission is expected in most types of grb progenitors , including massive star core collapse and neutron star mergers . this can impact the nucleosynthesis in different ways , either through its thermal and dynamical effects on the bro expansion , or through neutrino - induced interactions on the nucleons and nuclei ( meyer 1995 , meyer , mclaughlin & fuller 1998 ) . relatively realistic calculations including the effects mentioned so far may be a feasible goal for nucleosynthesis in the baryonic winds of neutron star mergers , employing outflow trajectories of density and temperature calculated in numerical simulations ( janka & ruffert 2002 ) . the considerable fraction of free neutrons remaining after bro nucleosynthesis ( 3.2 ) may induce some noteworthy effects before they decay into protons at a radius @xmath159 . for nucleosynthesis , an intriguing possibility is the operation of an `` external '' neutron capture process , i.e. heavy element synthesis by capture of the bro neutrons on any external baryonic matter with which the outflow interacts and mixes , especially when the bro is nonrelativistic . the external material may be the outlying stellar wind of the progenitor star , the surface of a binary companion , or the overlying stellar envelope in the case of a collapsar ( pruet et al . such processes may be externally seeded , as any nuclei like fe preexisting in the environs can constitute the seed for neutron captures . further interesting consequences of neutrons can be caused by their kinematic decoupling from the rest of the fireball plasma , such as modifications to the dynamical evolution of the fireball and/or blastwave ( e.g. derishev et al . 1999 , fuller et al . 2000 , beloborodov 2003a ) , characteristic neutrino and gamma - ray signatures ( bahcall & mszros 2000 , mszros & rees 2000 ) , and production of light elements by spallation ( pruet et al . 2002 ; see below ) . although we only dealt with thermal nuclear reactions in the expanding bro , additional , nonthermal nuclear processes can alter the nuclear composition in different ways in real grb - bros . these may actually dominate the production of some elements , particularly d , li , be and b. for the high @xmath9 grb jets , the observed short timescale variability of the grb emission may entail substantial fluctuations and dissipation by internal shocks within the outflow , which can lead to spallation of @xmath8he and other existing nuclei ( beloborodov 2003b ) . bros studied here , since they may not be as strongly time - varying as grb jets , and also because internal shocks in the bros would occur at much smaller radii where the temperature is still high and the outflow optically thick . ] appreciable spallation may also be induced by interaction of the outflow with external matter ( e.g. progenitor stellar wind , binary companion star , massive star envelope ) , affecting both the nuclei inside the bro and those composing the external baryons ( cameron 2001 ) . under circumstances where free neutrons become kinematically decoupled , bulk motions relative to the rest of the flow may develop and induce spallation by neutrons , either internally or externally ( pruet et al . 2002 , beloborodov 2003b ) . photodisintegration by nonthermal gamma - rays emitted in internal shocks can also take place outside the photosphere ( lemoine 2002 ) , although this may not be so significant for bros due to their large photospheric radii . virtually all of the above effects ( plus some other unique ones ) may become essential for scenarios in which the grb jet and associated bro initially propagate through the dense envelope of a massive star before emerging and generating a grb ( woosley 1993 , woosley et al . 2002b , wheeler et al . there is a multitude of deviations from the simple fireball picture due to the strong outflow - star interaction , such as reverse shock heating of the flow , expansion of the shocked material into a cocoon , entrainment and mixing with the stellar material through instabilities , breakout from the envelope and ensuing lateral expansion , etc . ( macfadyen & woosley 1999 , macfadyen et al . 2001 , mszros & rees 2001 , ramirez - ruiz et al . in fact , the expansion timescale at envelope breakout ( radii @xmath160 ) may be more important for nucleosynthesis than the expansion timescale at the central engine ( see discussions in pruet et al . 2002 , 2003 ) . these effects may also be highly variable and inhomogeneous in nontrivial ways ( aloy et al . 2000 , 2002 , zhang et al . quantitative modeling of nucleosynthesis within this complicated picture is a very interesting but challenging goal . note that the `` disk winds '' in collapsars have been suggested to be an important site for synthesis of @xmath161ni , which is essential for powering the light curve of associated supernovae / hypernovae ( macfadyen & woosley 1999 , pruet et al . 2003 , macfadyen 2003 ) . the physical properties of the bros considered here may actually bear a number of similarities to the magneto - rotational `` jets '' in core collapse supernovae originally studied by leblanc & wilson ( 1970 ) . if the collapsing core has sufficiently fast rotation and strong magnetic fields , such jets ( or related mechanisms ) may potentially power the supernova explosion ( bisnovatyi - kogan 1971 , meier et al . 1976 , symbalisty 1984 , khokhlov et al . 1999 , wheeler et al . 2002 ) . leblanc - wilson jets have also been discussed as promising @xmath0-process sites ( schramm & barkat 1972 , symbalisty , schramm & wilson 1985 , cowan , thielemann & truran 1991 ) , although no detailed nucleosynthesis calculations in this picture have been conducted . they are expected to expel @xmath162 of highly neutron - rich material from within the proto - neutron star as narrowly collimated ejections ( symbalisty 1984 ) . in the context of our fireball model , this should correspond roughly to an isotropic equivalent baryon load of @xmath163 . a significant fraction of the jet energy is also in magnetic form . in comparison , our fiducial @xmath83 outflow with isotropic equivalent mass @xmath164 and without magnetic fields has much lower density , higher entropy and shorter expansion timescale , leading to an @xmath2-process rather than an @xmath0-process . in light of the growing observational evidence for asymmetric supernova ejecta , ultra - strong magnetic field neutron stars and potential connection with grbs ( wheeler et al . 2002 and references therein ) , it should be worthwhile to reconsider in greater depth the physics and implications of such magneto - rotational outflows in core collapse supernovae ( woosley et al . 2002a , b , wheeler et al . note that a related @xmath0-process scenario has been discussed recently by cameron ( 2001 ) , where an @xmath0-process involving fission cycling takes place inside an accretion disk , and the products are subsequently ejected in jets . the ultimate answer to the problem of grb - bro nucleosynthesis is contingent on elucidating the physics of the grb central engine , including the progenitor s identity , the central engine configuration and the outflow formation mechanism , from which we may be a long way . nevertheless , we may hope to build increasingly realistic models , adding step by step different levels of complexity . comparison of such calculations to observations of various sites may then allow us to constrain the unknown properties of the grb source , as well as to clarify the role of grb - bros in the chemical evolution of the galaxy . further investigations of grb - bro nucleosynthesis in various directions should provide us with a rich scope of interesting work for the future . we are very grateful to the referee , stan woosley , for helpful and stimulating discussions before and after submission of this paper . many thanks also go to thomas janka for numerous insightful conversations , as well as grant mathews , brad meyer , wolfgang hillebrandt , claudia travaglio and hideyuki umeda for helpful comments . takahiro sasaqui is acknowledged for his help in the early stages of this study . one of the authors ( m. t. ) wishes to acknowledge the fellowship of the japan society for promotion of science ( jsps ) . audouze , j , & silk , j. 1995 , , 451 , l49 aloy , m .- a . , ibez , j .- m . , miralles , j .- a . & urpin , v. 2002 , , 396 , 693 aloy , m .- a . , mller , e. , ibez , j .- , mart , j .- m . & macfadyen , a. 2000 , , 531 , l119 bahcall , j. n. & mszros , p. 2000 , , 85 , 1362 beloborodov , a. m. 2003a , , 585 , l19 beloborodov , a. m. 2003b , , in press ( astro - ph/0210522 ) bisnovatyi - kogan , g. s. 1971 , soviet astron .- aj , 14 , 652 blake , j. b. & schramm , d. n. 1976 , , 209 , 846 blake , j. b. , woosley , s. e. , weaver , t. a. & schramm , d. n. 1981 , , 248 , 315 cameron , a. g. w. 2001 , , 562 , 456 christlieb , n. et al . 2002 , , 419 , 904 cowan , j. j. , thielemann , f .- k . & truran , j. w. 1991 , , 208 , 267 derishev , e. v. , kocharovsky , v. v. & kocharovsky , v. vl . 1999 , , 521 , 640 dermer , c. d. 2000 , in gamma - ray bursts 5th huntsville symposium , eds . r. m. kippen , r. s. mallozzi , and g. j. fishman , aip conf . 526 ( new york : aip ) p.431 ( astro - ph/0001389 ) dermer , c. d. , chiang , j. & mitman , k. e. 2000 , , 537 , 785 di matteo , t. , perna , r. & narayan , r. 2002 , , 579 , 706 frail , d. et al . 2001 , , 562 , l55 fryer , c. l. , woosley , s. e. & hartmann , d. h. 1999 , , 526 , 152 fuller , g. m. , pruet , j. & abazajian , k. 2000 , , 85 , 2673 goriely , s. & arnould , m. 2001 , , 379 , 1113 heise , j. , int zand , j. , kippen , r. m. & woods , p. m. 2001 , in gamma - ray bursts in the afterglow era , eds . e. costa , f. frontera & j. hjorth ( berlin : springer ) , p.16 ( astro - ph/0111246 ) huang , y. f. , dai , z. g. & lu , t. 2002 , , 332 , 735 israelian , g. , rebolo , r. , basri , g. , casares , j. & martin , e. l. 1999 , , 401 , 142 iwamoto , k. et al . 1998 , , 395 , 672 janka , h .- th . , eberl , t. , ruffert , m. & fryer , c. l. 1999 , , 527 , l39 janka , h .- th . & ruffert , m. 2002 , in stellar collisions , mergers and their consequences , asp conf . m. m. shara , ( san francisco : astronomical society of the pacific ) , p.333 ( astro - ph/0101357 ) khokhlov , a. m. et al . 1999 , , 524 , l107 kippen , r. m. et al . 2002 , to appear in proc . of the woods hole grb workshop ( astro - ph/0203114 ) leblanc , j. m. & wilson , j. r. 1970 , , 161 , 541 lemoine , m. 2002 , , 390 , l31 levinson , a. & eichler , d. 1993 , , 418 , 386 levinson , a. & eichler , d. 2000 , , 85 , 236 li , l .- x . & paczyski , b. 1998 , , 507 , l59 macfadyen , a. i. 2003 , to appear in `` from twilight to highlight - the physics of supernovae '' eso / mpa / mpe workshop ( astro - ph/0301425 ) macfadyen , a. i. & woosley , s. e. 1999 , , 524 , 262 macfadyen , a. i. , woosley , s. e. & heger , a. 2001 , , 550 , 410 maeda , k. & nomoto , k. 2003 , , submitted , astro - ph/0304172 mcwilliam , a. 1997 , , 35 , 503 meier , d. l. , epstein , r. i. , arnett , w. d. & schramm , d. n. 1976 , , 204 , 869 mszros , p. 2002 , , 40 , 137 mszros , p. & rees , m. j. 1992 , , 257 , 29p mszros , p. & rees , m. j. 1997 , , 482 , l29 mszros , p. & rees , m. j. 2000 , , 541 , l5 mszros , p. & rees , m. j. 2001 , , 556 , l37 meyer , b. s. 1994 , , 32 , 153 meyer , b. s. 1995 , , 449 , l55 meyer , b. s. 2002 , , submitted ( astro - ph/0207227 ) meyer , b. s. & brown , j. s. 1997 , , 112 , 199 meyer , b. s. , mclaughlin , g. c. , & fuller , g. m. 1998 , , 58 , 3696 murakami , t. , inoue , h. , nishimura , j. , van paradijs , j. & fenimore , e. e. 1991 , , 350 , 592 nakamura , t. 1998 , prog . phys . , 100 , 921 nakamura , t. et al . 2001 , , 555 , 880 narayan , r. , piran , t. & kumar , p. 2001 , , 557 , 949 nomoto , k. et al . 2002 , to appear in a massive star odyssey , from main sequence to supernova , iau symp 212 , eds . k. a. van der hucht , a. herrero & c. esteban ( astro - ph/0209064 ) orosz , j. a. et al . 2001 , , 555 , 489 paczyski , b. 1986 , , 308 , l43 paczyski , b. 1990 , , 363 , 218 paczyski , b. 1998 , , 494 , l45 panaitescu , a. & kumar , p. 2002 , , 571 , 779 piran , t. 1999 , , 314 , 575 podsiadlowski , p. et al . 2002 , , 567 , 491 popham , r. , woosley , s. e. & fryer , c. l. 1999 , , 518 , 356 pruet , j. , abazajian , k. & fuller , g. m. 2001 , , 64 , 063002 pruet , j. , guiles , s. and fuller , g. m. 2002 , , 580 , 368 pruet , j. , woosley , s. e. & hoffman , r. d. 2003 , , 586 , 1254 qian , y .- z . 2000 , , 534 , l67 qian , y .- z . 2003 , astro - ph/0301422 qian , y .- z . & woosley , s. e. 1996 , , 471 , 331 ramirez - ruiz , e. , celotti , a. & rees , m. j. 2002 , , 337 , 1349 rhoads , j. e. 2001 , in proceedings of the ninth marcel grossmann meeting , world scientific ( astro - ph/0103028 ) rossi , e. , lazzati , d. & rees , m. j. 2002 , , 332 , 945 rosswog , s. & ramirez - ruiz , e. 2002 , , 336 , l7 ruffert , m. , janka , h .- th . , takahashi , k. & schfer , g. 1997 , , 319 , 122 ruffert , m. & janka , h .- th . 1999 , , 344 , 573 ryan , s. g. , norris , j. e. & beers , t. c. 1996 , , 471 , 254 schramm , d. n. & barkat , z. 1972 , , 173 , 195 shigeyama , t. & tsujimoto , t. 1998 , , 507 , l135 symbalisty , e. m. d. 1984 , , 285 , 729 symbalisty , e. m. d. , schramm , d. n. & wilson , j. r. 1985 , , 291 takahashi , k. , witti , j. & janka , h .- 1994 , , 286 , 857 terasawa , m. 2002 , ph.d . thesis , university of tokyo terasawa , m. , sumiyoshi , k. , kajino , t. , mathews , g. j. & tanihata , i. 2001 , , 562 , 470 terasawa , m. , sumiyoshi , k. , yamada , s. , suzuki , h. & kajino , t. 2002 , , 587 , l137 thompson , t. a. , burrows , a. & meyer , b. s. 2001 , , 562 , 887 truran , j. w. , cowan , j. j. , pilachowski , c. a. & sneden , c. 2002 , , 114 , 1293 umeda , h. & nomoto , k. 2002 , , 565 , 385 vietri , m. & stella , l. 1998 , , 507 , l45 wheeler , j. c. , yi , i. , hflich , p. & wang , l. 2000 , , 537 , 810 wheeler , j. c. , meier , d. l. & wilson , j. r. 2002 , , 568 , 807 woosley , s. e. 1993 , , 405 , 273 woosley , s. e. , heger , a. & weaver , t.a . 2002a , rev . phys . , 74 , 1015 woosley , s. e. , wilson , j. r. , mathews , g. j. , hoffman , r. d. & meyer , b. s. 1994 , , 433 , 229 woosley , s. e. , zhang , w. & heger , a. 2002b , to appear in `` from twilight to highlight - the physics of supernovae '' eso / mpa / mpe workshop ( astro - ph/0211063 ) yoshida , a. et al . 1989 , , 41 , 509 zhang , b. & mszros , p. 2002 , , 571 , 876 zhang , w. , woosley , s. e. & macfadyen , a. i. 2003 , , 586 , 356

we study population of compact x - ray sources in the small magellanic cloud using the archival data of xmm - newton observatory . the total area of the survey is square degrees with the limiting sensitivity of erg / s/ , corresponding to the luminosity of erg / s at the smc distance . out of point sources detected in the 28 kev energy band , are background cxb sources , observed through the smc . based on the properties of the optical and near - infrared counterparts of the detected sources we identified likely hmxb candidates , and sources , whose nature is uncertain , thus , providing a lower and upper limits on the luminosity distribution of hmxbs in the observed part of the smc . the observed number of hmxbs is consistent with the prediction based on sfr estimates derived from the supernovae frequency and analysis of color - magnitude diagrams of the stellar population . if , on the contrary , the true value of the sfr is better represented by fir , h and uv based estimators , then the abundance of hmxbs in the smc may significantly ( by a factor of as much as ) exceed the value derived for the milky way and other nearby galaxies . the shape of the observed distribution at the bright end is consistent with the universal hmxb xlf . at the faint end , erg / sec , the upper limit on the luminosity function is consistent with while the lower limit is significantly flatter than the power law . [ firstpage ] x - rays : galaxies x - rays : binaries stars : neutron galaxies : individual : smc .

astro-ph0503477

x - ray observations of the small magellanic cloud ( smc ) revealed a surprisingly rich population of high mass x - ray binaries in this close neighbour of the milky way ( e.g. * ? ? ? * ; * ? ? ? at the time of writing , about 50 sources of this type have been identified in the smc with the majority of discovered systems being be / x - ray binaries . existence of such a rich population of hmxbs in a nearby galaxy opens unique possibility to study properties of population of hmxbs . as has been shown by @xcite , the x - ray luminosity function ( xlf ) of hmxbs obeys , to the first approximation , a universal power law distribution with the differential slope of @xmath10 , whose normalization is proportional to the star formation rate of the host galaxy . the validity of this universal hmxb xlf has been established in a broad range of the star formation rates and regimes and in the luminosity range @xmath11 . based on the asca observations of the small magellanic cloud and on the behavior of the integrated x - ray luminosity of distant galaxies located at redshifts @xmath12 observed by chandra in the hubble deep field north , @xcite tentatively suggested , that the hmxb xlf is not dramatically affected by the metallicity variations . study of the population of high mass x - ray binaries in the smc is of importance for several reasons : 1 . the magellanic clouds are known to have a significant under - abundance of metals @xcite . for example , the interstellar medium of the smc has a mean metallicity 0.6 dex lower than the local ism , the [ fe / h ] of young objects ( age@xmath13gyr ) in the smc is 0.5 dex lower than in the solar vicinity @xcite . the effects of the metallicity variations on the population of x - ray binaries are poorly understood . study of the smc , a galaxy with relatively well known chemical composition gives a unique opportunity to investigate such effects observationally . 2 . owing to the proximity of the smc , the weakest sources become reachable within a moderate observing time . indeed , the sensitivity of a typical chandra or xmm - newton observation , @xmath14 erg / s/@xmath2 corresponds to the luminosity of @xmath15 erg / s at the smc distance . this opens the possibility to study the low luminosity end of the hmxb xlf . in addition , the proximity of the smc allows to study the properties of its stellar population and the star formation history in details . [ cols="<,^,^,^,^,^ " , ] in the present paper we study population of x - ray binaries in the smc based on the archival xmm - newton data . the distance modulus of the smc is @xmath16 @xcite , corresponding to the distance of @xmath17 kpc . the paper is structured as follows . xmm - newton observations and data analysis are described in the section [ sec : data ] . in the section [ sec : nature ] we discuss nature of detected x - ray sources . in the section [ sec : ident ] we describe the hmxbs search procedure and its results . resulting hmxbs luminosity function is presented in the section [ sec : hmxbpop ] . the log(n)log(s ) of cxb in the direction of the smc is presented in the section [ sec : cxbsrc ] . ) map of the region around the small magellanic cloud . circles show fields of view of the xmm newton observations used for the analysis . , scaledwidth=50.0% ] we have selected 9 xmm - newton archival observations with the pointing direction towards the smc and with sensitivity better than @xmath18 erg / s/@xmath2 in the 210 kev energy band . these observations are listed in table [ tb : pnt ] . fig.[fig : fir_map ] shows their fields of view , overlayed on the far - infrared map ( iras , @xmath19 ) of the small magellanic cloud . the observations were processed with science analysis system ( sas ) v6.0.0 . after filtering out high background intervals we extracted images in the 28 kev energy band . this energy range was chosen to minimize the fraction of cataclysmic variables and foreground stars among detected sources , generally having softer spectra than high mass x - ray binaries . to improve the sensitivity of the survey , images from mos1 and mos2 detectors have been merged . if both mos and pn data were available , we used the data having higher sensitivity . this was determined comparing the exposure times and taking into account that the effective area of pn is @xmath20 two times higher than the effective area of each of mos cameras . the source detection was performed with standard sas tasks _ eboxdetect _ and _ emldetect_. the value of the threshold likelihood @xmath21 used in emldetect task to accept or reject detected source was chosen as follows . we simulated a number of images with poisson background ( without sources ) . each of the generated images was analyzed with the full sequence of the source detection procedures using different values of the emldetect threshold likelihood @xmath21 and for each trial value of @xmath21 the number of detected `` sources '' was counted . the final value of the threshold likelihood , @xmath22 , was chosen such that the total number of spurious detections was @xmath23 per 9 images . the obtained images were visually inspected and the source lists were manually filtered of spurious sources near bright sources and sources whose extent undoubtedly exceeded psf size . the final merged source list contains @xmath24 sources . their flux distribution is plotted in fig.[fig : lf0 ] . the 28 kev source counts were converted to the 210 kev energy flux assuming a power law spectrum with the photon index 1 and n@xmath25=5@xmath26@xmath27 . the energy conversion factors are ecf@xmath28=2.95@xmath29 erg/@xmath2 and ecf@xmath30=1.0@xmath29 erg/@xmath2 . we note that , as will be shown in sec . [ sec : cxb ] , significant part of sources in our sample are background agn . therefore , in sections [ sec : cxb ] and [ sec : cxbsrc ] ( figures [ fig : area ] , [ fig : lf0 ] and [ fig : cxb ] ) we used energy conversion factors calculated for photon index 1.7 , ecf@xmath28=2.27@xmath29 erg/@xmath2 and ecf@xmath30=8.0@xmath31 erg/@xmath2 , instead of values mentioned above . for several xmm observations we performed boresight correction using optical sources from the mcps catalogue @xcite . the correction was applied in case when two or more well - known sufficiently bright x - ray sources with optical counterparts were found . naturally , the point source detection sensitivity varies from observation to observation and , for a given observation , across the telescope field of view . the reasons of these variations are difference in the exposure times , deterioration of the point spread function of the telescope with the increase of the off - axis angle , presence of the diffuse emission and background variations . these factors affect the completeness of the survey at low fluxes . the conventional way to account for these effects without setting too high conservative completeness limit , is to calculate ( using the sensitivity maps of individual observations ) the survey area as a function of flux . after appropriate normalisation it represents the fraction of sources we are able to detect as a function of flux , and can be used to correct the detected number of low - flux sources : @xmath32 where a(s ) is the survey area at flux s and a@xmath33 is the total ( geometrical ) survey area . the corrected cumulative log(n)log(s ) distribution can be obtained as follows : @xmath34 where @xmath35 is the flux of the j - th source . the flux - dependent survey area calculated as described in @xcite is shown in fig . [ fig : area ] . from fig . [ fig : area ] one can see that the incompleteness effects become important at fluxes @xmath36erg / s/@xmath2 . in the high flux limit , the total area of the survey equals @xmath37 deg@xmath38 . the flux distribution , corrected for the incompleteness effects , is plotted as the thick histogram in fig.[fig : lf0 ] . the total number of sources with the flux @xmath39>2.3\cdot 10^{-14}$ ] erg / s/@xmath2 ( corresponding to the luminosity 10@xmath40erg / s at the smc distance ) is 151 . with account for the survey incompleteness ( eq . [ eq : lc ] ) this number corresponds to @xmath41 ( see thick histogram in fig . [ fig : lf0 ] ) . according to the cxb @xmath42 determined by @xcite , the total number of cxb sources expected in the field of 1.48 deg@xmath38 is @xmath43 . from the comparison of these numbers it is obvious that about 3/4 of sources in our sample are background agns . a significantly less important source of contamination is x - ray sources associated with foreground stars in the galaxy ( no known galactic x - ray binaries were located in the field of view of xmm observations ) . foreground stars are removed by our filtering procedure , as described in section [ sec : ident ] . given the limiting sensitivity of the survey , @xmath44 erg / s/@xmath2 , corresponding to the luminosity @xmath45 erg / s at the smc distance , the intrinsic smc sources are dominated by x - ray binaries . their total number is proportional to the stellar mass ( lmxb ) and star formation rate ( hmxb ) of the galaxy . the total stellar mass of the smc can be estimated from the integrated optical luminosity . according to rc3 catalog @xcite , the reddening corrected v - band magnitude of the smc equals @xmath46 , corresponding to the total v - band luminosity of @xmath47 @xmath48 . from the dereddened optical color @xmath49 @xcite and using results of @xcite , the v - band mass - to - light ratio is @xmath50 in solar units , giving the total stellar mass of the smc @xmath51m@xmath52 . for this stellar mass , using results of @xcite , about one lmxb with luminosity @xmath53 erg / s is expected in the entire galaxy . such a prediction agrees with observations as no lmxb candidates are known in the smc at present . ) . the thin and thick histograms show , respectively , the observed distribution and the distribution corrected for the incompleteness effects , as described in section [ sec : area ] . the thick grey line shows the log(n)log(s ) of cxb sources according to @xcite . the brightest source smc x-1 with flux s@xmath54erg / s/@xmath2 is situated far outside the flux range of the plot . , scaledwidth=50.0% ] determination of a star formation rate in the smc , a diffuse galaxy with low opacity is a nontrivial task . below we compare estimates of the star formation rate in the smc , obtained from different sfr indicators . the sfr values refer to the star formation rate of stars in the 0.1 - 100 m@xmath52 range , assuming salpeter s imf . the discussion of results of different star formation indicators could be also found in @xcite ( see also discussion of the sfr in the lmc by @xcite ) . @xcite based on isophot observations and calibration of @xcite obtained the star formation rate in the smc sfr(ir)=0.015@xmath55yr . @xcite , however , noted that this value underestimates the sfr of the smc because some stellar radiation , due to low opacity of the smc , can escape the star formation regions directly , and gave a more realistic estimate of @xmath56 with the total infrared luminosity of the smc @xcite l@xmath57erg / s and using the equation ( 5 ) from @xcite which accounts for the uncertainty mentioned above , we obtained @xmath58 note , that this calibration should be applied extremely cautious for such a low - luminosity galaxy @xcite . the h@xmath6 luminosity - based indicators can also underestimate the star formation rate in the smc due to the reasons mentioned above . based on h@xmath6 luminosity of the smc and applying extinction correction @xcite estimated the star formation rate of @xmath59 @xcite gave an conservative upper limit of @xmath60 an sfr estimator relatively free of this uncertainty is the one based on combined uv and fir emission . with the uv flux measured by the d 2b - aura satellite @xcite at @xmath61 corrected for foreground extinction and uncorrected isophot fir flux we obtain with the @xcite calibration @xmath62 this value is consistent with the sfr(uv ) derived from flux corrected for both foreground and internal extinction @xcite : @xmath63 @xcite , from comparison of discrete radio ( parkes telescope ) and x - ray ( rosat ) sources , estimated a number of snrs in the smc ( 14 ) . from the age radio flux density relation they estimated an snr birth rate of one snr in @xmath64 yrs . we convert this value to the sfr in the mass range 0.1 - 100m@xmath52 , assuming salpeter s imf and taking into account that only stars with m@xmath65 m@xmath52 can result in a radio snr : @xmath66 similar value has been obtained by @xcite who studied star formation history of the smc fitting the color - magnitude diagrams of stellar population with theoretical isochrones : @xmath67 this number corresponds to recent star formation which took place @xmath68myr ago . thus , the sfr values derived using different star formation rate proxies can differ by the factor of as much as @xmath7ten . the `` standart '' indicators ( ir , uv , h@xmath6 ) result in the rather low value of @xmath69 m@xmath52/yr , whereas the estimates based on the radio supernova rate and the analysis of the color - magnitude diagram point towards much larger star formation rates , in the @xmath70 m@xmath52/yr range . we note , that the methods based on the analysis of the color - magnitude diagrams are generally believed to be the most reliable and accurate . as the controversy between different sfr indicators can not be resolved at present , we shall assume in the following that the true value of sfr is somewhere in the range : @xmath71 for the star formation rate of sfr=@xmath72 yr@xmath73 , using hmxb - sfr relation of @xcite ( see also the discussion of the normalization in the hmxb - sfr calibration by * ? ? ? * ) we predict from @xmath74 to @xmath75 hmxbs with luminosities @xmath76erg s@xmath73 in the whole smc . this confirms that the population of x - ray binaries in the smc is dominated by hmxbs . to predict the number of hmxbs in our sample of x - ray sources , we estimate the star formation rate in the part of the smc covered by xmm pointings from iras infrared maps provided by _ skyview _ @xcite . calculating the far infrared flux according to the formula fir@xmath77 , where fir is flux in erg / s/@xmath2 and @xmath78 is flux in jy @xcite , and integrating the iras maps we obtained the total sfr@xmath79/yr . this number is a factor of @xmath80 smaller than eq.([eq : sfr ] ) because of the uncertainties mentioned in sec . [ sec : sfr ] . to make it consistent with our determination of the sfr in the smc ( eq.[eq : sfr ] ) , we simply multiply the fir - based values by the correction factor of @xmath81 . the thus calculated star formation rate of the part of the smc covered by xmm observations , is : @xmath82 with the above value of sfr we predict about @xmath83 hmxbs with luminosities @xmath76 erg s@xmath73 in the observed part of the smc . the error in this number accounts for uncertainty of the sfr estimate and does not include the poisson fluctuations . as demonstrated above , the population of compact x ray sources in the smc field is dominated by two types of sources background agn and high mass x - ray binaries in the smc . their log(n)log(s ) distributions in the flux range of interest can be described by a power law with the differential slopes @xmath84 ( cxb ) and @xmath10 ( hmxbs ) . due to significant difference in the slopes , their relative contributions depend strongly on the flux . at large fluxes , @xmath85 erg / s/@xmath2 ( @xmath53 erg / s ) the x - ray binaries in the smc prevail . on the contrary , in the low flux limit , e.g. near the sensitivity limit of our survey , @xmath86 erg / s/@xmath2 , the majority ( @xmath5 ) of the x - ray sources are background agn . this is illustrated by fig . [ fig : lf0 ] , showing the observed and corrected for incompleteness log(n)log(s ) distribution of all sources from the final source list . the corrected log(n)log(s ) distribution agrees at low fluxes with that of cxb sources . at high fluxes , there is an apparent excess of sources above the numbers predicted by the cxb sources log(n)log(s ) , due to the contribution of hmxbs . to filter out contaminating background and foreground sources , we use the same filtering procedure as described by @xcite . it is based on the fact that optical emission from hmxbs is dominated by their optical companions , whose properties , such as absolute magnitudes and intrinsic colors are sufficiently well known . high mass x - ray binaries are powered by accretion of mass lost from the massive early - type optical companion . the mechanism of accretion could be connected either to ( i ) strong stellar wind from an ob supergiant ( or bright giant ) or ( ii ) equatorial circumstellar disk around oe or be type star ( e.g. * ? ? ? * ; * ? ? ? * ) . in the case of small magellanic cloud where the only supergiant system smc x-1 is known at present time , the hmxbs with be companion seem to give the main contribution @xcite . given the large number of well known high mass x - ray binaries in the smc , the selection criteria for the optical counterparts of such systems can be determined straightforwardly . for example , @xcite studied the optical properties of be / x - ray pulsars in the smc . the majority of hmxbs with identified optical companions in their sample ( 32 out of 34 ) have apparent v - band magnitudes m@xmath87 . the exceptions are supergiant system smc x-1 with m@xmath88 and x - ray pulsar cxou j010042.8721132 with m@xmath89 ( ogle catalogue * ? ? ? * ) which has been previously classified by @xcite as a possible anomalous x - ray pulsar ( axp ) . this picture is consistent with the expectations based on the positions of possible optical counterparts on the hertzsprung - russel diagram , distance modulus of the smc , @xmath90 @xcite , and foreground and intrinsic extinction towards the smc , a@xmath91 ( for hot population ) @xcite . ob supergiants and bright giants ( luminosity classes i ii ) have absolute magnitudes m@xmath92 , corresponding to apparent magnitudes in the range m@xmath93 . the position of be stars in the h r diagram is close to the main sequence . apparent magnitudes m@xmath94 correspond to spectral classes earlier than b3b5 for main sequence stars ( b5b7 for giants ) . thus we can conclude , that hmxbs in the smc with main sequence be optical companions have spectral classes earlier than b3b5 ( b5b7 for giants ) , in agreement with expectations based on the optical properties of hmxbs in our galaxy . as potential optical counterparts of hmxbs belong to spectral classes earlier than @xmath7 b3b5 , their intrinsic optical and near - infrared colors are constrained by @xmath95 , @xmath96 . with account for the interstellar reddening , both apparent colors are expected to be @xmath97 . this is consistent with b - v colors of hmxbs from @xcite sample . in infrared part of the spectrum , however , be stars are known to exhibit an infrared continuum excess , caused by free - free and free - bound emission within the disk @xcite . therefore , j k colors of be stars can be higher than values qouted above . to avoid this uncertainty , in identification procedure we use mainly b v colors , applying j k identification criteria only in extreme cases ( in fact we used this criteria alone only for one source ) . we used the following optical and near - infrared catalogs : 1 . magellanic clouds photometric survey : the smc ( mcps ) @xcite 2 . guide star catalog , version 2.2.1 ( gsc2.2.1 ) @xcite 3 . the ccd survey of the magellanic clouds @xcite 4 . 2-micron all sky survey ( 2mass ) @xcite 5 . emission - line stars and pne in the smc @xcite 6 . a 2df survey of the small magellanic cloud @xcite as a first step , we cross - correlated the xmm - newton x - ray source list with mcps , gsc and @xcite optical catalogs , using search radius 4@xmath98 and the following selection criteria : @xmath99 , @xmath100 . if optical object was present in mcps and either gsc or @xcite catalogs , the preference was given to mcps , as it has a higher photometric accuracy ( especially at low magnitudes ) . the color limits are higher than possible colors of hmxbs optical companions ( b [email protected] - 0.2 ) and were chosen to account for limited photometric accuracy of optical catalogues . we found optical counterparts for 54 x - ray sources using these selection criteria . on the second stage , we cross - correlated all x - ray sources having optical counterparts with near - infrared catalog 2mass , catalog of @xcite and catalog of @xcite using the same search radius as before . then the following filtering procedure was applied . 1 . in case when source was present only in the gsc catalogue ( which has rather low photometric accuracy ) , we applied filtering criteria based on the near infrared data . in fact , there was only one source with j [email protected] rejected by this criteria . 2 . all x - ray sources with low x - ray - to - optical flux ratio @xmath102 were rejected . the optical flux was calculated from the v - band magnitude , f@xmath103 erg / s/@xmath104 . such low @xmath102 ratios are typical for foreground stars but not for x - ray binaries . all confirmed hmxbs in our sample have @xmath105 among which @xmath106 have @xmath107 . the choice of the search radius plays a crucial role in the procedure of search for potential hmxb candidates based on their optical companions . too small value of the search radius may lead to incompleteness of resulting hmxbs list ( especially at low luminosities ) , while too large value may lead to a significant fraction of chance coincidences . given the spatial density of optical stars ( whose optical colors and magnitudes satisfy our selection criteria ) in the mcps catalogue , the number of detected x - ray sources and the search radius of 4@xmath98 , we estimate the number of chance coincidences to be @xmath108 . on the other hand , a typical error on the position of a low - luminosity source is @xmath109 ( 1@xmath110 confidence ) . therefore , in ideal situation with the search radius of 4@xmath98 we miss only @xmath111 of low - luminosity sources . the real fraction of missed optical counterparts , however , could be somewhat higher as it was obtained assuming ideal astrometric accuracy . our choice of the search radius is further justified in the appendix . [ sec : indsrc ] after the filtering , 50 x - ray sources remained in the list of potential hmxb candidates ( table [ tb : hmxbcand ] ) . as follows from sec.[sec : searchrad ] , @xmath108 of these sources can be chance coincidences . therefore , we further split the filtered list into reliable hmxbs and the sources whose nature is less certain . we include in the list of reliable hmxb candidates only sources which have either detected x - ray pulsations or optical emission lines , or sources whose spectral class is well - determined based on their optical spectrum . the latter two identification criteria are mainly based on @xcite and @xcite catalogues . there were 32 sources satisfying at least one of these criteria . remaining 18 sources are considered to be sources of uncertain nature . the resulting likely hmxbs candidates and sources of uncertain nature are listed in table [ tb : hmxbcand ] . all known hmxbs located in the field of view of the xmm - newton observations are among these sources . comments on the individual sources : # 1 : to minimize the influence of pile - up on the flux of smc x-1 , for this source we used data from epic pn camera operating in small window mode . # 17 = rxj0049.27311 : optical colors consistent with hmxb nature . this source was identified with an bright h@xmath6 object by @xcite and is probably associated with nearby x - ray pulsar axj0049 - 732 ( p=9s ) . # 34 : optical colors consistent with hmxb nature , however , infrared colors are relatively high , j k=0.97 . # 35 : optical colors consistent with hmxb nature . @xcite classified this source as a possible anomalous x - ray pulsar ( p=8s ) . # 38 : this source is situated inside the ngc 330 stellar cluster . as a stellar density is high there , we consider the photometry to be unreliable . # 49 : optical colors consistent with hmxb nature . this source is situated in 24@xmath98 from source axj0048.27309 . @xcite based on the presence of emission - line object in @xcite catalogue within the asca error circle of 40@xmath98 ( @xmath112 confidence ) and x - ray spectrum classified this asca source as a possible be x - ray pulsar . the posititon of the xmm source is inconsistent with emission - line object in @xcite and therefore either optical identification of asca source was incorrect either xmm source is unrelated to asca source . therefore we consider this source as a source of uncertain nature . lcccccccl # & r.a . & @xmath113 ( 1 ) & @xmath114&@xmath115 & f@xmath116/f@xmath117 & i d & comments + & & & erg s@xmath73 & & + + 1 & 01 17 05.2 & -73 26 38 & @xmath118 & 13.2 & -0.14 & 0.55 & p , ob&smc x-1 p=0.717s ( 2 ) + 2 & 00 47 23.3 & -73 12 28 & @xmath119 & 16.0 & 0.08 & 0.80 & p & rx(ax)j0047.37312 p=263s ( 2 ) + 3 & 00 51 52.3 & -73 10 34 & @xmath120 & 14.5 & -0.07 & 0.14 & p , em & rxj0051.97311=axj0051.67311 p=172s ( 2 ) + 4 & 00 49 42.0 & -73 23 14 & @xmath121 & 14.9 & 0.15 & 0.12&p , em & rxj0049.77323=axj0049.57323 p=755.5s ( 2 ) + 5 & 00 57 50.4 & -72 07 55 & @xmath122 & 15.7 & -0.05 & 0.26&p , em & cxouj005750.3720756 p=152.3s ( 2 ) + 6 & 00 54 55.9 & -72 45 10 & @xmath123 & 15.0 & -0.03 & 0.14&p , em & axj0054.87244 p=500s(2,3 ) + 7 & 00 57 49.6 & -72 02 35 & @xmath124 & 15.7 & -0.12&0.19&p , em & rxj0057.87202=axj0058 - 72.0 p=280.4s ( 2 ) + 8 & 01 01 20.8 & -72 11 17 & @xmath125 & 15.6 & -0.19 & 0.15 & p , em & rxj0101.3 - 7211 p=455s ( 2 ) + 9 & 01 01 52.1 & -72 23 34 & @xmath125 & 14.9 & -0.01 & 0.084 & em , ob & axj0101.87223(2 ) + 10 & 00 55 18.5 & -72 38 52 & @xmath126 & 15.9 & 0.15 & 0.18 & p & p=702s ( 3 ) + 11 & 00 59 20.9 & -72 23 15 & @xmath127 & 15.0 & -0.01 & 0.075 & p&xmmuj005921.0722317 p=202s ( 4 ) + 12 & 00 50 57.3 & -73 10 09 & @xmath128 & 14.4 & 0.08 & 0.040 & em & rxj0050.97310=axj0050.87310 ( 2 ) + 13 & 00 50 44.6 & -73 16 05 & @xmath129 & 15.5 & -0.11 & 0.097&p , em & rxj0050.87316 axj0051733 p=323.2 ( 2 ) + 14 & 01 03 13.4 & -72 09 13 & @xmath130 & 14.8 & -0.06 & 0.053 & p , em & saxj0103.27209=axj0103.27209 p=345.2 ( 2 ) + 15 & 00 58 12.4 & -72 30 50 & @xmath131 & 14.9 & 0.11 & 0.038 & em@xmath132 & rxj0058.27231 ( 2,5 ) + 16 & 00 55 28.5 & -72 10 57 & @xmath133 & 16.8 & -0.12 & 0.21 & p & cxouj005527.9721058 p=34s ( 6 ) + 17 & 00 49 13.7 & -73 11 38 & @xmath134 & 16.4 & 0.19 & 0.10 & em & = ( ? ) axj0049732(p=9s ) rxj0049.27311 ( 2,7,10 ) + 18 & 00 48 34.3 & -73 02 31 & @xmath135 & 14.8 & 0.0 & 0.023 & em , ob & rxj0048.5 - 7302 ( 2 ) + 19 & 00 54 56.4 & -72 26 48 & @xmath136 & 15.3 & -0.05 & 0.023 & p , em & xtej0055724 p=59s ( 2 ) + 20 & 01 05 55.3 & -72 03 51 & @xmath137 & 15.7 & -0.06 & 0.028 & em & rxj0105.97203 axj0105.87203 ( 2 ) + 21 & 01 01 37.4 & -72 04 18 & @xmath138 & 16.3 & -0.16 & 0.044 & em & rxj0101.67204 ( 2 ) + 22 & 00 56 15.2 & -72 37 54 & @xmath139 & 14.6 & 0.13 & 0.0075 & em , ob & + 23 & 00 57 36.1 & -72 19 33 & @xmath140 & 16.0 & 0.01 & 0.027&em , p & cxouj005736.2721934 p=565s ( 2 ) + 24 & 00 57 24.0 & -72 23 57 & @xmath141 & 14.7 & -0.07&0.0068 & ob & + 25 & 00 49 29.9 & -73 10 58 & @xmath142 & 16.2 & 0.2 & 0.023 & em & = ( ? ) axj0049732(p=9s)(2,7 ) + 26 & 01 05 08.1 & -72 11 48 & @xmath142 & 15.7 & -0.08 & 0.016 & em & rxj0105.17211 axj0105722 p=3.34s ? ( 2 ) + 27 & 01 03 37.7 & -72 01 35 & @xmath143 & 14.7 & -0.11 & 0.0055 & em & rxj0103.67201 ( 2 ) + 28 & 01 19 39.0 & -73 30 14 & @xmath144 & 15.9 & -0.07 & 0.016 & em & ( 4 ) + 29 & 01 01 03.0 & -72 07 01 & @xmath145 & 15.8 & -0.07 & 0.013 & em&rxj0101.07206 p=304.5 ( 2 ) + 30 & 01 00 30.3 & -72 20 32 & @xmath146 & 14.6 & -0.06 & 0.0034 & em , ob & xmmuj010030.2722035 ( 2 ) + 31 & 00 56 05.8 & -72 21 57 & @xmath147 & 15.9 & -0.04&0.097 & p , em & xmmuj005605.2722200 p=140.1 ( 2 ) + 32 & 00 50 47.9 & -73 18 12 & @xmath148 & 15.6 & 0.12 & 0.0035 & em & + + 33 & 00 55 34.9 & -72 29 05 & @xmath149 & 14.7 & -0.04 & 0.031 & & + 34 & 00 48 18.8 & -73 21 00 & @xmath150 & 16.2 & 0.25 & 0.069 & & + 35 & 01 00 42.9 & -72 11 32 & @xmath150 & 17.7 & -0.03 & 0.28 & p & cxouj010043.1721134 axp ? p=8s ( 8) + 36 & 00 54 32.2 & -72 18 09 & @xmath151 & 16.6 & -0.08 & 0.067 & & + 37 & 00 54 03.6 & -72 26 30 & @xmath137 & 14.9 & 0.01 & 0.014 & & + 38 & 00 56 18.7 & -72 28 02 & @xmath152 & 15.3 & -0.40 & 0.012 & & ngc330 stellar cluster + 39 & 01 00 37.2 & -72 13 16 & @xmath153 & 16.7 & -0.17&0.026 & & + 40 & 00 48 33.4 & -73 23 55 & @xmath153 & 17.7 & 0.07 & 0.064 & & + 41 & 00 54 41.1 & -72 17 20 & @xmath146 & 15.7 & -0.11&0.0090 & & + 42 & 00 47 19.9 & -73 08 22 & @xmath147 & 17.3 & 0.08 & 0.036 & & + 43 & 00 53 35.8 & -72 34 24 & @xmath154 & 17.2 & -0.07 & 0.029 & & + 44 & 01 04 37.4 & -72 06 30 & @xmath155 & 18.0 & 0.13 & 0.048 & & + 45 & 01 00 22.9 & -72 11 27 & @xmath156 & 17.0 & -0.13 & 0.018 & & + 46 & 00 48 04.0 & -73 17 01 & @xmath156 & 17.0 & 0.21 & 0.016 & & + 47 & 00 49 05.4 & -73 14 10 & @xmath157 & 18.0 & 0.03&0.034 & & + 48 & 01 03 28.3 & -72 06 51 & @xmath157 & 16.5 & -0.15 & 0.0086 & & + 49 & 00 48 14.9 & -73 10 03 & @xmath158 & 15.3 & 0.26 & 0.0025 & & = ( ? ) axj0048.27309 ( 2,9 ) + 50 & 00 50 20.2 & -73 11 17 & @xmath158 & 17.0 & 0.11&0.012 & & + + \(1 ) 210 kev band , assuming distance of 60 kpc ; ( 2 ) see @xcite and references inside ; ( 3 ) @xcite ; ( 4 ) @xcite ; ( 5 ) @xcite ; ( 6 ) @xcite ; ( 7 ) @xcite ; ( 8) @xcite ; ( 9 ) @xcite ; ( 10 ) @xcite ; comments : p source shows pulsations ; em source has balmer emission lines in its optical spectrum ( based on @xcite catalogue , unless otherwise is noted ) ; ob source is present in @xcite catalogue and belongs to ob spectral class ; ) . two solid histograms correspond to upper and lower limits on xlf obtained with search method based on the optical properties of hmxbs . two dashed histograms correspond to upper and lower limits on xlf obtained with search method based on the x - ray hardness ratio . it is clear from this figure , that both methods give similar result for sources with luminosities @xmath159erg / s . the brightest source smc x-1 with luminosity l@xmath160erg / s is situated outside the luminosity range of the graph . , scaledwidth=50.0% ] according to the contribution of the xmm - newton survey area to the total sfr of the smc . the dashed histogram is the asca xlf from @xcite . the shaded area shows the luminosity distribution predicted from the `` universal '' hmxb xlf of @xcite extrapolated towards low luminosities . its width reflects the uncertainty of our knowledge of the sfr in the smc , as discussed in the section [ sec : sfr ] . , scaledwidth=50.0% ] the completeness of the list of hmxb candidates is defined by the following factors : 1 . the completeness of the optical catalogues . the initial search for optical counterparts is based on mcps and gsc2.2 catalogs . in general , the mcps catalogue is thought to be complete for v@xmath16120 @xcite . the gsc2.2 is a magnitude selected ( @xmath162 ) subset of the gsc - ii catalog ( http://www-gsss.stsci.edu/gsc/gsc2/gsc2home.htm ) . the latter is complete to @xmath163 at high galactic latitudes @xcite . + completeness of both catalogs is known to break down in the crowded regions . as no sensitivity maps for the optical catalogs exist , a quantitative estimate of the completeness of the initial counterpart search is impossible . however , the quoted completeness limits of both catalogs are @xmath164 mag better than chosen threshold of @xmath165 mag for the optical counterparts search . this suggests that the completeness of the optical catalogs is unlikely to be the primary limiting factor . + the mcps catalogue also has `` stripe - like '' gaps with zero stellar density . we checked that the majority of x - ray sources do not coincide with this `` gaps '' . for few sources which were close or inside the gaps we used two additional catalogs : ogle @xcite and ( if ogle had no data in this region ) usno - b @xcite . 2 . the efficiency of the initial search due to statistical and systematic uncertainties in the positions of x - ray and optical sources . this is probably one of the major limiting factors . as has been noted in sec [ sec : searchrad ] , the search radius of 4@xmath98 assuming ideal astrometric quality corresponds to detection of @xmath166 of the weakest sources , while for moderate luminosity and bright sources this fraction would be close to 100@xmath167 . in appendix we show , that in real case ( i.e. with all uncertainties taken into account ) , with the chosen value of the search radius we miss not more than few sources . the filtering procedure applied to the optical matches found in the initial search . this procedure is based mainly on observed optical and near - infrared properties of known hmxbs in the smc and milky way . the distribution of v - band magnitudes and b - v optical colors of hmxbs in the smc have maxima far inside our filtering limits ( see fig . [ fig : opticalproperties ] ) , meaning that we detect the vast majority of hmxbs unless some significantly different ( and previously undetected ) type of these systems exists . thus , we estimate the completeness of our hmxbs sample to be very high with the uncertainties in the positions of x - ray sources being the main factor of incompleteness . we note , that as this factor is flux - dependent , it affects most strongly the low - luminosity part of our sample . the identification method based on the properties of optical counterparts is not the only way to identify the population of hmxbs in the background of contaminating sources . for example , @xcite found that hardness ratios ( hrs ) of all known x - ray binary pulsars ( xbps ) in the magellanic clouds fall into narrow `` xbp region '' , hr@xmath168(asca)@xmath169 . in order to compare these two methods , we apply selection criteria based on hardness ratio to our sample of x - ray sources . we use xbps selection criteria similar to those described in @xcite . the hardness ratio is defined as @xmath170 , where h and s are count rates in the 2 - 7 kev and 0.7 - 2 kev energy bands respectively . all x - ray sources with hrs corresponding to photon indexes @xmath171 ( corresponding `` xbp region '' defined by @xcite ) are classified as x - ray binary pulsar candidates . x - ray sources which have slightly softer spectra , @xmath172 , are considered as sources of uncertain nature . higher photon indexes are untypical for hmxbs , especially for be / x - ray binaries which seem to constitute the majority of hmxbs population in the smc . luminosity distributions obtained using these two methods are shown in fig . [ fig : hardnessfilt ] . as is clear from fig . [ fig : hardnessfilt ] , the results of both methods agree for sources with l@xmath173erg / s , while for the low - luminosity sources the method based on hardness ratio gives significantly higher number of sources . this could be a result of large statistical uncertainty in the hardness ratios for low flux sources , leading to a `` leakage '' of background agn to the xbp region . in addition , the population of highly absorbed agn with truly hard spectra becomes important at low fluxes and compromises the simple hardness ratio based selection criteria . the incompleteness - corrected luminosity distribution of hmxbs in the smc is shown in fig . [ fig : lfhmxbxmmasca ] . in order to represent the hmxb population of the entire galaxy , the observed xlf has been scaled up according to the relative contribution of the area covered by the xmm - newton survey to the total sfr of the smc . the upper and lower histograms below @xmath174 erg / sec correspond to all sources from table [ tb : hmxbcand ] and to the likely hmxb candidates respectively . these two histograms provide upper and lower limits on the true x - ray luminosity function of hmxbs in the observed part of the smc . [ fig : lfhmxbxmmasca ] also shows the xlf of hmxb candidates in the smc obtained by @xcite based on the asca data . the asca xlf refers to the entire smc . we note that removal of the sources located outside xmm pointings does not change its shape significantly . the upper asca histogram in the figure corresponds to all sources from @xcite excluding snr , agn and foreground stars i.e. represents all sources which properties do not contradict to hmxb nature . the lower histogram shows the luminosity distribution of likely hmxb candidates confirmed x - ray binary pulsars and candidates and confirmed non - pulsating hmxbs and candidates ( see @xcite for details ) . as is evident from the figure , the xmm - newton and asca xlfs are consistent with each other . the k - s test for two distribution gives the probability of @xmath175 . the shaded area in fig . [ fig : lfhmxbxmmasca ] shows the universal hmxb xlf of @xcite and its extrapolation towards low luminosities . the width of the shaded area reflects uncertainties in our knowledge of the star formation rate in the smc as discussed in the section [ sec : sfr ] . as is clear from fig . [ fig : lfhmxbxmmasca ] , the observed abundance of hmxbs in the smc is consistent with the upper range of the predicted numbers , which is based on the sfr estimates derived from the supernovae rate and analysis of the stellar color - magnitude diagram . if , on the contrary , the true value of the sfr is better represented by fir , h@xmath6 and uv based estimators , then the abundance of hmxbs in the smc may exceed significantly , by a factor as much as @xmath710 , the value derived for the milky way and other nearby galaxies by @xcite . note that the analysis and interpretation of the smc data in @xcite was based on the star formation rate estimates obtained by @xcite from the supernovae rate . at the first glance the shape of the hmxb xlf obtained by xmm - newton appears to be more complex than a single slope @xmath176 power law in the entire @xmath177 erg / sec luminosity range . the maximum likelihood fit to the bright end of the distribution ( @xmath178erg / s ) with a power law results in the best fit value of the cumulative slope @xmath179 which is consistent with the slope of 0.6 . this is also confirmed by the kolmogorov - smirnov ( k - s ) test which gives probability of @xmath180 . thus , the bright end of the xlf is consistent with the universal hmxb xlf . this conclusion is further confirmed by the asca xlf containing by a factor of @xmath181 more sources . at low luminosities the uncertainties of hmxb identification procedure become of importance and we consider the upper and lower limits of the xlf separately . the upper limit on the xlf ( the upper histogram in fig . [ fig : lfhmxbxmmasca ] ) is consistent with a single slope power law in the luminosity range l@xmath182 erg / s with the k - s test probability of @xmath183 and the best fit value of the slope @xmath184 . for the lower histogram in fig . [ fig : lfhmxbxmmasca ] ( luminosity distribution of reliable hmxb candidates ) we obtain best fit slope of @xmath185 in the full luminosity range . the observed distribution , is consistent with a single slope power law the k - s test probability is @xmath186 . a power law approximation in the @xmath187 erg / sec luminosity range gives the best fit slope of @xmath188 . thus , the behavior of the xlf in the faint end is somewhat uncertain . however , based on the statistical arguments and the distribution of the optical magnitudes ( see sec . [ sec : searchrad ] and [ sec : hmxboptobserved ] ) , the majority of sources of uncertain nature are expected to be the result of chance coincidence . therefore the lower histogram in fig . [ fig : lfhmxbxmmasca ] should better represent the true hmxb xlf and the flattening of the xlf in the faint end can be real . it can result from the `` propeller '' effect , as suggested by @xcite . the distribution of v - band magnitudes of the hmxb candidates is shown in fig . [ fig : opticalproperties ] . in the case of the likely hmxb candidates it shows bimodal structure with two peaks at m@xmath189 and m@xmath190 . to estimate the statistical significance of the observed bimodality , we perform a test based on the bootstrap method @xcite . the obtained value of significance is rather low and does not allow us to make a definite conclusion ( with probability that the observed distribution is multimodal of 97% , i.e. significance slightly above 2@xmath110 ) . the origin of this bimodality , if real , is not clear . it could correspond to be / x - ray systems with main sequence and giant be secondaries . however , this explanation is not satisfactory as it would require a rather narrow distributions of the optical companions over spectral classes . the distribution of the m@xmath191 magnitudes of the uncertain hmxb candidates differs from those of likely candidates . this is consistent with expectations , that significant part of them should be a chance coincidences ( see sec . [ sec : searchrad ] ) . the log(n)log(s ) distributions of cxb sources , obtained after removal of likely hmxb candidates only and all hmxb candidates are shown in fig.[fig : cxb ] . we fit the resulting log(n)log(s ) distribution in the flux range @xmath192 erg @xmath27 s@xmath73 with a power law model @xmath193 , where @xmath194 erg / s/@xmath2 . best fit value of slope for the case with only likely hmxb candidates excluded is @xmath195 . for the case with all the hmxb candidates excluded we obtain @xmath196 . we also compare the resulting log(n)log(s ) distributions in the flux range @xmath192 erg @xmath27 s@xmath73 with cxb number counts obtained by @xcite . the kolmogorov - smirnov test accepts their model giving probabilities @xmath197 for the case with likely hmxb candidates excluded and @xmath198 for the case with all the hmxb candidates excluded . the number of sources at flux @xmath199 erg @xmath27 s@xmath73 predicted by @xcite , n@xmath200 , is also consistent with the observed numbers , n@xmath201 and n@xmath202 for the two above cases respectively . thus , the cxb log(n)log(s ) in the direction of the smc is consistent both in the shape and the normalization with results of other cxb surveys . based on the archival data of xmm newton observations , we studied the population of point - like sources in the field of the smc . the total area of the survey is @xmath0 sq.degr . with the limiting sensitivity of @xmath14 erg / s/@xmath2 ( fig.[fig : area ] ) , corresponding to the luminosity of @xmath203 erg / s at the smc distance . 1 . out of the 196 compact sources , detected in the 28 kev energy band , @xmath5 are cxb sources , observed through the smc ( section [ sec : cxb ] , fig.[fig : lf0 ] ) . 2 . based on the stellar mass and the star formation rate of the smc we demonstrate , that the majority of the intrinsic smc sources , detected in the 28 kev band are high mass x - ray binaries ( section [ sec : nature ] ) . the proximity of the smc and adequate angular resolution of xmm newton make it possible to reliably filter out the sources , whose properties are inconsistent with hmxb nature . based on the optical and infrared magnitudes and colors of the optical counterparts of the x - ray sources we identify 32 likely hmxb candidates ( 29 of which were previously known ) and 18 sources of uncertain nature ( section [ sec : ident ] , table [ tb : hmxbcand ] ) . the remaining @xmath204 sources , with a few exceptions , are background objects , constituting the resolved part of cxb . their flux distribution is consistent both in shape and normalisation with other determinations of the cxb log(n)log(s ) ( section [ sec : cxbsrc ] , fig.[fig : cxb ] ) . 4 . with these results we constrain lower and upper bounds of the luminosity distribution of hmxbs in the observed part of the smc . we compare these with the extrapolation towards low luminosities of the universal luminosity function of hmxbs , derived by @xcite ( fig . [ fig : lfhmxbxmmasca ] ) . + the observed number of hmxbs is consistent with the prediction based on sfr estimates derived from supernovae frequency and analysis of the stellar color - magnitude diagrams . if , on the contrary , the true value of the sfr is better represented by fir , h@xmath6 and uv based estimators , then the abundance of hmxbs in the smc may significantly ( by a factor of as much as @xmath710 ) exceed the value derived for the milky way and other nearby galaxies ( sections [ sec : sfr ] , [ sec : hmxbpop ] ) . + the shape of the observed distribution at the bright end is consistent with the @xmath176 power law of the universal hmxb xlf . at the faint end , @xmath205 erg / s , the upper limit of the luminosity function is consistent with the @xmath9 power law , while the lower limit is significantly flatter . 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 . this publication has made use of data products from the two micron all sky survey , guide star catalogue - ii . ps acknowledges the support of european association for research in astronomy ( mest - ct-2004 - 504604 marie curie - eara est fellowship at max - planck - institute for astrophysics ) . ps also acknowledges a partial support from the president of the russian federation grant ss-2083.2003.2 . bell e. , de jong r. , 2001 , apj , 550 , 212 bell e. , 2003 , apj , 586 , 794 m.j . coe , w.r.t . edge , j.l . galache , v.a . mcbride , 2004 , astro - ph/0410074 corbet r.h.d . , 1986 , mnras , 220 , 1047 cutri r. m. , skrutskie m. f. , van dyk s. , beichman c. a. , carpenter j. m. , chester t. , cambresy l. et al . , 2003 , ycat , 224 de vaucouleurs g. et al . , 1991 , third refererence catalog of bright galaxies . springer - verlag ( rc3 ) edge w. r. t. , coe m. j. , 2003 , mnras , 338 , 428 edge w. r. t. , coe m. j. , galache j. l. , mcbride v. a. , corbet r. h. d. , markwardt c. b. , laycock s. , 2004 , mnras , 353 , 1286 evans c.j . , howarth i.d . , irwin m.j . , burnley a.w . , harries t.j . , 2004 , mnras , 353 , 601 filipovic m. et al . , 1998 , a&a auppl . , 127 , 119 filipovic m. d. , pietsch w. , haberl f. , 2000 , a&a , 361 , 823 gehrz r. d. , hackwell j. a. , jones t. w. , 1974 , apj , 191 , 675 gilfanov m. , 2004 , mnras , 349 , 146 grimm , h .- j . , gilfanov , m.r . , sunyaev , r.a . 2003 , mnras , 339 , 793 haberl f. , pietsch w. , 2004 , a&a , 414 , 667 haberl f. , pietsch w. , schartel n. , rodriguez p. , corbet r. h. d. , 2004 , a&a , 420l , 19 harris j. , zaritsky d. , 2004 , aj , 127 , 1531 helou et al . , 1985 , apj , 298 , l7 kennicutt r , c . 1991 , in : haynes r.f . & milne d.k . 148 , the magellanic clouds , reidel , dordrecht , p.139 kennicutt r.c . , 1998 , ara&a , 36 , 189 kennicutt r.c . jr , bresolin f. , bomans d.j . , bothun g.d . , thompson i.b . , 1995 , aj , 109 , 594 lamb r. c. , fox d. w. , macomb d. j. , prince t. a. , 2002 , apj , 574 , 29 majid w. a. , lamb r. c. , macomb d. j. , 2004 , apj , 609 , 133 massey p. , 2002 , ycat , 2236 , 0 mcglynn t. , scollick k. , white n. , skyview : the multi - wavelength sky on the internet , mclean , b.j . et al . , new horizons from multi - wavelength sky surveys , kluwer academic publishers , 1996 , iau symposium no . 179 , p465 . meyssonnier n. , azzopardi m. , 1993 , a&as , 102 , 451 monet d.g . , levine s.e . , canzian b. , ables h.d . , bird a.r . , , guetter h.h . et al . , 2003 , aj , 125 , 984 moretti a. , campana s. , lazzati d. , tagliaferri g. , 2003 , apj , 588 , 696 morrison j. e. , mclean b. , gsc - catalog construction team , ii , 2001 , dda , 32.0603 negueruela i. , 1998 , a&a , 338 , 505 shtykovskiy p. , gilfanov m. , 2005 , a&a , 431 , 597 silverman b.w . , 1981 , j.r.statist.soc.b , 43 , 97 udalski a. , szymanski m. , kubiak m. et al . , 1998 , ycatp005004802u van paradijs , j. & mcclintock j.e . , 1995 , x - ray binaries , cambridge univ.press , p.58 vangioni - flam e. et al . , 1980 , a&a , 90 , 73 westerlund b. , `` the magellanic clouds '' , cambridge univ.press , 1997 wilke k. , klaas u. , lemke d. et al . , 2004 , a&a , 414 , 69 yokogawa j. , imanishi k. , tsujimoto m. , nishiuchi m. , koyama k. , nagase f. , corbet r.h.d . , 2000 , apjs , 128 , 491 yokogawa j. , imanishi k. , tsujimoto m. , koyama k. , nishiuchi m. , 2003 , pasj , 55 , 161 zaritsky d. , harris j. , thompson i.b . , grebel e.k . , massey p. , 2002 , aj , 123 , 855 below we consider the problem of cross - corellation of two catalogs ( e.g. of the list of detected x - ray sources and an optical catalogue ) . more specifically we concentrated on the case when only a fraction of x - ray sources has true optical counterparts in the optical catalogue and when the probability of chance coincidences is not negligible . in this case , the matches of x - ray sources with optical sources will include real optical counterparts and chance coincidences . we consider the behavior of the growth curves the dependence of the number of matches on the search radius . we show that using different dependences of the number of true matches and chance coincidences on the search radius , it is possible to estimate the number of true optical counterparts . the number of x - ray sources with optical matches inside radius r depends on the number of x - ray sources which have true optical counterpart inside r and number of chance coincidences of x - ray sources with field stars : @xmath206 where n - total number of x - ray sources , m - number of x - ray sources with true optical counterparts , @xmath207 - density of optical field stars ( i.e. excluding the true optical counterparts ) and @xmath208 is the distribution of distances between x - ray sources and their optical counterparts . the first term in the above equation is a number of chance coincidences , i.e. number of x - ray sources which have at least one field star inside its search radius . the second term corresponds to number of true matches falling inside search radius r. the exponent in this term excludes cases when both true optical counterpart and chance coincidence with field stars are present , as such matches were already taken into account in the first term . assuming that all x - ray sources have the same positional error @xmath110 ( hence @xmath209 ) , we can rewrite this equation in the following way : @xmath210 fitting the curve n@xmath211(r ) with this formula one can determine the number of x - ray sources with true optical matches , surface density of optical stars @xmath207 and error on the source positions . in a real situation , the density of optical stars @xmath207 varies from field to field and even from source to source . this can lead to significant deviations of n@xmath211(r ) curve behaviour from the eq . [ eq : appendix2 ] . there is no obvious way to overcome this problem completely . however , the density variations could be partially taken into account by replacing the mean density of optical objects in the first term of eq . [ eq : appendix2 ] by the local densities of optical objects around each individual x - ray source . the first term in this case , of course , should be replaced by the corresponding sum over all x - ray sources . the density @xmath207 in the second term , however , can not be straightforwardly calculated because it is related to x - ray sources with true optical counterparts , which we do not know a priori . thus , modifying the first term in equation [ eq : appendix2 ] we obtain : @xmath212 the local densities @xmath213 of optical objects should be calculated individually for each x - ray source . the areas over which these densities are calculated must be large enough to minimize the relative contribution of the true optical counterparts and poissonian errors . density @xmath207 corresponds now to the mean density of optical objects near x - ray sources which have true optical counterparts . fit parameters are the number of x - ray sources with true optical matches m , the value of positional error @xmath110 and the mean density of optical objects @xmath207 . fitting the n@xmath211(r ) curve using real list of xmm sources and optical sources from the mcps catalogue we obtain the best fit value for the total number of sources with true optical counterpart in our sample is @xmath214 , surprisingly close to the number of our likely hmxb candidates ( see fig . [ fig : nmatch2 ] ) . best fit value of positional error , @xmath215 , also agrees with typical values for x - ray sources from our list ranging from @xmath216 to @xmath217 . we note also , that from the eq . [ eq : appendix3 ] it follows that subtracting the first sum from @xmath218 curve and multiplying the remaining part by @xmath219 , we can obtain the dependence of the number of true counterparts among optical matches on the search radius . the resulting curve is shown on fig . [ fig : nmatch ] . it is clear from this figure , that almost all true optical counterparts of x - ray sources will be detected with the search radius of 4@xmath98 used in our hmxb identification procedure . . the individual components of this equation corresponding to spurious matches with field stars and true optical counterparts ( which do nt have simultaneous spurious matches ) are shown with dashed and dot - dashed lines respectively . , scaledwidth=45.0% ] ( see [ sec : appvardens ] ) . the vertical dashed line shows the search radius of @xmath220 arcsec . it is clear from this figure , that almost all true optical counterparts of x - ray sources fall within the chosen search radius.,scaledwidth=45.0% ] the deviation of positional errors distribution from the delta function together with the astrometric uncertainty can also modify n@xmath211(r ) curve . with the spread in positional errors accounted for , the equation [ eq : appendix3 ] becomes : where @xmath222 is the ( gaussian ) distribution of distances between the x - ray and optical sources for given @xmath110 and @xmath223 is a distribution of positional uncertainties of x - ray sources . to show that such a spread will not significantly change our results , we assumed that errors distribution follows the power law @xmath224 between @xmath225 and @xmath226 . in this case the free fit parameters are the number of x - ray sources with true counterparts m , optical sources density @xmath207 and slope @xmath227 .

we provide a simple example of dilaton stabilization in the framework of heterotic string theory . it requires a gaugino condensate and an up - lifting sector similar to the one postulated in type iib string theory . its signature is a hybrid mediation of supersymmetry breakdown with a variant of a mirage pattern for the soft breaking terms . the set - up is suited for the discussion of heterotic mssm candidates .

0802.1137

with the recent success of model building in the framework of the heterotic @xmath0 string theory @xcite it becomes important to reconsider the questions of moduli stabilization and supersymmetry breakdown . early attempts considered fluxes and gaugino condensates @xcite , race track superpotentials @xcite and/or khler stabilization @xcite . more recently , these questions have been studied in the framework of type iib string theory . while explicit model building towards the mssm ( the minimal supersymmetric extension of the standard model ) is more difficult in this framework , it is usually argued that the stabilization of moduli can be achieved quite easily . the reason is the appearance of two types of fluxes [ neveu - schwarz neveu - schwarz ( ns - ns ) and ramond ramond ( r - r ) ] @xcite while only one of them is present in the heterotic theory . with the inclusion of nonperturbative effects like gaugino condensation one might then stabilize all moduli @xcite . still , there is a problem with the adjustment of the vacuum energy @xmath1 to a small positive value . one has to postulate a so - called `` up - lifting '' sector that adjusts @xmath1 to the desired value @xcite . it turns out that this up - lifting sector has important consequences for the explicit pattern of supersymmetry breakdown @xcite . instead of modulus mediation one is led to a hybrid mediation scheme and a so - called `` mirage pattern '' of the soft breaking terms emerges @xcite . within the framework of the heterotic string , the importance of this up - lifting sector has not been fully appreciated so far . the present paper is an attempt to fill this gap . to illustrate the importance of this sector , we consider a simple example : a gaugino condensate in the absence of a flux background . this is known to lead to a so - called run - away potential with a supersymmetric vacuum at @xmath2 ( where @xmath3 denotes the dilaton field ) . attempts to stabilize the dilaton at a finite value include the consideration of nonperturbative correction to the khler potential @xcite . the resulting scheme is believed to provide a mediation of supersymmetry breakdown that is dominated by the dilaton . still the question of the fine - tuning of the vacuum energy has to be addressed . in the existing examples it seems again that an additional sector is needed to adjust the vacuum energy to the desired value @xcite , just like the up - lifting sector in the type iib case . to be explicit , we consider the model of barreiro , carlos and copeland ( bcc ) ref . @xcite that appears to be particularly suited for a class of realistic heterotic models @xcite . a stabilization of the dilaton can be achieved , but the actual value of @xmath1 turns out to be large and positive so that the desired `` up - lifting '' mechanism should provide a `` down - lift '' of @xmath1 . we here consider the mechanism of lebedev , nilles and ratz ( lnr ) @xcite to adjust @xmath1 . in the framework of type iib theory this is known as @xmath4-term uplifting @xcite and could be originated in dynamical schemes as considered in @xcite or schemes that might require the existence of additional branes and/or anti - branes @xcite . an explicit discussion of supersymmetry breakdown leads to similar conclusions as in the type iib case , the uplifting sector is important for the mechanism of supersymmetry ( susy ) breakdown and its mediation such that a variant of a mirage pattern emerges @xcite . there are , however , some quantitative differences between the heterotic and the type iib case that will be discussed in detail . so far it seems that the stabilization of the dilaton requires sizeable nonperturbative corrections to the khler potential while the uplifting sector adjusts @xmath1 . a closer inspection , however , reveals the surprising fact that one can switch off the corrections to the khler potential and still remain with a stabilized dilaton ! _ the uplifting sector alone is responsible for both , modulus stabilization and adjustment of the vacuum energy . _ stabilization of the dilaton neither requires nonvanishing flux nor corrections to the khler potential . the appearance of a gaugino condensate , however , remains crucial . this remarkable fact shows that modulus stabilization in heterotic string theory is not as difficult as usually assumed . one just needs a suitable uplifting sector very similar to the one postulated in the framework of type iib string theory . it also shows the importance of this uplifting sector for moduli stabilization and supersymmetry breakdown . moreover it leads to a hybrid mediation scheme and its signature is a mirage pattern of the soft supersymmetry breaking terms . the paper is organized as follows . in section 2 we shall discuss the model of ref @xcite with a gaugino condensate and nonperturbative corrections to the khler potential . we present the heterotic analogy of uplifting in the spirit of lnr @xcite , the adjustment of the vacuum energy of the metastable vacuum and the discussion of supersymmetry breakdown . this will be followed by a discussion of the soft supersymmetry breaking parameters , the appearance of a mirage pattern and the phenomenological properties of the set - up . section 4 will contain some concluding remarks . dilaton stabilization in the context of heterotic string theory can occur via a racetrack mechanism @xcite or by the means of nonperturbative corrections to the khler potential @xcite . the former option , consisting of at least two gaugino condensates , leads to a scenario where both the auxiliary field of the dilaton and the @xmath5 modulus are nonzero with typically @xmath6 . the latter option might require just one gaugino condensate , but sizeable nonperturbative corrections to the tree level khler potential . under certain circumstances this leads to a scenario with @xmath7 which is also known as the _ dilaton domination scenario_. casas @xcite has investigated the impact of an arbitrary khler potential on the low energy theory and also introduced possible string motivated nonperturbative corrections to the khler potential . the distinct feature of this ansatz was the fact that it became possible to stabilize the dilaton at phenomenologically acceptable values @xmath8 , although the minimum of the scalar potential turned out to be de sitter . in this work we shall investigate the impact of the presence of an up - lifting sector ( consisting of hidden sector matter fields @xcite ) on a set - up with just one gaugino condensate and the tree - level khler potential . but let us first repeat the arguments of @xcite and present the explicit model of bcc @xcite . consider @xcite @xmath9 where @xmath10 is the dedekind eta function ( convention as in @xcite ) , insuring the correct transformation under the @xmath11 target space modular invariance and @xmath12 if the condensing gauge group is @xmath13 . the khler potential at tree - level is given by @xmath14 in the supergravity language the scalar potential and the @xmath4 terms are expressed in terms of the function @xmath15 with @xmath16 and @xmath17 being the khler potential and the superpotential , respectively . the scalar potential is given by @xmath18 where @xmath19 , @xmath20 denote differentiation with respect to the fields and @xmath21 is the inverse khler metric . the @xmath4 terms are found to be @xmath22 using eqs.([eqn : super],[eqn : kahler ] ) we obtain @xmath23 with @xmath24 . in this work we will assume that the khler modulus is stabilized at one of the self - dual points of the @xmath25 function , leading to @xmath7 . from now on we will drop the @xmath5 dependence in khler potential , gauge kinetic function and superpotential and rescale @xmath26 . in order to provide the formation of a minimum and so to stabilize the dilaton we look at the stationary point condition @xmath27 which relates the derivatives of @xmath28 as @xmath29 if we only use the tree - level khler potential @xmath30 , eq.([eqn : oldgrelation ] ) becomes @xmath31_{{\re\mathfrak{e}\hspace{1pt}}{s}\simeq 2 } & = 0 , \label{eqn : oldwrong}\end{aligned}\ ] ] which can not be solved for any reasonable value of @xmath32 or @xmath33 , respectively . ) . ] ) . ] therefore one obtains a runaway potential for the dilaton ( fig.[fig : runaway ] ) . adding nonperturbative corrections to standard tree - level khler potential as motivated in @xcite @xmath34 \label{eqn : npkahler}\end{aligned}\ ] ] provides an extra contribution to eq.([eqn : oldwrong ] ) which also must be sizeable in order to allow for a solution . the explicit model of @xcite typically obtains a potential as displayed in fig.[fig : copeland ] . stabilization of the dilaton depends crucially on the presence of nonperturbative corrections to the khler potential . otherwise we would have a run - away behaviour of the dilaton and no stabilization . typically the vacuum energy turns out to be too large and a `` down - lift '' to a smaller value is necessary . an additional contribution to the scalar potential is therefore needed to give a solution to eq.([eqn : oldgrelation ] ) . the role of such an additional contribution should allow the adjustment of the vacuum energy to a zero ( or small ) value . _ we shall see that such an additional sector will stabilize the dilaton even in the absence of the nonperturbative corrections to the khler potential . _ we would like to investigate the interaction of the set - up described in the previous section with a new hidden sector designed to adjust the vacuum energy . this is the heterotic analogy of @xcite . the corresponding khler potential reads @xmath35 with @xmath36 being a hidden sector matter field , assumed to be a singlet under unbroken gauge symmetries . from eq.([eqn : scalarpot ] ) we obtain @xmath37 as obvious from eq.([eqn : newpot ] ) there are now new contributions to the scalar potential which , by properly adjusting the parameters of the new sector , may modify the shape of the scalar potential . let us for concreteness and simplicity consider the polonyi superpotential @xcite and focus on the possibility of having a minkowski vacuum . the superpotential for this choice is given by @xmath38 where @xmath39 and @xmath40 are constants . the scalar potential eq.([eqn : newpot ] ) becomes @xmath41 where we have introduced the quantities @xmath42 which denote the contribution of the dilaton and hidden sector matter field to susy breaking , respectively . the stationary point conditions now become @xmath43\stackrel{!}{=}0 , \label{eqn : stationarys}\\ \frac{{\partial}v}{{\partial}c } & = v\,{\overline{c } } + e^k \big [ -w_c\,\gamma^s + { \overline{w}}\,{\overline{\gamma}}^c + w_c\,{\overline{c}}\,\gamma^c -3w_c{\overline{w } } \big]\stackrel{!}{=}0 . \label{eqn : stationaryc}\end{aligned}\ ] ] in the following we show that eqs.([eqn : stationarys],[eqn : stationaryc ] ) are satisfied at @xmath44 to prove whether the stationary point corresponds to a minimum we have to calculate the hessian matrix for the scalar potential and subsequently evaluate the eigenvalues . in this analysis we can neglect the vacuum energy @xmath45 . up to the factor @xmath46}\sim\mathcal{o}(1)$ ] the second derivatives of @xmath47 are given by @xmath48 one can now consider two cases . for @xmath49 the susy breaking is dominated by the dilaton . however this does not correspond to a satisfactory choice as the contribution from the polonyi sector would then not be sufficient to solve eqs.([eqn : stationarys],[eqn : stationaryc ] ) . therefore it is more interesting to look at the so - called _ matter dominated _ susy breaking where @xmath50 . the leading order term in eqs.([eqn : vss][eqn : vscbar ] ) is @xmath51 . next to leading order terms are @xmath52 and @xmath53 . all other contributions are sub - leading and can be neglected in this analysis . in the matrix element @xmath54 contributes at the next to leading order . we however neglect it here to keep the hessian as simple as possible in order to obtain transparent equations for the eigenvalues . ] thus we arrive at @xmath55 with @xmath56 where @xmath57 . the indices of the hessian are defined as @xmath58 . the eigenvalues are given by @xmath59 and are all positive . this proves that the stationary point is a local minimum . moreover the spectrum consists of two heavy states with masses of order @xmath60 and two light states with masses of order @xmath61 . in consideration of a vanishing / small cosmological constant eq.([eqn : newpot ] ) can be parameterized as @xmath62 where @xmath63 is a fine - tuning parameter and @xmath64 . then the vacuum energy is @xmath65 the solution to first order in @xmath63 reads @xmath66 by choosing @xmath40 , which sets the scale of the polonyi field , one obtains @xmath67 , implying that the system under consideration can be used to construct a minkowski minimum ( or adjusting the vacuum energy to a small value ) . an example is presented in figs.[fig : sdirection ] and [ fig:3dplot ] . let us now have a look at the gravitino mass . it is given by @xmath68 and the polonyi part in the superpotential dominates . so we see that @xmath40 controls not only the mass of the polonyi field but also the gravitino mass . furthermore , for the dilaton we obtain @xmath69 the quantity @xmath70 can be estimated as follows . the stationary point conditions for the scalar potential eq.([eqn : newpot ] ) read @xmath71 the first term in each equation vanishes at the minimum and the remaining terms give relations among the derivatives of @xmath28 . the solution is given by @xmath72 with @xmath73 . we then arrive at @xmath74 the choice of a minkowski vacuum ( or small vacuum energy ) requires @xmath75 to have a value that cancels the contribution of @xmath76 in eq.([eqn : combinedfs ] ) and furthermore acts to suppress @xmath77 by @xmath32 , where @xmath78 for the polonyi field we have @xmath79 since @xmath73 . thus @xmath80 turns out to be the dominant part in susy breakdown whereas @xmath77 appears as being suppressed by a factor as given in eq.([eqn : suppression ] ) . the mass of the dilaton field is expressed as @xmath81 where the second derivative of the scalar potential has the behavior @xmath82 and thus @xmath83 and the mass of the dilaton is enhanced compared to @xmath84 . the mass of the polonyi field is @xmath85 which means that it is comparable to the gravitino mass . this is in accordance with the previous discussion , namely @xmath40 sets the scale of the polonyi field as well as that of the gravitino mass . and @xmath36 direction with minkowski minimum at @xmath87 . ] and @xmath36 direction with minkowski minimum at @xmath87 . ] if we compare this to the type iib case we , in fact , end up with very similar results ( although the starting point was quite different ) . in type iib one started with a supersymmetric theory in an anti - de sitter vacuum which then gets up - lifted to a minkowski vacuum with susy breakdown induced by the up - lifting sector . in the heterotic case we started with an unstabilized dilaton . the superpotential interaction involving matter fields provides the stabilization of the dilaton and also induces the breakdown of supersymmetry . finally a minkowski vacuum can be realized by properly adjusting the parameters of the polonyi sector . the hierarchical structure among @xmath88 , @xmath84 and gaugino masses @xmath89 is @xmath90 this is similar as the result obtained in the type iib case @xcite . in both cases , this little hierarchy has its origin in the appearance of the factor @xmath91 . it suppresses the modulus contribution to the soft mass terms such that loop induced effects become competitive to the tree level ones . the result is the so - called mirage pattern of soft mass terms . having presented the set - up we would like to analyze in detail the pattern of the emerging soft terms in the effective low energy theory . first we derive the soft breaking terms as boundary conditions valid at the grand unified theory ( gut ) scale with a suitable parameterization . we then evolve these soft terms down to the electroweak ( ew ) scale and impose several phenomenological constraints of theoretical and experimental nature . our analysis in section [ sub : combination ] was done in the framework of the polonyi model . following the lines of @xcite one can also construct generalized superpotentials for which the soft terms take a simpler form . in particular we are interested in a simple pattern for the gaugino masses . for a more general discussion see @xcite . in order to simplify the discussion we thus assume that the minimum of the scalar potential emerges at @xmath87 and @xmath92 . the full khler potential including interaction with observable matter fields is given by @xmath93 , \label{eqn : fullkahler}\end{aligned}\ ] ] where @xmath94 are the visible sector matter fields and @xmath95 describe the coupling between observable and hidden matter . furthermore we assume that the string threshold corrections to the gauge kinetic function involving the @xmath36 field are negligible at @xmath96 . the moduli / dilaton mediated contribution to the soft terms is then given by @xmath97 , \label{eqn : rawbc2 } \\ m^2_i & = m^2_{3/2 } - { \overline{f}}^{{\overline{\alpha } } } f^\beta { \partial}_{{\overline{\alpha}}}{\partial}_\beta \log\mathcal{k}_i , \label{eqn : rawbc3}\end{aligned}\ ] ] where @xmath19 and @xmath20 run over the susy breaking fields , @xmath98 are the gauge kinetic functions , @xmath99 and @xmath100 is the khler metric for the visible fields @xmath101 assuming that @xmath102 are independent of @xmath3 and @xmath36 we obtain @xmath103 with @xmath104 being the derivative of @xmath16 with respect to @xmath3 . the condition for having a minkowski vacuum gives a relation among @xmath77 and @xmath80 , namely , @xmath105 under these assumptions ( minimum at @xmath92 , @xmath106 independent of @xmath36 ) , the tree level soft terms for the gauginos and the @xmath107 parameters are independent of @xmath80 . as discussed in section [ sub : susybreaking ] , @xmath77 is suppressed . thus , we have to worry about loop suppressed contributions to the soft terms coming e.g. from the superconformal anomaly @xcite . the masses squared of the scalars , however , could behave differently . they do contain contributions from @xmath80 which dominate over @xmath77 . nevertheless the @xmath77 contribution as well as the anomaly part may be of interest if we consider @xmath108 . including anomaly mediated contributions into eqs.([eqn : rawbc1]-[eqn : rawbc3 ] ) the gut scale boundary values are given by @xmath109 with @xmath110 being the auxiliary field of the conformal compensator , @xmath111 are the beta function coefficients , @xmath112 gives the anomalous dimension and @xmath113 with @xmath114 being the renormalization scale . more details can be found in @xcite . for the gauginos eq.([eqn : metabc1 ] ) we obtain a similar result as in the type iib picture . they are split at the gut scale according to their beta function coefficients . since the evolution of the gauginos from the gut to the ew scale is governed by the same beta function coefficients the splitting disappears at an intermediate scale , leading to the _ mirage _ unification of the gaugino masses . the form of the @xmath107 terms and the scalar masses is very similar to the type iib case . for @xmath115 , @xmath80 dominates and we obtain matter dominated susy breaking with @xmath116 . the choice @xmath108 would suppress the @xmath80 contribution and make it comparable to the dilaton and anomaly mediated contributions . the problematic feature of anomaly mediation is the potential presence of tachyonic sleptons . due to the mixing between the dilaton and anomaly mediated contributions also the squarks might become tachyonic here . the presence of tachyons can be avoided by appropriately choosing @xmath95 . as already studied in the type iib case , the phenomenology of such a mixed mediation scheme depends crucially on the ratio between modulus / dilaton and anomaly contributions . we introduce the parameterization @xmath117 where @xmath118 measures the relative importance of dilaton and anomaly mediation and @xmath119 sets the scale of the soft mass terms . note that the limit @xmath120 corresponds to pure anomaly mediation whereas @xmath121 is pure dilaton domination . the last term in eq.([eqn : metabc3 ] ) is enhanced by @xmath122 with respect to the other terms . in order to compare its contribution with the remaining terms we use @xmath123 with @xmath124 corresponding to @xmath125 . then , the soft terms eqs.([eqn : metabc1]-[eqn : metabc3 ] ) take the form @xmath126 , \label{eqn : bc1 } \\ a_{ijk } & = m^{\phantom{2 } } \big [ -\varrho + { \left(\gamma_i + \gamma_j + \gamma_k\right ) } \big ] , \label{eqn : bc2 } \\ m^2_i & = m^2 \big [ \xi_i\varrho^2 - \dot{\gamma_i } + 2\varrho{\left(s_0+{\overline{s}}_0\right)}{\partial}_s\gamma_i + \eta^2_i \big ] . \label{eqn : bc3}\end{aligned}\ ] ] let us close this section by pointing out that in the heterotic case the anomaly mediated contributions to the @xmath107 parameters and to the scalar masses squared ( for @xmath127 ) are enhanced compared to the type iib situation , where the modulus mediated contribution contained a factor of 3 originating from @xmath128 as compared to @xmath129 . the soft terms in our set - up are described by two continuous parameters @xmath130 the three quantities @xmath131 and the @xmath132 parameters from the matter sector . in the following we will assume non - universal masses for sfermions and higgses @xcite and denote @xmath133 we use @xmath134 to fix the @xmath135 term . the requirement of correct electroweak symmetry breaking fixes the size of @xmath40 so its sign remains a free parameter . we will use @xmath136{gev}$ ] for the top quark mass and @xmath137 throughout our low energy analysis . for the calculation of the low energy data we use the public codes ` softsusy ` @xcite and ` micromegas ` @xcite . the gaugino soft terms eq.([eqn : bc1 ] ) approximately read @xmath138 the non - universality of the gaugino masses arises from the anomaly mediated contributions which are proportional to the beta function coefficients @xmath139 . at the gut scale the gauginos show a pattern @xmath140 because @xmath141 is suppressed by the negative contribution from anomaly mediation . depending on the value of @xmath118 this negative contribution to @xmath141 might become more or less important . at @xmath142 it might even lead to vanishing gluino mass . . for @xmath143 the squarks as well as sleptons in the particular range of interest@xmath144 appear to be tachyonic implying that the gut scale boundary conditions eqs.([eqn : bc1]-[eqn : bc3 ] ) might be ill - defined . introducing nonzero @xmath145 and @xmath146 provides positive contribution to the scalar masses squared . ] as already analyzed in the type iib case the scalar masses squared could become tachyonic due to the contributions from anomaly mediation . in pure anomaly mediation sleptons are tachyonic . due to the mixing between dilaton and anomaly mediation also squarks might become tachyonic here ( fig.[fig : inputscalars ] ) . for small @xmath118 the dilaton mediated contribution is too weak to cancel the negative anomaly contribution . nevertheless we can use the @xmath145 and @xmath146 parameters to avoid this problem . in oder to study a tachyon - free set - up for all values of @xmath118 we scan over @xmath147 and exclude tachyonic regions . this implies lower bounds on @xmath148 and @xmath149 . in fact , me might choose the values of @xmath145 and @xmath146 in such a way that the so - called `` mssm hierarchy problem '' can be avoided . correct electroweak symmetry breakdown ( ewsb ) requires @xmath150 at the ew scale . using the input soft terms eqs.([eqn : bc1]-[eqn : bc3 ] ) this condition can be rewritten at one - loop level as @xcite @xmath151 where we have considered @xmath152 and neglected terms with smaller numerical coefficients . if all the parameters on the right hand side of eq.([eqn : oneloopcondition ] ) are of order of magnitude of , no significant fine - tuning is needed . however , the soft masses are typically in the tev region . then , in order to obtain the correct value of @xmath153 , cancellations in eq.([eqn : oneloopcondition ] ) are required . one could , of course , adjust @xmath154 such that @xmath153 has the correct value but then @xmath154 might have to be very large . but one might also be interested in a situation where @xmath154 has a value of order of the weak scale . this would then require cancellations inside @xmath155 . the largest contribution to @xmath155 comes from the gluino . in order to keep @xmath155 small one would have to keep @xmath141 under control @xcite . as we have seen in section [ sub : gauginos ] the gluino mass might become small for small @xmath118 ( at @xmath156 it might even vanish ) . this is ruled out , as the gluino would be the lightest supersymmetric particle ( lsp ) . in addition if the gluino is light it can not provide the necessary renormalization group ( rg ) contribution to @xmath157 such that eq.([eqn : ewsbcondition ] ) will no longer be satisfied and consequently ewsb will not be realized around @xmath156 . thus larger values of @xmath118 are required . in order to achieve a cancellation within @xmath155 for moderate values of @xmath118 one has to adjust the masses of the scalars ( sfermions and higgses ) . here the freedom of choosing @xmath145 and @xmath146 enters the game . as is evident from eq.([eqn : oneloopcondition ] ) the contribution from @xmath157 is negative and thus by increasing @xmath157 one obtains a sizeable term that could cancel the contribution of the gluino @xmath141 . the contribution from squarks is positive and one has to keep their masses low . however , we can not choose @xmath145 too small , otherwise the squarks might become tachyonic at the gut scale . the essential lesson we learn from these considerations is to keep @xmath145 as low as possible and then adjust @xmath146 . if we want to keep @xmath158{gev})^2 $ ] a relation between the values of @xmath118 , @xmath84 , @xmath145 and @xmath146 has to be fulfilled . in that sense the `` mssm hierarchy problem '' can be avoided at the expense of a fine tuning of @xmath146 . after having excluded tachyons , we shall see that the parameter space of our set - up is further restricted by phenomenological constraints . these include the quest for correct ewsb , mass bounds from lep and the cosmological relic abundance of neutralino dark matter . * _ correct ewsb _ + the minimization of the mssm higgs scalar potential leads to eq.([eqn : ewsbcondition ] ) . here @xmath40 should be positive and @xmath157 should be negative at the ew scale . given a positive @xmath157 at the gut scale , it will be driven to negative values at the ew scale by the renormalization group evolution according to @xmath159 the rg evolution is most sensitive to the gluino mass which induces an increase of @xmath160 and @xmath161 . in a mirage mediation scheme as the one considered here a cancellation between dilaton and anomaly mediated contributions for the gluino mass occurs for small values of @xmath118 leading to an ultra - light gluino around @xmath156 . there the rg contribution from the gluino to eq.([eqn : higgsrg ] ) is too small and a satisfactory value of @xmath157 can not be obtained . _ the requirement of correct ewsb sets a lower bound on the @xmath118 parameter . _ * _ lep mass bounds _ + direct collider searches set lower bounds on the sparticle spectrum and higgs masses . most important and restrictive bounds are due to the lightest higgs boson mass @xmath162{gev}$ ] , the lightest chargino @xmath163{gev}$ ] and the lightest stop quark @xmath164{gev}$ ] @xcite . regions of parameter space violating one of these bounds are called _ below lep_. _ these constraints set a lower bound on @xmath84 . _ * _ neutralino dark matter _ + in susy models the weakly interacting neutralinos tend to be the lsps and they are perfect dark matter candidates ( under the assumption of @xmath165-parity conservation ) . in our model this is true throughout most of the parameter space . the four neutralinos of the mssm @xmath166 are superpositions of the neutral higgs fermions @xmath167 , @xmath168 and the fermionic partners of the ew gauge bosons @xmath169 , @xmath170 . the neutralino mass matrix can be diagonalized by an orthogonal matrix @xmath171 , such that the lightest neutralino is given by @xmath172 using this decomposition one defines @xmath173 we use the @xmath174 limit from the wmap collaboration on the neutralino dark matter relic abundance @xcite : @xmath175 we will require that the neutralinos annihilate efficiently enough to satisfy the bound eq . ( [ eqn : wmap ] ) and assume that the lsp abundance is thermal . in the remainder of this paper the regions of the parameter space that violate the upper wmap bound are called _ forbidden _ , those within the bounds are called _ favored _ and those below the lower bound are denoted as _ allowed_. in the latter case the correct cosmological abundance of dark matter could be achieved with additional dark matter particles and/or a nonthermal origin . in models where the contributions from modulus and anomaly mediation are comparable we experience the phenomenon of mirage unification at an intermediate scale . the same happens , of course , in the present case of mixed dilaton - anomaly mediation . * _ gauginos _ + below the gut scale the nonuniversality of the gaugino masses is given by the respective beta function coefficients . the renormalization group running of the gaugino masses is governed by the same coefficients , thus at an intermediate scale , the splitting disappears and the gauginos unify . since there is no physical threshold associated to this scale it is called _ mirage scale_. using the parameterization eq.([eqn : parameterrho ] ) the mirage scale is given by @xmath176 for @xmath177 the mirage scale is intermediate while for @xmath178 mirage unification occurs at the tev scale . the pattern of the gaugino masses at the gut scale is @xmath140 . at the ew scale this pattern becomes inverted @xmath179 and a compressed spectrum is obtained ( fig.[fig : mirage1 ] ) . * _ first and second generation scalars _ + + and @xmath180 seem to have a mirage unification point which is well approximated by eq.([eqn : miragescale ] ) and is robust ( for low @xmath134 ) under the variation of the parameters . ] + these sparticles behave in a similar way as the gauginos . the reason for this is that they are unaffected by the large yukawa coupling @xmath181 and string threshold corrections as well as khler anomalies are negligible . the renormalization group running of the first and second generation scalars is given by @xmath182 under these circumstances the first and second generation scalars unify at the same @xmath183 eq.([eqn : miragescale ] ) as the gauginos ( fig.[fig : mirage1 ] ) . * _ third generation scalars _ + here we have to distinguish between those whose rg running depends on the yukawa coupling @xmath181 and those that are unaffected by @xmath181 .. for large values of @xmath134 we also have to take into account the bottom quark yukawa coupling @xmath184 . ] the only third generation scalars that feel the effect of @xmath181 are @xmath185 , @xmath186 and @xmath157 . consequently they do not unify at the mirage scale eq.([eqn : miragescale ] ) . all other third generation scalars and @xmath187 are only affected by the smaller yukawa couplings @xmath188 and the structure of their rg running ( fig.[fig : mirage2 ] ) is very similar to eq.([eqn : rgfirstsecond ] ) , thus they ( partially ) share the mirage unification feature . for @xmath152 and@xmath190 . in the white region the low energy spectra are consistent with experimental and theoretical constraints . the brown strip satisfies the @xmath174 wmap constraint . the constraint from @xmath191 @xcite does not further restrict the allowed range of parameters . ] but with @xmath192 . the lep constraints are less restrictive whereas the chargino - lsp and the no - ewsb regions increase with larger @xmath134 . compared to fig.[fig : lowtan ] larger portions of the parameter space are attractive for the discussion of the relic abundance . ] as an illustrative example let us consider @xmath193 and @xmath194 . the corresponding parameter space is shown in fig . [ fig : lowtan ] . in this scenario there are no tachyons . however , correct ewsb and current lep bounds for @xmath195 , @xmath196 and @xmath197 put severe constraints on @xmath118 and @xmath84 . particularly we find that for @xmath152 only @xmath198 and @xmath199{tev}$ ] are allowed . the presence of the no - ewsb region appears because at @xmath142 the gluino contribution to the rg is too small to make @xmath157 negative at the ew scale . in contrast to type iib , in the heterotic case we find a region in the parameter space where a chargino is the lsp . this happens because the @xmath107 terms in the heterotic case are smaller than in type iib . terms is reduced by a factor of 3 . ] additionally , these reduced @xmath107 terms lead to a smaller intra - generational mixing and increase the masses of the stop quark and the stau lepton . therefore , stop or stau ( n)lsp are not realized in this scenario . for @xmath118 values close to the no - ewsb region we have @xmath200 and the neutralino lsp is higgsino - like . going to larger @xmath118 values the lsp becomes a mixed higgsino - bino state . from @xmath201 we have a mostly bino - like lsp and also @xmath202 . thus , in most of the parameter space the lsp is bino - like . the shaded/ strip in fig.[fig : lowtan ] shows the region of the parameter space which is _ favoured _ by the wmap results eq.([eqn : wmap ] ) . the region below the shaded/ strip is _ allowed _ ( lower abundance ) and that above ( and to the right ) of the shaded/ strip is _ forbidden _ ( too large relic abundance ) . for @xmath203{tev}$ ] and @xmath204 we are close to the @xmath205 lsp region and therefore we have @xmath206 coannihilation , which enhances the annihilation cross section and lowers the relic abundance . the neutralino in this region is higgsino - like . as @xmath118 increases , the neutralino becomes mixed higgsino - bino and the @xmath154 term increases . the coannihilation with the chargino gets reduced and the annihilation cross section decreases leading to a higher relic abundance , so that there ( @xmath207 ) we reach the shaded/ strip . when we proceed to increase @xmath118 , the @xmath208 becomes bino - like . however , we then reach @xmath209 and there the annihilation proceeds efficiently through the pseudo - scalar higgs exchange @xmath210 . this enhances the annihilation cross section and reduces the relic abundance . for @xmath211 the mass gap @xmath212 grows and the efficiency of the @xmath107-channel reduces . as there are no other coannihilation channels available the cross section decreases and the relic abundance becomes too large . for large values of @xmath134 , the lep mass constraint becomes less restrictive ( fig . [ fig : hightan ] ) . however , the no - ewsb region gets slightly bigger and the @xmath205 lsp region covers a larger part of the parameter space ( compared to the case of small @xmath134 ) . the composition of the neutralino lsp is similar to the @xmath152 situation . for low @xmath118 values ( close to the no - ewsb region ) the neutralino is higgsino - like . then , for larger @xmath118 it becomes more and more bino - like . the shaded/ strip , satisfying the wmap limits , differs significantly from the one discussed above . now , a larger part of the parameter space is consistent with the correct amount of dark matter . this is because for large @xmath134 the pseudo - scalar higgs exchange provides a sizable contribution to the annihilation cross section . for @xmath213{tev}$ ] and @xmath214 we have @xmath215 and chargino coannihilation enhances the annihilation cross section and lowers the relic abundance . when @xmath118 increases , @xmath154 gets larger ( the whole sparticle spectrum becomes heavier ) and the @xmath216 becomes bino - like . for @xmath217 the mass gap @xmath218 grows , thus the @xmath219 coannihilation channel no longer provides a sizable effect . as a result , the annihilation cross section decreases and the relic abundance grows above the upper wmap bound . at the same time the mass of the @xmath216 approaches the value @xmath220 and thus the pseudo - scalar higgs exchange begins to contribute . the cross section @xmath221 grows with @xmath222 and so this channel overcomes the decrease of the annihilation cross section caused by the bino component of the neutralino . for @xmath223 the production of the relic abundance is in the allowed range . for still larger @xmath118 finally this effect dies out and the relic abundance becomes too large . some points are selected from the the allowed parameter space in figs.[fig : lowtan ] and [ fig : hightan ] and the spectrum is analysed in detail . examples of the spectra are displayed in table [ tab : samplepoints ] . w2.5cmz1.2cmz1.2cmz1.2 cm & * a * & * b * & * c * + @xmath134 & 5 & 30 & 10 + @xmath118 & 6 & 10 & 6.5 + @xmath84 & 40 & 6 & 25 + @xmath145 & 4 & 4 & 5 + @xmath146 & 6 & 6 & 5 + @xmath224 & 1.040 & 0.211 & 0.676 + @xmath225 & 1.317 & 0.311 & 0.881 + @xmath141 & 2.391 & 0.743 & 1.697 + ( r)1 - 1(l)2 - 4 @xmath195 & 0.119 & 0.115 & 0.120 + @xmath226 & 2.118 & 0.468 & 1.408 + @xmath154 & 0.860 & 0.413 & 0.885 + ( r)1 - 1(l)2 - 4 @xmath227 & 0.850 & 0.204 & 0.665 + @xmath228 & 0.870 & 0.296 & 0.838 + @xmath229 & 0.855 & 0.296 & 0.836 + ( r)1 - 1(l)2 - 4 @xmath230 & 1.610 & 0.488 & 1.236 + @xmath231 & 2.110 & 0.694 & 1.578 + ( r)1 - 1(l)2 - 4 @xmath232 & 1.398 & 0.233 & 1.021 + @xmath233 & 1.522 & 0.347 & 1.112 + ( r)1 - 1(l)2 - 4 @xmath234 & 0.088 & 0.115 & 0.092 + as we have seen , dilaton stabilization in the framework of the heterotic string can be achieved quite easily , if we accept the existence of an up - lifting sector ( as postulated previously in type iib theory ) which in any case is needed to adjust the vacuum energy to an acceptable value . one just needs a gaugino condensate , while nontrivial background flux and/or nonperturbative corrections to the khler potential are not necessarily required . in that sense the heterotic mechanism of dilaton stabilization is somewhat similar to the one conjectured in the framework of m theory on @xmath235-manifolds ( which requires a racetrack scenario and an up - lifting sector @xcite ) . a comparison of these two scenarii will be the subject of future work , where we shall also investigate heterotic m theory with a gaugino condensate @xcite . such a ( partially sequestered ) up - lifting sector is thus common to many of the string schemes considered so far . while its exact origin has to be clarified ( branes , antibranes etc . ) , its importance for the resulting phenomenology can not be overestimated . it is the dominant source of supersymmetry breakdown , but as it is ( partially ) sequestered it leads to mediation schemes where tree - level moduli contributions compete with loop - effects from the up - lifting sector . the gravitino and the moduli fields become rather heavy , but the soft terms of the mssm particles are suppressed by a factor of the order of @xmath236 @xcite . these soft masses often show a characteristic pattern , known as the _ mirage pattern_. it leads to a rather compressed spectrum of masses and seems to be especially robust in the case of gauginos @xcite . this specific pattern has been investigated thoroughly in the framework of the type iib string . given the particularly successful attempts of realistic mssm model building in the heterotic theory @xcite , it is reassuring to see that a similar patterns seems to emerge here as well . we would like to thank oleg lebedev for valuable discussions . this work was partially supported by the european union 6th framework program mrtn - ct-2004 - 503069 `` quest for unification '' , mrtn - ct-2004 - 005104 `` forcesuniverse '' , mrtn - ct-2006 - 035863 `` universenet '' and sfb - transregio 33 `` the dark universe '' by deutsche forschungsgemeinschaft ( dfg ) . o. lebedev , h. p. nilles , s. raby , s. ramos - sanchez , m. ratz , k. s. vaudrevange , and a. wingerter , `` a mini - landscape of exact mssm spectra in heterotic orbifolds , '' _ phys . _ * b645 * ( 2007 ) 8894 , http://arxiv.org/abs/hep-th/0611095[hep-th/0611095 ] . o. lebedev , h. p. nilles , s. raby , s. ramos - sanchez , m. ratz , k. s. vaudrevange , and a. wingerter , `` the heterotic road to the mssm with r parity , '' http://dx.doi.org/10.1103/physrevd.77.046013[_phys . rev . _ * d77 * ( 2008 ) 046013 ] , http://arxiv.org/abs/0708.2691[arxiv:0708.2691 [ hep - th ] ] . k .- s . choi and j. e. kim , `` quarks and leptons from orbifolded superstring , '' _ springer : lecture notes in physics _ * 696 * ( 2006 ) 1 . derendinger , l. e. ibanez , and h. p. nilles , `` on the low - energy d = 4 , n=1 supergravity theory extracted from the d=10 , n=1 superstring , '' _ phys . lett . _ * b155 * ( 1985 ) 65 . m. dine , r. rohm , n. seiberg , and e. witten , `` gluino condensation in superstring models , '' _ phys . * b156 * ( 1985 ) 55 . j. p. derendinger , l. e. ibanez , and h. p. nilles , `` on the low - energy limit of superstring theories , '' _ nucl . phys . _ * b267 * ( 1986 ) 365 . n. v. krasnikov , `` on supersymmetry breaking in superstring theories , '' _ phys . * b193 * ( 1987 ) 3740 . j. a. casas , z. lalak , c. munoz , and g. g. ross , `` hierarchical supersymmetry breaking and dynamical determination of compactification parameters by nonperturbative effects , '' _ nucl . * b347 * ( 1990 ) 243269 . j. a. casas , `` the generalized dilaton supersymmetry breaking scenario , '' _ phys . * b384 * ( 1996 ) 103110 , http://arxiv.org/abs/hep-th/9605180[hep-th/9605180 ] . p. binetruy , m. k. gaillard , and y .- y . wu , `` dilaton stabilization in the context of dynamical supersymmetry breaking through gaugino condensation , '' _ nucl . * b481 * ( 1996 ) 109128 , http://arxiv.org/abs/hep-th/9605170[hep-th/9605170 ] . t. barreiro , b. de carlos , and e. j. copeland , `` on non - perturbative corrections to the kaehler potential , '' _ phys . * d57 * ( 1998 ) 73547360 , http://arxiv.org/abs/hep-ph/9712443[hep-ph/9712443 ] . k. dasgupta , g. rajesh , and s. sethi , `` m theory , orientifolds and g - flux , '' _ jhep _ * 08 * ( 1999 ) 023 , http://arxiv.org/abs/hep-th/9908088[hep-th/9908088 ] . s. b. giddings , s. kachru , and j. polchinski , `` hierarchies from fluxes in string compactifications , '' _ phys . _ * d66 * ( 2002 ) 106006 , http://arxiv.org/abs/hep-th/0105097[hep-th/0105097 ] . k. becker , m. becker , k. dasgupta , and p. s. green , `` compactifications of heterotic theory on non - kaehler complex manifolds . i , '' _ jhep _ * 04 * ( 2003 ) 007 , http://arxiv.org/abs/hep-th/0301161[hep-th/0301161 ] . g. lopes cardoso , g. curio , g. dallagata , and d. lust , `` heterotic string theory on non - kaehler manifolds with h- flux and gaugino condensate , '' _ fortsch . * 52 * ( 2004 ) 483488 , http://arxiv.org/abs/hep-th/0310021[hep-th/0310021 ] . s. gurrieri , a. lukas , and a. micu , `` heterotic on half - flat , '' _ phys . * d70 * ( 2004 ) 126009 , http://arxiv.org/abs/hep-th/0408121[hep-th/0408121 ] . s. kachru , r. kallosh , a. linde , and s. p. trivedi , `` de sitter vacua in string theory , '' _ phys . * d68 * ( 2003 ) 046005 , http://arxiv.org/abs/hep-th/0301240[hep-th/0301240 ] . k. choi , a. falkowski , h. p. nilles , m. olechowski , and s. pokorski , `` stability of flux compactifications and the pattern of supersymmetry breaking , '' _ jhep _ * 11 * ( 2004 ) 076 , http://arxiv.org/abs/hep-th/0411066[hep-th/0411066 ] . k. choi , a. falkowski , h. p. nilles , and m. olechowski , `` soft supersymmetry breaking in kklt flux compactification , '' _ nucl . * b718 * ( 2005 ) 113133 , http://arxiv.org/abs/hep-th/0503216[hep-th/0503216 ] . k. choi , k. s. jeong , and k .- okumura , `` phenomenology of mixed modulus - anomaly mediation in fluxed string compactifications and brane models , '' _ jhep _ * 09 * ( 2005 ) 039 , http://arxiv.org/abs/hep-ph/0504037[hep-ph/0504037 ] . m. endo , m. yamaguchi , and k. yoshioka , `` a bottom - up approach to moduli dynamics in heavy gravitino scenario : superpotential , soft terms and sparticle mass spectrum , '' http://dx.doi.org/10.1103/physrevd.72.015004[_phys . * d72 * ( 2005 ) 015004 ] , http://arxiv.org/abs/hep-ph/0504036[arxiv:hep-ph/0504036 ] . a. falkowski , o. lebedev , and y. mambrini , `` susy phenomenology of kklt flux compactifications , '' _ jhep _ * 11 * ( 2005 ) 034 , http://arxiv.org/abs/hep-ph/0507110[hep-ph/0507110 ] . h. baer , e .- k . park , x. tata , and t. t. wang , `` collider and dark matter searches in models with mixed modulus - anomaly mediated susy breaking , '' _ jhep _ * 08 * ( 2006 ) 041 , http://arxiv.org/abs/hep-ph/0604253[hep-ph/0604253 ] . o. lebedev , h. p. nilles , s. raby , s. ramos - sanchez , m. ratz , k. s. vaudrevange , and a. wingerter , `` low energy supersymmetry from the heterotic landscape , '' _ phys . * 98 * ( 2007 ) 181602 , http://arxiv.org/abs/hep-th/0611203[hep-th/0611203 ] . o. lebedev , h. p. nilles , and m. ratz , `` de sitter vacua from matter superpotentials , '' _ phys . * b636 * ( 2006 ) 126 , http://arxiv.org/abs/hep-th/0603047[hep-th/0603047 ] . m. gomez - reino and c. a. scrucca , `` locally stable non - supersymmetric minkowski vacua in supergravity , '' _ jhep _ * 05 * ( 2006 ) 015 , http://arxiv.org/abs/hep-th/0602246[hep-th/0602246 ] . d. lust , s. reffert , e. scheidegger , w. schulgin , and s. stieberger , `` moduli stabilization in type iib orientifolds . ii , '' _ nucl . phys . _ * b766 * ( 2007 ) 178231 , http://arxiv.org/abs/hep-th/0609013[hep-th/0609013 ] . e. dudas , c. papineau , and s. pokorski , `` moduli stabilization and uplifting with dynamically generated f - terms , '' _ jhep _ * 02 * ( 2007 ) 028 , http://arxiv.org/abs/hep-th/0610297[hep-th/0610297 ] . h. abe , t. higaki , t. kobayashi , and y. omura , `` moduli stabilization , f - term uplifting and soft supersymmetry breaking terms , '' _ phys . _ * d75 * ( 2007 ) 025019 , http://arxiv.org/abs/hep-th/0611024[hep-th/0611024 ] . o. lebedev , v. lowen , y. mambrini , h. p. nilles , and m. ratz , `` metastable vacua in flux compactifications and their phenomenology , '' _ jhep _ * 02 * ( 2007 ) 063 , http://arxiv.org/abs/hep-ph/0612035[hep-ph/0612035 ] . m. serone and a. westphal , `` moduli stabilization in meta - stable heterotic supergravity vacua , '' _ jhep _ * 08 * ( 2007 ) 080 , http://arxiv.org/abs/arxiv:0707.0497 [ hep - th][arxiv:0707.0497 [ hep - th ] ] . k. intriligator , n. seiberg , and d. shih , `` dynamical susy breaking in meta - stable vacua , '' _ jhep _ * 04 * ( 2006 ) 021 , http://arxiv.org/abs/hep-th/0602239[hep-th/0602239 ] . j. polchinski , `` string theory . 1 : an introduction to the bosonic string , '' . cambridge , uk : univ . ( 1998 ) 402 p. k. choi and h. p. nilles , `` the gaugino code , '' _ jhep _ * 04 * ( 2007 ) 006 , http://arxiv.org/abs/hep-ph/0702146[hep-ph/0702146 ] . l. randall and r. sundrum , `` out of this world supersymmetry breaking , '' _ nucl . * b557 * ( 1999 ) 79118 , http://arxiv.org/abs/hep-th/9810155[hep-th/9810155 ] . d. matalliotakis and h. p. nilles , `` implications of nonuniversality of soft terms in supersymmetric grand unified theories , '' _ nucl . * b435 * ( 1995 ) 115128 , http://arxiv.org/abs/hep-ph/9407251[hep-ph/9407251 ] . m. olechowski and s. pokorski , `` electroweak symmetry breaking with nonuniversal scalar soft terms and large tan beta solutions , '' _ phys . lett . _ * b344 * ( 1995 ) 201210 , http://arxiv.org/abs/hep-ph/9407404[hep-ph/9407404 ] . b. c. allanach , `` softsusy : a c++ program for calculating supersymmetric spectra , '' _ comput . * 143 * ( 2002 ) 305331 , http://arxiv.org/abs/hep-ph/0104145[hep-ph/0104145 ] . g. belanger , f. boudjema , a. pukhov , and a. semenov , `` micromegas : a program for calculating the relic density in the mssm , '' _ comput . * 149 * ( 2002 ) 103120 , http://arxiv.org/abs/hep-ph/0112278[hep-ph/0112278 ] . g. l. kane , j. d. lykken , b. d. nelson , and l .- t . wang , `` re - examination of electroweak symmetry breaking in supersymmetry and implications for light superpartners , '' _ phys . lett . _ * b551 * ( 2003 ) 146160 , http://arxiv.org/abs/hep-ph/0207168[hep-ph/0207168 ] . k. choi , k. s. jeong , t. kobayashi , and k .- okumura , `` little susy hierarchy in mixed modulus - anomaly mediation , '' _ phys . _ * b633 * ( 2006 ) 355361 , http://arxiv.org/abs/hep-ph/0508029[hep-ph/0508029 ] . o. lebedev , h. p. nilles , and m. ratz , `` a note on fine - tuning in mirage mediation , '' http://arxiv.org/abs/hep-ph/0511320[hep-ph/0511320 ] . collaboration , w. m. yao _ et al . _ , `` review of particle physics , '' _ _ * g33 * ( 2006 ) 11232 . collaboration , d. n. spergel _ et al . _ , `` wilkinson microwave anisotropy probe ( wmap ) three year results : implications for cosmology , '' _ astrophys . j. suppl . _ * 170 * ( 2007 ) 377 , http://arxiv.org/abs/astro-ph/0603449[astro-ph/0603449 ] . collaboration , s. chen _ et al . _ , `` branching fraction and photon energy spectrum for @xmath191 , '' _ phys . * 87 * ( 2001 ) 251807 , http://arxiv.org/abs/hep-ex/0108032[hep-ex/0108032 ] . b. acharya , k. bobkov , g. kane , p. kumar , and d. vaman , `` an m theory solution to the hierarchy problem , '' _ phys . rev . * 97 * ( 2006 ) 191601 , http://arxiv.org/abs/hep-th/0606262[hep-th/0606262 ] . b. s. acharya , k. bobkov , g. l. kane , p. kumar , and j. shao , `` explaining the electroweak scale and stabilizing moduli in m theory , '' _ phys . * d76 * ( 2007 ) 126010 , http://arxiv.org/abs/hep-th/0701034[hep-th/0701034 ] . horava , `` gluino condensation in strongly coupled heterotic string theory , '' _ phys . * d54 * ( 1996 ) 75617569 , http://arxiv.org/abs/hep-th/9608019[hep-th/9608019 ] . h. p. nilles , m. olechowski , and m. yamaguchi , `` supersymmetry breaking and soft terms in m - theory , '' _ phys . * b415 * ( 1997 ) 2430 , http://arxiv.org/abs/hep-th/9707143[hep-th/9707143 ] . o. loaiza - brito , j. martin , h. p. nilles , and m. ratz , `` @xmath237 ) , '' _ aip conf . * 805 * ( 2006 ) 198204 , http://arxiv.org/abs/hep-th/0509158[hep-th/0509158 ] . t. kobayashi , s. raby , and r .- j . zhang , `` constructing 5d orbifold grand unified theories from heterotic strings , '' _ phys . _ * b593 * ( 2004 ) 262270 , http://arxiv.org/abs/hep-ph/0403065[hep-ph/0403065 ] . s. forste , h. p. nilles , p. k. s. vaudrevange , and a. wingerter , `` heterotic brane world , '' _ phys . * d70 * ( 2004 ) 106008 , http://arxiv.org/abs/hep-th/0406208[hep-th/0406208 ] . w. buchmuller , k. hamaguchi , o. lebedev , and m. ratz , `` supersymmetric standard model from the heterotic string , '' _ phys . * 96 * ( 2006 ) 121602 , http://arxiv.org/abs/hep-ph/0511035[hep-ph/0511035 ] . j. e. kim , j .- h . kim , and b. kyae , `` superstring standard model from z(12-i ) orbifold compactification with and without exotics , and effective r- parity , '' _ jhep _ * 06 * ( 2007 ) 034 , http://arxiv.org/abs/hep-ph/0702278[hep-ph/0702278 ] .

utilizing large - scale monte - carlo simulations , we investigate an unconventional two - component classical plasma in two dimensions which controls the behavior of the norms and overlaps of the quantum - mechanical wavefunctions of ising - type quantum hall states . the plasma differs fundamentally from that which is associated with the two - dimensional xy model and abelian fractional quantum hall states . we find that this unconventional plasma undergoes a berezinskii - kosterlitz - thouless phase transition from an insulator to a metal . the parameter values corresponding to ising - type quantum hall states lie on the metallic side of this transition . this result verifies the required properties of the unconventional plasma used to demonstrate that ising - type quantum hall states possess quasiparticles with non - abelian braiding statistics .

1111.0135

key properties of physical systems can sometimes be understood by mapping them to seemingly unrelated ones . a powerful example of this was provided by laughlin , who observed that the squared norm of his @xmath0 fractional quantum hall trial wavefunction @xmath1 ( where @xmath2 is a complex coordinate in the two - dimensional plane ) could be expressed as the boltzmann weight of a two - dimensional one - component plasma @xcite : @xmath3 where @xmath4 and @xmath5 . this mapping allows properties such as quasiparticle charge and braiding statistics to be determined by appealing to the known properties of a one - component plasma . recently , a similar plasma mapping was established @xcite for ising - type quantum hall states , such as the moore - read ( mr ) @xcite , anti - pfaffian @xcite , and bonderson - slingerland ( bs ) hierarchy @xcite states , which are likely candidates to describe hall plateaus in the second landau level , in particular at filling fraction @xmath6 @xcite . in this case , the mapping is to a two - dimensional ( 2d ) two - component plasma , where the two species of particles , @xmath7 and @xmath8 , carry not only different values of charge , but also interact through two different interactions , both of the coulomb form , so the potential energy is : @xmath9 @xmath10 where @xmath11 . the @xmath8-particles interact with each other through the first coulomb - like interaction , @xmath12 , given in eq . ( and so does not depend on the @xmath13 coordinates ) . moreover , the @xmath8-particles interact with each other and with the @xmath7-particles through the second coulomb - like interaction , through which the @xmath7-particles also interact with each other , according to @xmath14 , given in eq . . note that @xmath14 is the 2d coulomb potential of the usual two - component plasma ( where the two species carry charge @xmath15 and @xmath16 , respectively ) . the @xmath8-particles carry charge @xmath17 for the first interaction and charge @xmath15 for the second interaction . the @xmath7-particles carry charge @xmath18 for the first interaction and charge @xmath16 for the second interaction . for a plasma with @xmath19 particles of each species , neutrality is satisfied using a uniform background density of type 1 charge , as in the second term in eq . . this unconventional plasma may be considered as an ordinary neutral two - component gas with positive and negative charges of magnitude @xmath15 , where the positive charges are given an additional charge of @xmath17 that is only felt by the other positive charges and not the negative charges . an illustration of the interactions between the two species in the system is shown in fig . [ fig : interactions ] . -particles only interact by the second coulomb - like interaction with charge @xmath16 , whereas the @xmath8-particles carry charge @xmath17 for the first coulomb - like interaction and @xmath15 for the second coulomb - like interaction . thus , the intraspecies interaction among the @xmath7-particles , shown in ( a ) , and the interspecies interaction between @xmath7 and @xmath8-particles , shown in ( b ) , are given by @xmath15 only , whereas the intraspecies interaction among the @xmath8-particles , shown in ( c ) , are determined by @xmath17 in addition to @xmath15 . interactions between the @xmath8-particles and the neutralizing background are omitted from the figure . ] we are thus led to consider a class of unconventional plasmas parametrized by @xmath20 and @xmath21 . as mentioned above , for mr ising - type states with filling @xmath0 , the relevant values are @xmath22 and @xmath23 . in this plasma mapping , the @xmath24 particles in the plasma correspond to the electrons in the mr wavefunctions and the @xmath13 particles correspond to screening operators ( fictitious particles ) . the case @xmath25 , @xmath23 is relevant for the plasma mapping @xcite of 2d chiral @xmath26-wave superconductors @xcite . we note that whenever @xmath25 , our model is a special case of the well - known 2d two - component plasma of equal and opposite charges @xcite . the screening properties of multi - component 2d plasmas with multiple coulomb interactions of this kind are also important for other physical systems , such as rotating multi - component bose - einstein condensates with interspecies current - current ( andreev - bashkin ) interaction @xcite and some multi - component superconducting systems @xcite . in these systems the screening properties and phase transitions determine superfluid and rotational responses . in this paper , we fix temperature to @xmath27 and consider the two most significant values of @xmath17 , namely @xmath28 . we investigate the screening and phase transition properties of these plasmas as a function of varying @xmath15 by performing a large - scale monte carlo simulation . here , a `` screening phase '' means that the system has a screening length which is finite , and exponentially decaying effective interactions . a system with logarithmic effective interactions is one where screening is defined to be absent . as a first check , we reproduce the well - known result that , for @xmath29 , there is a berezinskii - kosterlitz - thouless ( bkt ) phase transition at @xmath30 , as expected for a 2d two - component plasma of equal and opposite charges . for @xmath31 , the charges are unbound and the plasma screens , but for @xmath32 , the charges are bound into dipoles and the interaction is not screened . thus , for @xmath33 , the value relevant to 2d chiral @xmath26-wave superconductors , the plasma screens . for @xmath34 , we again find a bkt phase transition at @xmath30 , with a plasma screening phase for @xmath31 . the first coulomb - like interaction is deep within its screening phase and appears to have a negligibly small effect on the screening of the second interaction . in both cases , the critical values @xmath35 are obtained by a finite - size scaling fit of the monte carlo data to the bkt form . our findings demonstrate that the unconventional plasma which occurs in the mapping for both a chiral @xmath26-wave superconductor and the ising - type quantum hall states is clearly in the screening phase ( for both types of coulomb interaction ) and hence allows one to discern the non - abelian braiding properties of these states , as explained in ref . . the outline of this paper is as follows . in the introductory part of section [ sec : model ] , we present the model for the unconventional plasma we will be studying in this paper . in section [ sec : ising_type_qhs ] , we connect this to the ising - type of quantum hall states . in section [ sec : rotating_becs ] , we explain its connection to two - component , two dimensional , bose - einstein condensates . in section [ sec : spherical ] , we present a formulation of the model on a sphere . in section [ sec : mc_details ] , we give details of the monte - carlo simulations , and in section [ sec : results ] , we present our results for the screening properties , as well as our findings for the character of phase transition between the dielectric non - screening phase and the metallic screening phase . in section [ sec : conclusions ] , we present our conclusions . technical details on the derivation of a generalized dielectric constant is given in appendix [ app : general_dielectric_constant ] . in appendix [ app : fourth_order_derivative ] , we give a derivation of a relevant higher order response function that we use to characterize the metal - insulator transition . in appendix [ app : weber_minnhagen ] , we present technical details on the finite - size scaling we have used . the canonical partition function of the unconventional plasma is written @xmath36 where the potential energy @xmath37 is given by the 2d coulombic interactions @xmath38 similar to the study of the 2d two - component neutral coulomb gas @xcite , we introduce a short - range hard - core repulsion between all charges in the system . treating all charges as hard disks with the same diameter @xmath39 that limits the range of the hard - core repulsion , the interaction between charges of the same species is @xmath40 and the interaction between charges of different species is @xmath41 in eq . , @xmath42 are position vectors for the particles of component @xmath7 , and @xmath43 are position vectors for the particles of component @xmath8 . to ensure neutrality , the term @xmath44 includes the interaction of the @xmath17 charges of type 1 for the @xmath8-particles with a neutralizing background charge density . in ref . , this background is a uniform negatively charged 2d disk with charge density @xmath45 , where @xmath46 , that yields @xmath47 the particle - background and the background - background interaction also yields uninteresting constant terms , that are disregarded in eq . . we note that when @xmath29 we have the 2d two - component neutral coulomb plasma , which is well - studied both analytically and numerically @xcite . at low dipole density , this system will undergo a bkt transition , which is a charge - unbinding transition from a low - temperature state where charges of opposite signs form tightly bound dipoles to a high - temperature state in which a finite fraction of charges are not bound in dipoles , but rather form a metallic state . in the low - temperature phase , this coulomb gas is an insulator and the dielectric constant @xmath48 ( see for instance refs . and appendix [ app : general_dielectric_constant ] for a formal definition of @xmath48 ) , is finite . in the high - temperature phase , the existence of free charges , yields a conductive gas with an infinite value of @xmath48 . at the critical temperature @xmath49 , when tightly bound dipoles starts to unbind , there is a universal jump in the inverse dielectric constant from a non - zero value in the insulating phase to zero in the metallic phase , @xmath50 the screening properties that follow are that the coulomb gas is able to perfectly screen test charges in the metallic phase when there are free charges in the system , whereas there is no screening in the insulating dielectric phase . in this work , we will focus our attention on the low dipole density regime , so we will not go into details of the physics in the 2d two - component neutral coulomb gas at higher densities . however , we note that when density is increased , the critical point of the bkt transition is shifted towards lower temperatures @xcite . another well - studied case is when @xmath51 , for which the model reduces to the 2d one - component plasma ( for the @xmath8-particles only ) . early numerical studies of this system found a weak first - order melting transition at @xmath52 from a state where the charges form a triangular lattice with quasi - long - range translational and long - range orientational order to a fluid plasma state @xcite . these results were , in a sense , contrasting with the defect - mediated melting theory of kosterlitz - thouless - halperin - nelson - young ( kthny ) that predicts melting from a solid to a liquid via two bkt - transitions and an intermediate hexatic phase with no translational order and quasi - long - range orientational order @xcite . other studies of 2d melting point in favor of the kthny theory @xcite , suggesting that the nature of melting transition may depend on details in the interatomic potential , or that finite - size effects and lack of equilibration might lead to erroneous conclusions in earlier works . there are also studies that argue for the absence of a phase transition to a low - temperature solid phase in the 2d one - component plasma with repulsive logarithmic interactions because the crystalline state would be unstable to proliferation of screened disclinations for any @xmath53 @xcite . the unconventional 2d two - component plasma studied here is mapped to inner products of trial wavefunctions for the mr quantum hall states using conformal field theory ( cft ) methods , as explained in ref . . in particular , this mapping utilizes the coulomb gas description of cfts @xcite together with a procedure for replacing holomorphic - antiholomorphic pairs of contour integrals in screening charge operators for 2d integrals @xcite . the mr states wavefunctions can be written as a product of correlation functions of fields from the ising and u(1 ) cfts . in particular , the mr ground - state wavefunction for @xmath19 electrons is @xmath54 where the pfaffian of an antisymmetric matrix @xmath55 is given by @xmath56 here , @xmath57 is the symmetric group , @xmath58 is one of the permutation elements in @xmath57 , and @xmath59 is the signature of @xmath58 . the @xmath60 portion of this wavefunction is produced from the correlation function of @xmath61 fields in the ising cft , while the laughlin - type portion @xmath62 is produced from the correlation function of vertex operators in the u(1 ) cft . the laughlin - type portion of the mr wavefunctions can be mapped to charges of type 1 , similar to laughlin s plasma mapping . the mentioned cft methods provide identities such as @xmath63 which allow the pfaffian portion of the mr wavefunctions to be mapped to charges of type 2 . this allows one to write the norm of the mr ground - state wavefunction as the partition function of the unconventional 2d two - component plasma of eq . ( [ eqn : two - coulomb - ints ] ) @xmath64 with @xmath65 and @xmath33 . more generally , one can also construct a similar , but more complicated mapping between inner products of wavefunctions of the mr states with quasiparticles , as explained in ref . . in this case , the quasiparticles map to fixed `` test '' objects in the plasma that carry electric charge of type 1 and can carry both electric and magnetic charges of type 2 ( and also changes the number of screening operators , i.e. @xmath7-particles in the plasma , to maintain neutrality ) . ( the charges of type 1 and 2 carried by the quasiparticles are typically some fractions of the charges @xmath17 and @xmath15 carried by the @xmath8-particles . ) strictly speaking , the right - hand - side of eq . ( [ eq : pf_map ] ) is divergent for @xmath66 ( since the integrand diverges as @xmath67 as a @xmath7-particle approaches a @xmath8-particle ) . it can be made well - defined ( and equal to the left - hand - side ) by replacing @xmath67 with @xmath68 , evaluating the integrals for @xmath69 and then analytically - continuing @xmath70 to @xmath71 . on the other hand , we regularize the divergences of eq . ( [ eq : norm ] ) in this paper by using a hard - core repulsion that forbids the particles from approaching each other closer than a distance @xmath39 , i.e. replacing @xmath37 in this expression with that of eq . ( [ eq : model_s ] ) . it should not matter how we regularize the divergence in eq . ( [ eq : norm ] ) as long as the probability for @xmath8-particles and @xmath7-particles to sit right on top of each other has measure zero . as we will see in this paper , this is true for @xmath72 , in which case the configurational entropy to be gained by having @xmath8-particles and @xmath7-particles separate overcomes the energy gained by having them on top of each other . we refer to this as an `` entropic barrier '' for putting @xmath8-particles and @xmath7-particles on top of each other . in contrast , in eq . ( [ eq : pf_map ] ) , where only the @xmath73s are integrated over and the @xmath24 coordinates are fixed , regularization by a simple hard - core repulsion does not appear to be a suitable alternative to analytic continuation . in this case , since the @xmath24 coordinates are fixed , the entropic barrier is lower . equivalently , there are fewer integrals to compensate the inverse powers . thus , in eq . ( [ eq : pf_map ] ) , a simple hard - core cutoff will not reproduce the left - hand - side , and one must use the analytic continuation procedure described above . in a rotating frame , a bose - einstein condensate in the london limit is described by the uniformly frustrated xy model , @xmath74 ^ 2 , \label{eq : frustrated_3dxy}\ ] ] where @xmath75 for a condensate with mass @xmath76 , phase @xmath77 and density @xmath78 , and @xmath79 where @xmath80 is the angular velocity of the rotation . in 3d , this model is frequently used to describe the melting of vortex - line lattices in extreme type - ii superconductors and superfluids @xcite . by a duality transformation , the model in eq . can be rewritten in terms of vortex fields @xmath81 to yield @xcite @xmath82\frac{\rho}{q^2}\nonumber\\ & \times\left[l(-{\textbf{q}})-(2\pi)^2f\delta(-{\textbf{q}})\right ] , \label{eq : frustrated_3dxy_dual}\end{aligned}\ ] ] where @xmath83 is the vortex number density and @xmath84 is the fundamental quantum unit of vorticity . this is a one - component 2d classical coulomb plasma where charges correspond to nonzero values in the vortex field @xmath85 and the quantity @xmath86 now plays the role as the neutralizing background number density . extending to two components , a model for a rotating two - component bose - einstein condensate with a generic andreev - bashkin drag interaction @xcite reads @xmath87 where now @xmath76 , @xmath78 and @xmath77 is given an index that denotes the component and @xmath88 is the drag density . this model has recently been studied in three dimensions @xcite . by a duality transformation , we arrive at the following 2d coulomb plasma @xmath89\frac{r_{ij}}{q^2}\nonumber\\ & \times\left[l_j(-{\textbf{q}})-(2\pi)^2f_j \delta(-{\textbf{q}})\right ] , \label{eq : rotating_2comp_bec_dual}\end{aligned}\ ] ] where @xmath90 , @xmath91 , @xmath92 is the vortex field of component @xmath93 , @xmath94 and an implicit sum over repeated component indices @xmath93 , @xmath95 is assumed . by setting @xmath96 such that @xmath97 , and absorbing a factor @xmath98 in the density coefficients , we see that the two - component bose - einstein condensate in eq . with @xmath99 , @xmath100 and @xmath101 corresponds to the unconventional two - component coulomb plasma in eq . . thus , the unconventional coulomb plasma has a counterpart in a two - component bose - einstein condensate with a negative non - dissipative drag interaction . however , note that in order to preserve a fixed number of charges when going from the plasma description in eq . to the phase description in eq . , we have to fix the number of vortices to only include rotationally induced vortices . in principle , in the bec problem , the system can thermally excite vortex - antivortex pairs , but that process can be substantially suppressed by going beyond the phase only model in eq . and introducing an additional energy penalty associated with vortex cores . computer simulations of coulomb interactions are generally difficult to perform due to the long - ranged nature of the interaction . several techniques have been presented to deal with the complications that arise @xcite . we have performed large - scale monte - carlo simulations of the system described in eqs . and on a spherical surface . for other simulations on a spherical surface , see refs . . this may seem like a brute - force approach since the workload of the simulations scales as @xmath102 . however , the benefit is that there are no boundaries , the implementation is relatively easy , and there is no need to constrain the particles to move on a lattice . however , one must also be aware that simulation results may differ due to effects induced by topology . for instance , the triangular crystalline ground state of a 2d one - component plasma will necessarily include a number of dislocations and disclinations on a sphere . these defects are not present in the ground state when the one - component plasma is located on the plane @xcite . we consider a sphere with radius @xmath103 , with origin defined as the center of the sphere such that all particle position vectors @xmath42 and @xmath43 are radial vectors with fixed magnitude @xmath103 in three dimensions . the distance between the particles is measured along the chord @xcite @xmath104 where @xmath105 is the chord angle between the two particles at @xmath106 and @xmath107 with unit vectors @xmath108 and @xmath109 , respectively . we may now rewrite the model in eq . on the surface of a unit sphere as @xmath110 , \label{eq : model_s_sphere}\end{aligned}\ ] ] with interactions given by @xmath111 and @xmath112 note that the interaction @xmath44 in eq . between the neutralizing background and the excess charge of type 1 becomes a constant term on the sphere , so we disregard it in eq . . the dimensionless density of particles on the sphere is given by the packing fraction @xmath113 where @xmath114 is the area of a hard disk of diameter @xmath39 on the sphere of area @xmath115 . in the simulation , we use a unit sphere with @xmath116 . as explained in appendix [ app : general_dielectric_constant ] , in order to account for screening properties when particles interact by two interactions simultaneously , we measure a general inverse dielectric constant , @xmath117 , given by @xmath118 where @xmath119 is a type specific inverse dielectric constant , @xmath120 and @xmath121 are type - dependent weights for the contributions of the different @xmath122 ( which are determined by the values of both types of charge carried by the test particles for which screening is being measured ) , and where @xmath123 and @xmath124 are the dipole moments for charges of type 1 and type 2 , respectively , given by @xmath125 @xmath126 note that the type 2 inverse dielectric constant , @xmath127 , is the same dielectric constant as was used when studying the two - component neutral coulomb plasma on a spherical surface @xcite . in addition to measuring the screening properties , the inverse dielectric constant may be used to identify the existence of a bkt - transition if it exhibits a universal discontinuous jump at the critical point , according to eq . . in addition to the inverse dielectric constant , we also measure the fourth - order modulus , @xmath128 @xcite . this quantity may be used to verify a discontinuous jump in the inverse dielectric constant without making any _ a priori _ assumptions regarding the character of the phase transition . as explained in detail in appendix [ app : fourth_order_derivative ] , a negative @xmath128 at the phase transition in the thermodynamic limit implies that the inverse dielectric constant jumps to zero discontinuously . as for the inverse dielectric constant , we use a general fourth - order modulus to account for the two interactions , @xmath129 where @xmath130 . \label{eq : specific_fourth_order_modulus_two_interactions}\end{aligned}\ ] ] the explicit derivation of eqs . and , is given in appendix [ app : fourth_order_derivative ] . the monte - carlo updating scheme consists of trial moves for one or two particles at the same time , to a randomly chosen new location on the surface of the sphere . the change in the action eq . was calculated and the move was accepted or rejected according to the metropolis - hastings algorithm @xcite . the trial moves were performed in three different ways . the first way was to move a single particle to a new random location uniformly over the total surface . the second way was to move a single particle to a new random location uniformly within some short distance , adjusted to yield a high acceptance rate . the last trial move was to move a nearest - neighbor pair of one @xmath8-particle and one @xmath7-particle together , to a random new location uniformly within some short distance , adjusted to yield a high acceptance rate , and with a random new orientation . in order to straightforwardly ensure detailed balance , we additionally required the two particles to mutually be nearest - neighbors both in the old and the new configuration . and to ensure ergodicity , the pair - move must be mixed with a number of single - particle moves . all of these moves were found to be essential in order to have fast thermalization as well as short autocorrelation times for the cases considered here . pseudorandom numbers were generated by the mersenne - twister algorithm @xcite and the sampled data were postprocessed using ferrenberg - swendsen reweighting techniques @xcite . motivated by its relevance to the fractional quantum hall effect ( in particular , the @xmath131 mr state ) , we focus on analyzing the screening properties of this system at @xmath34 ( @xmath132 ) @xcite . we also perform simulations in the neutral two - component coulomb gas case at @xmath29 ( @xmath133 ) in order to provide a check on the numerics , as well as for comparison with the @xmath34 case . furthermore , the system is also studied for a number of values of the packing fraction , @xmath134 to extract the screening properties in the low - density limit . for the two cases of @xmath17 and the values of @xmath15 studied below , the quantities @xmath135 and @xmath136 were found to be zero , within statistical uncertainty , and except for a small finite - size effect when system size @xmath19 , was small . thus , we focus on the results for @xmath127 as this was the only term in eq . that contributed to the general inverse dielectric constant , @xmath137 . this means that screening properties of particles that interact with charges of both types , are determined by the charges of type 2 , only . note also that when @xmath138 , the unconventional coulomb plasma will screen test particles with charge of type 1 , only . in fig . [ fig : effect_of_eta_and_n ] , we plot @xmath127 in the relevant range of @xmath139 when the two - component neutral coulomb gas ( @xmath29 ) is known to have a bkt transition . at small values of @xmath139 , the system is in the screening phase where @xmath140 . the reason for the @xmath141 sign rather than an equal sign is that there is a mainly size - dependent offset from @xmath142 , because perfect screening is not possible with a small number of charges . for large @xmath139 there is a phase in which charges of different components form tightly bound dipoles and the coulomb gas turns into an insulator where @xmath143 . here , there is a mainly density - dependent offset from @xmath144 because the polarizability of the system increases with density , since the hard - core diameter @xmath39 yields a minimum distance between the charges in the dipoles . the plot in fig . [ fig : effect_of_eta_and_n ] indeed shows that the charge - unbinding transition is dependent on the number of particles in the system , as well as the size of the hard disk charges . when @xmath19 increases , the onset of a finite value in @xmath127 moves to higher values of @xmath139 . however , when we reduce @xmath134 , the value of @xmath139 at onset of @xmath127 becomes smaller . thus , this figure illustrates that understanding the behavior in both limits @xmath145 as well as @xmath146 is not straightforward . for the model in eq . with @xmath29 and @xmath147 . results are presented for three different values of packing fraction @xmath134 and three different values of system size @xmath19 . ] in fig . [ fig : effect_of_eta_and_n - m_2 ] , results for the same case as in fig . [ fig : effect_of_eta_and_n ] are presented , but with @xmath34 . the results for @xmath29 and @xmath34 are very similar , both qualitatively and quantitatively . thus , the screening properties with respect to charge of type 2 of the unconventional coulomb plasma when @xmath34 are very similar to the well - studied two - component neutral coulomb gas . for the model in eq . with @xmath34 and @xmath147 . results are presented for three different values of packing fraction @xmath134 and three different values of system size @xmath19 . ] to get a qualitative picture of the type 2 charge binding of the unconventional plasma , three snapshots of the charge configuration when @xmath34 , @xmath148 , and @xmath149 is given in fig . [ fig : charge_grid ] . when @xmath150 , deep into the screening phase of the system ( see fig . [ fig : effect_of_eta_and_n - m_2 ] ) , most charges are free and only a small fraction of the charges may be said to form closely bound dipoles . at @xmath33 , which is the relevant value for the ising - type quantum hall states , the system is closer to the unbinding transition and a larger fraction ( though not all ) of the particles are bound in dipoles . at @xmath151 , deep in the type 2 insulating region , all particles form closely bound dipoles and the ability to screen type 2 test charges is lost . when @xmath34 , @xmath148 and @xmath149 . red markers represent @xmath7-particles , while blue markers represent @xmath8-particles . the marker diameters are about @xmath152 times larger than hard disk diameter @xmath39 . ] although it is clear from figs . [ fig : effect_of_eta_and_n],[fig : effect_of_eta_and_n - m_2 ] that there is a transition between a screening phase and an insulating phase , it is not easy to spot the transition point in the curves in these figures , which look rather smooth . therefore , we must make some assumptions about the nature of the transition in order to identify it . for the case @xmath29 , where the transition is known to be a bkt transition , it is natural to follow a method that was proposed in ref . . at the bkt critical point , @xmath127 scales logarithmically with @xmath19 for large @xmath19 . it taked the following finite - size scaling form : @xmath153 , \label{eq : finite_size_scaling}\ ] ] where @xmath154 is the value of @xmath155 when @xmath145 and @xmath156 is an undetermined constant . least - squares curve - fitting to eq . may be performed for various sizes @xmath19 with @xmath156 and @xmath154 as free parameters at fixed values of @xmath139 . the critical point is then estimated as the value of @xmath139 which exhibits the best fit to eq . . additionally , for a bkt - transition , the value of @xmath154 obtained at the best fit , must correspond with the universal jump condition , @xmath157 , cf . details of this procedure are given in appendix [ app : weber_minnhagen ] . for @xmath34 , motivated by the similarity between figs . [ fig : effect_of_eta_and_n ] , [ fig : effect_of_eta_and_n - m_2 ] , we assume that the transition is also a bkt transition . we again look for the @xmath139 value at which the system best fits eq . . since we are able to find a value at which there is a very good fit to this form , we conclude that our assumption was justified . in fig . [ fig : critical_qq_weber_minnhagen_method ] , we present results for the critical coupling @xmath35 for four different densities @xmath158 , for @xmath29 and @xmath34 . the results for @xmath29 reproduce the main features of the two - component coulomb gas , namely that @xmath159 when density is low and that @xmath35 increases when density increases . these results also correspond well with earlier results in refs . and . when @xmath34 , we find that the behavior of the critical temperature is very similar to the @xmath29 case , within statistical uncertainty . in addition , in fig . [ fig : universal_jump_weber_minnhagen_method ] , results for the corresponding value of the parameter @xmath154 at the critical point is presented . the values for both @xmath29 and @xmath34 are close to the universal value of @xmath160 for the bkt - transition . since the results for @xmath161 ( the standard coulomb - plasma bkt - transition case ) and @xmath34 are essentially the same , we suggest that the charge - unbinding transition for the unconventional coulomb plasma indeed is a bkt - transition in the sense that the type 2 inverse dielectric constant @xmath127 exhibits logarithmic finite - size scaling and a discontinuous jump with a universal value , as predicted by the bkt renormalization equations . found by curve fitting to eq . with two free parameters . results are presented for four values of the packing fraction @xmath134 and for @xmath29 and @xmath34 . fourteen system sizes in the range @xmath162 have been used . ] and for @xmath29 and @xmath34 . fourteen system sizes in the range @xmath162 have been used . ] as an additional verification of the discontinuous jump in the bkt - transition , we also study the fourth - order modulus @xmath163 , presented in eqs . and . as for the general inverse dielectric constant , we found that the only contributing term in the sum of eq . is the term with all indices equal to 2 , @xmath164 . illustrating the typical behavior of this quantity , results for @xmath164 for a number of sizes when @xmath148 and @xmath34 are presented in fig . [ fig : typical_bahavior ] . typically , @xmath164 exhibits a dip at a value of the coupling that can be associated with the transition . as explained in appendix [ app : fourth_order_derivative ] , a negative and finite dip in the limit when @xmath145 signals the discontinuous jump in @xmath127 that is a characteristic feature of a bkt - transition . to this end , the size of the dip in @xmath164 is plotted as a function of inverse system size @xmath165 in fig . [ fig : dip_fourth_order_modulus ] in the case when @xmath148 . the size of the dip @xmath166 decreases when @xmath19 increases towards the thermodynamic limit . however , assuming power - law dependence of @xmath166 , the positive curvature in the log - log plot indicates a nonzero value of @xmath166 when @xmath167 , verifying a discontinuous jump in @xmath127 , as expected for a bkt - transition . again , we find that the results for @xmath34 are very similar to @xmath29 . as a function of coupling @xmath139 for five different system sizes @xmath19 , when @xmath34 and @xmath148 . ] as a function of inverse system size @xmath165 . the packing fraction is @xmath148 , and results for @xmath29 and @xmath34 are shown . the inset shows the results on a log - log scale . system sizes in the range @xmath168 are used . ] we also associate the coupling value of the minimum in the dip in @xmath164 with the critical point and the results are shown in fig . [ fig : dip_critical_qq ] in the case when @xmath148 . clearly , the position of the dip moves towards higher values of @xmath139 when the system size increases . however , the evolution towards @xmath169 is too slow to make a sharp determination of @xmath139 in this limit as also noted before @xcite . with this method , we are not able to verify that @xmath170 , as was found above in fig . [ fig : critical_qq_weber_minnhagen_method ] for this density . . the packing fraction is @xmath148 , and results for @xmath29 and @xmath34 are shown . the inset shows the results on a log - log scale . system sizes in the size @xmath168 are used . ] by assuming a universal value of the discontinuous jump for a bkt - transition , we may determine the critical point of the bkt - transition using eq . with only one free parameter as described in appendix [ app : weber_minnhagen ] . the results are given in fig . [ fig : critical_qq_ofp_weber_minnhagen_method ] . the critical values of @xmath139 are very similar to what was obtained in fig . [ fig : critical_qq_weber_minnhagen_method ] , but are determined with greater accuracy . for both cases , the critical point appears at higher @xmath139 when density increases . however , @xmath35 is systematically lower at @xmath34 compared to @xmath29 . found by curve - fitting to eq . with one free parameter . results are presented for five values of the packing fraction @xmath134 and for two values of @xmath17 . ] for the range of small densities that we have investigated , the monte - carlo results for the unconventional coulomb plasma with @xmath34 are rather conclusive . this plasma undergoes a charge - unbinding transition that should be regarded as a bkt - transition in the sense that the inverse dielectric constant of type 2 exhibits the well - established signatures of a bkt - transition . specifically , there is a density - dependent critical point @xmath35 that separates a phase where particles of different species form bound pairs at high values of @xmath139 from a phase where particles of different species are free at low values of @xmath139 . for test particles carrying type 2 charge , the high-@xmath139 phase is unscreened , whereas the low-@xmath139 phase is screened . and @xmath34 when @xmath171 , @xmath172 , and @xmath149 . red markers are @xmath7-particles and blue markers are @xmath8-particles . the marker diameters are about 2.5 times larger than hard disk diameter @xmath39 . ] and @xmath34 with @xmath173 , @xmath174 in the range @xmath175 . ] the results presented so far show that the behavior when @xmath29 and @xmath34 are quite similar . however , in the phase with bounded dipoles , when charges of type 2 are not screened , the cases @xmath161 and @xmath176 behave rather differently . we first consider the case when @xmath29 . when charges are bound , this system consists of @xmath19 dipoles that interact by dipole - dipole interactions . consequently , these dipoles tend to form clusters with increased dipole strength , i.e. higher values of the coupling or the density @xcite . in fig . [ fig : charge_grid_bounded_phase ] , a snapshot of a @xmath29 configuration with @xmath149 , @xmath171 and @xmath172 is shown , where some dipoles are seen to form clusters . in the case when @xmath34 , the type 2 interactions are effectively reduced to dipole - dipole interactions , similar to the @xmath29 case . _ however , the logarithmic interactions of type 1 charges remain_. neglecting the weaker dipole - dipole interactions among dipoles of type two , the dipoles now essentially form elementary constituents with charge @xmath17 interacting logarithmically . effectively , the two - component unconventional plasma is reduced to a one - component plasma where the particles carry charge of type 1 and a ( neutral ) dipole of type 2 . when @xmath177 this plasma is in the liquid state , i.e. the tightly bound dipoles do not form an ordered state with a broken translational or orientational symmetry . also , the logarithmic interaction of type 1 charge will prevent the dipoles from forming clusters . a snapshot of the state with bounded dipoles when @xmath34 is shown in fig . [ fig : charge_grid_bounded_phase ] and the qualitative difference from the case when @xmath29 is clearly seen . quantitatively , this is seen by the behavior of @xmath127 , presented in fig . [ fig : inverse_dielectric_constant_highdens ] . when @xmath29 , dipole - dipole interactions at short distances will reduce the fluctuations in the dipole moment resulting in a weakly increasing @xmath127 inside the bounded phase . on the other hand , when @xmath34 the logarithmic interaction of type 1 charge will keep the dipoles at some distance from each other , thus the fluctuations of a dipole are not much restricted by the surrounding dipoles . moreover , the strength of the dipoles increases with @xmath139 and a reduction in @xmath127 follows . the qualitative difference between the cases @xmath29 and @xmath34 is an effect due to the minimum separation of charges at finite density originating with the hard cores , and it will vanish in the limit @xmath178 . we have shown that the unconventional coulomb plasma analyzed in this paper , where particles can carry two distinct types of coulombic charge , will screen test particles with charges of both types for the case most relevant for the plasma analogy of ising - type fractional quantum hall states , i.e. when there is one species of particles that carry type 1 charge @xmath34 ( @xmath132 ) and type 2 charge @xmath179 and another species of particles that carry only type 2 charge @xmath16 . for test particles carrying both types of charge , screening will cease to occur at @xmath30 in the limit of small density , when @xmath34 . for higher values of @xmath139 , the system will continue to screen test particles that carry only type 1 charge , but will not be able to screen test particles with type 2 charge . one striking feature of these results is that @xmath35 and the critical behavior at this point hardly seem to depend on @xmath17 when density is small . this implies that the role of the type 1 interaction ( which corresponds , in quantum hall wavefunction language , to the laughlin - jastrow factor which accounts for the filling fraction of the system ) is simply to maintain the @xmath24-particles in a liquid state . since its critical point is very far away , the type 1 interaction leads to a weak , smooth dependence on @xmath17 . the physics in the transition at @xmath35 is then dominated by the type 2 interaction . we therefore conjecture that our results hold for all reasonable values of @xmath180 not only @xmath133 and @xmath181 , the cases which we have studied here , but also @xmath182 ( which may be relevant to ultra - cold trapped bosons ) and larger values of @xmath180 , possibly all the way up to or near the critical value @xmath183 , below which the one - component plasma of eq . ( [ eqn : laughlin - plasma ] ) is in the metallic phase @xcite . we acknowledge useful discussions with k. brkje , j. s. hye , i. b. sperstad , b. svistunov , and m. wallin . e. v. h , e. b. , and v. g. thank nordita for hospitality during the initial stage of this work . e. b. , p. b. , and a. s. thank the aspen center for physics for hospitality and support under the nsf grant @xmath184 . e. v. h. thanks ntnu for financial support . e. b. was supported by knut and alice wallenberg foundation through the royal swedish academy of sciences fellowship , swedish research council and by the national science foundation career award no . v. g. was supported by nsf grant no . phy-0904017 . c. n. was supported in part by the darpa quest program . a. s. was supported by the norwegian research council under grant no . 205591/v30 ( frinat ) . the work was also supported through the norwegian consortium for high - performance computing ( notur ) . in the unconventional plasma with two components that interact with two different coulomb - like interactions , we are free to insert test particles that may interact with different charge strength through both interactions simultaneously . here , we will generalize the inverse dielectric constant for such test particles . for consistency , we will also perform the derivation on the surface of a sphere by expanding in spherical harmonics . for a similar derivation , but with one interaction only and on a planar geometry , see refs . . when an external test charge field is inserted in the system , the free energy in the system will change according to the effective interaction among the test charges , @xmath185 = \int \mathrm{d}\omega \int \mathrm{d}\omega^{\prime } \sum_{\mu,\nu } \delta q_\mu ( \theta , \phi ) u^{\textnormal{eff}}_{\mu \nu}(\hat{{\textbf{r}}}\cdot\hat{{\textbf{r}}}^{\prime } ) \delta q_\nu ( \theta^{\prime},\phi^{\prime } ) . \label{eq : free_energy_increase_by_effective_interaction}\ ] ] here , the effective interaction between charges of type @xmath186 and @xmath187 , is assumed to be of the form @xmath188 , @xmath189 is the test charge field for charges of type @xmath186 , and the integrations are over the solid angle @xmath190 . to correctly model the test particles as carrying charge of different types , we write @xmath191 where @xmath192 is a type dependent factor that accounts for the relative strength of charges of different types . for instance , the choice @xmath193 describes the test charges corresponding to quasiholes in the mr state , as given in eq . ( 125 ) in ref . , which map to particles in the plasma that carry charge @xmath194 of type 1 and charge @xmath195 of type 2 . moreover , in eq . @xmath196 is a common charge factor for all types such that @xmath197 is the total charge of type @xmath186 carried by a test particle ( which means that @xmath198 in the example above ) , and @xmath199 is the density field of the test particles . it is now convenient to expand the interaction and density fields in spherical harmonics . the test particle density field is expanded by @xmath200 where @xmath201 and @xmath202 are the associated legendre polynomials . the coefficients are given by @xmath203 the effective interaction is expanded by using the addition theorem for spherical harmonics , @xmath204 here , @xmath205 are the legendre coefficients of the interaction , given by @xmath206 where @xmath207 is the legendre polynomial of order @xmath81 . now . is written @xmath185 = \delta q^2 \sum_{l=0}^{\infty } \frac{4\pi}{2l+1 } \sum_{\mu,\nu } a_\mu u^{\textnormal{eff}}_{\mu \nu , l } a_\nu \sum_{m =- l}^{l } \rho_{l}^{m * } \rho_{l}^{m}. \label{eq : free_energy_increase_by_effective_interaction_1}\ ] ] hence , in the limit when the test charge field is infinitesimal , @xmath208 , we find that @xmath209}{\partial \delta q^2}\bigg|_{\delta q = 0 } = \sum_{l=0}^{\infty } \frac{8\pi}{2l+1 } \sum_{\mu , \nu } a_\mu u^{\textnormal{eff}}_{\mu \nu , l } a_\nu \sum_{m =- l}^{l } \rho_{l}^{m * } \rho_{l}^{m}. \label{eq : effective_second_derivative_free_energy}\ ] ] this derivative can also be calculated by inspection of the partition function of the system perturbed with the external test charge field . with @xmath210 = -\ln z[\delta q]$ ] and a potential energy on the form @xmath211 = v_0 + v_1[\delta q]$ ] where @xmath212 is the potential energy of the unperturbed system and @xmath213 $ ] is the contribution due to the test charge field , we find that @xmath214}{\partial \delta q^{2}}\bigg|_{\delta q = 0 } = & \left \langle \frac{\partial^2 v_1 [ \delta q]}{\partial \delta q^{2}}\bigg|_{\delta q = 0 } \right \rangle \nonumber \\ & - \left \langle \left(\frac{\partial v_1 [ \delta q]}{\partial \delta q}\bigg|_{\delta q = 0 } \right)^2 \right \rangle . \label{eq : bare_second_derivative_free_energy}\end{aligned}\ ] ] here , we have also used that @xmath215/\partial \delta q|_{\delta q = 0 } = 0 $ ] , and the brackets denote statistical average with respect to the unperturbed system . the test charges @xmath216 will interact with each other as well as with the charge field @xmath217 . as for the test charge field , the charge field is expanded according to eq . to yield @xmath218 = & \int \mathrm{d}\omega \int \mathrm{d}\omega^{\prime } \sum_{\mu } \left[q_\mu ( \theta , \phi ) + \delta q_\mu ( \theta , \phi)\right]\nonumber \\ & \times u(\hat{{\textbf{r}}}\cdot\hat{{\textbf{r}}}^{\prime } ) \delta q_\mu ( \theta^{\prime},\phi^{\prime})\nonumber\\ = & \sum_{l=0}^{\infty } \frac{4\pi}{2l+1 } u_{l } \sum_{\mu } a_\mu \sum_{m =- l}^{l } \delta q \ , \rho_{l}^{m * } \left(q_{\mu , l}^{m } + a_\mu \delta q \ , \rho_{l}^{m } \right ) , \label{eq : perturbed_part_of_action}\end{aligned}\ ] ] where @xmath219 is the bare interaction , expanded by eq . with coefficients @xmath220 . performing the derivatives in eq . yields @xmath214}{\partial \delta q^2}\bigg|_{\delta q = 0 } = & \sum_{l=0}^{\infty}\frac{8\pi}{2l+1}u_l \sum_{\mu,\nu } a_\mu \delta_{\mu \nu } a_\nu \sum_{m =- l}^{l } \rho_{l}^{m * } \rho_{l}^{m}\nonumber\\ & - \sum_{l=0}^{\infty}\frac{4\pi}{2l+1 } u_l \sum_{l^{\prime}=0}^{\infty } \frac{4\pi}{2l^{\prime}+1 } u_{l^{\prime } } \sum_{\mu,\nu } a_\mu a_\nu \nonumber\\ & \quad \times \sum_{m =- l}^{l } \sum_{m^{\prime}=-l^{\prime}}^{l^{\prime } } \rho_{l}^{m * } \rho_{l^{\prime}}^{m^{\prime } } \left \langle q_{\mu , l}^{m } q_{\nu , l^{\prime}}^{m^{\prime } * } \right \rangle . \label{eq : bare_second_derivative_free_energy_1}\end{aligned}\ ] ] we introduce the dielectric function @xmath221 by @xmath222 and by comparing eqs . and , the inverse dielectric function is found to be @xmath223 moreover , since the bare interaction is only dependent on the distance between the charges , @xmath224 , we have that @xmath225 , which yields @xmath226 additionally , the property that the bare interaction is distance dependent , only , yields an interaction @xmath220 that is independent of @xmath76 . hence , the correlator @xmath227 must be @xmath76 independent as well , @xmath228 . the dielectric function thus reads @xmath229 the dielectric constant @xmath230 is now found in the long wavelength limit of the dielectric function . on a spherical surface , this corresponds to setting @xmath231 in the dielectric function , i.e. @xmath232 . thus , the dielectric constant is @xmath233 so far , only a few assumptions are made regarding the bare interaction @xmath219 and the charge field @xmath217 . to apply eq . for the system under consideration in this paper , we invoke @xmath234 to find @xmath235 by eq . . moreover , the charge field is modeled as point charges in a uniform background @xmath236 where @xmath237 is the uniform background ensuring charge neutrality for charges of type @xmath186 , @xmath238 is the charge of type @xmath186 in particle @xmath93 and the sum is over all @xmath19 particles of the unperturbed system . now , using eq . , the actual coefficient of the charge field is found to be @xmath239 where @xmath240 is the total dipole moment for charges of type @xmath186 . finally , by inserting these results in eq . , the inverse dielectric constant is found to be @xmath241 where @xmath242 by assuming isotropy . when there are test charges with multiple interactions , there are multiple contributions to the change in free energy as seen in eq . . to account for all contributions to the increase in free energy , we construct a generalized dielectric constant by @xmath243 notice that even though there is no bare interaction between charges of different type , there may be nonzero cross terms in eq . , as charges of different type are constrained to be together within the same particle . in ref . a method of verifying the discontinuous character of the bkt - transition was introduced , by examining a higher - order term in the free energy expansion in the xy - model when the system is perturbed with an infinitesimal phase twist . similarly , in ref . , the method was applied in a two - dimensional logarithmic plasma . here , we show that the same idea also applies when we perturb a logarithmic coulomb plasma on a spherical surface with an infinitesimal test charge field with multiple types of coulomb interactions . consider a system with particles interacting with different charges of multiple types , as previously described . we now _ choose _ to perturb this system with a neutral distribution of test charge of multiple types , which has the form @xmath244 , i.e. a similar test particle density field as given in eq . but with @xmath245 being the only nonzero coefficient in the spherical harmonics expansion . this is a convenient choice because it corresponds to the most long - waved nonuniform test charge configuration on the surface of a sphere , and hence , the prefactor of the second - order term in the free energy expansion will be proportional to the inverse dielectric constant , as we will see below . the test charges yield a contribution to the potential energy as given by the @xmath231 and @xmath246 term in eq . , @xmath247 = \frac{4\pi}{3 } u_{1 } \sum_{\mu } a_\mu \delta q \ , \rho_1 ^ 0 \left(q_{\mu,1}^{0 } + a_\mu \delta q \rho_1 ^ 0 \right ) . \label{eq : perturbed_part_of_action_l1}\ ] ] we now consider how the system responds to the test charges by a taylor expansion of the free energy in the test charge field around @xmath248 , @xmath249 = & \frac{\partial f[\delta q]}{\partial \delta q}\bigg|_{\delta q=0 } \delta q + \frac{\partial^2 f[\delta q]}{\partial\delta q^2}\bigg|_{\delta q=0}\frac{\delta q^2}{2!}\nonumber\\ & + \frac{\partial^3 f[\delta q]}{\partial\delta q^3}\bigg|_{\delta q=0}\frac{\delta q^3}{3 ! } + \frac{\partial^4 f[\delta q]}{\partial\delta q^4}\bigg|_{\delta q=0}\frac{\delta q^4}{4 ! } + \ldots . \label{eq : free_energy_expansion}\end{aligned}\ ] ] the change in the free energy @xmath250 $ ] must be invariant to @xmath251 , and hence , all odd - order derivatives in eq . are zero . from appendix [ app : general_dielectric_constant ] ( see eqs . , and ) , the second - order free energy derivative is found to be @xmath252}{\partial\delta q^2}\bigg|_{\delta q=0 } = \frac{8\pi}{3 } ( \rho_1 ^ 0)^2 u_1 \epsilon_{(a_1 , a_2 , ... ) } ^{-1}. \label{eq : appendix_second_order_derivative}\ ] ] the fourth - order derivative is @xmath253}{\partial\delta q^4}\bigg|_{\delta q=0 } = & \ , 3\left\langle \left(\frac{\partial v_1[\delta q]}{\partial\delta q}\bigg|_{\delta q=0}\right)^2\right\rangle^2\nonumber\\ & - \left\langle \left(\frac{\partial v_1[\delta q]}{\partial\delta q}\bigg|_{\delta q=0}\right)^4\right\rangle\nonumber\\ = & \left(\frac{4\pi}{3 } \rho_1 ^ 0 u_1\right)^4 \sum_{\mu,\nu,\rho,\sigma}a_{\mu}a_{\nu}a_{\rho}a_{\sigma}\nonumber\\ \times \big[3\big\langle q_{\mu,1}^0q_{\nu,1}^0\big\rangle & \big\langle q_{\rho,1}^0q_{\sigma,1}^0\big\rangle-\big\langle q_{\mu,1}^0q_{\nu,1}^0q_{\rho,1}^0q_{\sigma,1}^0\big\rangle\big ] . \label{eq : appendix_fourth_order_derivative}\end{aligned}\ ] ] where brackets denote a statistical average with respect to the unperturbed action . inserting eqs . and in eq . yields @xmath249 = \frac{8\pi}{3 } ( \rho_1 ^ 0)^2 u_1 \big[&\epsilon_{(a_1 , a_2 , ... ) } ^{-1}\frac{\delta q^2}{2!}\nonumber\\ & + \gamma_{(a_1 , a_2 , ... ) } \frac{\delta q^4}{4 ! } + \ldots \big ] , \label{eq : free_energy_expansion_1}\end{aligned}\ ] ] where @xmath254 and @xmath255 . \label{eq : specific_fourth_order_term}\end{aligned}\ ] ] now , inserting @xmath245 and assuming the charge field in eq . and a logarithmic bare interaction , @xmath256 , yields @xmath257 , \label{eq : specific_fourth_order_term_1}\end{aligned}\ ] ] where @xmath242 by assuming isotropy . when @xmath248 , the free energy of the system has a global minimum , and hence , the right - hand side of eq . must be greater or equal to zero . now , if @xmath258 approaches a nonzero negative value at the critical point in the thermodynamical limit , the general inverse dielectric constant must simultaneously have a nonzero positive value for the ground state to be stable . however , since @xmath259 in the screening phase , it follows that @xmath260 must exhibit a discontinuous jump at the critical point . hence , investigation of @xmath258 may be used to verify a discontinuity in the inverse dielectric constant , which is a necessary requirement for observing a bkt - transition . the finite - size scaling relation of the bkt - transition has been used throughout this article to verify the universal jump in @xmath127 and to provide estimates for the critical coupling @xmath35 . here , some details to the curve fitting procedure and the goodness of fit measure are presented . least - squares curve fitting of the monte - carlo results for @xmath127 to eq . may be performed with both @xmath261 and @xmath156 as free parameters @xcite . if the transition is of the bkt - type , a good fit to eq . should be obtained at the critical point . in addition , when @xmath261 is free , no _ a priori _ assumption on the value of the universal jump is made , thus a resulting value of @xmath261 that corresponds to the universal jump of the bkt - transition should be obtained . however , with two free parameters , higher quality of the monte - carlo statistics is required to single out when they system is closely obeying the behavior of eq . . we have employed the marquardt - levenberg algorithm minimizing @xmath262 to the nonlinear fitting function in eq . . specifically , @xmath262 is the sum of squared weighted residuals , @xmath263 where @xmath78 is the number of system sizes @xmath264 , @xmath265 is the value of the inverse dielectric constant @xmath127 obtained from the monte - carlo simulation at system size @xmath264 , and @xmath266 is the corresponding error . for a good fit , we expect the weight - normalized residuals , @xmath267 to be gaussian - distributed with mean @xmath268 and variance @xmath269 . thus , to measure the goodness of the fit , we use the anderson - darling test statistic @xmath270 for the data set @xmath271 to arise from a normal distribution with @xmath268 and @xmath272 : @xmath273+\ln[\phi(y_{n+1-i})]\right\ } , \label{eq : anderson_darling}\ ] ] where @xmath274 is the standard normal cumulative distribution function and where the data set @xmath271 is ordered from low to high values . a smaller value of @xmath270 essentially means a better fit between the data and the fit function . to illustrate the method , monte - carlo results for @xmath127 at fourteen different system sizes and the corresponding curve - fit according to eq . are given in fig . [ fig : curvefit_finite_size ] for three different values of @xmath139 . here , @xmath172 and @xmath29 . clearly , at @xmath275 , the fit between the data and the fit function is better than for the two other cases . moreover , in fig . [ fig : curvefit ] the corresponding results for the goodness of fit parameter as well as the results for the parameter @xmath261 as a function of @xmath139 are shown . indeed , the minimum in @xmath270 indicates a critical region where the data seem to follow the logarithmic finite size scaling of @xmath127 given in eq . . also note that this region coincides with a value of @xmath276 close to the universal jump value of 4 . with the minimum of @xmath270 as a measure of the critical point and with error estimates obtained by the jackknife method , we find that @xmath277 and that @xmath278 , less than 2% off the universal number . the results in figs . [ fig : critical_qq_weber_minnhagen_method ] and [ fig : universal_jump_weber_minnhagen_method ] are found by repeating this procedure for different values of @xmath134 and @xmath17 . for fourteen different system sizes in the range @xmath279 at three different values of the coupling @xmath139 . the best fit according to the fit function in eq . with two free parameters , is given as the corresponding solid line in all three cases . the packing fraction is @xmath172 and @xmath29 . ] and the corresponding free parameter @xmath280 obtained when curve fitting to the critical finite - size relation given in eq . . the results are given as a function of @xmath139 . system sizes @xmath19 , and @xmath134 and @xmath17 are the same as in fig . [ fig : curvefit_finite_size ] . error estimates are obtained by the jackknife method . ] the procedure described in detail above with two free parameters , may be performed with a fixed value of @xmath281 and with @xmath156 as the only free parameter . if the transition is of the bkt - type , a good fit to eq . should be obtained at the critical point . this is a rather well - used method to determine the critical point of a bkt - transition @xcite . with only one free parameter , @xmath35 will be determined with greater accuracy compared to the case when there are two free parameters . refs . and used @xmath262 as a goodness of fit parameter . we also tried this , and the results for the critical coupling as well as the corresponding parameter @xmath261 were consistent with @xmath270 results within statistical uncertainty . however , we found that error estimates were clearly underestimated with @xmath262 , probably due to over - fitting . the parameter @xmath156 in the finite - size scaling relation ( eq . ) is density dependent @xcite . specifically , @xmath156 increases when @xmath134 decreases . hence , at the critical point , the finite - size scaling slows down when @xmath134 is lowered . therefore , larger systems @xmath19 or better statistics are required to resolve the critical scaling when @xmath134 is small . in particular , curve fitting to eq . was also performed for @xmath282 in addition to the densities presented in figs . [ fig : critical_qq_weber_minnhagen_method ] and [ fig : universal_jump_weber_minnhagen_method ] . however , in this case the statistics were not good enough to resolve a clear minimum in @xmath270 . also note that there are higher - order corrections@xcite to eq . that are not taken into account in this work .

we model messaging activities as a hierarchical doubly stochastic point process with three main levels , and develop an iterative algorithm for inferring actors relative latent positions from a stream of messaging activity data . each of the message - exchanging actors is modeled as a process in a latent space . the actors latent positions are assumed to be influenced by the distribution of a much larger population over the latent space . each actor s movement in the latent space is modeled as being governed by two parameters that we call confidence and visibility , in addition to dependence on the population distribution . the messaging frequency between a pair of actors is assumed to be inversely proportional to the distance between their latent positions . our inference algorithm is based on a projection approach to an online filtering problem . the algorithm associates each actor with a probability density - valued process , and each probability density is assumed to be a mixture of basis functions . for efficient numerical experiments , we further develop our algorithm for the case where the basis functions are obtained by translating and scaling a standard gaussian density . = 1 social network ; multiple doubly stochastic processes ; classification ; clustering 62m0 , 60g35 , 60g55

1205.5920

communication networks are presenting ever - increasing challenges in a wide range of applications , and there is great interest in inferential methods for exploiting the information they contain . a common source of such data is a corpus of time - stamped messages such as e - mails or sms ( short message service ) . such messaging data is often useful for inferring a social structure of the community that generates the data . in particular , messaging data is an asset to anyone who would like to cluster actors according to their _ similarity_. a practitioner is often privy to messaging data in a _ streaming _ fashion , where the word _ streaming _ describes a practical limitation , as the practitioner might be privy only to the incoming data in a fixed summarized form without any possibility to retrieve past information . it is in the practitioner s interest to transform the summarized data so that the transformed data is appropriate for detecting _ emerging _ social trends in the source community . we mathematically model such streaming data as a collection of tuples of the form @xmath0 of time and actors , where @xmath1 and @xmath2 represent actors exchanging the @xmath3-th message and @xmath4 represents the occurrence time of the @xmath3-th message . there are many models suitable for dealing with such data . the most notable are the cox hazard model , the doubly stochastic process ( also known as the cox process ) , and the self - exciting process ( although self - exciting processes are sometimes considered as special cases of the cox hazard model ) . for references on these topics , see @xcite , @xcite and @xcite . all three models are related to each other ; however , the distinctions are crucial to statistical inference as they stem from different assumptions on information available for ( online ) inference . to transform @xmath5 data to a data representation more suitable for clustering actors , we model @xmath5 as a ( multivariate ) doubly stochastic process , and develop a method for embedding @xmath5 as a stochastic process taking values in @xmath6 for some suitably chosen @xmath7 . for statistical inference when there is information available beyond @xmath5 , the cox - proportional hazard model is a natural choice . in @xcite and @xcite , for instance , instantaneous intensity of messaging activities between each pair of actors is assumed to be a function of , in the language of generalized linear model theory , known covariates with unknown regression parameters . more specifically , in @xcite , the authors consider a model where @xmath8 with @xmath9 and @xmath10 representing independent counting processes , e.g. , @xmath9 are bernoulli random variables and @xmath10 are random variables from the exponential family . on the other hand , in @xcite , a cox multiplicative model was considered where @xmath11 . the model in @xcite posits that actor @xmath12 interacts with actor @xmath13 at a baseline rate @xmath14 modulated by the pair s covariate @xmath15 whose value at time @xmath16 is known and @xmath17 is a common parameter for all pairs . in @xcite , it is shown under some mild conditions that one can estimate the global parameter @xmath17 consistently . in @xcite , the intensity is modeled for _ adversarial _ interaction between _ macro _ level groups , and a problem of nominating unknown participants in an event as a missing data problem is entertained using a self - exciting point process model . in particular , while no explicit intensity between a pair of actors ( gang members ) is modeled , the event intensity between a pair of groups ( gangs ) is modeled , and the spatio - temporal model s chosen intensity process is self - exciting in the sense that each event can affect the intensity process . when data @xmath5 is the only information at hand , a common approach is to construct a time series of ( multi-)graphs to model association among actors . for such an approach , a simple method to obtain a time series of graphs from @xmath18 is to `` pairwise threshold '' along a sequence of non - overlapping time intervals . that is , given an interval , for each pair of actors @xmath12 and @xmath13 , an edge between vertex @xmath12 and vertex @xmath13 is formed if the number of messaging events between them during the interval exceeds a certain threshold . this is the approach taken in @xcite , @xcite and @xcite , to mention just a few examples . the resulting graph representation is often thought to capture the structure of some underlying social dynamics . however , recent empirical research , e.g. , @xcite , has begun to challenge this approach by noting that changing the thresholding parameter can produce dramatically different graphs . another useful approach when @xmath5 is the only information available is to use a doubly stochastic process model in which count processes are driven by latent factor processes . this is the approach taken explicitly in @xcite and @xcite , and this is also done implicitly in @xcite . in @xcite and @xcite interactions between actors are specified by proximity in their latent positions ; the closer two actors are to each other in their latent configuration , the more likely they exchange messages . using our model , we consider a problem of clustering actors `` online '' by studying their messaging activities . this allows us a more geometric approach afforded by embedding @xmath19 data to an @xmath20 representation for some fixed dimension @xmath21 . in this paper , we propose a useful mathematical formulation of the problem as a filtering problem based on both a multivariate point process observation and a population latent position distribution . as a convention , we assume that a vector is a column vector if its dimension needs to be disambiguated . we denote by @xmath22 the filtration up to time @xmath16 that models the information generated by undirected communication activities between actors in the community , where `` undirected '' here means we do not know which actor is the sender and which is the receiver . we denote by @xmath23 the space of probability measures on @xmath6 . for a probability density function defined on @xmath6 , @xmath24 denotes the probability density function that is proportional to @xmath25 where the normalizing constant does not depend on @xmath26 . the set of all @xmath27 matrices over the reals is denoted by @xmath28 . for each @xmath29 matrix @xmath30 , we write @xmath31 . given a vector @xmath32 , we write @xmath33 for its euclidean norm . let @xmath34 and @xmath35 . for each @xmath36 and @xmath37 , we write @xmath38 for the hadamard product of @xmath36 and @xmath39 , i.e. , @xmath40 denotes component - wise multiplication . given vectors @xmath41 in @xmath6 , the gram matrix of the ordered collection @xmath42 is the @xmath43 matrix @xmath44 such that its @xmath45- entry @xmath46 is the inner product @xmath47 of @xmath48 and @xmath49 . given a matrix @xmath50 , @xmath51 is the column vector whose @xmath52-th entry is the @xmath52-th diagonal element of @xmath53 . with a slight abuse of notation , given a vector @xmath32 , we will also denote by @xmath54 the @xmath43 diagonal matrix such that its @xmath52-th diagonal entry is @xmath55 . we always use @xmath56 for the number of actors under observation and @xmath21 for the dimension of the latent space . we denote by @xmath57 the @xmath56-fold product @xmath58 of @xmath6 . an element of @xmath57 will be written in bold face letters , e.g. @xmath59 . similarly , bold faced letters will typically be used to denote objects associated with the @xmath56 actors collectively . an exception to this convention is the identity matrix which is denoted by @xmath60 , where the dimension is specified only if needed for clarification . also , we write @xmath61 as the column matrix of ones . with a bit of abuse of notation , we also write @xmath61 for an indicator function , and when confusion is possible , we will make our meaning clear . our actors under observation are assumed to be a subpopulation of a bigger population . that is , we observe actors @xmath62 that are sampled from a population for a longitudinal study . we are not privy to the actors latent features that determine the frequency of pairwise messaging activities , but we do observe messaging activities @xmath63 . a notional illustration of our approach thus far is summarized in figure [ fig : hierarchalmodeldiagram ] , figure [ fig : subpop_historgrams ] , and figure [ fig : kullback - leibler divergence - simulation ] . in both figure [ fig : subpop_historgrams ] and figure [ fig : kullback - leibler divergence - simulation ] , @xmath64 represents the ( same ) initial time when there was no cluster structure , and @xmath65 and @xmath66 represent the emerging and fully developed latent position clusters which represent the object of our inference task . for a more detailed diagram . ] [ [ population - density - process - level ] ] population density process level + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the message - generating actors are assumed to be members of a community , which we call the population . the aspect of the population that we model in this paper is its members distribution over a latent space in which the proximity between a pair of actors determines the likelihood of the pair exchanging messages . the population distribution is to be time - varying and a mixture of component distributions . the latent space is assumed to be @xmath6 for some @xmath7 , and the population distribution at time @xmath16 is assumed to have a continuous density @xmath67 . to be more precise , we assume that the ( sample ) path @xmath68 is such that for each @xmath69 , @xmath70 where * @xmath71 is a smooth _ sample _ path of a stationary ( potentially degenerate ) diffusion process taking values in @xmath72 , * @xmath73 is a probability density function on @xmath6 with convex support with its mean vector being the zero vector and its covariance matrix being a positive definite ( symmetric ) matrix , * @xmath74 is a smooth _ sample _ path of an @xmath6-valued ( potentially non - stationary or degenerate ) diffusion process , * @xmath75 is a smooth _ sample _ path of a stationary ( potentially degenerate ) diffusion process taking values in @xmath76 . note that it is implicitly assumed that @xmath77 , and additionally , we also assume that for each @xmath78 and @xmath79 , the @xmath80-th moment @xmath81 of the @xmath52-th coordinate of @xmath67 is finite , i.e. , @xmath82 . in this paper , we take @xmath83 and @xmath75 as exogenous modeling elements . however , for an example of a model with yet further hierarchical structure , one _ could _ take a cue from a continuous time version of the classic `` co - integration '' theory , e.g. , see @xcite . the idea is that the location @xmath84 of the @xmath3-th center is non - stationary , but the inter - point distance between a combination of the centers is stationary . more specifically , one _ could _ further assume that there exist @xmath85 matrix @xmath86 , @xmath87 matrix @xmath88 and @xmath89 matrix @xmath90 such that * @xmath91 is the @xmath92 dimensional zero matrix , * @xmath93 is a @xmath94 dimensional brownian motion , * @xmath95 is a stationary diffusion process . thus , the position of centers are unpredictable , but the relative distance between each pair of centers are as predictable as that of a stationary process . @xmath96)$ ] and @xmath97 0.1 in @xmath98 @xmath99 @xmath100 0.3 , the lightly - colored outer histogram represents the latent position distribution @xmath67 for the full population , and the darkly - colored inner histogram represents the distribution of the latent positions of actors under _ consideration_. the illustrated temporal order is @xmath101 . , title="fig : " ] 0.3 , the lightly - colored outer histogram represents the latent position distribution @xmath67 for the full population , and the darkly - colored inner histogram represents the distribution of the latent positions of actors under _ consideration_. the illustrated temporal order is @xmath101 . , title="fig : " ] 0.3 , the lightly - colored outer histogram represents the latent position distribution @xmath67 for the full population , and the darkly - colored inner histogram represents the distribution of the latent positions of actors under _ consideration_. the illustrated temporal order is @xmath101 . , title="fig : " ] [ [ actor - position - process - level ] ] actor position process level + + + + + + + + + + + + + + + + + + + + + + + + + + + + figure [ fig : subpop_historgrams ] sketches the connection between actors and populations . we first define a process for a single actor . to begin , for each @xmath16 , given @xmath67 and @xmath102 , we write @xmath103 where @xmath104 the formulation here for the @xmath105 and @xmath106 is based on a quadratic taylor - series approximation of a so - called `` bounded confidence '' model studied in @xcite . here , the value of @xmath107 represents the confidence level of actor @xmath12 on its current position and @xmath108 represents the visibility of other actors position by actor @xmath12 . roughly speaking , an actor with a _ small _ value of @xmath107 and a _ large _ value of @xmath108 will be influenced _ greatly _ by actors that are positioned _ both near and far _ in the latent space whereas an actor with a _ large _ value of @xmath107 and a _ small _ value of @xmath108 will be influenced only a _ small _ amount by actors that are _ nearby _ in the latent space . for further discussion on our motivation for the form of @xmath109 , see appendix [ appendix : motivationfora ] . for each actor @xmath12 , the deterministic path @xmath110 is assumed to be continuous , taking values in a compact subset of @xmath111 . it is assumed that given @xmath112 , each actor s latent position process @xmath113)$ ] is a diffusion process whose generator is @xmath109 , and moreover we assume that @xmath114 are mutually independent . for each @xmath16 , let @xmath115 where each @xmath116 is assumed to be a column vector , i.e. , a @xmath117 matrix . in other words , the @xmath12-th row of @xmath118 is the transpose of @xmath116 . @xmath97 , @xmath119)$ ] , and @xmath120)$ ] 0.1 in compute @xmath121 and @xmath122 compute the non - negative definite symmetric square root @xmath123 of @xmath124 @xmath98 @xmath125 standardnormalvector @xmath126 @xmath100 [ [ messaging - process - level ] ] messaging process level + + + + + + + + + + + + + + + + + + + + + + + denote by @xmath127 the number of messages sent _ from _ actor @xmath12 _ to _ actor @xmath13 . also , denote by @xmath128 the number of messages exchanged _ between _ actor @xmath12 _ and _ actor @xmath13 . note that @xmath129 . for each actor @xmath12 , we assume that the path @xmath130 is deterministic , continuous and takes values in @xmath76 . for each @xmath16 , we assume that @xmath131 & = ( \lambda_{t , i } \lambda_{t , j}/2 ) p_{t , i\rightarrow j}(\bm x_{t})dt + o(dt).\end{aligned}\ ] ] for our algorithm development and experiment @xmath132 in section [ sec : numericalexperiments ] , we take @xmath133,\end{aligned}\ ] ] but for experiment @xmath134 in section [ sec : numericalexperiments ] , we take @xmath135 . next , by way of assumption , for each pair , say , actor @xmath12 and actor @xmath13 , we eliminate the possibility that both actor @xmath12 and actor @xmath13 send messages concurrently to each other . more specifically , we assume that @xmath136 \\ & = ( \lambda_i(t ) \lambda_j(t)/2 ) ( p_{t , i\rightarrow j}(x_{t , i } ) + p_{t , j \rightarrow i}(x_{t , j } ) ) dt + o(dt).\end{aligned}\ ] ] for future reference , we let @xmath137 @xmath138 , @xmath139 and @xmath140 0.1 in @xmath141 @xmath142 @xmath143 \leftarrow ( t , i , j)$ ] @xmath144 @xmath145 @xmath100 0.3 0.3 0.3 we denote by @xmath146 the conditional distribution of @xmath147 given @xmath22 , i.e. , for each @xmath148 , @xmath149.\end{aligned}\ ] ] for the rest of this paper , we shall assume that the ( random ) measure @xmath150 is absolutely continuous with respect to lebesgue measure with its density denoted by @xmath151 . that is , @xmath152 for @xmath153 . denote by @xmath154 the @xmath12-th marginal posterior distribution of @xmath146 , i.e. , for each @xmath155 , @xmath156 $ ] , and let @xmath157 denote its density . in theorem [ thm : exactposterior ] , the _ exact _ formula for updating the posterior is presented , and in theorem [ thm : simplifyingposteriorupdaterule ] , our _ working _ formula used in our numerical experiments is given . we develop our theory for the case where @xmath158 and @xmath159 are the same for all actors for simplicity , as generalization to the case of each actor having different values for @xmath107 and @xmath108 is straightforward but requires some additional notational complexity . [ thm : exactposterior ] for each @xmath160 and @xmath161 , @xmath162 where @xmath163 is an @xmath164 matrix such that for each @xmath165 , @xmath166 and for each @xmath167 , @xmath168 , and @xmath169 is an @xmath164 matrix such that for each pair @xmath170 , @xmath171 and for each pair @xmath167 , @xmath172 . hereafter , for developing algorithms further for efficient computations , we make the assumption that for each @xmath16 , @xmath173 where @xmath174 denotes the joint density for actors @xmath12 , @xmath13 and @xmath52 . [ thm : simplifyingposteriorupdaterule ] for each function @xmath175 , we have @xmath176 replacing @xmath177 with a dirac delta generalized function , theorem [ thm : simplifyingposteriorupdaterule ] states that for each @xmath178 , @xmath179 where @xmath180 denotes the formal adjoint operator of @xmath181 . for use only within algorithm [ algo : estimateactorposterior ] , @xmath182 @xmath138 , @xmath139 , @xmath67 and @xmath183 , 1 \le u_\ell < v_\ell \le n\}$ ] 0.1 in compute @xmath184 from @xmath185 using @xmath186 @xmath187 @xmath188 @xmath189 @xmath190 @xmath191 update @xmath184 using @xmath192 and @xmath186 in update @xmath193 using @xmath192 and @xmath194 in @xmath145 the projection filter is an algorithm which provides an approximation to the conditional distribution of the latent process in a systematic way , the method being based on the differential geometric approach to statistics , cf . @xcite . when the space on which we project is a mixture family , as in @xcite , the projection filter is equivalent to an approximate filtering via the galerkin method , cf . following this idea , starting from theorem [ thm : simplifyingposteriorupdaterule ] , we obtain below in theorem [ thm : projfilterzakai ] the basic formula for our approximate filtering algorithm . to be more specific , consider a set of probability density functions @xmath195 . then , let @xmath196 be the space of all probability density functions that can be written as a probability - weighted sum of @xmath197 . that is , @xmath198 if and only if @xmath199 for some probability vector @xmath200 on indices @xmath201 . in particular , for deriving our algorithms , we will assume that for some systematic choice of @xmath202 , each probability density under consideration is a member of @xmath203 . among many possible choices for @xmath204 in theorem [ thm : projfilterzakai ] are a multivariate haar wavelet basis and a multivariate daubechies basis . on the other hand , a gaussian mixture model is pervasively used throughout statistical inference tasks such as clustering and classification in algorithms such as @xmath52-means clustering . as such , we develop our algorithms with an eye towards use with other gaussian mixture model - based algorithms . in appendix [ appendix : mixtureprojectionformula ] , we further develop our algorithm under the assumption that @xmath205 where @xmath73 is the standard gaussian density function defined on @xmath206 , @xmath207 , and the finite sequence @xmath208 is to be chosen judiciously prior to implementing the algorithm . preparing for our next result in theorem [ thm : projfilterzakai ] , we let @xmath209 be the symmetric matrix such that @xmath210 , and for each @xmath52 , let @xmath211 be the symmetric matrix such that its @xmath45-entry @xmath212 is @xmath213 . collectively , we denote by @xmath214 the three - way tensor whose @xmath215 entry is @xmath212 . let @xmath216 be the matrix such that its @xmath45-entry @xmath217 is @xmath218 , where @xmath219 is the differential operator such that @xmath220 with @xmath221 [ thm : projfilterzakai ] suppose that for each @xmath16 , @xmath12 and @xmath26 , @xmath222 . let @xmath223 denote the column vector whose @xmath52-th entry is @xmath224 . then , @xmath225 in our application , our final analysis is completed by clustering the posterior distributions . instead of working directly with posteriors , an infinite - dimensional object , we propose to work with objects in an euclidean space each of which represents a particular actor . however , given @xmath185 and @xmath226 , using their mean vectors or their kl distance for clustering can be uninformative . for example , if @xmath227 , then their mean vectors would be the same and their kl distance would be zero . however , if @xmath227 is the density of , say , a normal random vector such that its mean is zero and its covariance matrix is @xmath228 for a large value of @xmath229 , then concluding that actor @xmath12 and actor @xmath13 are similar could be misleading . to alleviate such situations in a clustering step of our numerical experiments , we propose using a multivariate statistical technique called classical multidimensional scaling ( cmds ) to obtain a lower dimensional representation of @xmath230 . more specifically , we achieve this by representing each actor as a point in @xmath6 , where the configuration is obtained by solving the optimization problem @xmath231 where @xmath232 denotes a strictly decreasing function defined on @xmath233 taking values in @xmath234 . for example , one can take @xmath235 where @xmath236 is chosen so that @xmath237 $ ] for all possible values @xmath238 of @xmath239 . another possibility among many others is to choose @xmath240 if @xmath241 is chosen so that @xmath242 $ ] for all possible values @xmath238 of @xmath243 . we denote by @xmath244 the set of solutions to the optimization problem . given a vector @xmath245 of @xmath56 probability densities on @xmath6 , it can be shown that the solution set @xmath244 is not empty and is closed under orthogonal transformations . in the classical embedding literature , ensuring continuous embeddings is neglected as it is not relevant to their applications . however , for our work , this is crucial as we study their evolution through time , i.e. , ideally , we would like to see that a small change in time corresponds to a small change in latent location . in this section , we propose an extension to the cmds algorithm to remedy the aforementioned non - uniqueness issue , and show that the resulting algorithm ensures continuity of embeddings . in our numerical experiments , for each @xmath245 , we choose a particular element @xmath246 of the solution set @xmath244 so that @xmath246 depends on @xmath245 in a consistent manner . a matrix @xmath247 with full column rank and a symmetric matrix @xmath248 such that @xmath249 @xmath250 @xmath251 any @xmath252 classical mds solution of @xmath53 compute a singular value decomposition @xmath253 of @xmath254 return @xmath255 by a dissimilarity matrix , we shall mean a real symmetric non - negative matrix whose diagonal entries are all zeros . first , fix @xmath21 such that @xmath256 . then , for each @xmath164 dissimilarity matrix @xmath53 , define @xmath257 where @xmath258 . the elements of @xmath259 are known as _ classical multidimensional scalings _ , and as discussed in @xcite , it is well known that @xmath259 is not empty provided that the rank of @xmath260 is at least @xmath21 . our discussion in this section concerns making a selection @xmath261 from @xmath259 so that the map @xmath262 is continuous over the set of dissimilarity matrices such that @xmath260 is of rank at least @xmath21 . let @xmath53 be a dissimilarity matrix such that the rank of @xmath260 is at least @xmath21 . we begin by choosing an element of @xmath259 , say @xmath263 , through classical dimensional scaling . let @xmath264 be the eigenvalue decomposition of @xmath263 , where @xmath265 and @xmath266 is the diagonal matrix whose entries are the eigenvalues in non - increasing order . by the rank condition , we have @xmath267 . first , we formally write @xmath268 where ( i ) : : @xmath269 is the @xmath270 matrix with its @xmath271 entry @xmath272 , ( ii ) : : @xmath273 is the @xmath43 diagonal matrix whose @xmath12-th diagonal entry is @xmath274 . dependence of @xmath275 on @xmath53 will be suppressed in our notation unless needed for clarity . now , if the diagonals of @xmath273 are distinct , then @xmath275 is well defined . however , in general , due to potential geometric multiplicity of an eigenvalue , our definition of @xmath275 can be ambiguous . this is the main challenge in making a continuous selection and we resolve this issue in our following discussion . for our remaining discussion , without loss of generality , we may assume that for each dissimilarity matrix @xmath53 , @xmath275 is well - defined by making an arbitrary choice if there is more than one cmds solution . note that the mapping @xmath276 may not be a continuous selection . we now remedy this . first , fix an @xmath270 matrix @xmath277 and let @xmath278 where @xmath279 runs over all @xmath43 real orthogonal matrices . then , define @xmath280 algorithm [ algo : cmdscc ] yields the solution @xmath261 and the proof of the following theorem , theorem [ thmstat : continuousembedding ] , can be found in appendix [ thmstat : contembedding : proof ] . [ thmstat : continuousembedding ] suppose that the @xmath270 matrix @xmath277 is of full column rank . then , the mapping @xmath281 yields a well - defined continuous function on the space of dissimilarity matrices such that @xmath260 is of rank at least @xmath21 . , @xmath282 , and @xmath283 ) or exogenous modeling element @xmath284 . the arrows are associated with _ influence _ relationships , e.g , @xmath285 reads @xmath67 influences @xmath286 . ] here we discuss some insightful facts related to our model given the assumption stated in the last section . first , we have the following : [ thm : elliptic ] fix @xmath287 and suppose that @xmath288 for a.e . @xmath289 . the operator @xmath109 is elliptic , i.e. , for each @xmath290 , the matrix @xmath291 is positive definite . note that for each @xmath292 , @xmath293 note that @xmath294 and @xmath295 for each @xmath296 , and that @xmath297 away from a subspace of @xmath6 whose lebesgue measure is zero . it follows that @xmath298 for each @xmath299 , whence @xmath300 is positive definite . now , we further examine algorithm [ algo : populationprocess ] , algorithm [ algo : latentprocess ] , algorithm [ algo : messagingactivities ] , algorithm [ algo : estimateactorposterior ] , algorithm [ algo : cmdscc ] , and discuss some technical points behinds these algorithms . in algorithm [ algo : latentprocess ] , existence and uniqueness of @xmath301 follows from theorem [ thm : elliptic ] . the continuity of @xmath302 follows from theorem [ thm : elliptic ] and @xcite . in algorithm [ algo : messagingactivities ] , for simulating a sample path of @xmath303 , we use the so - called time - change property ; that is , we use the fact that the process given by @xmath304 is a unit - rate simple poisson process , where @xmath305 and @xmath306 and @xmath307 . for simulation , we use its dual result , i.e. , @xmath308 is a point process whose intensity process is @xmath309 , where @xmath310 is a path of a unit - rate simple poisson process . also , we note that for computation of @xmath309 , online inference is a necessary part of our simulation in algorithm [ algo : messagingactivities ] ; that is , we need to compute @xmath311.$ ] in algorithm [ algo : messagingactivities ] and algorithm [ algo : estimateactorposterior ] , _ near - infinitesimally - small _ means @xmath97 so small that the likelihood of having more than one event during a time interval of length @xmath97 is practically negligible . also , by standardnormalvector in algorithm [ algo : latentprocess ] and unitexponentialvariable in algorithm [ algo : messagingactivities ] , we mean generating , respectively , a single normal random vector with its mean vector being zero and its covariance matrix being the identity matrix , and a single exponential random variable whose mean is one . in our experiments , we hope to detect clusters with accuracy and speed similar to that possible if the latent positions @xmath312 were actually observed even though we use only information in @xmath313 estimated from information contained in @xmath22 . we denote the end - time for our simulation as @xmath314 . there are two simulation experiments presented in this section , and the computing environment used in each experiment is reported at the end of this section . [ [ experiment-1 ] ] experiment 1 + + + + + + + + + + + + we take @xmath315 and we assume that for each @xmath316 $ ] and actor @xmath317 , @xmath318 and @xmath319 . for the population process we take for each @xmath316 $ ] @xmath320 where @xmath321 . \end{cases } \end{aligned}\ ] ] then , we also consider @xmath322 , where @xmath323 . \end{cases } \end{aligned}\ ] ] there is only one population density ; in other words , @xmath324 . note that even with _ one population center _ , we can have _ more than one empirical mode _ for the subpopulation . one of these modes is near zero , and another mode is near one . the reason for this is that because of the value of @xmath325 and @xmath326 , when an actor is too far away from the mode @xmath327 of the population process , the population process affects the actors on its _ tail _ only by negligibly small amount . in figure [ fig : experimentone - sixactorpath ] and figure [ fig : experimentone - perfect - filtering - case ] , a sample path of the true latent position of each of eight actors is illustrated in black lines . it is apparent that in the @xmath328 , all eight actors are equally _ informed _ of the population mode shift , but in the @xmath329 case , only the last three were able to adapt to the change , and the first five actors are surprised by the abrupt change at time @xmath330 . our simulation is discretized . our unit time is @xmath331 , and in figure [ fig : experimentone - sixactorpath ] , each tick in the horizontal axis corresponds to an integral multiple of @xmath97 . the jump term in our update formula is quite sensitive to the number of actors being considered . as such , for updating the jump term , we further discretized @xmath97 into @xmath332 subintervals for numerical stability of our update iterations . for @xmath333 , each unit interval is associated with @xmath334 sub - iterations , and the total number of the ( main ) iteration is @xmath335 , and we use @xmath336 instead of @xmath337 in each @xmath3-th subiteration of each main iteration staring at time @xmath16 . to implement our mixture projection algorithm , we take @xmath338 . the initial position of the @xmath333 actors are sampled from the initial population distribution @xmath339 . we take @xmath340 . the discretized version @xmath341 of @xmath109 is illustrated in figure [ fig : levelplot(at)example ] . for inference during our experiment , we have dropped the second order term and used only the first order term to keep the cost of running our experiment low . on the other hand , for simulating the actors latent positions , we have used both the first and second order term of @xmath109 . the value of @xmath342 gives the first part of the change in @xmath343 . note that in both figure [ fig : levelplot(at)one ] and figure [ fig : levelplot(at)zero ] , the entries that are _ sufficiently far off _ from the diagonals are near zero . for @xmath344 , the time plot of the number of messages produced during interval @xmath345 $ ] is given in figure [ fig : experimentone - num - msg ] , and shows transient behaviors of varying degrees of messaging intensity over the interval . our set up for @xmath346 produced a simulation sample output of observing @xmath347 messages amongst the @xmath333 actors in unit time once the population center changed _ abruptly _ from @xmath348 to @xmath349 at the start of the @xmath350-th unit time interval , i.e. , @xmath351 . in other words , after @xmath352 , a single unit time is roughly associated with the amount of time during which the whole subpopulation of eight actors exchanges around @xmath353 messages , or equivalently , during which each pair of actors exchange around @xmath354 messages . on the other hand , in both @xmath328 and @xmath329 for the interval @xmath355=[0,5]$ ] , the subpopulation messaging rate is relatively constant at the rate of @xmath350 messages over each unit interval , and this is expected as all eight actors are tightly situated around @xmath132 . 0.45 for the discretized version @xmath341 of @xmath109 , used in the simulation experiment for two particular cases , where the horizontal axis is associated with the rows of @xmath356 and the vertical axis is associated with the columns of @xmath356.,title="fig : " ] 0.45 for the discretized version @xmath341 of @xmath109 , used in the simulation experiment for two particular cases , where the horizontal axis is associated with the rows of @xmath356 and the vertical axis is associated with the columns of @xmath356.,title="fig : " ] case . the number of messages per @xmath97 across the time interval @xmath357 $ ] for the subpopulation of actors @xmath358 . ] case . the _ sample _ path of the true and estimated latent position of each of eight actors used for experiment 1 in a black solid line and in a dashed red line , respectively . ] the _ sample _ path of the true and estimated latent position of each of eight actors used for experiment 1 respectively in a black solid line and in a dashed red line . ] our experiments for @xmath328 and @xmath329 both show that the filtered positions for all eight actors are close to the exact positions . + [ [ experiment-2 ] ] experiment 2 + + + + + + + + + + + + in this experiment , for each @xmath16 , we have used the empirical distribution of @xmath359 to obtain an estimate @xmath360 of @xmath67 by partitioning the latent space into sufficiently small intervals , where we place a uniform kernel of height equal to the proportion of @xmath361 that lies in that interval . our inference is on @xmath362 . recall that @xmath56 denotes the size of the subpopulation . the number @xmath363 is the size of the _ full _ population . this set - up is closer to the motivation for our work , the bounded confidence model , @xcite , and the connection with our model in this paper is made in appendix [ appendix : motivationfora ] . in theory , the general setup in experiment 1 is comparable to the setup in experiment 2 when @xmath364 in experiment 2 is taken to be @xmath365 . we set @xmath366 , @xmath367 , @xmath368 , and @xmath369 . we take the clustering based on @xmath370 as the _ ground truth_. note that @xmath371 here is comparable to @xmath108 in experiment 1 , or more generally , in our model . we set up the simulation to observe roughly 3000 messages amongst the @xmath56 actors in unit time . this translates to @xmath372 per actor per unit time . note that this is a rough estimate as the messaging intensity is time - dependent and stochastic . in figure [ fig : mdsvsexact ] , we have snapshots of @xmath373 and those of @xmath374 for a single simulation run . denote as the latency @xmath375 where the dependency on our choice for a clustering algorithm is suppressed in our notation and for some @xmath376 , @xmath377 : \operatorname*{mari}(\kappa({\bm x}_s ) , \kappa({\bm x}_t ) ) \ge 1-\varepsilon , \text { for a.e.~ } s \in [ t , t]\},\\ & \widehat{\zeta } \equiv \inf\ { t \in [ 0,t ] : \operatorname*{mari}(\kappa(\psi^*(\mathbf p_s ) ) , \kappa({\bm x}_t ) ) \ge 1-\varepsilon , \text { for a.e.~ } s \in [ t , t]\},\end{aligned}\ ] ] where @xmath378 denotes a moving average of the adjusted rand index ( c.f . @xcite and @xcite ) and we fix @xmath379 to be a @xmath52-means clustering algorithm for concreteness . we use the latency as a performance measure for a clustering algorithm @xmath379 under our framework . for our projection , we use a haar basis , i.e. , a set of simple step functions , where the width of the intervals used in the experiment is @xmath380 . also , unlike in experiment 1 , we take @xmath381 these changes require us to modify our algorithm slightly . however , the necessary modifications are straightforward , and we leave the details to the reader . it is important to note that we do not assume knowledge of the latent position of any individual , @xmath382 ; instead , we use only our knowledge of the overall population . as the number @xmath364 gets larger , as shown in @xcite , the dependence among @xmath383 diminishes , agreeing more closely with the model we specified in our framework . we investigate the behavior of our algorithm for small , medium and large values of @xmath364 , showing robustness of our framework in the face of limited information . recall that figure [ fig : mdsvsexact ] shows results for @xmath366 . figure [ fig : latency ] compares the latency for @xmath384 and @xmath366 . the clarity and accuracy of the clustering suffers with significant reductions in information used to estimate the priors @xmath67 . in figure [ fig : mdsvsexact ] , we present snapshots of @xmath385 and @xmath386 for a single simulation run . note that @xmath387 is a cmds embedding of a dissimilarity matrix based on the posteriors @xmath388 . the colors denote the final cluster membership as determined from @xmath52-means clustering with @xmath389 . it is clear that the emerging cluster structure of the @xmath387 lags slightly behind that of @xmath118 in both accuracy and clarity ; comparing the middle two figures , we can see that there are a few data points misclassified at time @xmath65 . indeed , figure [ fig : ari_ma ] shows that the clustering based on the embedded positions mirrors that possible with the true but unobserved latent positions with a small latency . + [ [ computing - environment ] ] computing environment + + + + + + + + + + + + + + + + + + + + + for experiment 1 , we used r 2.14.1 ( 64 bit ) under ubuntu 12.0.4 on an intel core i7 cpu 870 @ 2.93 ghz @xmath390 8 machine with 16 gb ram . for a single run for @xmath391 , @xmath392 and @xmath393 actors , our experiment took @xmath394 , @xmath395 and @xmath396 seconds respectively . for experiment 2 , we used a red hat linux cluster with 24 nodes with 24 @xmath390 2.5 mhz cpus and 132 gb memory each . each monte carlo replicate took a single slot . a single replicate took approximately 3000 seconds . ) distribution for @xmath397 monte carlo experiments . the translucent grey histogram is based on @xmath384 , and the cross - hatch shaded histogram is based on @xmath366 . the latency is defined as the difference between the time @xmath398 at which the moving average of the predictive ari maintained a level of @xmath399 for all @xmath400 and the time @xmath401 at which true locations moving average of the ari maintained a level of @xmath399 for all @xmath402 . the latency can be negative , but is generally small and positive . , scaledwidth=70.0% ] -means clusterings of the embedded @xmath403 and @xmath312 against the @xmath52-means clustering of @xmath389 . note that @xmath65 ( cf . figure [ fig : mdsvsexact ] ) and @xmath404 are nearly identical . , scaledwidth=70.0% ] versus @xmath405 at times @xmath406 . the size of the population used to estimate the prior was @xmath407 . the first row shows the cmds embedded positions ( @xmath408 ) , and the second row shows the latent positions . due to our cmds embedding procedure ( with rotation ) , the 1-dimensional embedding is the first coordinate of a 2-dimensional embedding . we show a 2-dimensional embedding for illustration purposes . , scaledwidth=70.0% ] we have described a strategy for clustering actors based on messaging activities . our analysis is completed by clustering a cmds embedding of posteriors . we have presented ways to simplify posterior analysis on two levels . the first level allows us to obtain an estimate of the posteriors in an online manner . the second level allows us to reduce our analysis to studying diffusion processes , which is often a starting point for addressing the optimal stopping problem . we have illustrated in our numerical experiments that the assumptions used to derive our two simplified approaches are mild enough to be useful for our inference task at hand , i.e. , clustering . we believe that our framework has potential for tackling the problems faced by the social network practitioner regarding emergence of structure . we intend to develop a measure of confidence for our inferred latent positions . this will be crucial to many applications , as it will provide the decision - maker with information about whether to act or to wait for more data to increase the confidence in the inferred positions . a measure of confidence would therefore be a way to establish a stopping rule . noting that we took the parameters of our model to be exogenous , we will need to explore robustness of our inference to incorrect parameter choices and then make explicit an algorithm for parameter estimation . making our algorithm more scalable is also an area of our interest . these areas of future work will be key to applying our framework on substantial problems . this work is partially supported by a national security science and engineering faculty fellowship ( nsseff ) , by the acheson j. duncan fund for the advancement of research in statistics , and by the johns hopkins university human language technology center of excellence ( jhu hlt coe ) . we also thank dr . youngser park for his technical assistance . our work in this paper is in part influenced by a so - called bounded confidence model in @xcite which focuses on establishing a propagation of chaos property of the interacting particles model studied there . when denoting the actors latent positions @xmath409 $ ] , in the bounded confidence model , the _ opportunities _ for ( latent ) position changes that each actor experiences is modeled as a simple poisson process . when there is a change at time @xmath16 , the change is assumed to involve precisely two actors , say , actor @xmath12 and actor @xmath13 , such that their position @xmath410 and @xmath411 differs by at most @xmath371 . this yields an inhomogeneity in the rate at which actors change their locations . then , the exact amount of change is specified by the following formula : @xmath412 where @xmath413 is a fixed constant . roughly speaking , upon interaction , actor @xmath12 keeps @xmath414 percent of its original position , and is allowed to be influenced by @xmath415 percent of the original position of actor @xmath13 , and vice versa . fix constants @xmath416 and @xmath417 . then , define @xmath418 by letting for each @xmath419 and @xmath177 , @xmath420 studied in @xcite particularly is the interaction between @xmath67 and @xmath421 where @xmath67 is the empirical distribution of @xmath422 . as shown in @xcite , the bounded confidence model has an appealing feature that the parameter space for the underlying parameters @xmath423 and @xmath371 can be partitioned according to the type of consensus that the population eventually reaches , namely , a total consensus and a partial consensus . in a total consensus regime , for sufficiently large @xmath16 , everyone is expected to gather tightly around some fixed common point @xmath424 $ ] . on the other hand , in a partial consensus regime , ( depending on @xmath423 and @xmath371 ) , there is a finite collection of distinct values in @xmath425 $ ] separated by at least @xmath371 , to exactly one of which each actor s position is attracted . in particular , the ( asymptotic ) position of actors yields a partition of the actor set when the exact locations of @xmath426 are known . generally , @xmath427)$ ] is contracting toward for some closed convex non - empty disjoint subsets @xmath428 and @xmath429 of @xmath425 $ ] in the sense that for some @xmath316 $ ] , @xmath430 for each @xmath431 and @xmath432 ) = 1 $ ] . in our adaptation , for analytic tractability , we replace the indicator function @xmath433 with @xmath434 , take @xmath67 to be an exogenous modeling element , and take @xmath435 to be potentially time dependent , yielding the operator @xmath436 the second numerical experiment in section [ sec : numericalexperiments ] focuses on the case where the community starts with no apparent clustering but as time passes , each actor becomes a member of exactly one of clusters , where each cluster is uniquely identified by a closed convex subset of the latent space @xmath6 . in this work , we use a model that that captures the action in up to the second order . to begin , note that @xmath437 where @xmath438 and @xmath439 denote respectively the gradient and the hessian of @xmath177 at @xmath26 , and h.o.t . denotes the higher order terms . suppose that @xmath67 is given . now , we have @xmath440 where @xmath441 and @xmath442 are given by the following : @xmath443 dropping the term associated with h.o.t . , we obtain the following : @xmath444 for each @xmath445 , we see that @xmath446 we first consider the second term of the right side of . @xmath447 now , we have that @xmath448 and that @xmath449 hence , @xmath450 next , for the first term of the right side of , we have @xmath451 in summary , for each @xmath452 , we have @xmath453 and our claim follows from this . this section contains two formulas to be used in the next section . our result and proof in lemma [ lem : product - density - formula ] is stated in the same notation as in lemma [ lem : algebra - formula ] . recall that @xmath454 . [ lem : algebra - formula ] let @xmath455 and @xmath456 . then , @xmath457 where @xmath458 is the gram matrix for @xmath459 and @xmath460 is the matrix whose @xmath45-entry is @xmath461 . let @xmath462 and for each @xmath3 , let @xmath463 . first , note that @xmath464 now , @xmath465 our claim follows from this . [ lem : product - density - formula ] let @xmath73 be the standard multivariate normal density defined on @xmath6 . also , fix a sequence @xmath466 , and a sequence @xmath467 . @xmath468 using lemma [ lem : algebra - formula ] , we see that @xmath469 here , we assume , as done in theorem [ thm : projfilterzakai ] , that @xmath470 and @xmath471 , where for simplicity , we have written @xmath472 . in this section , we fix @xmath73 to be the standard multivariate normal density defined on @xmath6 and recall that @xmath473 . also , we fix @xmath207 , and a sequence @xmath474 . [ lem : micro - b - a ] fix @xmath475 , @xmath289 and @xmath207 . for each @xmath52 , @xmath476 let @xmath477 where to simplify the expression of @xmath478 , we have used the fact that @xmath479 also , note that @xmath480 using lemma [ lem : product - density - formula ] with @xmath481 , @xmath482 and @xmath483 , we see that @xmath484 then , for our claim in , it is enough to see that @xmath485 next , we show our claim in . hereafter , to ease our notation , we write @xmath486 for @xmath478 . first , for @xmath487 , we have @xmath488 and hence , @xmath489 on the other hand , for @xmath490 , we have @xmath491 and so , we have @xmath492 our claim in follows . for lemma [ lem : b ] and lemma [ lem : a ] , by @xmath493 , we denote the gram matrix for @xmath494 , and define @xmath495 to be as in lemma [ lem : algebra - formula ] for @xmath496 , @xmath497 , @xmath498 . let @xmath499 to simplify our notation , we let @xmath500 define and note @xmath501 also , denote by @xmath502 the multivariate normal density defined on @xmath6 such that its mean vector is @xmath503 and its covariance matrix is @xmath504 . for @xmath505 and @xmath506 , we write @xmath507 and note that in particular , @xmath508 starting from , it is easy to see that @xmath509 and as a matter of definition , we have @xmath510 lemma [ lem : b ] and lemma [ lem : a ] are associated , respectively , with the first and the second terms appearing in the right side of . [ lem : b ] for each @xmath511 and @xmath512 , we have @xmath513 to ease our notation , we first let @xmath514 it follows that @xmath515 we compute @xmath516 instead of directly working with . first , we observe that @xmath517 and that @xmath518 using lemma [ lem : algebra - formula ] on the third equality , we see that @xmath519 continuing with the calculation , @xmath520 putting together , , and , and plugging in the full expression for @xmath521 , we see that @xmath522 our claim follows after summing over @xmath52 and replacing @xmath523 with its full expression . [ lem : a ] for each @xmath524 , @xmath16 and @xmath12 , @xmath525 note that @xmath526 where @xmath527 we first compute the diagonal terms , i.e. , the @xmath490 cases . note that @xmath528 and also that @xmath529 next , we compute the off - diagonal terms , i.e. , the @xmath487 cases . first , using our calculation just above , we see that we note that @xmath530 our claim follows from this after combining them together , and simplifying the combined term into a matrix notation . here , we will take the convention that @xmath531 is organized as a matrix . by the @xmath12-th row of @xmath118 , we mean @xmath532 . let @xmath533 = \int { \rho}_t(d\bm x ) e^{\imath \langle \bm v , \bm x \rangle},\end{aligned}\ ] ] where for each @xmath534 and @xmath535 , @xmath536 in other words , @xmath537 is the ( random ) conditional characteristic function of @xmath118 . note @xmath538 also , let , for each @xmath534 and @xmath539 , @xmath540.\end{aligned}\ ] ] for each @xmath541 , @xmath542 denotes the function obtained by fixing all other indices different from the @xmath12-th actor indices but letting the @xmath12-th actor indices to be free , and if @xmath542 is in the domain of the operator @xmath543 , with some abuse of notation , we write : @xmath544 similarly , for each @xmath545 , let @xmath546.\end{aligned}\ ] ] in other words , @xmath547 denotes the conditional characteristic function of the @xmath12-th row @xmath548 of @xmath312 , and also , let , for @xmath549 , and @xmath550 , @xmath551.\end{aligned}\ ] ] note that the definition of @xmath552 is actually independent of a particular choice of vertex @xmath12 as they are all identically distributed . one can prove the next result by directly following @xcite , but one needs to adapt to the fact that the underlying process can now be a time - inhomogeneous non - linear markov process . the proof details are left to the reader . for a survey of similar techniques , see also @xcite and @xcite . [ prop : donaldsnyder ] for each @xmath553 and @xmath161 , @xmath554 our proof of theorem [ thm : exactposterior ] is by brute force calculation , starting from proposition [ prop : donaldsnyder ] . in particular , our claim in theorem [ thm : exactposterior ] follows from proposition [ prop : donaldsnyder ] by directly applying lemma [ lemmaaa ] , lemma [ lemmaab ] and lemma [ lemmaac ] which we list and prove now . [ lemmaaa ] for each @xmath555 , @xmath556 fix @xmath16 , @xmath12 , @xmath229 and @xmath26 . then , for each @xmath557 , we have : @xmath558 = 1 + \int_{0}^{\varepsilon}\mathbb{e}\left[\mathcal{a}\left(\mu_{t+s}\right)e^{\imath \langle v , \cdot - x\rangle}\left(x_{t+s , i}\right)\left|x_{t , i}=x\right.\right]ds.\end{aligned}\ ] ] we have @xmath559 and hence , @xmath560 \right| \le 1.\end{aligned}\ ] ] it follows that @xmath561 ds \\ = & \ \mathbb{e}\left[\mathcal{a}\left(\mu_{t}\right ) e^{\imath \langle v , \cdot - x\rangle}(x_{t } ) \left|x_{t}=x\right . \right]\\ = & \ \mathcal{a}\left(\mu_{t}\right ) e^{\imath \langle v , \cdot - x\rangle}(x).\end{aligned}\ ] ] [ lemmaab ] for each @xmath562 , we have : @xmath563 fix @xmath564 and note : @xmath565 and that @xmath566 d\bm v\\ & = \frac{1}{2\pi } \int \lim_{\varepsilon\rightarrow0 } \frac{1}{\varepsilon } \mathbb{e}\left[e^{\imath \langle \bm v , \bm x_{t+\varepsilon } - \bm y\rangle}- e^{\imath \langle \bm v , \bm z -\bm y \rangle } \left|\bm x_t = \bm z\right . \right ] d\bm v.\end{aligned}\ ] ] treating @xmath567 as a generalized function ( i.e. a tempered distribution ) , we have : @xmath568 d\bm v \right ) d\bm y \right)\\ & = \int \rho_t(d\bm z ) \lim_{\varepsilon\rightarrow0 } \frac{1}{\varepsilon } \left ( \mathbb{e}\left [ \int f(\bm y ) \left ( \frac{1}{2\pi } \int e^{\imath \langle \bm v , \bm x_{t+\varepsilon } - \bm y\rangle } d\bm v \right ) d\bm y \left|\bm x_t = \bm z\right . \right ] - \int f(\bm y ) \left ( \frac{1}{2\pi } \int e^{\imath \langle \bm v , \bm z -\bm y \rangle } d\bm v \right ) d\bm y \right)\\ & = \int \rho_t(d\bm z ) \lim_{\varepsilon\rightarrow0 } \frac{1}{\varepsilon } \left ( \mathbb{e}\left [ \int f(\bm y ) \delta_0(\bm x_{t+\varepsilon } - \bm y ) d\bm y \left|\bm x_t = \bm z\right . \right ] - \left ( \int f(\bm y ) \delta_0(\bm z -\bm y ) d\bm y \right ) \right)\\ & = \int \rho_t(d\bm z ) \lim_{\varepsilon\rightarrow0 } \frac{1}{\varepsilon } \left ( \mathbb{e}\left [ f(\bm x_{t+\varepsilon } ) \left|\bm x_t = \bm z\right . \right ] - f(\bm z ) \right)\\ & = \int \rho_t(d\bm z ) ( \mathcal{a}\left(\mu_t\right)f)(\bm z).\end{aligned}\ ] ] [ lemmaac ] for each @xmath562 , we have : @xmath569 note @xmath570 + recall that for each @xmath541 , @xmath542 denotes the function obtained by fixing all other indices different from the @xmath12-th actor indices but letting the @xmath12-th actor indices to be free . fix @xmath571 . let @xmath541 be such that @xmath572 for all @xmath573 . for each @xmath161 , @xmath574 then , the claimed formula follows from our assumption in . suppose that @xmath575 as @xmath576 and that for each @xmath577 , @xmath578 satisfies the rank condition , i.e. , @xmath579 is of rank at least @xmath21 . note that each @xmath580 is a non - empty compact subset of @xmath581 since @xmath582 for any real orthogonal matrix @xmath279 . in particular , for sufficiently small @xmath583 , we may assume that @xmath584 } \| \xi_d^*(m_\varepsilon ) \|_f^2 < \infty$ ] . it is enough to show that for each arbitrary convergent subsequence of @xmath585 , @xmath586 consider an arbitrary convergent subsequence of @xmath585 . we begin by observing some linear algebraic facts . first , any sequence of real orthogonal matrices has a convergent subsequence whose limit is also real orthogonal . next , since both @xmath587 and @xmath277 are of rank @xmath21 , there exists a unique real orthogonal @xmath43 matrix @xmath588 such that @xmath589 and in fact , @xmath590 where @xmath591 is a singular value decomposition of @xmath592 , and @xmath593 is the corresponding _ unique right factor _ in the polar decomposition of @xmath592 . note that this implies the well - definition part of our claim on @xmath594 . also , since @xmath575 , we have that @xmath595 for relevant linear algebra computation details for these facts , see ( * ? ? ? 69 , pg . 370 , pg . 412 , and pg . now , by taking a subsequence if necessary , we also have that for some @xmath270 matrix @xmath596 such that @xmath597 , @xmath598 . then , @xmath599 next , note that if @xmath600 has distinct diagonal elements , then we also have @xmath601 so that @xmath602 . on the other hand , more generally , i.e. , even when there are some repeated diagonal elements , we can find a @xmath43 matrix @xmath603 such that @xmath604 . to see this , note that the @xmath12-th column of @xmath605 is also an eigenvector of @xmath606 for the eigenvalue @xmath607 , and @xmath608 , and hence it follows that for some @xmath43 real orthogonal matrix @xmath609 , we have @xmath610 . moreover , exploiting the block structure of @xmath600 owing to algebraic multiplicity of eigenvalues , we can in fact choose @xmath609 so that @xmath611 . then , @xmath612 now , we have @xmath613 where @xmath614 is a @xmath43 real orthogonal matrix and and implicitly the limit was taken along a further subsequence when necessary . moreover , @xmath615 in summary , we have : @xmath616 by definition of @xmath617 , along with the facts that ( i ) all of the convergent subsequences share the common limit , ( ii ) each subsequence has a convergent subsequence , and ( iii ) @xmath618 and @xmath277 have of full column rank @xmath21 , we have .

we present a novel-way high - dimensional graphical model called the generalized root model ( grm ) that explicitly models dependencies between variable sets of sizewhere is the standard pairwise graphical model . this model is based on taking the-th root of the original sufficient statistics of any univariate exponential family with positive sufficient statistics , including the poisson and exponential distributions . as in the recent work with square root graphical ( sqr ) models which was restricted to pairwise dependencies we give the conditions of the parameters that are needed for normalization using the radial conditionals similar to the pairwise case . in particular , we show that the poisson grm has no restrictions on the parameters and the exponential grm only has a restriction akin to negative definiteness . we develop a simple but general learning algorithm based on-regularized node - wise regressions . we also present a general way of numerically approximating the log partition function and associated derivatives of the grm univariate node conditionals in contrast to which only provided algorithm for estimating the exponential sqr . to illustrate grm , we model word counts with a poisson grm and show the associated-sized variable sets . we finish by discussing methods for reducing the parameter space in various situations .

1606.00813

most standard graphical models are restricted to pairwise dependencies between variables . for example , the ising model for binary data and the multivariate gaussian for real - valued data are popular pairwise graphical models . however , real - world data often exhibits triple - wise , or more generally @xmath0-wise dependencies . for example , the words _ deep _ , _ neural _ and _ network _ often occur together in recent research papers note that this _ triple _ of words refers to something more specific than any of the two words without the third word , i.e. if a document only contains _ neural _ and _ network _ but not _ deep _ , then this may be a more classical paper about shallow neural networks . in the biological domain , genetic , metabolic and protein pathways play an important role in studying the development of diseases and possible interventions . these pathways are known to be complex and involve many genes or proteins rather than just simple pairwise interactions . thus , we seek to begin bridging this gap between pairwise models and complex real - world data that contain complex @xmath0-wise interactions by defining a class of @xmath0-wise graphical models called generalized root models ( grm ) , which can be instantiated for any @xmath4 and any univariate exponential family with positive sufficient statistics including the gaussian ( using the @xmath5 sufficient statistic ) , poisson and exponential distributions . we estimate the graphical model structure and parameters using @xmath3-regularized node - wise regressions similar to previous work @xcite . however , unlike previous work , because the log partition function of the grm node conditionals is not known in closed - form even for the previous work considering the pairwise case@xcite we develop a novel numerical approximation method for the grm log partition function and related derivatives . in addition , we present a newton - like optimization algorithm similar to @xcite to solve the node - regressions which significantly reduces the number of numerical log partition function approximations needed compared to gradient descent . finally , we demonstrate the grm model and parameter estimation algorithm on real - world text data . this paper generalizes the square root graphical model ( sqr ) from @xcite , which only considers pairwise dependencies . @xcite followed the idea of constructing a joint distribution by defining the form of the node - conditional distributions as in @xcite but introduced the idea of taking the square root of the sufficient statistics @xmath6 to form a pairwise term @xmath7 which is linear @xmath8 rather than the pairwise term @xmath9 in @xcite which is quadratic @xmath10 . this elegant modification allowed for arbitrary _ positive _ and _ negative _ dependencies in the poisson sqr graphical model whereas the poisson graphical model in @xcite only permitted negative dependencies a crucial limitation of the poisson models from @xcite . while @xcite proposed three modifications to the original poisson models as defined in @xcite , the modifications lead to distributions with either gaussian - esque thin tails or truncated distributions which required unintuitive cutoff points where the probability mas may concentrate near the corners of the distribution @xcite . though sqr models have great promise , sqr models are limited to pairwise dependencies , and @xcite did not provide an estimation algorithm for the poisson sqr model because the node conditional log partition function is not known in closed form . thus , this paper extends the sqr model class to include @xmath0-wise interactions where @xmath11 and , in addition , instantiates a concrete approximation algorithm for the node conditional log partition function and associated derivatives . in a somewhat different direction , latent variable models provide an implicit and indirect way of modeling complex dependencies . generally , though the explicit dependencies in latent variable models are only pairwise , many variables can be related implicitly through a latent variable . for example , mixture models associate a discrete latent variable with every instance which implicitly introduces dependencies . other more complex latent variable models such as topic models @xcite can introduce even more implicit dependencies in interesting ways . while latent variable models have proven to be practically effective in helping to model complex dependencies , the development of grm models in this paper is distinctive and somewhat orthogonal to latent variable models . as opposed to implicitly modeling dependencies through latent variables , the grm model explicitly models dependencies between observed variables . thus , the discovered dependencies have an intuitive and obvious explanation in terms of the observed data variables . in addition , grm models can be seen as complementary to latent variable models because grm models can be used as base distributions for these latent variable models . for example , @xcite explore using count - valued graphical models in mixtures and topic models . thus , grms can provide new components from which to build more interesting models for real - world situations . finally , node - conditional models such as grm can be estimated using convex optimization problems , which often have theoretical guarantees @xcite whereas latent variable models often require optimizing a non - convex function and struggle with theoretical guarantees . [ [ notation ] ] notation + + + + + + + + let @xmath12 and @xmath13 be the number of dimensions and data instances respectively . let @xmath14 denote the set of nonnegative real numbers and @xmath15 denote the set of nonnegative integers . unless indicated otherwise , we denote vectors with boldface lower case letters ( e.g. @xmath16 , @xmath17 ) and their corresponding scalar values as normal lower case letters ( e.g. @xmath18 , @xmath19 ) . we denote the standard basis vectors as @xmath20^t$ ] and the ones vector as @xmath21^t$ ] . let @xmath22 and @xmath23{\bm{x}}$ ] to be the entry - wise power and @xmath24-th root of the vector @xmath25 . we denote tensors ( or multidimensional arrays ) with parenthesized superscripts as @xmath26 where @xmath0 is the order of the tensor . for example , @xmath27 is a matrix , @xmath28 is a three dimensional tensor , and @xmath29 is a @xmath0-th order tensor . we index tensors using brackets and subscripts , e.g. @xmath30_{1,2,3}$ ] is a scalar value in the multidimensional array at index @xmath31 . we define @xmath32_{\vi } \in \r^{\vmax \times^{\ell-1}}$ ] to be a sub tensor created by fixing the last index to @xmath33 and letting the others vary in matlab colon indexing notation , this would be @xmath34 . for example , if @xmath35 , then @xmath30_{\vi } \in \r^{\vmax \times \vmax}$ ] is a matrix corresponding to the @xmath33-th slice of the tensor @xmath36 . we define @xmath37 to be the outer product operation . for example , @xmath38 and @xmath39 , where @xmath40_{\vi_1\vi_2\vi_3 } = \x_{\vi_1}\x_{\vi_2}\x_{\vi_3}$ ] . for more general sizes , we denote a @xmath0-th outer product to be @xmath41 such that there are @xmath0 copies of @xmath16 and the result is a @xmath0-th order tensor . we define @xmath42^t$ ] . we also denote the inner product operation of two tensors as @xmath43 . with the notation given in the previous section , we will define the grm model . first , let the sufficient statistic and log base measure of a univariate exponential family be denoted as @xmath6 and @xmath44 respectively . we will also define the domain ( or support ) of the random variable to be @xmath45 and it s corresponding measure to be @xmath46 , which is either the counting measure or lebesgue measure depending on whether @xmath47 is discrete or continuous . let us denote a new @xmath24-th root sufficient statistic @xmath48{\t(\x)}$ ] except in the case when @xmath49 where @xmath50 is an even positive integer . if @xmath49 , then we simplify @xmath51 ( rather than the usual @xmath52 ) . for example , if @xmath53 , then @xmath54 ( rather than @xmath55 ) . as in @xcite , this nuanced definition is necessary to recover the multivariate gaussian distribution . however , for notational simplicity , we will merely write @xmath23{\x}$ ] for @xmath56 throughout the paper . note that @xmath57{\x}$ ] for the poisson and exponential grm models . using this simplified notation , we can define the generalized root model for @xmath58 as : @xmath59{\xvec } \ , \circ^{\ell } \right\rangle + \textstyle{\sum_{\vi } } \b(\x_\vi ) - \a(\multipall ) \right ) \label{eqn : full - model } \\ \a(\multipall ) & = \log \int_\domain \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \left\langle \multip^{(\ell)}_{(\j ) } , \sqrt[\j]{\xvec } \ , \circ^{\ell } \right\rangle + \textstyle{\sum_{\vi } } \b(\x_\vi ) \right ) \mathrm{d}\mu(\xvec ) \ , , \label{eqn : full - model - a}\end{aligned}\ ] ] where @xmath60 is the joint log partition function , @xmath61 , @xmath62 are super symmetric tensors of order @xmath63 which are zero whenever two indices are the same . more formally , letting @xmath64 be an index permutation : @xmath65_{\vi_1,\cdots,\vi_\ell } & = [ a^{(\ell)}]_{\pi(\vi_1,\cdots,\vi_\ell ) } & \forall \pi(\cdot ) , \\ \left[a^{(\ell)}\right]_{\pi(\vi_u,\vi_v,\cdots,\vi_\ell ) } & = 0 & \forall \{(u , v , \pi(\cdot ) ) : u \neq v , \vi_u = \vi_v \ } \end{array } \right\ } \ , . \label{eqn : parameter - constraints}\end{aligned}\ ] ] note that the non - zeros of @xmath62 define @xmath63-sized variable sets ( or cliques ) of the underlying graphical model . we now consider several special cases of this model to build some understanding of the grms connection to previous models . the independent model is trivially recovered if @xmath66 : @xmath67 . [ [ square - root - graphical - model ] ] square root graphical model @xcite + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + another special case is the previous sqr models ( i.e. @xmath2 ) from @xcite by taking ( using the notation from @xcite ) @xmath68 , @xmath69 and @xmath70 , where @xmath71 is a column vector of the diagonal entries and @xmath72 has the same off - diagonal entries as @xmath73 but is zero along the diagonal . thus , the sqr model can be written as : @xmath74{\xvec } \rangle + \langle \multip^{(2)}_{(2 ) } , \sqrt[2]{\xvec } \circ \sqrt[2]{\xvec } \rangle + \textstyle{\sum_{\vi } } \b(\x_\vi ) - \a ( \multipall ) \right ) \ , .\end{aligned}\ ] ] [ [ simplified - model - with - only - strongest - interaction - terms ] ] simplified model with only strongest interaction terms + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we consider another special case such that only the strongest interaction ( i.e. when @xmath75 ) terms are non - zero : @xmath76{\xvec } \,\circ^\j \rangle \,+ \ , \textstyle{\sum_{\vi } } \b(\x_\vi ) - \a ( \multipall ) \right ) \ , . \label{eqn : simplified - model}\end{aligned}\ ] ] this restricted parameter space forces @xmath24-wise dependencies to only be through the @xmath24-th root term . for example , pairwise interactions are only available through the sufficient statistic @xmath77{\x_\vi\x_\vit}$ ] and ternary interactions are only available through the sufficient statistic @xmath78{\x_\vi\x_\vit\x_\vir}$ ] . without this restriction interactions would be allowed through multiple terms , e.g. pairwise interactions would be allowed through multiple sufficient statistics @xmath77{\x_\vi\x_\vit } , \sqrt[3]{\x_\vi\x_\vit},\cdots,\sqrt[\k]{\x_\vi\x_\vit}$ ] . thus , this simplified model is more interpretable and easier to learn while still retaining the strongest @xmath24-wise interaction terms . for our experiments , we assume this simplified model unless specified otherwise . as in @xcite , we derive both the _ node _ conditionals and the _ radial _ conditional distributions . an illustration of these two types of univariate conditional distributions can be seen in fig . [ fig : conditional - illustration ] . this _ conditional distribution is critical for the parameter estimation that will be described in later sections ; whereas the _ radial _ conditional distributions are critical for showing the normalization of grm models . the node conditionals are as follows ( see appendix for full derivation ) : @xmath79 where @xmath80 is all other variables except @xmath18 , @xmath81_{\vi } , \ell \sqrt[\j]{\xvec } \,\circ^{\ell-1 } \right\rangle$ ] . this is a univariate exponential family with sufficient statistics @xmath82 , natural parameters @xmath83 and base measure @xmath84 . note that this reduces to the original exponential family if the interaction terms @xmath85 . as in @xcite , we define the _ radial _ conditional distribution by fixing the unit direction @xmath86 of the sufficient statistics but allowing the scaling @xmath87 to be unknown . thus , we get the following _ radial _ conditional distribution ( see appendix for derivation ) : @xmath88 where @xmath89 is the set of possible ratios , @xmath90{\uvec } \ , \circ^{\ell } \rangle$ ] are the exponential family parameters , @xmath91 are the corresponding sufficient statistics and @xmath92 is the base measure . thus , the radial conditional distribution is a univariate exponential family ( as in @xcite ) . the previous exponential and poisson graphical models @xcite could only model negative dependencies . however , we generalize the results from the pairwise sqr model in @xcite and show that grm normalization for any @xmath0 puts little to no restriction on the value of the parameters thus allowing both positive and negative dependencies . for our derivations , let @xmath93 be the set of unit vectors in the positive orthant . the grm log partition function @xmath60 can be decomposed into nested integrals over the unit direction and over the scaling @xmath94 : @xmath95{\radx\uvec } \ , \circ^{\ell } \right\rangle + \textstyle{\sum_{\vi } } \b(\radx\usca_\vi ) \right ) \mathrm{d}\measure(\radx ) \ , \mathrm{d}\uvec \notag \\ & = \log \!\ ! \int\limits_\uset \!\ ! \int\limits_{\radset(\uvec ) } \!\!\!\ ! \exp\left ( \sum_{r\in\mathcal{r } } \natp_r(\uvec ) \radx^r + \tilde{\b}_{\uvec}(\radx ) \right ) \mathrm{d}\measure(\radx ) \ , \mathrm{d}\uvec\end{aligned}\ ] ] where @xmath96 , and @xmath97 and @xmath45 are the measure and domain ( or support ) of the random variable . because @xmath98 is bounded , the joint distribution will be normalizable if the radial conditional distribution is normalizable generalizing the results from @xcite for @xmath11 . informally , the radial conditional distribution converges if the asymptotically largest term of @xmath99 is monotonically decreasing at least linearly . we give several examples in the following paragraphs . [ [ gaussian - grm ] ] gaussian grm + + + + + + + + + + + + for the gaussian grm , we take the gaussian univariate distribution with sufficient statistic @xmath53 and @xmath100 . when @xmath101 ( i.e. the standard multivariate gaussian ) , the largest radial conditional term is @xmath102 where @xmath103 . note that the radial conditional ( i.e. a univariate gaussian ) is normalizable only if @xmath104 for all @xmath105 , which is equivalent to the positive definite condition on the gaussian inverse covariance matrix . we can also consider a gaussian - like model with @xmath106 . in this case , we have that @xmath107 and we need @xmath108 . note that the gaussian grm models for @xmath11 are novel models to the authors best knowledge . [ [ exponential - grm ] ] exponential grm + + + + + + + + + + + + + + + because the exponential distribution also has a constant base measure like the gaussian , the asymptotically largest term is @xmath109 and thus we must have that @xmath108 . however , unlike the gaussian , in the case of the exponential distribution @xmath98 is only positive @xmath3-normalized vectors . this is a significantly weaker condition on the parameters than for a gaussian and allows strong positive and negative dependencies . [ [ poisson - grm ] ] poisson grm + + + + + + + + + + + for the poisson distribution , the base measure is the asymptotically largest term @xmath110 . thus , as in @xcite , the parameters can be arbitrarily positive or negative because eventually the base measure will ensure normalizability . note that this is true for arbitrarily large @xmath0 . as in @xcite , we solve a set of independent @xmath3-regularized node - wise regressions for each node based on the node conditional distributions in sec . [ sec : node - conditionals]using a newton - like method for convex optimization with an non - smooth @xmath3 penalty as in @xcite . more specifically we take the log likelihood of the node conditionals and add an @xmath3 penalty on all interaction terms : @xmath111 where @xmath112_{\vi } , \ell \sqrt[\j]{\xvec_\ii } \,\circ^{\ell-1 } \right\rangle$ ] and @xmath113 is an entry - wise sum of absolute values . note that this is trivially decomposable into @xmath12 subproblems and can thus be trivially parallelized to improve computation speed . we use the newton - like method as in @xcite to greatly reduce computation . the initial innovation from @xcite was that the hessian only needed to be computed over a _ free _ set of variables each newton iteration because of the @xmath3 regularization which suggested sparsity of the parameters . yet , the number of newton iterations was very small compared to gradient descent . in the case of grm models , whose bottleneck is the computation of the gradient of @xmath114 ( at least under our current implementation though it might be possible to significantly reduce this bottleneck ) , this newton - like method provides even more benefit because the gradient only has to be computed a small number of times ( roughly 30 ) in our case rather than the several thousand times that would be needed for running thousands of proximal gradient descent steps for the same level of convergence . in the next section , we derive the gradient and hessian for the smooth part of the optimization as a function of the gradient and hessian of the node conditional log partition function @xmath115 . then , we develop a general method for bounding the log partition function @xmath115 and associated derivatives even though usually no closed - form exists . [ [ notation - for - gradient - and - hessian ] ] notation for gradient and hessian + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath116 be the vectorized form of a tensor . for example , the vectorized form of a @xmath117 matrix is formed by stacking the matrix columns on top of each other to form one long @xmath118 vector . also , let @xmath119 $ ] be analogous to the normal set notation @xmath120 except that the bracket and vertical line notation creates a vector from all the elements concatenated to together . this is similar to a list comprehension in python . for our gradient and hessian calculations , we define the following variable transformations and give them as examples of this notation : @xmath121_{\vi } \right ) \midgiven \ell \leq \j \right ] : \j \in \{1,2,\cdots,\k\}\big\ } \\ & = \bigg\ { \underbrace{\left[\vectorize\left ( \left[\multip^{(1)}_{(1)}\right]_{\vi } \right)\right]}_{\regpvec_{1\vi } } , \underbrace{\left[\vectorize\left ( \left[\multip^{(1)}_{(2)}\right]_{\vi } \right ) , \vectorize\left ( \left[\multip^{(2)}_{(2)}\right]_{\vi } \right)\right]}_{\regpvec_{2\vi } } , \underbrace{\left[\vectorize\left ( \left[\multip^{(1)}_{(3)}\right]_{\vi } \right ) , \cdots\right]}_{\regpvec_{3\vi } } , \cdots , \underbrace{\left [ \cdots , \vectorize\left ( \left[\multip^{(\k)}_{(\k)}\right]_{\vi } \right)\right]}_{\regpvec_{\k\vi } } \bigg\ } , \\ { \mathcal{z}}_{\vi\ii } & = \big\ { \left [ \vectorize\left ( \ell \sqrt[\j]{\xvec_{\vi\ii } } \ , \circ^{\ell-1 } \right ) \midgiven \ell \leq \j \right ] : \j \in \{1,2,\cdots,\k\}\big\ } \\ & = \bigg\ { \underbrace{\left[\vectorize\left ( \sqrt[1]{\xvec_{\ii } } \ , \circ^{0 } \right)\right]}_{\zvec_{1\vi } } , \underbrace{\left[\vectorize\left ( \sqrt[2]{\xvec_{\ii } } \ , \circ^{0 } \right ) , \vectorize\left ( 2\sqrt[2]{\xvec_{\ii } } \ , \circ^{1 } \right)\right]}_{\zvec_{2\vi } } , \underbrace{\left[\vectorize\left ( \sqrt[3]{\xvec_{\ii } } \ , \circ^{0 } \right ) , \cdots\right]}_{\zvec_{3\vi } } , \cdots , \underbrace{\left [ \cdots , \vectorize\left ( \k \sqrt[\k]{\xvec_{\ii } } \ , \circ^{\k-1 } \right)\right]}_{\zvec_{\k\vi } } \bigg\ } \ , .\end{aligned}\ ] ] with this notation , we have that @xmath122 . because each node regression is independent , we focus on solving one of the @xmath12 subproblems for a particular @xmath33 using the notation from above : @xmath123 where @xmath124{\x_{\vi\ii } } + \a([\regp_{\j\vi}^t \zvec_{\j\vi\ii } \given \j \in \{1,\cdots,\k\})$ ] . for notational simplicity , we suppress the dependence on @xmath33 and @xmath125 in the derivations of the gradient and hessian of @xmath126 ( the gradient and hessian are merely the sum over all instances ) . with this simplified notation , the gradient and hessian are as follows ( as functions of @xmath114 , @xmath127 and @xmath128 ) : @xmath129{\x } + \frac{\partial \a}{\partial \natp_{\j } } \right)\zvec_\j \midgiven \j \in \{1,2,\cdots,\k \ } \right ] \ \nabla^2 f({\mathcal{b}}\given \x , { \mathcal{z } } ) & = \left [ \begin{array}{l } \left[\frac{\partial^2 \a}{\partial \natp_1 \partial\natp_\j } \zvec_\j \circ \zvec_1 \midgiven \j \in \{1,2,\cdots,\k\ } \right ] , \vspace{0.5em } \\ \left[\frac{\partial^2 \a}{\partial \natp_2 \partial\natp_\j } \zvec_\j \circ \zvec_2 \midgiven \j \in \{1,2,\cdots,\k\}\right ] , \\ \hspace{5em } \vdots \\ \left[\frac{\partial^2 \a}{\partial \natp_\k \partial\natp_\j } \zvec_\j \circ \zvec_\k \midgiven \j \in \{1,2,\cdots,\k\}\right ] \end{array } \right ] \ , .\end{aligned}\ ] ] note how the gradient and hessian are simple functions of @xmath130 and the derivatives of @xmath115 . thus , we develop bounded approximations for @xmath115 , @xmath131 and @xmath132 next . because the node conditional distributions are not standard distributions , we must either derive the closed - form log partition function as done with the specific case of the exponential sqr model in @xcite , or we must numerically approximate the log partition function and its first and second derivatives . to the authors best knowledge , even for the simplified sqr model with @xmath2 , no closed - form solution to log partition function exists for sqr node conditionals except for the discrete , gaussian and exponential sqr models . thus , we seek a general way to estimate the log partition function and associated derivatives for any univariate exponential family ; we also provide a concrete realization of this approximation method for the poisson grm case . [ [ derivatives - of - anatpvec - reformulated - as - expectations ] ] derivatives of @xmath115 reformulated as expectations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we first note that the gradient and hessian of @xmath115 are merely functions of particular expectations a well - known result of exponential families : @xmath133 \\ \nabla^2 \a(\natpvec ) & = \left [ \begin{array}{c } \left [ \e(\x^{\frac{1}{\j}+\frac{1}{2 } } ) - \e(\x^{\frac{1}{\j}})\e(\x ) \given \j \in \{1,\cdots,\k\ } \right ] \\ \left [ \e(\x^{\frac{1}{\j}+\frac{1}{2 } } ) - \e(\x^{\frac{1}{\j}})\e(\x^{\frac{1}{2 } } ) \given \j \in \{1,\cdots,\k\ } \right ] \\ \hspace{3em } \vdots \\ \left [ \e(\x^{\frac{1}{\j}+\frac{1}{\k } } ) - \e(\x^{\frac{1}{\j}})\e(\x^{\frac{1}{\k } } ) \given \j \in \{1,\cdots,\k\ } \right ] \\ \end{array } \right ] \ , .\end{aligned}\ ] ] thus , we need to compute expectations for at most @xmath134 functions of the form @xmath135 . [ [ definition - of - ma - to - unify - approximations ] ] definition of @xmath136 to unify approximations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to develop our approximations under a unified framework , let us define the following function @xmath136 and its subfunctions denoted @xmath137 and @xmath138 : @xmath139 by simple inspection , we see that @xmath140 and @xmath141 . thus , by approximating @xmath136 , we can approximate all the necessary derivatives . if @xmath142 , then this is simply the log partition function of the base exponential family , which is usually known in closed form . if @xmath143 ( as we will develop in the next sections ) , then we can create a modified @xmath144 and @xmath145 such that @xmath146 and @xmath147thus also allowing us to use the machinery of the base exponential family to compute the needed integrals . [ [ overall - approach - to - bounding - ma ] ] overall approach to bounding @xmath136 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our approach splits the integral into @xmath148 integrals which bound the integral over different subdomains of the domain . we will choose the subdomains in appropriate way to minimize error , which will be described in a future section . for each subdomain , we will form linear upper and lower bounds for @xmath138 so that we can then use the cdf function of the base exponential family to approximate the integrals over these subdomains . first , we will describe how to compute linear upper and lower bounds to @xmath138 so that the integrals reduce to the original exponential family . because we can determine the concavity of each region of @xmath138 , we can form linear upper and lower bounds using the theory of convexity . the secant line and the first - order taylor series approximation form upper and lower bounds or vice versa depending on concavity . we can bound the tails of @xmath138 with a constant function or taylor series approximation as appropriate . see appendix for details on linear approximations for @xmath138 . if @xmath138 is upper and lower bounded by a linear functions , i.e. @xmath149 , then we can form a modified functions of @xmath137 that will be upper and lower bounds of @xmath150 : @xmath151 assuming @xmath152 and @xmath153 are valid parameters , we can then use the original exponential family cdf which is usually known in closed form to compute the needed integrals . now that we have linear upper and lower bounds for @xmath138 , we can upper and lower bound @xmath136 using the cdf of the original exponential family to compute the needed integrals ( see appendix for more derivation ) : @xmath154 where the domain is split into disjoint subdomains , i.e. @xmath155 , @xmath156 are either @xmath157 or @xmath158 depending on whether the upper or lower bound is needed , @xmath159 and @xmath160 are the log partition function and cdf of the original exponential family . note that assuming @xmath159 and @xmath160 are available in closed form as is the case for the poisson distribution this approximation can be computed in @xmath161 time . [ [ algorithm - to - find - appropriate - subdomains - domain_i ] ] algorithm to find appropriate subdomains @xmath162 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we need that every subdomain has a constant concavity ( i.e. either concave or convex over the subdomain ) in order to use taylor series and secant line bounds ( and a constant bound for the tails ) . thus , we use the following algorithm to find subdomains to help minimize the difference between the upper and lower bounds ( an illustration of the method can be seen in fig . [ fig : approx - illustration ] . ) : 1 . find all real roots of @xmath163 , denoted @xmath164 so we know the inflection points ( which will define the regions of constant concavity ) . 2 . use inflection points and endpoints of domain ( e.g. @xmath165 and @xmath166 for poisson ) to define the initial subdomains . 3 . compute initial bounds for these subdomains using eqn . [ eqn : m - approx ] . 4 . split the subdomain with the largest difference between upper and lower bounds ( i.e. the subdomain with the largest error ) . recompute bounds for the two new subdomains formed by splitting the largest error subdomain . repeat previous two steps until @xmath167 domains have been obtained . note that the roots of @xmath163 can be solved by expanding to a polynomial and then computing the eigenvalues of the companion matrix . for example if @xmath168 , then @xmath169 . note that the zeros of this function are equal to the zeros of @xmath170 . thus , we can let @xmath171 and form the polynomial function @xmath172 . we can then solve the zeros of this polynomial by forming the companion matrix and solving for the eigenvalues . however , we only need the real zeros and we do not care about multiplicity so it may be faster to use a direct root finding algorithm instead though we have not explored this option . we computed the poisson grm model on two datasets : classic3 and grolier encyclopedia articles . the classic3 dataset contains 3893 research abstracts from library and information sciences , medical science and aeronautical engineering . the grolier encyclopedia dataset contains 5000 random articles from the grolier encyclopedia . we set @xmath106 , @xmath173 and @xmath174 for our experiments . we chose 10 interval endpoints ( i.e. 9 subdomains ) for our approximations . note that this means there are at least @xmath175 possible parameters . we give the top 10 positive parameters for individual , edge - wise and triple - wise combinations . the top 50 ( unless there are less than 50 non - zeros ) of both negative and positive dependencies for single , pairwise and triple - wise dependencies can be found in the appendix . these results illustrate that our model and algorithm can find interesting pairwise and triple - wise words . the timing for these experiments using prototype code in matlab on tacc maverick cluster ( https://portal.tacc.utexas.edu/user-guides/maverick ) was 2653 seconds for the classic3 dataset and 5975 seconds for the grolier dataset . given the extremely large number of parameters to be optimized , this gives evidence that grm models are computationally tractable while still wanting for some improvement . while it may seem at first that this model is impractical for even @xmath176 , we suggest some practical ideas for reducing the parameter space . first , if some parameters are known or expected a priori to be non - zero , we could only allow those parameters to be non - zero . for example , known genetic pathways could be encoded as @xmath0-wise cliques . thousands of known pathways could be added which would only incur thousands of parameters , which is very small relative to all possible parameters . second , the optimization could proceed in a stage - wise fashion such that the first a model is fit for @xmath66 , then this model is used to choose which parameters to allow in the next model of @xmath101 , etc . for example , we could first train a model with only pairwise parameters ( @xmath177 ) . then , we could find all triangles in the discovered graph and only add these parameters for training a model with @xmath106 . this heuristic would significantly reduce the number of possible parameters if the parameters are assumed to be sparse ( as is usually the case with @xmath3-regularized objectives ) . third , the tensors could be constrained to be low - rank and thus only @xmath178 values for each tensor would be needed . for example , we could assume that the pairwise tensors are low - rank matrices . for higher order tensors , a similar idea could hold , e.g. @xmath179 , where @xmath180 is @xmath181 . we generalize the previous sqr @xcite model to include factors of size @xmath11 . we study this general distribution by giving the node and radial conditional distributions , which provides simple conditions for normalization of the grm class of models . we then develop an approximation technique for estimating the node - wise log partition function and associated derivatives for the poisson case note that @xcite only provided an algorithm for approximating the exponential sqr model . finally , we qualitatively demonstrated our model on two real world datasets . this work was supported by nsf ( dge-1110007 , iis-1149803 , iis-1447574 , iis-1546459 , dms-1264033 , ccf-1320746 ) and aro ( w911nf-12 - 1 - 0390 ) . @xmath182{\xvec_{\vi0 } + \x_\vi\evec_\vi } \ , \circ^{\ell } \rangle + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \langle \multip^{(\ell)}_{(\j ) } , ( \sqrt[\j]{\xvec_{\vi0 } } + \sqrt[\j]{\x_\vi\evec_\vi } ) \ , \circ^{\ell } \rangle + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \left\langle \multip^{(\ell)}_{(\j ) } , \sum_{m=0}^\ell \binom{\ell}{m } \left(\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell - m}\right ) \circ \left(\sqrt[\j]{\x_\vi\evec_\vi } \ , \circ^{m}\right ) \right\rangle + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \sum_{m=0}^\ell \binom{\ell}{m } \left\langle \multip^{(\ell)}_{(\j)},\left(\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell - m}\right ) \circ \left(\sqrt[\j]{\x_\vi\evec_\vi } \ , \circ^{m}\right ) \right\rangle + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \sum_{m=0}^\ell \binom{\ell}{m } \left\langle \left[\multip^{(\ell)}_{(\j)}\right]_{\mathbf{i}(\vi , m)},\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell - m } \right\rangle \x_\vi^{m/\j } + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \sum_{m=1}^\ell \binom{\ell}{m } \left\langle \left[\multip^{(\ell)}_{(\j)}\right]_{\mathbf{i}(\vi , m)},\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell - m } \right\rangle \x_\vi^{m/\j } + \b(\x_\vi ) \right ) \tag{$m=0 $ is constant}\\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \binom{\ell}{1 } \left\langle \left[\multip^{(\ell)}_{(\j)}\right]_{\mathbf{i}(\vi,1)},\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell-1 } \right\rangle \x_\vi^{1/\j } + \b(\x_\vi ) \right ) \tag{$m\geq 2 $ are all zero since subtensors are zero by construction}\\ & \propto \exp\left ( \sum_{\j=1}^\k \left ( \sum_{\ell=1}^\j \ell \left\langle \left[\multip^{(\ell)}_{(\j)}\right]_{\mathbf{i}(\vi,1)},\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell-1 } \right\rangle \right ) \x_\vi^{1/\j } + \b(\x_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \natp_{\j\vi } \x_\vi^{1/\j } + \b(\x_\vi ) \right ) \ , , \end{aligned}\ ] ] where @xmath183_{\vi},\sqrt[\j]{\xvec_{\vi0 } } \,\circ^{\ell-1 } \right\rangle \right)$ ] . see notation section for definition of @xmath184_{\vi}$ ] . this is a univariate exponential family with sufficient statistics @xmath82 , natural parameters @xmath83 , and base measure @xmath84 . this recovers the sqr node conditional from @xcite with @xmath101 . as in @xcite , we define the _ radial _ conditional distribution by fixing the unit direction @xmath86 of the sufficient statistics but allowing the scaling @xmath87 to be unkown . thus , we get the following _ radial _ conditional distribution : @xmath185{\radx\uvec } \ , \circ^{\ell } \rangle + \textstyle{\sum_{\vi } } \b(\radx\usca_\vi ) \right ) \\ & \propto \exp\left ( \sum_{\j=1}^\k \sum_{\ell=1}^\j \langle \multip^{(\ell)}_{(\j ) } , \sqrt[\j]{\uvec } \ , \circ^{\ell } \rangle \radx^{\frac{\ell}{\j } } + \textstyle{\sum_{\vi } } \b(\radx\usca_\vi ) \right ) \\ & \propto \exp\left ( \sum_{r\in\mathcal{r } } \natp_r(\uvec ) \radx^r + \tilde{\b}_{\uvec}(\radx ) \right ) \ , , \end{aligned}\ ] ] where @xmath186 is the set of possible ratios , @xmath90{\uvec } \ , \circ^{\ell } \rangle$ ] are the exponential family parameters , @xmath91 are the corresponding sufficient statistics , and @xmath92 is the base measure . thus , the radial conditional distribution is a univariate exponential family . [ [ taylor - series - linear - bound ] ] taylor series linear bound + + + + + + + + + + + + + + + + + + + + + + + + + + upper bound if concavity = -1 and lower bound if concavity = 1 : @xmath188 [ [ secant - linear - bound ] ] secant linear bound + + + + + + + + + + + + + + + + + + + upper bound if concavity = 1 and lower bound if concavity = -1 : @xmath189 [ [ tail - bounds ] ] tail bounds + + + + + + + + + + + we know there are only a finite number of inflection points so let us take the @xmath47 value for the last inflection point , denoted @xmath190 . by simple asymptotic analysis , we know that the largest non - zero term will dominate eventually . let s assume w.l.o.g . that @xmath191 dominates , @xmath192 , then we can take @xmath193 , @xmath194 . ] and @xmath195 . then , we know that after the last inflection point , the concavity will be negative . in addition , we know that the @xmath196 as @xmath197 . the function must be monotonically increasing after the last inflection point . proof by contradiction : suppose the monotonicity is negative after the last inflection point . then , because the @xmath138 is a continuous function and @xmath196 as @xmath197 , the function must eventually have a positive monotonicity . yet this would switch from negative monotonicity to positive monotonicity after the last inflection point . however , this would be an inflection point that is greater than the assumed last inflection point which leads to a contradiction . the case where @xmath198 can be proved similarly . thus , we can use a constant function for an upper bound if concavity = 1 . and we can use a constant function as a lower bound if concavity = -1 . a taylor series approximation forms an upper or lower bound depending on concavity . < < < largest 1-tuples , j = 1 , ell = 1 > > > -0.6256 information -0.6906 flow -0.8097 library -1.1693 pressure -1.4066 system -1.4090 theory -1.4209 results -1.4248 data -1.4597 patients -1.5737 found -1.5791 method -1.6292 cells -1.6566 analysis -1.7161 given -1.7198 use -1.7389 number -1.7525 used -1.7550 study -1.7860 made -1.7884 effect -1.8054 time -1.8089 body -1.8371 research -1.8563 cases -1.8958 normal -1.9246 effects -1.9436 present -1.9690 discussed -1.9803 shock -1.9899 presented -2.0112 wing -2.0135 surface -2.0275 large -2.0295 case -2.0447 obtained -2.0603 new -2.0695 paper -2.0776 libraries -2.0834 high -2.0884 problems -2.1160 methods -2.1162 well -2.1311 development -2.1414 general -2.1417 growth -2.1659 problem -2.2109 jet -2.2112 terms -2.2313 systems -2.2416 form < < < positive 2-tuples , j = 2 , ell = 2 > > > 4.9827 boundary + layer 4.2583 heat + transfer 3.9493 tunnel + wind 3.3190 edge + leading 3.1510 bone + marrow 3.0395 angle + attack 2.8711 skin + friction 2.5638 growth + hormone 2.3182 plate + flat 2.2675 shock + wave 2.2548 mach + numbers 2.1277 number + mach 2.1047 number + reynolds 2.0561 agreement + good 2.0306 attack + angles 2.0102 document + documents 1.9377 cells + cell 1.7727 journals + journal 1.5436 library + libraries 1.5291 lift + drag 1.5090 wing + wings 1.4283 shells + cylindrical 1.4192 buckling + shells 1.4126 temperature + thermal 1.4080 free + stream 1.3980 ratio + aspect 1.3892 equations + differential 1.3721 boundary + layers 1.3675 point + stagnation 1.2517 shock + waves 1.2495 heat + temperature 1.2285 reynolds + transition 1.2000 wings + aspect 1.1793 temperature + temperatures 1.1554 thin + shells 1.1335 science + scientific 1.1279 cells + marrow 1.1188 numbers + reynolds 1.1004 cylinder + circular 1.0910 renal + kidney 1.0907 pressure + pressures 1.0384 high + speed 1.0294 layer + laminar 1.0266 information + retrieval 1.0232 patients + therapy 1.0159 patients + cancer 1.0127 jet + nozzle 0.9848 group + groups 0.9660 experimental + theoretical 0.9528 buckling + stress < < < negative 2-tuples , j = 2 , ell = 2 > > > -0.8428 flow - library -0.6400 information - pressure -0.5896 flow - cells -0.5885 pressure - library -0.5737 library - patients -0.5570 flow - system -0.5559 information - cells -0.5533 information - heat -0.5185 information - patients -0.5184 flow - patients -0.4941 theory - patients -0.4902 information - normal -0.4695 information - found -0.4600 information - effect -0.4587 library - theory -0.4489 library - normal -0.4479 library - cells -0.4213 library - effects -0.4097 flow - retrieval -0.4071 flow - growth -0.3925 library - found -0.3857 library - cases -0.3667 information - wing -0.3653 information - case -0.3580 pressure - cells -0.3548 flow - information -0.3383 flow - subject -0.3364 results - library -0.3316 information - effects -0.3242 information - temperature -0.3170 information - surface -0.3143 flow - children -0.3131 library - obtained -0.2957 flow - book -0.2904 flow - research -0.2903 information - mach -0.2863 theory - cells -0.2858 library - effect -0.2824 information - equations -0.2817 flow - literature -0.2780 flow - index -0.2754 flow - buckling -0.2745 analysis - patients -0.2692 information - cases -0.2650 information - shock -0.2626 information - boundary -0.2583 information - method -0.2566 information - high -0.2524 library - body -0.2522 information - ratio < < < positive 3-tuples , j = 3 , ell = 3 > > > 0.5067 layer + skin + friction 0.3171 information + retrieval + storage 0.3149 pressure + number + mach 0.3118 layer + plate + flat 0.2672 flow + given + case 0.2411 flow + plate + flat 0.1759 number + mach + investigation 0.1390 number + mach + conducted 0.1340 wing + ratio + aspect 0.1317 number + based + reynolds 0.1100 pressure + ratio + jet 0.1072 heat + transfer + coefficients 0.0973 system + retrieval + user 0.0972 boundary + layer + experiments 0.0926 mach + free + stream 0.0862 pressure + layer + gradient 0.0825 heat + temperature + coefficient 0.0716 pressure + supersonic + base 0.0711 boundary + shock + interaction 0.0709 boundary + layer + distance 0.0709 layer + shock + interaction 0.0678 theory + experimental + experiment 0.0594 flow + fluid + steady 0.0578 flow + boundary + present 0.0554 flow + body + revolution 0.0537 flow + case + form 0.0524 flow + body + shape 0.0511 information + data + base 0.0479 boundary + layer + found 0.0478 cells + bone + marrow 0.0462 flow + theory + approximation 0.0457 data + retrieval + base 0.0430 results + number + higher 0.0420 layer + temperature + compressible 0.0417 number + mach + static 0.0403 boundary + injection + mass 0.0397 number + mach + approximately 0.0393 flow + hypersonic + region 0.0365 theory + wing + wings 0.0360 growth + human + hormone 0.0359 number + mach + lower 0.0357 heat + transfer + blunt 0.0337 number + mach + increasing 0.0337 number + boundary + increasing 0.0328 boundary + layer + measurements 0.0306 number + boundary + reynolds 0.0305 flow + body + conditions 0.0300 information + field + science 0.0293 flow + number + based 0.0290 flow + data + experimental < < < negative 3-tuples , j = 3 , ell = 3 > > > -0.3490 boundary - layer - conditions -0.2025 number - mach - numbers -0.0907 boundary - layer - wing -0.0566 flow - number - numbers -0.0548 boundary - layer - time -0.0450 layer - shock - laminar -0.0433 number - mach - solution -0.0353 boundary - layer - jet -0.0274 heat - transfer - jet -0.0265 boundary - solutions - turbulent -0.0236 flow - mach - reynolds -0.0208 pressure - number - numbers -0.0101 boundary - layer - flutter -0.0019 flow - mach - velocity -0.0014 number - mach - problems .... < < < largest 1-tuples , j = 1 , ell = 1 > > > -1.6202 american -1.7936 century -1.8188 john -1.8830 called -1.8866 city -1.9162 world -1.9543 life -2.0359 united -2.1299 system -2.1328 university -2.1390 family -2.1473 time -2.1591 war -2.1858 include -2.1870 english -2.2457 water -2.2485 history -2.2559 de -2.2694 form -2.3326 major -2.3442 national -2.3523 french -2.3537 william -2.3708 art -2.3808 found -2.4045 name -2.4049 modern -2.4255 music -2.4315 power -2.4433 king -2.4445 social -2.4455 british -2.4596 usually -2.4718 charles -2.4784 south -2.4923 law -2.4995 north -2.5030 repr -2.5165 species -2.5247 theory -2.5383 human -2.5520 ft -2.5530 black -2.5566 government -2.5660 west -2.5773 york -2.5777 church -2.5841 school -2.5890 development -2.5899 common < < < positive 2-tuples , j = 2 , ell = 2 > > > 8.7140 km + mi 3.9800 language + languages 2.9617 china + chinese 2.6237 plants + plant 2.5229 deg + temperatures 2.5152 music + musical 2.4147 spanish + spain 2.1495 novel + novels 2.1059 art + painting 2.0869 poetry + poet 2.0738 agricultural + agriculture 2.0492 war + civil 2.0024 literature + literary 1.8595 french + france 1.8405 german + germany 1.7779 culture + cultural 1.7453 china + asia 1.7368 india + asia 1.7088 system + systems 1.6836 city + york 1.6818 west + east 1.5932 africa + african 1.5932 deg + mm 1.5137 southern + northern 1.4987 architecture + building 1.4788 style + architecture 1.4431 body + blood 1.4359 role + played 1.4313 sea + ocean 1.4304 cells + blood 1.4118 science + scientific 1.4074 century + centuries 1.3808 population + sq 1.3723 social + society 1.3611 italian + renaissance 1.3611 music + opera 1.3600 ocean + pacific 1.3569 cause + disease 1.3548 cities + urban 1.3419 war + army 1.3308 united + countries 1.3065 animals + animal 1.2998 church + christian 1.2991 art + museum 1.2907 education + schools 1.2809 programs + program 1.2715 deg + temperature 1.2550 world + war 1.2515 party + leader 1.2477 government + federal < < < negative 2-tuples , j = 2 , ell = 2 > > > -0.2449 life - languages -0.2179 century - species -0.2041 city - species -0.1575 war - species -0.1156 city - sq -0.0941 century - june -0.0911 war - languages -0.0771 war - example -0.0725 city - theory -0.0718 city - common -0.0707 city - system -0.0684 city - called -0.0666 century - cells -0.0623 war - cells -0.0501 american - eng -0.0436 city - english -0.0420 century - president -0.0415 american - ft -0.0379 art - america -0.0367 city - found -0.0359 city - form -0.0340 city - development -0.0338 war - form -0.0333 war - usually -0.0301 called - deg -0.0279 war - forms -0.0247 city - time -0.0241 city - united -0.0237 war - human -0.0199 century - july -0.0199 century - party -0.0152 war - theory -0.0151 american - king -0.0128 city - family -0.0119 century - south -0.0112 called - eng -0.0109 city - life -0.0103 american - deg -0.0102 american - east -0.0080 american - city -0.0071 war - water -0.0058 city - usually -0.0053 american - cells -0.0051 form - university -0.0046 city - body -0.0039 city - process -0.0029 century - american -0.0027 ft - english -0.0008 city - cells < < < positive 3-tuples , j = 3 , ell = 3 > > > 0.3126 american + city + york 0.2773 city + population + center 0.2549 population + deg + mm 0.1971 major + population + persons 0.1641 ft + sea + level 0.1523 american + south + america 0.1515 deg + sq + consists 0.1351 city + deg + july 0.1330 war + civil + union 0.1173 population + deg + elected 0.1170 american + united + english 0.1140 war + congress + program 0.1054 population + sq + persons 0.0986 american + french + british 0.0972 language + includes + languages 0.0965 world + war + japanese 0.0923 major + time + changes 0.0899 century + world + laws 0.0879 life + human + stage 0.0875 north + south + president 0.0870 city + population + university 0.0845 century + world + war 0.0830 km + mi + discovered 0.0813 war + united + received 0.0770 city + river + historical 0.0760 century + history + short 0.0744 century + english + story 0.0738 war + south + union 0.0721 american + war + congress 0.0698 world + united + david 0.0676 century + history + active 0.0674 century + history + wide 0.0669 war + united + america 0.0668 war + army + june 0.0664 city + km + population 0.0664 government + national + rise 0.0663 century + form + appeared 0.0649 war + united + caused 0.0644 century + world + separate 0.0642 united + people + continued 0.0638 city + united + urban 0.0638 century + time + applied 0.0633 world + united + building 0.0629 world + south + iron 0.0621 world + population + rate 0.0614 city + university + center 0.0605 century + time + studied 0.0599 american + war + people 0.0586 north + km + fish 0.0582 world + william + series < < < negative 3-tuples , j = 3 , ell = 3 > > > -0.2560 city - population - york -0.1206 km - mi - america -0.1076 american - km - mi -0.0586 km - mi - york -0.0518 war - north - example -0.0429 km - mi - social -0.0413 km - mi - family -0.0294 km - mi - own -0.0293 km - mi - style -0.0263 city - population - style -0.0262 km - mi - theory -0.0253 city - center - sq -0.0208 km - mi - law -0.0207 km - mi - human -0.0182 km - mi - example -0.0157 city - population - greek -0.0145 km - mi - water -0.0125 called - km - mi -0.0114 england - language - languages -0.0099 km - mi - greek -0.0093 time - km - mi -0.0042 km - mi - english -0.0025 mi - population - america -0.0021 population - america - sq -0.0012 war - km - mi -0.0003 american - city - center ....

we study a conjecture by fendley , ludwig and saleur for the nonlinear conductance in the boundary sine - gordon model . they have calculated the perturbative series of twisted partition functions , which require particular ( unphysical ) imaginary values of the bias , by applying the tools of jack symmetric functions to the `` log - sine '' coulomb gas on a circle . we have analyzed the conjectured relation between the analytically continued free energy and the nonlinear conductance in various limits . we confirm the conjecture for weak and strong tunneling , in the classical regime , and in the zero temperature limit . we also shed light on this special variant of the-method and compare it with the real - time keldysh approach . in addition , we address the issue of quantum statistical fluctuations . quantum transport , full counting statistics , quantum impurity problem , coulomb gas , jack polynomials 05.60.gg , 05.40.ca , 71.10.pm , 73.50.td

1003.0366

quantum brownian motion ( qbm ) of a particle or charge in a tilted periodic potential is one of the basic subjects of nonequilibrium quantum - statistical mechanics and a key model for a multitude of quantum transport phenomena in condensed matter @xcite . of particular interest is a duality symmetry between the weak- and strong - binding representations of the model @xcite . there is a close correspondence of the qbm model with seemingly different models : ( i ) a resistively shunted josephson junction @xcite , ( ii ) a quantum impurities in a 1d wire @xcite , ( iii ) tunneling of edge currents through constrictions in fractional quantum - hall devices @xcite , and ( iv ) a coherent one - channel conductor in a resistive electromagnetic environment @xcite . in fact , all these models are directly related to each other . the calculation of the dynamics may be done with the use of the feynman - vernon influence functional method @xcite or the equivalent nonequilibrium keldysh technique . in this theory , the reduced density matrix ( rdm ) is expressed as a double path sum over all possible paths on the rdm - plane . the path sum problem is formally equivalent to the grand - canonical sum of a charge gas with complex - valued interactions . of particular interest is the scaling or field theory limit . until now the respective real - time coulomb gas has not been solved analytically barring special cases . one may also consider the system in imaginary time compactified on a circle of length @xmath1 , where @xmath2 is the temperature . in the field theory limit , the charges of the respective coulomb gas interact with a `` log - sine '' long range interaction . the compactified coulomb gas has been solved analytically for particular unphysical values of the bias with the tool of jack symmetric functions @xcite . the consideration of nonequilibrium quantum transport requires ( i ) analytic continuation of the calculated partition function to the physical bias , and ( ii ) a conjectured relation between the analytically continued partition function and transport quantities @xcite . here we analyze this route in different parameter regimes , and we compare the findings with results available by different methods . our paper is organized as follows . in section [ sgtm ] we introduce several related models , which may describe nonequilibrium quantum transport of a particle or a charge , josephson junction dynamics , scattering off an impurity embedded in a luttinger liquid , and conductance of one - channel conductors in a resistive electromagnetic environment . in section [ scgr ] we review the solution of the `` log - sine '' coulomb gas on a circle with winding numbers . in section [ scond ] we study the conjecture for the nonlinear conductance in the weak - tunneling regime , in the classical regime , and in the zero temperature limit . in section [ snoise ] we compare the results of the imaginary time approach with those of the real - time keldysh method . we also address the issue of statistical fluctuations . section [ sconcl ] is devoted to concluding remarks . the boundary sine - gordon ( bsg ) model is a one - dimensional quantum field with gapless bulk excitations and a sinusoidal boundary interaction . the model is integrable and can be treated by a variety of powerful techniques . it constitutes a generic quantum transport model which displays a wide range of interesting characteristics , and it is a realistic model to describe seemingly different physical situations . in fact , all these models are closely related to each other . they have in common that a nonlinear quantum degree of freedom is coupled to a harmonic field . a quantum brownian particle moving in a periodic potential is a key model for many transport phenomena in condensed matter . of particular interest is the case of strict ohmic dissipation in which the regions of weak and strong corrugation are related by an exact self - duality . the dissipative dynamics of a particle moving in a tilted , sinusoidally corrugated potential of periodicity length @xmath3 may be described by the translational - invariant hamiltonian with bilinear environmental coupling . ] @xmath4 & & \!\!\!\!\!+ \ , { \displaystyle \sum_\alpha \omega_\alpha\left(b_\alpha^\dagger- \frac{\lambda_\alpha}{\omega_\alpha}\frac{x}{x_0^{}}\right ) \left(b_\alpha^{}- \frac{\lambda_\alpha}{\omega_\alpha}\frac{x}{x_0^{}}\right ) \;. } \end{array}\ ] ] the `` weak - binding '' form ( [ hwb ] ) is the convenient representation for the perturbative expansion of transport quantities in the corrugation strength @xmath5 . of particular interest is the normalized mobility @xmath6 ( particle transport ) or normalized conductance @xmath7 ( charge transport ) , where @xmath8 and @xmath9 are mobility and conductance in the absence of the corrugation . we have @xmath10 . in the sequel we use the language of charge transport . in the tight - binding regime @xmath11 , where @xmath12 is the distance of low - lying levels in the individual well regimes , and @xmath13 is the potential drop between neighboring wells , only the lowest energy band is occupied . then the system is effectively reduced to a tight - binding lattice . the tight - binding or tunneling hamiltonian with transfer matrix element @xmath14 and translational invariant coupling is @xmath15 & & \!\!\!\!\ ! + \,{\displaystyle \sum_\alpha \omega_\alpha\left(b_\alpha^\dagger- \frac{\mu_\alpha}{\omega_\alpha}\frac{y}{y_0}\right ) \left(b_\alpha^{}- \frac{\mu_\alpha}{\omega_\alpha}\frac{y}{y_0}\right ) } \ ; , \end{array}\ ] ] where @xmath16 is the tb lattice constant , and @xmath17 is the position in tb representation . since we postulate the same tilting force in both models , the potential drops between neighboring sites are related by @xmath18 in the scaling or field theory limit , the spectral densities of the environmental couplings , @xmath19 g_{\rm tb}^{}(\omega ) & = & \sum_\alpha\mu_\alpha^2 \,\delta(\omega-\omega_{\alpha } ) \ ; , \end{array}\ ] ] are of strict ohmic form @xmath20 g_{\rm tb}(\omega ) & = & 2 k_{\rm tb}\ , \omega \ ; . \end{array}\ ] ] the dimensionless so - called kondo parameters @xmath21 and @xmath22 are related to the ohmic viscosity @xmath23 by @xmath24 in the ohmic regime ( [ gft ] ) , the tb and wb model ( [ hwb ] ) and ( [ htb ] ) are related by an exact self - dual symmetry , as noted first by schmid for the equilibrium density matrix at @xmath25 @xcite . later , the duality was proven to hold also at finite @xmath2 and for real time dynamics @xcite . in the self - dual mapping we have @xmath26 and the normalized conductance in the wb model @xmath27 is directly related to the corresponding quantity in the associated tb model ( [ htb ] ) , @xmath28 the self - duality relation ( [ dual2 ] ) exchanges weak and strong corrugation , as well as the regimes of weak and strong environmental coupling . the explicit calculation of ( [ dual2 ] ) may be executed with the use of the feynman - vernon influence functional method @xcite , or directly by unitary transformations of the wb and tb hamiltonians @xcite . computations are conveniently performed in the discrete tb model . with use of the self - duality , the results can then easily be transferred to the dual wb model ( see below ) . with regard to the partition function , we consider the system in imaginary time compactified on a circle of length @xmath1 . the partion functions @xmath29 and @xmath30 can be expanded in even powers of @xmath14 and @xmath31 , respectively . the perturbative series of the partition function is equivalent to the grand - canonical sum of a one - dimensional gas of positive and negative charges @xmath32 with overall neutrality . the charges represent the forward and backward moves in the tb or wb potential . all effects of the environmental coupling are described by the imaginary - time charge - charge correlator @xmath33\,q(0)\,\rangle } _ { } \\[4 mm ] \ ! & = & \!{\rm e}^{-\,w(\tau ) } _ { } \ ; , \end{array}\ ] ] where @xmath34 means quantum statistical equilibrium average . the spectral representation of the charge interaction is @xmath35}{\sinh(\frac{\omega}{2 t } ) } \ ; .\ ] ] in the scaling limit ( [ gft ] ) we have @xmath36 \;,\ ] ] where @xmath37 is the uv - frequency cutoff in the spectral density @xmath38 . while in a two - state system the charges alternate , in our case of infinitely many states the charges are unordered . in the term of order @xmath39 or @xmath40 , we have @xmath41 positive and @xmath41 negative charges , and we must integrate over the locations of the charges in the interval @xmath1 . the grand - canonical series of the tb partition function is @xmath42 & & \!\times \;{\displaystyle \exp\left\{\sum_{m>\,k\,=\,1}^n\left[w(\tau_m^{}-\tau_k^ { } ) + w(\tau'_m-\tau'_k ) \right ] \right\ } } \\[5 mm ] & & \!\quad \times\ ; { \displaystyle \exp\left\{-\sum_{k,\,m\,=\,1}^n w(|\tau_m^{}-\tau'_k| ) \right\ } } \\[5 mm ] & & \!\qquad\times \ ; { \displaystyle \exp\left\{-\;\epsilon\sum_{k\,=\,1}^n(\tau_k^{}-\tau'_k)\right\ } } \ ; , \end{array}\ ] ] and a similar expression holds for @xmath30 . according to the dc josephson effect , a super current of cooper pairs flows through a superconducting tunnel junction . at finite voltage @xmath43 , there is a supercurrent only if the excess energy @xmath44 can drain off to the electromagnetic environment . quasiparticle excitations can be disregarded when the voltage @xmath43 is small compared to the gap voltage . for weak josephson coupling , @xmath45 , the charge on the junction is fairly well defined . the relevant minimal model of the junction - plus - environment entity is @xcite @xmath46 & & \!\!\!\!\ ! + \;{\displaystyle \sum_{\alpha\,=\,1}^n \bigg [ \,\frac{q_\alpha^2}{2c_\alpha^ { } } \,+ \ , \frac{1}{e^2 _ { } } \frac{1 } { 2 l_\alpha^{}}\left(\frac{\psi}{2 } \,- \,ev_{\rm x}^{}t \,- \ , \psi_\alpha^{}\right)^2\,\bigg ] } \ ; . \end{array}\ ] ] the first term represents the charging energy , the second term accounts for tunneling of cooper pairs through the junction , and the third term describes the coupling to the electromagnetic environment , which is modeled by a set of lc - circuits , and it also includes the applied voltage . in the limit @xmath47 , the capacitances @xmath48 and inductances @xmath49 provide an arbitrary linear admittance @xmath50 where @xmath51 . the total impedance seen by the junction is @xmath52 after tracing out the electromagnetic modes , one finds for the imaginary - time phase correlator in thermal equilibrium @xmath53\,\psi(0)\,\rangle } _ { } \ ; \equiv \ ; { \rm e}^{-\,w_\psi(\tau ) } _ { } \ ; , \ ] ] where @xmath54 has the spectral representation ( [ corrspec ] ) with @xmath55 and where @xmath56 is the resistance quantum for cooper pairs . for a resistive electromagnetic environment , @xmath57 $ ] with @xmath58 , and @xmath59 = e_{\rm c}/(\pi\rho)$ ] , we get @xmath60 \,+ \ , 2\rho\,\zeta \ ; .\ ] ] the first term in eq . ( [ wpsi ] ) describes the resistive environment in the scaling limit [ cf . ( [ corr1 ] ) ] , and the quantity @xmath61 encapsulates details of the total impedance at high frequencies @xcite , @xmath62 & & { \displaystyle \;+ \ , \int_0^\infty \!\frac{{\rm d}\omega}{\omega}\left[\,\frac { { \rm re}\,z_{\rm t}(\omega)}{\rho r_{\rm q } } - \frac{1 } { 1+(\omega/\omega_{\rm r}^{})^2 } \,\right ] } \ ; . \end{array}\ ] ] the second term in ( [ wpsi ] ) leads to adiabatic renormalization of the tunneling coupling @xmath63 the model ( [ jos2 ] ) with ( [ wpsi ] ) directly corresponds to the model ( [ htb ] ) with ( [ gft ] ) . the correspondence relations are @xmath64 in the opposite limit of large josephson coupling , @xmath65 , the phase of the bcs condensate is fairly well defined and localized in one of the wells of the josephson potential . then it is pertinent to turn to the bloch band representation of the josephson junction and to choose a basis of discrete phase states , @xmath66 where @xmath67 is the quasi - charge . assuming that the junction dynamics is confined to the lowest energy band of width @xmath5 , we arrive at the hamiltonian @xcite @xmath68 & & + \ ; { \displaystyle \sum_\alpha \bigg [ \frac{\overline{q}_\alpha^{\,2}}{2\overline{c}_\alpha^ { } } \,+ \ , \frac{1}{e^2_{}}\frac{1}{2\overline{l}_\alpha^ { } } \left(\frac{\overline{\psi}}{2 } \ , - \ , \overline{\psi}_\alpha^{}\right)^2\bigg ] } \ ; . \end{array}\ ] ] for a current - biased junction we have @xmath69 . since charge and phase are interchanged when turning from the charge representation([jos2 ] ) to the phase representation ( [ josdual ] ) , the environmental effects are now captured by the charge autocorrelation function @xmath70 \overline{q}(0)\,\rangle$ ] . the role of @xmath71 in ( [ gpsi ] ) is now taken by @xmath72 , where @xmath73 is the admittance . thus we obtain for a strictly resistive environment @xmath74 \ ; , \ ] ] where @xmath75 is the required cutoff . the ground state of 1d interacting spinless electrons is a tomonaga - luttinger liquid ( tll ) , which is characterized by a gapless collective sound mode . the low - energy modes of the 1d interacting electron liquid is conveniently treated in the framework of bosonization . an extremely useful field - theoretical formulation of a tll has been given by haldane @xcite . the creation operator for spinless fermions may be expressed in terms of bosonic phase field operators @xmath76 and @xmath77 as @xmath78\ } \exp[{\rm i}\,\sqrt{\pi}\phi(x)]$ ] . at long wave lengths only the terms @xmath79 are important . they represent the right- and left - moving parts of the electron field . the bosonic field operators obey the equal - time commutation relation @xmath80 = - \,{\rm i}\ , { \rm sgn}(x - x')/2 $ ] . the generic interaction of the tll universality class is effectively given by a @xmath81-potential . this yields the interaction term @xmath82 . in the tll , the electron interaction is characterized by a single dimensionless parameter @xmath83 , where @xmath84 is the fermi velocity . excitations of the liquid are described by the generic harmonic hamiltonian @xmath85\ ; , \ ] ] where @xmath86 is the sound velocity . consider now a single strong point - like impurity , or equivalently a weak link or tunnel junction . a hopping or tunneling term which transfers electrons across the impurity or weak link may be included in the original electronic hamiltonian by adding @xmath87 $ ] . expressed in terms of the bosonic fields , the hopping term induces a jump of the @xmath88-field at the impurity , @xmath89 $ ] . in the @xmath88-representation , the impurity is then described by @xmath90 \ ; .\ ] ] here we have included a voltage term . the model is rounded off with addition of the harmonic liquid in the right / left ( @xmath91 ) lead . we then arrive at the weak - link hamiltonian ( @xmath88-model ) @xcite @xmath92 since the nonlinear degree of freedom is only at @xmath93 , we may integrate out the fluctuations of @xmath94 away from @xmath93 . with the fourier ansatz @xmath95 , where @xmath96 is a matsubara frequency , the euclidean action is minimized when @xmath97 . the resulting euclidean influence action , which includes all effects of the excitations in the leads on the tunneling degree of freedom , is @xmath98 . average with the weight function @xmath99 yields for the charge correlator @xmath100 with the charge interaction @xmath101 & = & { \displaystyle \frac{2}{g } \ , \ln\left[\ , \frac{\omega_{\rm c}^{}}{\pi t}\,\sin(\pi t \tau ) \,\right ] } \ ; . \end{array}\ ] ] a single weak impurity is modeled by the hamiltonian @xmath102 where @xmath103 is the scattering potential . for a point - like scatterer at @xmath93 , the @xmath104-backscattering contribution takes the form @xmath105 $ ] , where @xmath106 . with the lead and voltage term added , the weak - impurity or strong - tunneling hamiltonian ( @xmath107-model ) is @xmath108 \,+\ , e v_{}^{}\overline{\theta}/\sqrt{\pi } \,+\ , h_{\rm l}^{}(g ) \;.\ ] ] one may again integrate out the luttinger modes away from the scatterer . this leads to the influence action for the scattering mode @xmath109 , and the charge interaction in the coulomb gas representation takes the form @xmath110 \ ; .\ ] ] the @xmath88-model is self - dual to the @xmath107-model , and vice versa . in the mapping the regimes of a weak and a strong barrier are interchanged . we also see that the @xmath111- and @xmath112-representations of the impurity model directly correspond to the above tb- and wb - representations of the brownian particle in the washboard potential . the kondo parameter @xmath113 is related to the electronic interaction parameter @xmath114 by @xmath115 since the correspondences are valid for the hamiltonians and the effective actions @xcite , they hold not only for transport quantities like the current but also for all quantum statistical fluctuations . in addtion , the correspondences also hold for the full counting statistics of tunneling of edge currents in fractional quantum hall ( fqh ) systems , where the fractional filling factor @xmath116 takes the role of @xmath114 ( see e.g. ref . @xcite ) . the 1d fermion - boson correspondence suggests that a coherent one - channel conductor under influence of a resistive electromagnetic environment can be mapped on a quantum impurity embedded in a tll . the mapping has been analyzed by safi and saleur @xcite . a mesoscopic conductor in contact with an electromagnetic environment forms a quantum system violating ohm s law . transfer of energy from the electrons to the environment leads to dynamical coulomb - blockade , which reduces the current at low voltage.the luttinger parameter for a coherent conductor in the absence of the electromagnetic environment is @xmath117 . following the treatment given in the preceding subsection , and augmenting the phase in the tunneling term ( [ tunint ] ) with the phase @xmath118 of the electromagnetic environment , the relevant weak - tunneling hamiltonian is found to read @xmath119 \\[2 mm ] & & \!\!\!\!\ ! + \ ; h_{\rm l,+}^{}(g=1)\ , + \ , h_{\rm l,-}^{}(g=1)\ , + \ , h_{\rm env}^{}[{\cal q},\varphi ] \ ; . \end{array}\ ] ] it is convenient to combine the electronic phase with the electromagnetic phase into the auxiliary phase @xmath120 . following the route sketched in the previous subsection , one finds for the phase correlation function for a resistive environment @xmath121\,\overline{\chi}(0)\,\rangle = w_{\overline{\chi}}(\tau)$ ] with @xmath122 \,+\ , 2\rho\,\zeta \ ; .\ ] ] here @xmath123 , and @xmath124 is the resistance quantum . the hamiltonian of the corresponding strong - tunneling conductor is @xmath125 \,+\ , h_{\rm l}^{}(g=1 ) \\[2 mm ] & & \!\!\!\!\ ! -\ , ( ev_{\rm a}^ { } \,+\,\dot\varphi)\,\overline{\theta}/\sqrt{\pi } \,+\,h_{\rm env}^{}[{\cal q},\varphi ] \ ; , \end{array}\ ] ] with @xmath126 the applied voltage and @xmath118 the fluctuating phase of the environment . upon integrating out the electronic modes away from @xmath93 , and also the environmental modes for a strictly resistive environment , the effective action takes the form @xmath127 . evidently , if we turn to the coulomb gas representation , this leads to the charge interaction @xmath128 \,+\ , 2 \zeta/\rho \ ; .\ ] ] again , we have added both in ( [ wchi ] ) and in ( [ wvartheta ] ) a term which captures the details of the resistive environment at high frequencies as given in eq . ( [ hifr ] ) . this term leads to an adiabatic renormalization of the couplings @xmath14 and @xmath31 , as in eq . ( [ dressed ] ) . thus we find that the two representations of the coherent conductor are self - dual to each other . furthermore , they can be mapped on the models described before . in the mapping to the brownian particle model and to the quantum impurity system we have @xmath129 these correspondence relations hold both for the weak- and strong - tunneling representations . in the sequel , we choose from the above diverse models the tight - binding or weak - tunneling model ( [ htb ] ) . with the duality mapping ( [ dual1 ] ) and ( [ dual2 ] ) and the above correspondences between the various models , subsequent findings can be transferred easily to the weak - binding regime and to both regimes of the other models . the change of integration variables , @xmath130 , maps the coulomb gas ( [ coul1 ] ) with ( [ corr1 ] ) on the unit circle . the perturbative series of the partition function ( [ coul1 ] ) may be written as a power series in the effective fugacity @xmath131 the tight - binding series takes the form @xmath132 where @xmath133 & & \!\!\!\ ! \times \;\;{\displaystyle \left| \frac{\prod_{i < j } \sin ( \frac{u_i - u_j}{2 } ) \sin ( \frac{u'_i - u'_j}{2 } ) } { \prod_{i , j}\sin(\frac{u_i - u'_j}{2 } ) } \right|^{2k}\!\ ! { \rm e}^{{\rm i } p\sum_{i}(u_i -u'_i ) } } \ ; , \end{array}\ ] ] and where @xmath134 represents the scaled physical bias , @xmath135 changing integration variables @xmath136 and @xmath137 we get @xmath138 & \times & { \displaystyle \frac{[\,\delta(z)\overline{\delta(z)}\,]^k [ \,\delta(z')\overline{\delta(z')}\,]^k } { \prod_{i , k } [ \,(1-z_i^{}\bar{z}'_k ) ( 1-z'_k\bar{z}_i^ { } ) \,]^k } \,\left ( \frac{z_1\cdots z_n}{z'_1\cdots z'_n}\right)^p } \ ; , \end{array}\ ] ] where @xmath139 is the @xmath41-variable vandermonde determinant . unfortunately , the multiple integrals in ( [ icoef1 ] ) or ( [ ico ] ) can not be evaluated for general @xmath113 in the regime @xmath140 and for general complex @xmath134 . to advance , we must put @xmath113 rational and @xmath134 integer . integer @xmath134 may be regarded as a winding number due to a magnetic charge located at the origin . for rational @xmath113 , we may expand the integrand in terms of jack polynomials @xcite @xmath141 here , the function @xmath142 is a symmetric polynomial in the set of variables @xmath143 , and @xmath144 is a partition of an integer , @xmath145 . for positive integer @xmath134 , we have @xmath146 where @xmath147 means the partition @xmath148 where @xmath134 columns of length @xmath41 have been added in the respective young tableau . the multiple integrals in ( [ ico ] ) can be executed by use of the orthogonality relation of the jack polynomials , since it involves the vandermonde determinant . in the end , one arrives at the totally ordered @xmath41-fold series expression @xmath149 with @xmath150 & & \times\ ; { \displaystyle \frac{\gamma ( j k+p+m ) } { \gamma ( 1-k + j k + p + m ) } } \;. \end{array}\ ] ] it has been suggested in ref . @xcite to use ( [ zexpan ] ) with ( [ icoef2 ] ) and ( [ ejot ] ) to define @xmath151 for complex @xmath134 . furthermore , it has been conjectured that this is the unique analytic continuation . below , we confirm this conjecture in various limiting cases where analytic expressions are also available by different methods . for instance , we find from the significant limit @xmath152 that the free energgy @xmath153 has the proper behavior in all orders of @xmath14 , as @xmath154 . consider now the cumulant expansion @xmath155 the first three cumulant coefficients are @xmath156 for the analysis of the limit @xmath157 in the cumulant expressions , it will be pivotal to rearrange the multiple sums in @xmath158 into totally ordered sums as in ( [ icoef2 ] ) . this leads to the split - up @xmath159 where @xmath160 is an @xmath161-fold ordered sum . the first cumulant is @xmath162 the second cumulant has two contributions , @xmath163 \mathcal{c}_4^{(2)}(p ) & = & { \displaystyle \sum_{j\,=\,0}^\infty \sum_{k\,=\,0}^j \big[\ , e_2^ { } ( j ) - e_1^ { } ( j)\,\big ] e_1^ { } ( k ) \;. } \end{array}\ ] ] the three contributions of the third cumulant are @xmath164 \mathcal{c}_6^{(2)}(p ) & = & { \displaystyle \sum_{j\,\ge \,k\,=\,0}^\infty \big[\ , e_2^ { } ( j ) -e_1^ { } ( j)\,\big]\,\big[\ , e_1^ { } ( j)+ e_1^{}(k)\,\big]\ , e_1^ { } ( k ) \ ; , } \\ \mathcal{c}_6^{(3)}(p ) & = & { \displaystyle \sum_{j\,\ge\ , k\,\ge\ , \ell\,=\,0}^\infty \big\{\ , \big[\ , e_3^ { } ( j)- e_1^ { } ( j)\,\big]\ , e_2^ { } ( k)\ , e_1^ { } ( \ell ) } \\ & & \hspace{3em } { \displaystyle -\ ; 2\big[\ , e_2^ { } ( j)- e_1^ { } ( j)\,\big ] \ , e_1^ { } ( k)\,e_1^ { } ( \ell)\ , \big\ } \;. } \end{array}\ ] ] the integrals in the expression ( [ icoef1 ] ) , which defines the perturbative coefficients @xmath165 , diverge at short distances when @xmath166 . correspondingly , the multiple series ( [ icoef2 ] ) for the coefficient @xmath165 diverges in the regime @xmath167 for all @xmath41 . interestingly , the coefficient @xmath158 in the cumulant series ( [ cumexpan ] ) is regular for @xmath113-values even when @xmath165 is singular . this is because some of the divergences in the coulomb integrals are cancelled when taking the connected part @xmath158 . the analysis of the ordered @xmath161-fold sums @xmath168 in ( [ cum2 ] ) shows that the cumulant coefficient @xmath158 is nonsingular in the range @xmath169 ( see below for details ) . in ref . @xcite a conjecture was made which relates the nonlinear conductance directly to the partition function @xmath170 or to the free energy @xmath171 . the conjecture for the conductance of the tb model ( [ htb ] ) is @xmath172 & = & \!{\displaystyle \frac{k}{2p } \frac{x}{t}\frac{\rm d}{{\rm d}x}\big ( \mathcal{f}_{\rm tb}(x , p ) - \mathcal{f}_{\rm tb}(x ,- p)\big ) \ ; , } \end{array}\ ] ] where @xmath134 is given in eq . ( [ pdef ] ) . the conjecture is based on the fact that the free energy is real only for @xmath134 integer , and that the continuation of @xmath173 is not even in @xmath134 , so that the free energy aquires an imaginary part when @xmath134 is complex . the respective relation for the wb model ( [ hwb ] ) is @xmath174 where @xmath175 . use of the perturbative cumulant expansion ( [ cumexpan ] ) in the relation ( [ cond1 ] ) yields the weak - tunneling series of the conductance , @xmath176 the perturbative tunneling contribution of order @xmath41 is @xmath177 \,x^{2n}_{}\ ; , \ ] ] where @xmath178 . we now prove the conjecture in diverse parameter regimes . the series that defines @xmath179 can be summed in analytic form @xmath180 & = & { \displaystyle \frac{\sin(k \pi ) \,\gamma(1 - 2 k)}{\pi } \;\frac{\gamma(k+p)}{\gamma(1-k+p ) } } \ ; . \end{array}\ ] ] with this form the perturbative conductance ( [ bsc1 ] ) in order @xmath181 is found as @xmath182 correspondingly , the perturbative conductance through a voltage - biased josephson junction arising from incoherent cooper pair tunneling is @xmath183 the expressions ( [ bsc2 ] ) and ( [ pertcond ] ) are exactly the same as the ones obtained from the non - equilibrium real - time approach @xcite . let us study the equivalence in some detail . in the weak tunneling limit , the normalized conductance is given in terms of the golden rule transition rates @xmath184 representing incoherent forward / backward tunneling transitions to adjacent wells ( see section [ snoise ] for notation ) . we have @xmath185 where @xmath186 is given in ( [ pdef ] ) . the rates may be calculated directly from the standard real - time double - path representation in order @xmath181 , which is pictorially sketched in the right diagrams of fig . alternatively , one may calculate the rates with the standard @xmath187-method @xcite . here , the free energy of order @xmath181 , @xmath188 , ( see left diagrams of fig . [ f1 ] ) is analytically continued to a complex value . the imaginary part of the free energy is obtained by distorting the integration path of the charge correlation integral from imaginary to real time . in this way we get @xmath189 taking either road we obtain the unique expression @xcite @xmath190 \ ; , \ ] ] where @xmath191 is the real - time noise correlation integral @xmath192 & = & { \displaystyle \frac{\gamma(1 - 2k)\,\gamma(k+p)}{2\pi\,\gamma(1-k+p ) } } \ ; . \end{array}\ ] ] upon combining ( [ condgr ] ) with ( [ krel ] ) we get @xmath193 \ ; , \ ] ] which is in agreement with the expression ( [ bsc2 ] ) . imaginary - time charge diagrams representing @xmath194 or the connected part @xmath188 ( left ) , and analytic continuation to real - time forward / backward path ( right ) . diagram ( a ) represents the forward rate @xmath195 , and diagram ( b ) the backward rate @xmath196 . each graph has fixed charge order along the forward / backward path . enqueuing the double path ( a ) into a single path yields two contributions , @xmath197 and @xmath198 , while diagram ( b ) yields @xmath199 and @xmath200 . these are taken into account by the integration limits @xmath201 and @xmath202 in ( [ imfint ] ) . ] consider next the conjecture ( [ cond1 ] ) with ( [ zexpan ] ) and ( [ icoef1 ] ) . this yields the expression @xmath203 \;.\ ] ] the function @xmath204 coincides with the cumulant @xmath205 . it is instructive that @xmath204 can be written as the imaginary - time noise correlation integral @xmath206 the analytically continued integrand of ( [ intreal ] ) , @xmath207 , is free of singularities in the half - strip @xmath208 of the complex plane @xmath209 . accordingly , the closed contour integral along the edges of the half - strip vanishes . this entails that the noise integrals @xmath204 and @xmath191 satisfy the relation @xmath210\,j_1(k , p ) \;.\ ] ] for integer @xmath134 , this relation reduces to @xmath211 hence the expressions ( [ condreal ] ) and ( [ condim ] ) coincide for integer @xmath134 . evidently , the result ( [ condreal ] ) of the real - time calculation is also valid for complex @xmath134 . this confirms that the expression ( [ condim ] ) , which is based on the thermodynamic method and has been calculated for integer @xmath134 , can be analytically continued to complex @xmath134 . unfortunately , the real - time approach is not practical for perturbative contributions @xmath212 with @xmath213 , since the relevant coulomb gas multiple integrals for charges distributed on the forward / backward paths can not be carried out in analytic form . the loophole now is to take the imaginary time - route to calculate the free energy for integer @xmath134 , as sketched in section [ scgr ] . with the conjecture that the resulting expression ( [ zexpan ] ) with ( [ icoef2 ] ) can be analytically continued to complex @xmath134 , we then get the conductance ( [ cond1 ] ) . let us now see whether this line of argument is correct in all orders of @xmath214 . in the regime @xmath215 , the leading contribution to the coefficient @xmath165 in the multiple series ( [ icoef2 ] ) is the summand labelled with @xmath216 . furthermore , for @xmath215 the product term @xmath217 given in ( [ ejot ] ) reduces to the form @xmath218 $ ] . with these truncations the series coefficient @xmath219 takes the concise form @xmath220 interestingly , the resulting perturbative series of the partition function ( [ zexpan ] ) can be summed in closed form , @xmath221 \ ! \ ! & = & \ ! \!\gamma(1+p / k)\ , x^{-p / k } i_{p / k}(2x ) \;. \end{array}\ ] ] the function @xmath222 is a modified bessel function with index @xmath116 . applying functional relations of the bessel functions @xcite the tunneling susceptibility @xmath223 takes the analytic form @xmath224 with this the expression ( [ cond1 ] ) for the nonlinear conductance takes the concise analytic form @xmath225 upon employing recursion relations of the bessel functions , this expression can be transformed into @xmath226 this form reduces to the linear conductance @xmath227 with the substitution @xmath228 , @xmath229 , and @xmath230 , the expression ( [ cond2 ] ) corresponds to the result of ivanchenko and zilberman for the conductance of a classic overdamped josephson junction @xcite . for the standard derivation of the expression ( [ cond2 ] ) within a real - time coulomb gas approach , resulting in an infinite continued fraction expression , we refer to refs . @xcite . in the weak - tunneling model ( [ htb ] ) , the regime @xmath215 corresponds to the classical regime of the brownian particle model ( [ hwb ] ) , as the damping parameter @xmath231 is very large . the respective conductance is independent of @xmath232 and is @xmath233 & = & { \displaystyle \frac{\sinh(\frac{\epsilon}{2t})}{\frac{\epsilon}{2t}}\,\frac{1}{|i_{{\rm i}\,q}(2x)|^2 } } \ ; , \end{array}\ ] ] as follows from the selfduality ( [ dual2 ] ) . in the strict classical limit we have @xmath234 . weak adiabatic quantum fluctuations may be taken into account ( i ) in the wb model by an effective corrugation strength @xmath235 , and ( ii ) in the model ( [ jos2 ] ) by an effective josephson energy @xmath236 @xcite , respectively . consider next the zero temperature limit of the expression ( [ cond5 ] ) . as @xmath154 , both the argument and the modulus of the bessel function become very large . in this case the uniform asymptotic expansion of the modified bessel function applies @xcite . the resulting expression for the nonlinear conductance of the weak - corrugation model ( [ hwb ] ) is @xmath237 the expressions ( [ cond5 ] ) and ( [ cond6 ] ) coincide with the solutions of the classical smoluchowski equation in corresponding regimes @xcite . for @xmath238 , the conductance ( [ cond6 ] ) is zero because the sliding motion of the overdamped particle comes to rest at locations with zero slope . the particle rests there forever , since thermal and quantum fluctuations are absent in the classical regime at @xmath25 . for @xmath239 , the potential in ( [ hwb ] ) is sloping down everywhere , so that the particle is continuously sliding . transcribing ( [ cond6 ] ) into the regime of classical phase diffusion in the josephson junction , which corresponds to the regime @xmath240 , yields for the nonlinear conductance at @xmath25 @xmath241 \ ; , \ ] ] where @xmath242 is the critical current . in conclusion , our analysis shows that the above imaginary - time approach to nonequilibrium transport in the limit @xmath243 ( and @xmath244 in the dual model ) reproduces all known expressions calculated from an elaborate real - time method . the limit @xmath154 constitutes a crucial proof whether the analytic continuation from integer @xmath134 to complex @xmath134 in the expression ( [ zexpan ] ) with ( [ icoef2 ] ) is correct for general @xmath113 in all orders of @xmath214 . for general @xmath134 , the @xmath161-fold ordered sum in @xmath160 can not be done in analytic form . the simplest term @xmath246 is a generalized hypergeometric series of the form @xmath247 . as @xmath157 , the ordered sums in @xmath160 turn into ordered integrals which can be solved in analytic form . with the substitution @xmath248 , we have the mapping @xmath249 and the asymptotic expansion of the @xmath250 emerges as @xmath251 & & \!\!\!\!+{\displaystyle\left . \frac{1}{p^2}\left[b_{21}(j)\frac{u^2+u^2}{u^2\ , u^2 } + b_{22}(j)\frac{1}{u\,u}\right ] + \cdots \ ! \right\ } } \ ; , \end{array}\ ] ] where @xmath252 . the expansion coefficients are @xmath253 \label{expancoef } b_{21}(j ) & = & { \textstyle \frac{1}{24}}\ , k(k-1)(k-2)\big[\,3k(2j-1)^2 - 1 \,\big ] \ ; , \\[1 mm ] \nonumber b_{22}(j ) & = & { \textstyle \frac{1}{4}}\ , k^2(k-1)^2 ( 2j-1)^2 \ ; .\end{aligned}\ ] ] from this we find that each term of @xmath160 behaves asymptotically as @xmath254 . however , there are formidable cancellations by adding up the individual terms , as we immediately see from the expressions ( [ c2 ] ) - ( [ c6123 ] ) . as a result , as far as the cumulant @xmath158 is concerned , subleading terms in the curly bracket of ( [ asy ] ) become the leading ones . the analysis shows that there is an extra factor @xmath255 which results from the relevant subleading terms . thus , in reality , the asymptotic power law is @xmath256 , and hence @xmath257 . consider now first the partial cumulant coefficient @xmath258 in the asymptotic regime @xmath157 we have @xmath259^{(k-1)n } \\ \nonumber \!\!&=&\!\ ! \frac{p^{2(k-1)n + 1}}{n\,\gamma^{2n}(k ) } \frac{\gamma(1-n+n k)\,\gamma(2n-1 - 2n k)}{\gamma(n- n k ) } \ ; .\end{aligned}\ ] ] the integral in the first line of ( [ cumzero ] ) is convergent in the regime @xmath260 . on the other hand , the total cumulant @xmath173 is regular down to @xmath261 , as follows from the integrals in ( [ icoef1 ] ) , i.e. , @xmath262 . in comparison with @xmath173 , the expression ( [ cumzero ] ) has @xmath263 additional singularities located at @xmath264 , where @xmath265 . the spurious singularities are contained in the term @xmath266 . the analysis now gives that the sum of all partial cumulants in ( [ cum2 ] ) gives @xmath267 times a polynomial @xmath268 which is of order @xmath269 in @xmath113 , @xmath270 the polynomial @xmath268 provides zeros just at the locations of the spurious singularities , so that @xmath173 is indeed regular in the range @xmath271 . with the limiting value @xmath272 , which follows from ( [ sublead ] ) , we thus get the unique expression @xmath273 & = & { \displaystyle \frac{(-1)^{n-1}}{(n-1)!}\ , \frac{\gamma(n k)}{\gamma(1-n + n k ) } } \ ; . \end{array}\ ] ] insertion of ( [ acoef ] ) into ( [ cumzero ] ) yields the asymptotic form of the cumulant coefficients in the regime @xmath152 , @xmath274 in order to corroborate these considerations and the result ( [ cumres ] ) , we calculate @xmath275 and @xmath276 in [ sapp ] explicitly . the cumulant @xmath173 has simple poles at @xmath277 , where @xmath278 . these poles are due to the gamma function term @xmath279 in @xmath173 . consider next the analytic continuation provided in the expression ( [ bsc1 ] ) . most interestingly , this produces a trigonometric factor @xmath280 $ ] , which has zeros just at the poles of @xmath281 . as a result , the perturbative conductance contribution @xmath212 turns out to be regular for all @xmath282 . it may be written as @xmath283 a crucial point now is that temperature cancels out in the combined expression @xmath284 . finally , it is useful to combine the tunneling coupling @xmath14 and cutoff @xmath37 to the universal kondo scale @xmath285 as @xcite @xmath286 with use of this scale in the expression ( [ pertcontr ] ) , the weak - tunneling series of the conductance at zero temperature and general @xmath113 takes the concise form @xmath287 this is exactly the expression for the conductance at @xmath25 found with the thermodynamic bethe ansatz by fendley , ludwig and saleur @xcite . the series ( [ tbseries ] ) is appropriate for @xmath288 in the regime @xmath289 , where @xmath290}_{}\,\epsilon_0^{}$ ] . for @xmath288 and @xmath291 , the appropriate starting point would be the relation ( [ condwb ] ) . in this case , the resulting expression for the conductance is the strong - tunneling series @xmath292 expressed in terms of the parameters of the wb model , the kondo scale @xmath285 is @xmath293 here the extra factor @xmath294 compared to ( [ epstb ] ) accounts for the mapping @xmath295 in the self - duality relation ( [ dual2 ] ) . for @xmath296 and fixed @xmath297 , the regimes of weak tunneling , eq . ( [ tbseries ] ) , and strong tunneling , eq . ( [ wbseries ] ) , are exchanged . we conclude this subsection with the remark that there are contour integral representations of the conductance of which ( [ tbseries ] ) is the taylor series and ( [ wbseries ] ) the asymptotic series , or vice versa @xcite . the conductance @xmath298 and the current @xmath299 are first moments of the transition rates , @xmath300 higher moments of the current give information on the statistical fluctuations of the transport process , @xmath301 a knowlegde of all the connected moments ( [ noise ] ) allows then a reconstruction of the population probability distribution . the dynamics of the population probability @xmath302 of site @xmath41 is governed by the master equation @xmath303 \ ; .\ ] ] the characteristic function @xmath304 with initial state @xmath305 is found from the master equation to read @xmath306 \ ; .\ ] ] the moments of the probability distribution follow from the characteristic function by differentiation , @xmath307 the cumulant expansion @xmath308 leads us to the connected moments ( [ noise ] ) of the current . it is evident from ( [ noise ] ) that knowledge of all moments is tantamount to knowledge of all rates @xmath309 in all orders of @xmath214 or @xmath14 . the rate @xmath309 is built up by paths which start out at an arbitrary initial state of the reduced density matrix ( rdm ) , say @xmath310 , then may run through any number off - diagonal and diagonal states of the rdm , and finally end in diagonal state @xmath311 . the minimal number of moves , each of weight @xmath312 , is @xmath313 . since the paths may make arbitrary detours , the transition rate @xmath309 is a series of partial rates @xmath314 , where the second index indicates that the transition consists of @xmath315 moves , or @xmath315 charges in the charge representation , @xmath316 in the last form , the dependence on the effective fugacity @xmath214 is exposed . for paths , which have interim visits of diagonal states , the reducible part must be subtracted ( see ref . @xcite ) . the partial rates @xmath314 may be calculated via analytical continuation of the @xmath317th cumulant of the partition function @xmath318 . in the analytical continuation the @xmath315 charges representing the cumulant @xmath319 may be arbitrarily partitioned on the forward and backward path , as sketched for the case @xmath320 in fig . [ f2 ] . the charge correlation factor for @xmath161 time - ordered charges @xmath321 on the forward path @xmath322 and @xmath317 time - ordered charges @xmath323 on the backward path @xmath324 is @xmath325 & = & { \displaystyle \exp\left\ { \ ; \sum_{j\,=\,2}^m \sum_{i\,=\,1}^{j-1}\ , u_ju_i q(t_j - t_i ) \right . } \\[5 mm ] & & \qquad\;\;+\ ; { \displaystyle \left . \sum_{j\,=\,2}^\ell\sum_{i\,=\,1}^{j-1}\ , \upsilon_j \upsilon_i q^\ast(t'_j - t'_i ) \right . } \\[5 mm ] & & \qquad\;\ ; - \ ; { \displaystyle \left . \sum_{i\,=\,1}^\ell \sum_{j\,=\,1}^m \,\upsilon_i u_j q(t'_i -t_j ) \ ; \right\ } } \\[6 mm ] & & \times\;{\displaystyle \exp\left\{{\rm i}\,\epsilon\left ( \sum_{j\,=\,1}^m u_j\ , t_j\,-\ , \sum_{i\,=\,1}^\ell \upsilon_i\ , t'_i\right ) \ ; \right\ } } \ ; , \end{array}\ ] ] where @xmath326+{\rm i}\,\pi k\,{\rm sgn}(t)$ ] , and where the last line represents the correlations due to the bias . for a charge sequence contributing to @xmath314 , we have the constraints @xmath327 and @xmath328 . the term @xmath329 represents the sum of all @xmath330-fold coulomb integrals with the correlator ( [ infl ] ) which are compatible with the constraints . the six imaginary - time charge diagrams representing @xmath331 or the connected part @xmath332 ( left ) , and the various possibilities to partition the charges on the forward / backward path . a charge on the backward path changes sign when the time axis is reversed . the encircled charges are the charges @xmath333 in ( [ infl ] ) . six diagrams each contribute to @xmath334 and to @xmath335 , and one diagram each represents @xmath336 and @xmath337 . the other sixteen charge sequences , which all contribute to @xmath338 , are omitted . upon enqueuing the various charge sequences along the forward / backward path into a single path , the graphs @xmath339 and @xmath340 each split up into 6 different time - ordered charge sequences , and each of the residual 12 graphs splits up into 4 different time - ordered charge sequences . ] the property @xmath341 in the correlator ( [ infl ] ) ensures detailed balance between the forward and backward partial rates @xmath314 @xcite , @xmath342 actually , detailed balance already holds pairwise between particular subsets with fixed time - order along the keldysh contour . the subset pairs are related by charge conjugation and time - reversal . for instance , in order @xmath343 there are detailed balance relations between ( f.2 ) and ( a.2 ) , between ( c.1 ) and ( a.1 ) , between ( e.1 ) and ( b.1 ) , and between ( f.1 ) and ( d.1 ) . ( cf . fig . [ f2 ] and ref . @xcite for details ) . unfortunately , the coulomb integrals determining the various contributions to the partial rate @xmath314 can not be evaluated in analytic form for general @xmath2 and general @xmath113 . however at @xmath25 , one may use besides the detailed balance relations for the subset pairs also scaling properties of the integrals because of the logarithmic interactions at @xmath25 . these together yield a multitude of relations between different coulomb integrals of same order @xmath317 @xcite . as a result of the analysis , one finds formidable cancellations in the sum of the various coulomb integrals of given order @xmath317 , and only those give rate contributions at @xmath25 , where the charges @xmath344 and @xmath345 are all positive , i.e. , only direct paths without detours , @xmath346 since @xmath347 at @xmath25 , it is straightforward to deduce from the perturbative series ( [ tbseries ] ) the analytic expression @xmath348 expression ( [ charfun ] ) with ( [ kzero ] ) yields an analytic form of the characteristic function . in addition , we get the concise moment relation @xmath349 thus , all statistical properties of the quantum transport process at @xmath25 can be deduced directly from the current @xcite . let us finally return to finite temperature . evidently , the analytic continuation rule in eq . ( [ cond1 ] ) can not provide information about individual tunneling rates . rather we find with use of ( [ curr ] ) and ( [ rateseries ] ) @xmath350 \;= \ ; \sum_{\ell\,=\,1}^\infty \sum_{n\,=\,1}^\ell \frac{n}{\ell } \left ( k_{n,\ell}^{(+ ) } - k_{n,\ell}^{(-)}\right ) \ ; .\ ] ] this yields for the term of order @xmath351 @xmath352 \;=\ ; \sum_{n\,=\,1}^\ell \frac{n}{\ell}\left ( k_{n,\ell}^{(+ ) } - k_{n,\ell}^{(-)}\right)\ ] ] thus , the analytically continued perturbative series of the free energy yields for each term , say @xmath353 , a particular linear combination of partial rates of same order @xmath354 , as given in eq . ( [ fcontrel ] ) . in order to calculate individual partial tunneling rates , which are required in the full counting statistics at finite temperature , an analysis of the coulomb integrals in real - time is indispensable . we have discussed a variety of seemingly different physical models which display a wide range of interesting characteristics and can be realized experimentally . these models have in common that they can be mapped on a one - dimensional quantum field with gapless bulk excitations and a sinusoidal boundary interaction . the massless boundary sine - gordon model represents an impurity in a luttinger liquid . when the luttinger liquid modes are integrated out , the model describes a quantum - dissipative particle moving in a tilted sinusoidal potential . all these models can be successfully treated by a variety of interesting and powerful techniques . in most cases , the task of calculating nonlinear transport and full counting statistics requires application of nonequilibrium techniques . for general model parameters , this is a challenging endevour . despite the fact that the systems are integrable with help of the thermodynamic bethe ansatz , this method can not be used to calculate the full counting statistics , apart from the current . to gather full information about the statistical fluctuations , a real time calculation of the individual tunneling rates is essential . here we pursued a different route . we seized a suggestion by fendley , ludwig and saleur , who calculated the twisted partition function of a `` log - sine '' coulomb gas on a ring , and analyzed a conjectured relation with the nonlinear conductance which involves analytic continuation in the bias and coupling strength . we studied the weak - tunneling limit , the classical regime , and the zero temperature case and found in all cases agreement with results available from a non - equilibrium real - time calculation . with the substitution ( [ asy1 ] ) and use of the asymptotic expansion ( [ asy ] ) , the expressions ( [ c412 ] ) and turn into @xmath357 where @xmath358 h(u ) & = & k(k-1)\big({\textstyle \frac{1}{u}+\frac{1}{1+u}}\big ) \;.\end{aligned}\ ] ] the function @xmath359 originates from the terms of order @xmath360 in the curly bracket of ( [ asy ] ) . the integrals ( [ int41 ] ) and ( [ int42 ] ) are convergent in the regime @xmath361 . observing that @xmath362 where the prime indicates differentiation with respect to the variable , the double integral can be reduced by partial integration to a single integral , which is a beta function . in the end , we get @xmath363 this yields @xmath364 . thus , in the sum of @xmath365 and @xmath366 the singularity at @xmath367 is abrogated , and we end up with the cumulant expression @xmath368 which is regular in the range @xmath369 . as @xmath152 , way may use ( [ asy1 ] ) and the asymptotic series ( [ asy ] ) in the expressions ( [ c6123 ] ) . we then obtain @xmath370 \label{intc62 } \mathcal{c}_6^{(2)}(p ) & = & g\ , \frac{p^{6k-5}}{\gamma^6 ( k)}\left\ { \int_0^\infty \!\!{\rm d}u \ , \left ( f^2(u ) \right ) ' \int_0^{\,u}\!\ ! { \rm d}\upsilon \,f(\upsilon ) \right . \\ \nonumber & & \qquad\qquad + \left . ! { \rm d}u \ , f\,'(u)\int_0^{\,u}\!\!{\rm d}\upsilon \,f^{\,2}(\upsilon ) \right\ } \ ; , \\[2 mm ] \label{intc63 } \mathcal{c}_6^{(3)}(p ) & = & k^2\ , \frac{p^{6k-5}}{\gamma^6 ( k)}\\ \nonumber & & \times \;\left\ { \int_0^\infty\!\!{\rm d}u \ , f\,''(u ) \int_0^{\,u}\!\!{\rm d}\upsilon \ , f(\upsilon ) \int_0^{\,\upsilon } \!\!{\rm d}w \ , f(w ) \right . \\ \nonumber & & + \left . 2 \int_0^\infty\!\!{\rm d}u \ , f\,'(u ) \int_0^{\,u}\!\!{\rm d}v \ , f\,'(\upsilon ) \int_0^{\,\upsilon}\!\!{\rm d}w \ , f(w ) \right\ } \ ; .\end{aligned}\ ] ] here we have again employed the relation ( [ frel ] ) . the integrals in ( [ intc61 ] ) - ( [ intc63 ] ) are convergent in the region @xmath371 . the double integrals ( [ intc62 ] ) and triple integrals ( [ intc63 ] ) may be transformed by partial integrations into single integrals , which again are representations of the beta function . we get @xmath372 the sum of these terms yields @xmath373 with @xmath374 the prefactor ( [ prefac3 ] ) just cancels the singularities of @xmath375 located at @xmath376 and at @xmath377 . while @xmath375 is regular in the range @xmath378 , the resulting cumulant expression @xmath379 is regular in the extended regime @xmath380 .

thermal emission from cold dust grains in giant molecular clouds can be used to probe the physical properties , such as density , temperature and emissivity in star - forming regions . we present the scuba-2 shared - risk observations at 450 m and 850 m of the orion a molecular cloud complex taken at the james clerk maxwell telescope ( jcmt ) . previous studies showed that molecular emission lines can contribute significantly to the measured fluxes in those continuum bands . we use the harpco j=3 - 2 integrated intensity map for orion a in order to evaluate the molecular line contamination and its effects on the scuba-2 maps . with the corrected fluxes , we have obtained a new spectral index map for the thermal emission of dust in the well - known integral - shaped filament . furthermore , we compare a sample of 33 sources , selected over the orion a molecular cloud complex for their highco j=3 - 2 line contamination , to 27 previously identified clumps in omc-4 . this allows us to quantify the effect of line contamination on the ratio of 850 m to 450 m flux densities and how it modifies the deduced spectral index of emissivity for the dust grains . we also show that at least one spitzer - identified protostellar core in omc-5 has aco j=3 - 2 contamination level of . furthermore , we find the strongest contamination level ( ) towards a young star with disk near omc-2 . this work is part of the jcmt gould belt legacy survey . [ firstpage ] stars : formation stars : protostars submillimetre : ism ism : molecules instrumentation : detectors

1601.01989

although interstellar dust grains represent only a small fraction ( @xmath6 ) of the total mass in molecular clouds , they are important sites for the production of molecular hydrogen ( h@xmath7 ) and other complex molecules . these molecules play a role in the formation and the chemistry of pre - stellar cores @xcite . therefore , studying the physical properties of interstellar dust , such as temperature and emissivity , should inform us about the processes leading to the birth of stars and their stellar systems . specifically , the emissivity represents the efficiency at which interstellar dust grains re - emit as thermal emission in the far infrared the starlight they absorbed . the emissivity spectral index @xmath3 is a parameter related to the composition and size distribution of the dust mixture , which can often be characterized by a power law @xcite . with the current tools at our disposal to probe molecular clouds in the infrared , submillimetre and radio wavelengths , it is possible to achieve a more complete description of interstellar dust in regions of star formation . the gould belt is a ring - like configuration of nearby o - type stars of @xmath8 pc diameter centered about 200 pc from the sun @xcite . it is associated with a number of active star formation regions such as the orion nebula and ngc 1333 . the jcmt gould belt survey @xcite is an international effort to expand the study of cold interstellar matter in these regions with the help of scuba-2 , a continuum instrument sensitive to thermal emission from dust grains @xcite , and harp , an heterodyne instrument useful for studying molecular line emission @xcite . the orion a molecular cloud is a well known star formation region located at a distance between 400 and 500 pc from our solar system @xcite . along with orion b , it is one of two giant molecular clouds in the orion superbubble @xcite . it contains the famous orion nebula ( messier 42 ) and a collection of smaller molecular complexes . even though orion objects have been observed extensively for decades , many questions remain unanswered . for example , it is still unclear which mechanism , turbulence or magnetism , is the most effective at regulating star formation in parts of the cloud @xcite . in this work , we present the scuba-2 shared risk observations of orion a taken in february 2010 at the jcmt . the maps at 450 @xmath0 m and 850 @xmath0 m are used to evaluate the spectral index of emissivity in different regions of the orion molecular cloud complex . we know from previous studies @xcite that molecular contamination can be particularly important at these wavelengths and in this region . in order to evaluate its effects on the determination of the dust grain emissivity , we use harp integrated intensity maps from @xcite of the contaminating @xmath1co j=3 - 2 emission . this paper is divided in four main sections . first , we describe a simple method to evaluate the spectral index of emissivity with scuba-2 observations . then we present the observations , followed by the results . finally , we discuss the importance of quantifying molecular contamination in the context of understanding star formation in orion . as mentioned in the previous section , the emissivity @xmath9 ( @xmath10 g@xmath11 ) at a given frequency @xmath12 ( hz ) represents the efficiency at which dust grains re - emit as thermal emission the energy they absorbed at higher frequencies . this parameter is directly related to the size distribution and physical properties of dust mixtures . since thermal emission from cold dust grains dominates the continuum in the submillimetre regime , the emissivity can be approximated from scuba-2 observations , assuming an optically thin medium , by using a simple power - law @xcite . if @xmath13 is the emissivity at a reference frequency @xmath14 , the spectral index of emissivity @xmath3 is defined by : @xmath15 a more detailed look at interstellar dust emissivity models is presented by @xcite , but our simplified approach still provides a useful estimation of the size distribution and composition of the dust grains @xcite . . for a given temperature , the theoretical ratio of flux densities ( @xmath16 @xmath17 ) is plotted as a function of the spectral index , as defined from equation [ eq:01 ] , for an ideal case where there is no molecular contamination ( @xmath18 @xmath19 ) . the solid line represents the temperature - independent case , obtained with the rayleigh - jeans approximation for dust thermal emission.,width=317 ] if dust grains are in thermal equilibrium with the radiation field , which is a reasonable assumption for large enough grains in molecular clouds @xcite , we can suppose that their emission follows a modified form of planck s law @xmath20 ( erg s@xmath11 @xmath21 hz@xmath11 sr@xmath11 ) for a blackbody at a temperature @xmath22 ( k ) . modeling the 3d dust distribution in star - forming regions from only a handful of observations is , however , far from trivial @xcite . in order to simplify the analysis included in this paper , we assume that the properties of dust grains are identical along any given line of sight . this permits the use of the dust column density @xmath23 ( g @xmath21 ) to represent the distribution in the plane of the sky . we also introduce the fraction @xmath24 of contamination at a given frequency . this umbrella factor represents the total contribution to the measured intensity @xmath25 from sources other than the dust s thermal emission . from these assumptions , we can write the relationship between the total measured intensity @xmath25 and the dust thermal emission at a given frequency as : @xmath26 although this equation is defined as a flux density in units of frequency ( @xmath27 jansky sr@xmath11 ) , the notation used from this point on will use wavelengths @xmath28 ( @xmath0 m ) instead to relate with the language commonly used for far - infrared astronomy . this will simplify the discussion without modifying the units used to quantify scuba-2 observations . we can now define the spectral index @xmath2 as the slope of continuum thermal emission between two measured flux densities : @xmath29 if we define @xmath30 as the difference between spectral indices @xmath2 after and before correction for molecular contamination , we obtain this simple relation from equation [ eq:02 ] : @xmath31^{-1}.\ ] ] the spectral index @xmath2 is the measurable quantity from which we evaluate the emissivity spectral index @xmath3 . if we combine equations [ eq:01 ] and [ eq:02 ] at any given temperature @xmath22 and assume the dust emissivity from equation [ eq:04 ] , we find that there exists a direct relationship between these spectral indices : @xmath32^{-1}.\ ] ] for this two - wavelength method , we know from equation [ eq:03 ] that any effect contamination has on the measured spectral index @xmath2 is going to propagate linearly to the deduced emissivity spectral index @xmath3 . furthermore , any difference calculated for one of the spectral indices is going to be identical for the other ( @xmath33 ) for a given dust temperature @xmath22 . this means that it is possible to evaluate the effect of molecular contamination on the emissivity spectral index @xmath3 with no information about the column density @xmath23 or the average temperature @xmath22 if it is assumed fixed . for this study , we use scuba-2 maps of orion a at 850 @xmath0 m and 450 @xmath0 m . we define the measured flux densities per unit of telescope beam area as @xmath34 and @xmath35 ( jansky beam@xmath11 ) respectively . if those flux densities are properly calibrated , normalized and beam - matched ( see section 3 ) , they can be directly substituted for @xmath36 and @xmath37 in equation [ eq:02 ] without any loss of information . in the rayleigh - jeans approximation ( @xmath38 ) , the flux density @xmath20 in equations [ eq:01 ] and [ eq:03 ] can be written as @xmath39 , where @xmath40 is boltzmann s constant ( erg k@xmath11 ) and @xmath41 is the speed of light ( cm s@xmath11 ) . this allows for a temperature - independent determination of the spectral index @xmath3 through a simple ratio of flux densities . in this regime , equation [ eq:03 ] is simplified as follows : @xmath42 . it is , however , true only in regions with high dust temperatures ( @xmath43 k ) , which is not the case in most prestellar cores ( @xmath44 10 - 15 k , see * ? ? ? the influence of temperature on the emissivity spectral index @xmath3 is shown in figure [ fig:01 ] . it includes the specific case of the rayleigh - jeans approximation , without including contamination , as well as the ones calculated from the more general planck law for a given series of temperatures ( 10 k to 50 k ) . from figure [ fig:01 ] , it is clear that the main purpose of the rayleigh - jeans approximation is to establish a lower limit on the spectral index @xmath3 by directly measuring the spectral index @xmath2 introduced in equation [ eq:02 ] , without any prior assumption about temperature or density . in order to evaluate the effect of molecular contamination on the spectral index of emissivity , we postulate that the fraction @xmath45 from equation [ eq:02 ] comes entirely from @xmath1co j=3 - 2 rotational line emission . we use a molecular emission map converted from harp spectroscopic data using the technique presented by @xcite . it is important to note that this conversion depends on the effective transmission of the telescope , and therefore varies with the atmospheric conditions during the observations . although it is not expected to be the only contributing factor , this @xmath1co j=3 - 2 molecular line is the most likely culprit of molecular contamination at 850 @xmath0 m as it is the brightest and most common emission line over the entire region except in omc-1 s bright core . away from hot cores , the @xmath1co j=3 - 2 line can be 5 to 10 times brighter in the 850 @xmath0 m band than all other molecular lines combined ( see * ? ? ? similarly , we would expect the main driver of molecular contamination at 450 @xmath0 m to be the emission from the @xmath1co j=6 - 5 rotational line . @xcite extrapolated from @xmath1co j=3 - 2 observations that the molecular contamination @xmath46 is , in most cases , unlikely to be significant when compared to the dust thermal emission at this wavelength . this can be verified in omc-1 using the @xmath1co j=6 - 5 integrated main - beam brightness temperature map from @xcite . with the conversions factors from @xcite , it is possible to estimate the @xmath1co j=6 - 5 contamination toward some key regions in our 450 @xmath0 m observations : orion kl , orion south and the orion bar . we estimate the contamination to be @xmath47 , @xmath47 and @xmath48 of the total 450 @xmath0 m flux density for these regions respectively . these estimates are likely to be upper limits since they do not take into account instrumental differences , such as sensitivity to large - scale emission , between champ@xmath49 @xcite , harp and scuba-2 . it is therefore reasonable , at least for omc-1 , to assume the @xmath1co j=6 - 5 line contamination to be negligible for scuba-2 observations at 450 @xmath0 m . since other molecular lines in that observation band are generally much weaker than the @xmath1co j=6 - 5 line , except for hot cores @xcite , we limit this study only to the effects of contamination in the 850 @xmath0 m band . the coefficient @xmath45 can only decrease the 850 to 450 @xmath0 m flux ratio , therefore the spectral index of emissivity will be _ underestimated _ if contamination from molecular emission is not taken into account . this may be particularly important in regions of strong molecular emission relative to the dust continuum , such as a low dust column density region containing molecular outflows @xcite . the submillimetre common - user bolometer array ( scuba-2 ) is a 10,000 pixel cryogenically cooled continuum camera capable of simultaneously observing in the 450 and 850 @xmath0 m atmospheric windows @xcite . the heterodyne array receiver programme ( harp ) is a 16-element detector providing high - sensitivity and high - resolution spectral line measurements between 325 and 375 ghz ( 799 and 922 @xmath0 m ) . the auto - correlation spectral imaging system ( acsis ) is a digital interferometric spectrometer that can use harp measurements to create large - scale velocity maps @xcite . shared - risk observations of the orion a molecular cloud complex were taken on february 13 2010 at the james clerk maxwell telescope ( jcmt ) with scuba-2 . the measured 225 ghz atmospheric opacity during the observations varied between @xmath50 . the complementary spectroscopic observations of the @xmath1co j=3 - 2 molecular line were obtained between february 9 and 16 in 2007 with harp @xcite . the opacity varied between @xmath51 . the three maps cover a field of approximately 60@xmath52 by 60@xmath52 . they include most of the orion a molecular clouds along the integral - shaped filament , such as omc-1 to omc-3 , as well as omc-4 and the northernmost part of omc-5 . the @xmath1co j=3 - 2 emission line map used in this study was reduced independently from the scuba-2 shared - risk observations @xcite . the orion a shared - risk observations were reduced using the starlink software map - making capabilities @xcite . the shared - risk programme for scuba-2 was completed with only one array ( instead of the final four ) of the detector in each continuum band , leading to less sensitive maps than what can be produced by the full detector ( e.g. * ? ? ? * mairs et al . in preparation ) . because of differences in the way large - scale emission is handled between the shared - risk and the full - array reductions , the @xmath1co j=3 - 2 contamination analysis presented in this study applies only to small - scale structures ( @xmath53 ) . contamination levels for individual sources may change when larger spatial scales are included . as mentioned in section 2 , some calibration is needed before the scuba-2 and harp maps can be used together . because of the way the telescope s dish distributes the incident energy over the detectors , the effective beam shape for each scuba-2 bandpass is approximated by two gaussian components . the full - width at half - maximum ( fhwm ) of these main and secondary beams are , respectively , 7.9@xmath54 and 25@xmath54 at 450 @xmath0 m with relative amplitudes of 0.94 and 0.06 . at 850 @xmath0 m , they are 13.0@xmath54 and 48@xmath54 with relative amplitudes of 0.98 and 0.02 @xcite . the process of beam - matching consists of convolving each map with the other s effective beam . this is done to ensure that the 450 and 850 @xmath0 m maps are comparable by degrading their quality in similar ways . the @xmath1co j=3 - 2 emission map is treated slightly differently . the map is first filtered with a gaussian ( fwhm = 1.0@xmath52 ) in order to subtract excess large - scale background @xcite . this is made necessary by harp s sensitivity to larger spatial scales than the 850 @xmath0 m shared - risk observations . this large - scale emission can represent a significant proportion of the @xmath1co j=3 - 2 integrated intensity ( @xmath55 in some parts of the cloud ) . it is then convolved with the 450 @xmath0 m primary and secondary beams to degrade it in a similar way to the 850 @xmath0 m map . this was done to treat co data in a manner as close as possible to the 850 @xmath0 m data . once the beam - matching process is complete , an automated linear interpolation transforms the 450 @xmath0 m and @xmath1co j=3 - 2 maps to share the same pixel - scale ( 4.0@xmath54 ) and associated celestial coordinates as the 850 @xmath0 m map . before the three maps can be used with equation [ eq:02 ] , they first need to be properly normalized . since we are working with peak photometry ( jy beam@xmath11 ) instead of aperture photometry ( jy arcsec@xmath56 ) , this calibration must be included in the beam - matching process . we use the effective beam widths @xmath57 and @xmath58 from @xcite to calculate the new effective mixed beam width @xmath59 created by convolving the 450 and 850 @xmath0 m beams together . after applying the appropriate filtering , each map is multiplied by a correction factor calculated from these effective beams to compensate for the change in beam area . once this normalization is complete , the three maps effectively share the same units and can be directly compared . the noise levels were calculated after the beam matching process , for an effective 17.6@xmath54 beam width , by measuring the standard deviation of the background emission in each map . this was done by selecting a 276@xmath54 by 172@xmath54 low - emission region centered at ( 5:34:19.550 , -5:42:41.91 ) . to calculate the spectral index map , only pixels with a flux density above 1@xmath60 were used , which is 280 mjy beam@xmath11 at 450 @xmath0 m and 20 mjy beam@xmath11 at 850 @xmath0 m . m in jy per unit of 850 @xmath0 m beam area . the contours trace two levels of the @xmath1co j=3 - 2 molecular line contamination as a percentage of the corresponding 850 @xmath0 m flux density ( 10@xmath61 in green and 30@xmath61 in blue ) . the 60 submillimetre sources selected for this study ( see tables [ tab:01 ] and [ tab:02 ] ) are identified as follows : clumps are black plus symbols , protostars are blue triangles and young stars with disks are red diamonds . for ease of reading , only the locations of high @xmath1co contamination ( i.e. , table [ tab:01 ] ) are labelled.,width=317 ] the shared - risk observations for orion a are shown in figure [ fig:02 ] for the 850 @xmath0 m and 450 @xmath0 m bands . the measured @xmath1co j=3 - 2 integrated intensity for a value of 20 k km s@xmath11 is shown as the blue contour in the left panel of figure [ fig:02 ] . this figure also shows the emissivity spectral index @xmath2 map from equation [ eq:02 ] after the @xmath1co j=3 - 2 contamination was removed from the 850 @xmath0 m map . an estimate of the regional variations in emissivity can be obtained from equation [ eq:03 ] in the rayleigh - jeans approximation : @xmath62 . as stated previously , this approximation generally leads to an incomplete representation of low temperature regions by underestimating the value of the spectral index @xmath3 . however , it still shows some key features such as the low emissivity region in omc-1 north - west of the trapezium stars . this is a region with well - known contamination from molecular lines other than the @xmath1co j=3 - 2 line @xcite . the orion bar also shows low spectral index regions similar to those measured in previous studies @xcite . the @xmath1co j=3 - 2 molecular line contamination is shown in figure [ fig:06 ] overlaid on the 850 @xmath0 m continuum map . the green contours trace three levels of the @xmath1co j=3 - 2 line emission as a percentage of the corresponding 850 @xmath0 m flux density ( 10@xmath61 , 20@xmath61 and 30@xmath61 ) . to study molecular contamination in more detail , we chose 33 sources around the integral - shaped filament selected by eye for their higher level of @xmath1co j=3 - 2 line emission . this represents most sources with a contamination @xmath63 in equation [ eq:02 ] . each area of high contamination is clearly labelled in figure [ fig:06 ] . the continuum properties of these sources were checked individually to verify their plausibility as clump candidates . we took care to select sources away from possibly problematic regions in the 450 and 850 @xmath0 m maps . this is made necessary because the contrast between intensities in the scuba-2 maps can create bowling effects that artificially decrease the flux density around very bright regions ( omc-1 , for example ) . we also use 27 clumps identified by @xcite in omc-4 . as these authors suspected , we confirm that these clumps have very low levels of @xmath1co j=3 - 2 contamination and thus can be used as a control group . the positions of all 60 sources are indicated in the 450 @xmath0 m continuum map of figure [ fig:02 ] , as well as in figure [ fig:06 ] ( see caption ) . their properties , shown in tables [ tab:01 ] and [ tab:02 ] , are obtained by averaging the values in each small 12@xmath54 by 12@xmath54 map box centered on the peak position of the source . these boxes have roughly the same angular area as the effective 850 @xmath0 m beam . table [ tab:01 ] lists the 33 sources selected for their higher levels of @xmath1co j=3 - 2 molecular line contamination . their ids have been attributed in order of decreasing right ascension . table [ tab:02 ] lists the 27 reference clumps in omc-4 . the i d for each clump is the same as given by @xcite . co j=3 - 2 molecular line emission as a percentage ( @xmath61 ) of the corresponding average flux density @xmath64 ( jy beam@xmath11 ) measured directly from the scuba-2 850 @xmath0 m shared - risk observation map . the dashed line sets the percentage of contamination for a @xmath1co j=3 - 2 line emission at 20 k km s@xmath11 . the empty circles are the 33 sources chosen in this study for their high level of molecular contamination . the crosses are reference sources identified by @xcite in omc-4 with an average flux density lower than 0.3 jy beam@xmath11 . since no reference source above 0.3 jy beam@xmath11 shows significant contamination , the horizontal axis is truncated at that value to facilitate the viewing of this figure . the uncertainties for all sources are identified with plain black lines . _ bottom _ : histogram showing the distribution of contaminated ( black bars ) and reference ( gray bars ) sources as a function of their average flux density @xmath64 ( jy beam@xmath11 ) . the categories have a width of 0.02 jy beam@xmath11 , with a range going from 0.00 to 0.30 jy beam@xmath11.,title="fig:",width=317 ] co j=3 - 2 molecular line emission as a percentage ( @xmath61 ) of the corresponding average flux density @xmath64 ( jy beam@xmath11 ) measured directly from the scuba-2 850 @xmath0 m shared - risk observation map . the dashed line sets the percentage of contamination for a @xmath1co j=3 - 2 line emission at 20 k km s@xmath11 . the empty circles are the 33 sources chosen in this study for their high level of molecular contamination . the crosses are reference sources identified by @xcite in omc-4 with an average flux density lower than 0.3 jy beam@xmath11 . since no reference source above 0.3 jy beam@xmath11 shows significant contamination , the horizontal axis is truncated at that value to facilitate the viewing of this figure . the uncertainties for all sources are identified with plain black lines . _ bottom _ : histogram showing the distribution of contaminated ( black bars ) and reference ( gray bars ) sources as a function of their average flux density @xmath64 ( jy beam@xmath11 ) . the categories have a width of 0.02 jy beam@xmath11 , with a range going from 0.00 to 0.30 jy beam@xmath11.,title="fig:",width=317 ] co j=3 - 2 molecular line emission as a percentage ( @xmath61 ) of the corresponding average 850 @xmath0 m flux density @xmath64 , as a function of the measured ( i.e. , uncorrected for co contamination ) 850 to 450 @xmath0 m average flux ratio @xmath65 . the symbols have the same meaning as in figure [ fig:03 ] . for each source , these two quantities are used in equation [ eq:02 ] to deduce the spectral index @xmath2 before and after contamination correction . the uncertainties for all sources are identified with plain black lines . _ bottom _ : histogram showing the distribution of contaminated ( black bars ) and reference ( gray bars ) sources as a function of their average flux ratio @xmath65 . the histogram bins have a width of 0.02 , with a range going from 0.00 to 0.26.,title="fig:",width=317 ] co j=3 - 2 molecular line emission as a percentage ( @xmath61 ) of the corresponding average 850 @xmath0 m flux density @xmath64 , as a function of the measured ( i.e. , uncorrected for co contamination ) 850 to 450 @xmath0 m average flux ratio @xmath65 . the symbols have the same meaning as in figure [ fig:03 ] . for each source , these two quantities are used in equation [ eq:02 ] to deduce the spectral index @xmath2 before and after contamination correction . the uncertainties for all sources are identified with plain black lines . _ bottom _ : histogram showing the distribution of contaminated ( black bars ) and reference ( gray bars ) sources as a function of their average flux ratio @xmath65 . the histogram bins have a width of 0.02 , with a range going from 0.00 to 0.26.,title="fig:",width=317 ] the @xmath1co j=3 - 2 contamination in each clump expressed as a percentage of their average 850 @xmath0 m flux density is plotted versus their average 850 @xmath0 m flux density per unit of effective 850 @xmath0 m beam area ( @xmath58 ) in figure [ fig:03 ] ; the same contamination percentage is given in figure [ fig:04 ] as a function of the 850 to 450 @xmath0 m flux ratio . although we use the beam matched ( @xmath59 ) values to calculate the contamination and the ratio of flux densities ( and thus the spectral indices ) , it is more useful for future studies to describe the sources presented in this paper by their @xmath64 flux density per unit of effective 850 @xmath0 m beam area . the @xmath1co j=3 - 2 contamination as a percentage of the 850 @xmath0 m flux density remains essentially unchanged between the matched ( @xmath66 ) and effective ( @xmath67 ) beams . this list of 60 sources in orion a was compared to a catalogue of known young stellar objects identified by @xcite using the spitzer space telescope . we confirm the presence of three protostars ( blue triangles in figure [ fig:06 ] ) in our sample , including one ( c24 ) , showing a @xmath1co j=3 - 2 contamination of @xmath4 , in the northernmost part of omc-5 . furthermore , five young stars with disks lie within continuum sources along the integral - shaped filament , including three in regions with strong @xmath1co j=3 - 2 line emission . the most highly contaminated source in this study ( c23 ) , with a level of @xmath5 , is associated with a young star with a disk located between omc-2 and omc-3 ( identified in figure [ fig:06 ] ) . it is unlikely that this object alone is responsible for such a strong @xmath1co j=3 - 2 line emission . some possible explanations include that this region is a shock knot , or that it is surrounded by outflows from past or ongoing star formation . the objects in our sample that have also been detected with the spitzer space telescope are identified on figure [ fig:06 ] ( see caption ) and listed in tables [ tab:01 ] and [ tab:02 ] . additional notes on astronomical objects found near these sources of interest are presented in appendix b. to check for signs of molecular outflows , the wing criterion from @xcite ( @xmath68 k at @xmath69 km s@xmath11 from line centre ) was applied to the spectrum of each source from tables [ tab:01 ] and [ tab:02 ] . this criterion is reached for most sources in both the contaminated and reference samples ; the spectra often exhibit wide @xmath1co j=3 - 2 molecular lines . the wing criterion is not met in only a few cases : five sources in the contaminated sample ( c12 , c17 , c19 , c32 , c33 ) and one in the reference sample ( 06 ) . all the spitzer - identified young stellar objects listed in tables [ tab:01 ] and [ tab:02 ] also fit the wing criterion with wide @xmath1co j=3 - 2 lines . it is therefore likely that molecular outflows play a major role in @xmath1co j=3 - 2 line contamination at 850 @xmath0 m . before we study the effect of molecular line contamination in more detail , we can deduce some general guidelines concerning @xmath1co j=3 - 2 emission in the sources shown on figure [ fig:03 ] . first of all , all the identified sources with a contamination level above @xmath70 seem to be concentrated below a threshold of 300 mjy beam@xmath11 in the 850 @xmath0 m scuba-2 shared - risk map . such a limit is to be expected because the optically thick @xmath1co j=3 - 2 line emission saturates much faster than the mostly optically thin dust thermal emission . the @xmath1co j=3 - 2 line emission is therefore more likely to become a problem toward regions of lower column density . this also explains why peaks in @xmath1co j=3 - 2 line emission do not necessarily lead to high levels of contamination in scuba-2 maps . the omc-1 central @xmath1co j=3 - 2 emission peak is an example of this situation ; it is completely dominated by the total 850 @xmath0 m peak flux density : it produces a contamination @xmath71 . this does not mean that the total molecular contamination in the central part of omc-1 is negligible ; it is a region known for its multitude of molecular species @xcite which may contribute up to @xmath72 of its 850 @xmath0 m intensity @xcite . there is however no easy way to guess which sources will be contaminated by molecular emission without having the corresponding spectroscopic observations . even among sources with the same classification , as we see in spitzer - identified protostars and young stars with disks in our sample , there can be wide differences in the levels of @xmath1co j=3 - 2 line emission . the sources that do show a strong molecular contamination must be corrected as their deduced physical properties depend on an accurate measurement of flux density . the most direct effect of @xmath1co j=3 - 2 molecular line emission is its contribution to the total 850 @xmath0 m flux density measured by scuba-2 , therefore inflating the resulting 850 to 450 @xmath0 m flux ratio . following equation [ eq:02 ] and figure [ fig:01 ] , this systematically leads to an underestimation of the spectral indices @xmath2 and @xmath3 . the size of this deviation depends on the fraction of molecular line contamination in the measured 850 @xmath0 m flux density , as shown in equation [ eq:05 ] . the percentage of @xmath1co j=3 - 2 molecular line emission and the measured 850 to 450 @xmath0 m flux ratios are shown in figure [ fig:04 ] for a sample of 60 sources in orion a. the largest spectral index difference ( @xmath73 ) is associated with an object ( c23 ) with a @xmath1co j=3 - 2 line contamination percentage of @xmath74 . its average 850 to 450 @xmath0 m flux ratio is reduced from @xmath75 to @xmath76 , thus increasing the deduced spectral index @xmath2 from @xmath77 to @xmath78 . m average flux ratio @xmath65 . the number of sources for each bin is given before ( black bars ) and after ( gray bars ) contamination correction is applied to the measured ratio . the histogram bins have a width of 0.02 , with a range going from 0.00 to 0.26 . _ top _ : the histogram is limited to the 33 sources chosen in this study for their high level of molecular contamination . _ bottom _ : the histogram including all 60 sources shown in figure [ fig:04].,title="fig:",width=317 ] m average flux ratio @xmath65 . the number of sources for each bin is given before ( black bars ) and after ( gray bars ) contamination correction is applied to the measured ratio . the histogram bins have a width of 0.02 , with a range going from 0.00 to 0.26 . _ top _ : the histogram is limited to the 33 sources chosen in this study for their high level of molecular contamination . _ bottom _ : the histogram including all 60 sources shown in figure [ fig:04].,title="fig:",width=317 ] the effects of molecular contamination on emissivity can also be understood by looking at the histograms in figure [ fig:05 ] . by using the flux ratio as a proxy for the spectral index @xmath3 , following the same relation as the one shown in figure [ fig:01 ] , we directly quantify the difference @xmath1co j=3 - 2 line subtraction has on the emissivity of both the total source selection and the contaminated subgroup.the top panel of figure [ fig:05 ] quantifies how correcting the 850 @xmath0 m flux density influences the 850 to 450 @xmath0 m flux ratio for the 33 most highly contaminated sources in our sample . by combining equation [ eq:02 ] with the measurements shown in figure [ fig:04 ] , we calculate the spectral index @xmath2 before and after contamination correction for all 60 sources . the results are presented in tables [ tab:01 ] and [ tab:02 ] . the average uncorrected ratio of these sources leads to a spectral index @xmath79 . the corresponding average correction is @xmath80 . this results in an average corrected value @xmath81 for the contaminated sources . in comparison , the spectral index for the omc-4 control group gives a value @xmath82 . the uncertainties are obtained from the standard deviations of the ratios . although they are statistical uncertainties , we have just enough sources in each sample ( 33 and 27 ) for them to be significant . if these two samples are indeed comparable , as their similar standard deviations and their distribution of ratios in figure [ fig:04 ] would suggest , then this indicates that the contamination correction brings the average spectral indices @xmath2 to levels similar to those in the uncontaminated clumps of omc-4 . the similarity between the two samples could be explained by the selection of sources away from the ridge in omc-1 ( see figure [ fig:02 ] ) , known to be warmer and denser than the rest of the integral shaped filament . a kolmogorov - smirnov test was applied to both the contaminated and the total subgroups from figure [ fig:05 ] in order to verify the significance of the contamination correction on these distributions . the subgroup with the 33 contaminated sources has a @xmath83 probability that the uncorrected and corrected distributions are statistically identical . since it represents the group most affected by molecular contamination , this is a confirmation that contamination correction has a significant effect on the measured flux ratios . for the complete selection , this test gives a @xmath84 probability . unsurprisingly , this only means that , as more uncontaminated sources are added to the sample tested , the less statistically significant the average contamination becomes . quantifying the molecular line contamination in submillimetre observations , such as in scuba-2 850 @xmath0 m maps , is a necessary step in the study of cold interstellar dust . through the construction of accurate spectral energy distributions , it is possible to learn much about the physical properties of this important mass tracer in giant molecular clouds . as an example , models using observations carried out with the herschel space observatory , hereafter herschel , benefit strongly from the addition of scuba-2 850 @xmath0 m measurements . while herschel observations efficiently probe the peak blackbody emission of cold dust grains , scuba-2 wavelengths help constrain the rayleigh - jeans tail of the distribution . this leads to a better determination of density , temperature and spectral index @xmath3 than herschel results alone @xcite . furthermore , both the jcmt and herschel have significant overlapping data in the gould belt . a good example of what can be achieved with a multifaceted approach is the work of @xcite . these authors combined observations from the planck and herschel space telescopes in order to build optical depth and temperature maps for the orion molecular cloud complex . they also included in their fits the spectral index @xmath3 map from @xcite at a resolution of @xmath85 . from this optical depth map , and with near - infrared extinction measurements from 2mass , @xcite obtained a column density map for the region . finally , they found a temperature @xmath86 k in dense regions along the spine of orion a , which we will use as a reference for the examples used in this discussion . since dust thermal emission depends strongly on the emissivity of the grains , we focus specifically on how molecular contamination influences this property . this is why we use the spectral index @xmath3 instead of @xmath2 for the rest of this discussion . we have already given in section 2 a relation between the spectral indices @xmath2 and @xmath87 obtained from equation [ eq:03 ] in the rayleigh - jeans approximation : @xmath88 . this allows us to use these indices interchangeably to calculate differences ( @xmath33 ) , assuming the dust temperature @xmath22 is fixed . the rayleigh - jeans approximation also provides us with a reliable lower limit for the more realistic emissivity spectral index @xmath3 . [ cols="^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] @xmath89 right ascension and declination of the sources for epoch j@xmath90 + @xmath91 average flux density and uncertainty ( in mjy ) per unit of effective 850 @xmath0 m beam area ( beam width of @xmath92 ) + @xmath93 @xmath1co j=3 - 2 molecular emission and uncertainty as a percentage of the 850 @xmath0 m flux density + @xmath94 ratio of 850 to 450 @xmath0 m flux densities corrected for @xmath1co j=3 - 2 contamination and uncertainty + @xmath95 spectral index corrected for @xmath1co j=3 - 2 contamination as shown in equation [ eq:02 ] and average uncertainty + @xmath96 spectral index difference after correction for @xmath1co j=3 - 2 line emission + @xmath97 young stellar object : protostar ( p ) or young star with disk ( ysd ) + [ tab:02 ] from this perspective alone , the average rayleigh - jeans emissivity spectral index @xmath87 obtained in section 4 after correction for @xmath1co j=3 - 2 contamination ( @xmath98 ) could be interpreted as at the lower limit for cores and filaments . while our spectral index map in figure [ fig:02 ] is very similar to the one presented by @xcite , the mean corrected indices for the contaminated and control samples ( @xmath99 ) are in fact slightly higher than expected from using the temperature - independent rayleigh - jeans approximation . even if this approximation is useful to establish lower limits , we showed in figure [ fig:01 ] that it underestimates the spectral index toward cold regions . as an example , a prestellar core with a real emissivity spectral index @xmath100 at a temperature @xmath101 k would have , using equation [ eq:03 ] , an associated spectral index @xmath102 ( @xmath103 ) . this illustrates the importance of knowing the temperature for finding a reliable value of the emissivity spectral index @xmath3 . a possible explanation for the statistical uncertainties in the average measured rayleigh - jeans emissivity spectral index @xmath98 is that we are comparing objects of different natures . since the properties of most sources in our sample are unknown , they could very well cover a large intrinsic range of emissivity spectral indices @xmath3 as well as dust temperatures @xmath22 . this spread of measured spectral indices could also be explained in part by the uncertainties on the measured ratios , as they are shown in figure [ fig:04 ] . since the ratio method presented in this work is sensitive to small flux variations , these uncertainties are mostly driven by the noise level in the 450 @xmath0 m flux density map . furthermore , if the 450 @xmath0 m map is more susceptible to large - scale fluctuations than its 850 @xmath0 m counterpart , it could contribute to the artificial decrease in the measured ratios , thus increasing the average measured spectral indices for each sample . although this leads to a large range of values for the rayleigh - jeans emissivity spectral index @xmath87 , we know from equations [ eq:05 ] and [ eq:03 ] that the correction @xmath104 for each source depends on the more accurate harp measurements and on the smoother 850 @xmath0 m map . we are confident that this measured effect of molecular contamination on the emissivity spectral index @xmath3 is a reliable estimate , even if the spectral indices themselves are subject to large uncertainties and probably also intrinsic spread . the variation in the average spectral index due to @xmath1co j=3 - 2 line contamination correction @xmath105 illustrates the different scenarios possible in sources contaminated by molecular line emission . for some , it only acts as a correction leading to a slightly more accurate emissivity spectral index @xmath3 . there are however cases where the difference is dramatic enough to change the physical description of the dust mixture in a submillimetre source . as an example , the most highly contaminated source in the sample ( c23 ; @xmath74 ) , associated with a spitzer - identified young star with disk , gives a large correction @xmath106 that brings it on par with the average rayleigh - jeans spectral index measured in omc-4 ( @xmath99 ) . this @xmath107 could mean the difference between grain growth up to centimetre sizes typical of protostellar disks ( @xmath108 ) , and the emissivity spectral index expected in collapsing cores ( @xmath109 ) @xcite . the other contaminated spitzer objects ( c14 , c22 , c24 ; @xmath110 ) all share a correction @xmath111 comparable to the average difference @xmath112 obtained from the entire contaminated sample . as discussed in the previous section , objects of these types ( protostars and young stars with disks ) are likely to be associated with molecular outflows , which would explain their @xmath1co j=3 - 2 line contamination levels @xcite . the dust temperature @xmath22 was assumed fixed up to this point for the analysis of the sources presented in this paper . if we instead suppose a fixed emissivity spectral index @xmath100 @xcite , it is possible to rewrite equations [ eq:05 ] and [ eq:03 ] to evaluate numerically the effect of molecular line contamination on the temperature determination ( @xmath113 ) . for the previous example of a theoretical submillimetre source with an emissivity spectral index @xmath100 and a temperature @xmath101 k , a contamination level of @xmath114 ( @xmath115 ) would then lead to a measured temperature @xmath116 k. this temperature determination is however sensitive to the slightest variation in the 850 to 450 @xmath0 m flux ratio . it becomes unreliable as the temperature increases and the thermal emission approaches the rayleigh - jeans regime . while the 850 to 450 @xmath0 m flux ratio method is relevant to evaluate the effects of molecular contamination on the emissivity spectral index @xmath3 , it is insufficient to properly derive the physical properties of cold interstellar dust grains . it requires some prior assumptions about the emissivity or the temperature , and it is insensitive to line - of - sight density variations @xcite . a more accurate determination of those properties requires a multi - wavelength approach @xcite . even then , molecular line contamination can remain a concern in spectral energy distributions built using scuba-2 850 @xmath0 m observations . as an example , the dust temperature @xmath22 in cold interstellar regions can be well constrained by fitting the peak thermal emission with herschel observations alone . on the other hand , the determination of the emissivity spectral index @xmath3 benefits from the addition of accurate scuba-2 850 @xmath0 m measurements . in their study of the b1 clump in the perseus molecular cloud , @xcite showed that combining observations from both herschel and scuba-2 lead to a better fit than using the herschel data alone for the spectral energy distribution . the degeneracy between the spectral index @xmath3 and the temperature @xmath22 is a serious challenge when attempting to fit accurate spectral energy distributions . @xcite have shown that noise from observations can induce an artificial anti - correlation between @xmath3 and @xmath22 when using a @xmath117 fitting method . this is confirmed by @xcite , who then introduced a hierarchical bayesian fitting technique as an alternative to the @xmath117 method . the hierarchical bayesian technique recovers the parameters of an artificial distribution with more accuracy ; it is less sensitive to statistical noise and calibration uncertainties . it has also been successfully tested on herschel observations of cb244 . furthermore , @xcite compared several fitting methods and found they all exhibited some level of bias when estimating the relationship between @xmath3 and @xmath22 . in particular , the hierarchical bayesian model presented in @xcite , while more precise than the @xmath117 method , is nonetheless biased toward a flat @xmath118 relation in sources with a low signal - to - noise ratio . it is important to point out other possible sources of contamination in the scuba-2 observations . although generally much fainter than @xmath1co lines , the contribution from other molecular lines , such as those from @xmath119co and c@xmath120o @xcite , can add up to a significant portion of the total flux at both 450 and 850 @xmath0 m toward hot cores . the orion kl hot core in omc-1 is famous for having a forest of molecular lines @xcite contributing a significant fraction ( @xmath121 ) of its total flux density at 850 @xmath0 m @xcite . this is also true at 450 @xmath0 m , where the corresponding molecular lines contribute up to @xmath114 of the total flux density toward orion kl @xcite . another common source of contamination is the continuum free - free emission from electrons in the interstellar medium . while it generally represents only a small fraction ( @xmath122 ) of the submillimetre emission at 850 @xmath0 m in the galactic plane @xcite , it can become non - negligible toward regions with strong photo - ionization . as an example , free - free emission in extended hii optically thin regions typically follow a gentle , almost flat , power - law @xcite . it is therefore likely to be measurable even at submillimetre wavelengths . however , this is not necessarily true for partially - thick jets or stellar winds where the free - free spectrum can be steeper @xcite . in those cases , the contamination will depend on the properties of the observed object . one such example is the mwc 297 herbig b1.5 star in the serpens molecular cloud , which exhibits free - free contamination of @xmath123 and @xmath124 at 450 and 850 @xmath0 m respectively @xcite . several regions in the orion a molecular cloud show a strong free - free continuum . @xcite fitted the dust thermal emission and the free - free emission toward three locations in omc-1 . the highest free - free contamination they found at 850 @xmath0 m ( @xmath125 ) is toward the free - free emission peak near the trapezium stars . for orion kl and orion south , the free - free emission has only a small contribution ( @xmath126 of the total flux density at 850 @xmath0 m ) . in each case , the free - free component at 450 @xmath0 m is negligible ( @xmath127 ) when compared to the dust thermal emission . @xcite also estimated free - free contamination in omc-1 observations at 1110 @xmath0 m . this contamination reaches up to @xmath128 in the hii region south - east of the trapezium stars , and @xmath129 in the orion bar . if we use these results as upper limits for the free - free contamination at 850 @xmath0 m , it is possible to estimate its effect on the spectral index @xmath3 that would be calculated from scuba-2 observations . if free - free emission is negligible at 450 @xmath0 m , we know from equation [ eq:05 ] that contamination levels of @xmath130 and @xmath131 at 850 @xmath0 m would overestimate the spectral index by @xmath132 and @xmath133 respectively . however , considering the strong free - free emission toward the hii region , it is possible given a flat power - law that this contribution is not negligible at 450 @xmath0 m after all . if the 450 @xmath0 m emission has some component of free - free emission , this would reduce the bias in the measurement of beta , but would not eliminate it completely . we have presented the scuba-2 shared - risk observations for the orion a molecular cloud complex at 450 and 850 @xmath0 m . they were combined with harp spectroscopic measurements in order to evaluate the effects of molecular contamination on the emissivity spectral index @xmath3 . we studied a list of 33 sources along the integral - shaped filament chosen for their high @xmath1co j=3 - 2 line emission . at least four young stellar objects identified with the spitzer space telescope [ three young stars with disks ( c14 , c22 , c23 ) and one protostar ( c24 ) ] show significant levels of contamination in their measured 850 @xmath0 m flux densities . from the analysis of these contaminated sources , we have concluded that @xmath1co j=3 - 2 line contamination leads to an average underestimation @xmath105 in affected sources , with some individual sources displaying larger differences ( up to @xmath106 ) . while there is no obvious continuum property to easily identify contaminated sources , these results illustrate the need to subtract @xmath1co j=3 - 2 line emission from scuba-2 850 @xmath0 m continuum maps where the corresponding harp measurements are available . the properties shared by the contaminated sources seem to be their peak emission threshold ( @xmath134 mjy beam@xmath11 ) and their more likely location toward regions of low dust column density . emissivity is a crucial parameter for the description of dust grains in the interstellar medium , as it is related directly to size distributions in dust mixtures . measuring it accurately is a necessary step in order to reach , with far - infrared and submillimetre observations , a more complete understanding of the physical processes leading to the formation of stars and their planetary systems . special thanks to john bally , patrice beaudoin , jonathan gagn , maryvonne grin and sarah sadavoy for enlightening discussions . we would like to thank the referee for their suggestions , which helped improve the discussion presented in this paper . this study has been achieved thanks to the support of the staff at the joint astronomy centre and the members of the jcmt gould belt survey team . this work would not have been possible without the support of the national sciences and engineering research council of canada . this research has made use of the simbad database , operated at cds , strasbourg , france . 99 arab h. , abergel a. , habart e. , bernard - salas j. , ayasso h. , dassas k. , martin p. g. , white g. j. , 2012 , a&a , 541 , 19 arzoumanian d. et al . , 2011 , a&a , 529 , l6 bally j. , 2008 , in reipurth b. , ed . , handbook of star forming regions : volume i , the northern sky . asp monograph publications , provo , ut , p. 459 buckle j. v. et al . , 2009 , mnras , 399 , 1026 buckle j. v. et al . , 2012 , mnras , 422 , 521 chapin e. , dempsey j. , jenness t. , scott d. , thomas h. , tilanus r. , 2010 , the scuba-2 sro data reduction cookbook , starlink cookbook 19 chapin e. , gibb a. , jenness t. , berry d. s. , scott d. , tilanus r. , 2015 , sub - millimetre user reduction facility , starlink user note 258 dempsey j. t. , friberg p. , jenness t. , tilanus r. p. j. , thomas h. s. , holland w. s. , bintley d. , berry d. s. , chapin e. l. , chrysostomou a. , davis g. r. , gibb a. g. , parsons h. , robson e. i. , 2013 , mnras , 430 , 2534 dicker s. r. , mason b. s. , korngut p. m. , cotton w. d. , compigne m. , devlin m. j. , martin p. g. , ade p. a. , benford d. j. , irwin k. d. , maddalena r. j. , mcmullin j. p. , sheperd d. s. , sievers a. , staguhn j. g. , tucker c. , 2009 , apj , 705 , 226 di francesco j. , johnstone d. , kirk h. , mckenzie t. , ledwosinska e. , 2008 , apjs,175 , 277 dobbs c. l. , krumholz m. r. , ballesteros - paredes j. , bolatto a. d. , fukui y. , heyer m. , low m - m . m. , ostriker e. c. , vzquez - semadeni e. , 2014 , in beuther h. et al . , eds , protostars and planets vi , formation of molecular clouds and global conditions for star formation . university of arizona , tucson , p.3 drabek e. , hatchell j. , friberg p. , richer j. , graves s. , buckle j. v. , nutter d. , johnstone d. , di francesco j. , 2012 , mnras , 426 , 23 draine b. t. , 2003 , ara&a , 41 , 241 goldsmith p. f. , bergin e. a. , lis d. c. , 1997 , apj , 491 , 615 gould b. g. , 1879 , results argentine natl . , 1 , 1 graves s. , 2011 , phd thesis , univ . cambridge groesbeck t. d. , 1995 , phd thesis , california institute of technology hatchell j. , fuller g. a. , richer j. s. , 2007 , a&a , 472 , 187 hatchell j. , wilson t. , drabek e. , curtis e. , richer j. , nutter d. , di francesco j. , ward - thompson d. , jcmt gbs consortium , 2013 , mnras , 429 , l10 hildebrand r. h. , 1983 , quarterly journal of the royal astronomical society , 24 , 267 holland w. s. et al . , 2013 , mnras , 430 , 2513 johnstone d. , bally j. , 1999 , apj , 510 , l49 johnstone d. , bally j. , 2006 , apj , 653 , 383 johnstone d. , boonman a. m. s. , van dishoeck e. f. , 2003 , a&a , 412 , 157 juvela m. , montillaud j. , ysard n. , lunttila t. , 2013 , a&a , 556 , a63 kelly b. c. , shetty r. , stutz a. m. , kauffmann j. , goodman a. a. , launhardt r. , 2012 , apj , 752 , 55 lequeux j. , 2005 , the interstellar medium . springer , heidelberg lis d. c. , serabyn e. , keene j. , dowell c. d. , benford d. j. , phillips t. g. , hunter t. r. , wang n. , 1998 , apj , 509 , 299 lombardi m. , bouy h. , alves j. , lada c. j. , 2014 , a&a , 566 , a45 megeath s. t. , gutermuth r. , muzerolle j. , kryukova e. , flaherty k. , hora j. l. , allen l. e. , hartmann l. , myers p. c. , pipher j. l. , stauffer j. , young e. t. , fazio g. g. , 2012 , apj , 144 , 192 mezger p. g. , henderson a. p. , 1967 , apj , 147 , 471 natta a. , testi l. , calvet n. , henning th . , waters r. , wilner d. , 2007 , in reipurth d. , jewitt d. , keil k. , eds , protostars and planets v , dust in proto - planetary disks : properties and evolution . university of arizon , tucson , p.767 nutter d. , ward - thompson d. , 2007 , mnras , 374 , 1413 ossenkopf v. , henning th . , 1994 , a&a , 291 , 943 oster l. , 1961 , reviews of modern physics , 33 , 525 palmeirim p. et al . , 2013 , a&a , 550 , a38 peng t.c . , wyrowski f. , zapata l. a. , gsten r. , menten k. m. , 2012 , a&a , 538 , a12 planck collaboration xix . , 2011 , a&a , 536 , a6 planck collaboration xi . , 2014 , a&a , 571 , a11 planck collaboration int . xxiii . , 2015 , a&a , 580 , a13 poidevin f. , bastien p. , matthews b. c. , 2010 , apj , 716 , 893 reynolds s. p. , 1986 , apj , 304 , 713 rumble d. et al . , 2015 , mnras , 448 , 1551 sadavoy s. et al . , 2013 , apj , 767 , 126 salji c. , richer j. s. , buckle j. v. , di francesco j. , hatchell j. , hogerheijde m. , johnstone d. , kirk h. , ward - thompsion d. , jcmt gbs consortium , 2015 , mnras , 449 , 1782 schilke p. , benford d. j. , hunter t. r. , lis d. c. , phillips t. g. , 2001 , apjs , 132 , 281 schnee s. , enoch m. , noriega - crespo a. , sayers j. , terebey s. , caselli p. , foster j. , goodman a. , kauffmann j. , padgett d. , rebull l. , sargent a. , shetty r. , 2010 , apj , 708 , 127 schnee s. , mason b. , di francesco j. , friesen r. , li d. , sadavoy s. , stanke t. , 2014 , mnras , 444 , 2303 serabyn e. , weisstein e. w. , 1995 , apj , 451 , 238 shetty r. , kauffmann j. , schnee s. , goodman a. a. , 2009 , apj , 696 , 676 shetty r. , kauffmann j. , schnee s. , goodman a. a. , ercolano b. , 2009 , apj , 696 , 2234 shirley y. l. , nordhaus m. k. , grevich j. m. , evans n. j. ii , rawlings j. m. c. , tatematsu k. , 2005 , apj , 632 , 982 schlafly e. f. et al . , 2014 , apj , 786 , 29 steinacker j , baes m. , gordon k. d. , 2013 , annual review of astronomy and astrophysics , 51 , 63 ward - thompson d. et al . , 2007 , the publications of the astronomical society of the pacific , 119 , 855 wenger m. , ochsenbein f. , egret d. , dubois p. , bonnarel f. , borde s. , genova f. , jasniewicz g. , lalo s. , lesteven s. , monier r. , 2000 , a&as , 143 , 9 weingartner j. c. , draine b. t. , 2001 , apj , 548 , 296 wiseman j. , ho p. t. p. , 1998 , apj , 502 , 676 the full members of the jcmt gould belt survey consortium ( july 2015 ) are : p. bastien , d.s . berry , d. bresnahan , h. broekhoven - fiene , j. buckle , h. butner , m. chen , a. chrysostomou , s. coud , m.j . currie , c.j . davis , j. di francesco , e. drabek - maunder , a. duarte - cabral , m. fich , j. fiege , p. friberg , r. friesen , g.a . fuller , s. graves , j. greaves , j. gregson , j. hatchell , m.r . hogerheijde , w. holland , t. jenness , d. johnstone , g. joncas , h. kirk , j.m . kirk , l.b.g . knee , s. mairs , k. marsh , b.c . matthews , g. moriarty - schieven , j.c . mottram , c. mowat , k. pattle , j. rawlings , j. richer , d. robertson , e. rosolowsky , d. rumble , s. sadavoy , c. salji , h. thomas , n. tothill , s. viti , d. ward - thompson , g.j . white , j. wouterloot , j. yates , and m. zhu . several sources listed in tables [ tab:01 ] and [ tab:02 ] have also been linked to spitzer - identified young stellar objects from the @xcite catalogue . for completeness , this section lists known objects found in the simbad database within @xmath92 of these scuba-2 submillimetre sources @xcite . these associated objects are listed in roman numerals under each source presented in the following subsections . even though the sources in the contaminated sample ( see table [ tab:01 ] ) were identified based on their strong @xmath1co j=3 - 2 line contamination , there are young stars within one 850 @xmath0 m effective beam - width ( @xmath92 ) of their central position . these are typical variable stars , some of them with significant reddening . there are also two herbig - haro emission jets in the surroundings of the contaminated sources . 1 . v * v2560 ori is an orion - type variable star with a spectral type of m6 located @xmath135 mostly south of c14 with @xmath136 and @xmath137 ; 2 . 2mass j05353920 - 0516355 is an irregular - type variable star located @xmath138 north - east of c14 with @xmath139 and @xmath140 . 1 . mgm2012 1539 is an irregular - type variable star with a spectral type of m5.5 located @xmath141 north - west of c22 with @xmath142 and @xmath143 ; 2 . v * v1514 ori is an orion - type variable star with a spectral type m4.5e located @xmath144 south - west of c22 with @xmath145 and @xmath146 ; 3 . v * my ori ( parenago 1963 ) is an orion - type variable star with a spectral type m5e located @xmath147 south - east of c22 with @xmath148 and @xmath146 ; 4 . hh 561 is an herbig - haro object located @xmath149 south - east of c22 . 1 . tkk 544 is an infrared source located @xmath150 north - west of c23 with @xmath151 and @xmath152 , and it is somewhat nebulous on the 2mass image ; 2 . 2mass j05351649 - 0506003 ( mgm2012 2363 ) is a pre - main sequence star also located @xmath150 north - west of c23 with @xmath153 and @xmath154 ; 3 . jbv2003 sk1-omc3 is a submillimetre source identified with scuba by @xcite and whose peak is located @xmath155 east of c23 ; 4 . hh 357 is an herbig - haro object located @xmath156 south - west of c23 . the reference sample ( see table [ tab:02 ] ) consists of sumillimetre clumps found in omc-4 and listed in @xcite . they were selected as a control group for the contaminated sample because of their low @xmath1co j=3 - 2 line contamination in the 850 @xmath0 m band . 1 . isoy j053444.05 - 054125.7 is a young stellar object located @xmath160 west of source 01 ; 2 . jcmtsf j053443.8 - 054126 , a submillimetre source identified with scuba , is a young stellar object candidate listed in @xcite and whose peak is located @xmath161 west of source 01 . 1 . mgm2012 1216 is a pre - main sequence star located @xmath162 north - east of source 02 ; 2 . jcmtse j053449.8 - 054614 is a submillimetre source identified with scuba , listed in @xcite and whose peak is located @xmath162 mostly west of source 02 . 1 . mgm2012 1400 is a young stellar object located @xmath163 of source 24 ; 2 . jcmtsf j053507.9 - 053556 , a submillimetre source identified with scuba , is a young stellar object candidate listed in @xcite and whose peak is located @xmath164 mostly west of source 24 ; 3 . 2mass j05350800 - 0535537 is a star located @xmath164 north - west of source 24 with @xmath165 and @xmath166 . 1 . v * v2260 ori is an orion - type variable star with a spectral type k4-m0 located @xmath157 west of source 39 with @xmath167 and @xmath168 ; 2 . nw2007 orionan-0535149 - 53307 is a young stellar object candidate listed in @xcite and located @xmath169 west of source 39 .

we examine images of the little things sample of nearby dwarf irregular ( dirr ) and blue compact dwarf ( bcd ) galaxies to identify distinct young regions in their far outer disks . we use these data , obtained with the _ galaxy evolution explorer _ satellite , to determine the furthest radius at which _ in situ _ star formation can currently be identified . the knots are found at distances from the center of the galaxies of 1 to 8 disk scale lengths and have ages of myrs and masses of 20 m_ to m_ . the presence of young clusters and ob associations in the outer disks of dwarf galaxies shows that dirrs do have star formation taking place there in spite of the extreme nature of the environment . most regions are found where the hi surface density is m_pc , although both the hi and dispersed old stars go out much further . this limiting density suggests a cutoff in the ability to form distinct ob associations and perhaps even stars . we compare the star formation rates in the regions to the average rates expected at their radii and beyond from the observed gas , using the conventional correlation for gas - rich regions . the localized rates are typically 10% of the expected average rates for the outer disks . either star formation in dirrs at surface densities pc occurs without forming distinct associations , or the kennicutt - schmidt relation over - predicts the rate beyond this point . in the latter case , the stellar disks in the far - outer parts of dirrs result from scattering of stars from the inner disk .

1603.00495

star formation in dwarf irregular galaxies ( dirrs ) is difficult to understand because gas densities are low , but this is particularly true in the outer parts where the gas densities are even lower than in the central regions . there , the standard 2-dimensional @xcite gravitational instability parameter for the gas , suggested by @xcite as an important parameter for star formation , predicts that the gas is highly stable against collapse into star - forming clouds @xcite . yet , @xmath0 emission , a signpost of recent star formation , is detected even into the outer disks . in a deep - imaging study , we obtained @xmath8-band and _ galaxy evolution explorer _ ( _ galex _ ; * ? ? ? * ) @xmath0 images of 4 dirr galaxies and one blue compact dwarf ( bcd ) , and @xmath9-band images for 3 of the dirrs @xcite . through surface photometry we traced stellar disks in these images to @xmath10 mag arcsec@xmath6 . we found that the stellar surface brightness in both @xmath8 and @xmath0 continues exponentially as far out as we can measure ( see also * ? ? ? the presence of @xmath0 emission into the outer disk suggests a continuity of star formation with radius even though the outer gas is stable against two - dimensional gravitational collapse in the conventional model @xcite . in spiral galaxies , numerical simulations show that spiral arms are capable of scattering stars into the outer disks @xcite , and this could be a significant source of the stellar populations there . however , dirrs do not have spiral arms . are outer disks of dirrs populated primarily through stars scattered from the inside out or from _ in situ _ star formation ? are the outer parts of dirrs from halos and not disks @xcite ? if star formation is taking place in the outer parts of dirrs , then there should be young @xmath0-bright star clusters , not just diffuse @xmath0 emission . we have previously identified clusters and associations to a radius of a few disk scale lengths @xcite . here we use @xmath0 images to search for young star clusters and ob associations in the far outer regions of a larger sample of nearby dirr galaxies . since scattering takes longer than the ages of young clusters and strong scattering may destroy clusters ( but see * ? ? ? * ) , this search is a good test for _ in situ _ star formation in outer disks . we also examine the environments for star formation in the far outer parts of these galaxies ( [ environ ] ) . the pressure and average density are low , gravity is weak , and the dynamical time is long . most of the identified regions occur near the radius where the average gas surface density drops below @xmath11 pc@xmath6 , suggesting a cutoff or slowdown at lower densities . the gas and distributed stellar population go further than this . to test whether the outer disk might be built up from dispersed regions like those we observe , we compare in [ outer ] the star formation rates ( sfrs ) in the @xmath0 regions to the average rates expected in the outer disks from the kennicutt - schmidt relation @xcite . the sample of galaxies is taken from little things ( local irregulars that trace luminosity extremes , the h@xmath4i nearby galaxy survey , * ? ? ? this is a multi - wavelength survey of nearby ( @xmath12 mpc ) dirr galaxies and blue compact dwarfs ( bcds ) , which builds on the things project , whose emphasis was on nearby spirals @xcite . the galaxies and a few key parameters are listed in table [ tab - data ] . the galaxies haro 29 , haro 36 , mrk 178 , and viizw 403 , which appear at the end of the table , are classified as bcds . ngc 3738 , classified as a dirr , has characteristics , such as a central concentration of stars , star formation , and gas , that are similar to bcds , and so we will include it with the bcds when that group is singled out . the little things sample was chosen to be relatively isolated because we are investigating internal processes of star formation . none of the galaxies are companions to large spiral galaxies , members of dense clusters , or obviously engaged in an interaction with another galaxy . most are , however , members of groups and so other galaxies are around . in addition a few galaxies and particularly the bcds ( ngc 1569 , ic 10 , haro 29 , haro 36 , mrk 178 , viizw 403 ) are undergoing bursts of star formation that may have been triggered by some sort of external process , including mergers @xcite . @xcite gives the nearest neighbor to each galaxy and their separation . the non - starburst little things galaxies are 110 - 1100 kpc from their nearest known neighbor , with a median separation of 480 kpc . the starburst sub - sample is at a comparable range of distances 240 - 900 kpc with a median separation of 450 kpc . therefore , we expect that for most of the galaxies the extra - galactic environment is not a factor in determining the extent of star - formation in the systems . lcccrrccc cvnidwa & 3.6 & [email protected] & 0.56@xmath13 0.49 & 12:38:37.8 & + 32:45:40 & [email protected] & 2.0@xmath13 0.9 & 0.9@xmath13 0.8 + ddo 43 & 7.8 & [email protected] & 1.46@xmath13 0.53 & 7:28:16.7 & + 40:47:01 & [email protected] & 2.2@xmath13 0.3 & 1.3@xmath13 0.5 + ddo 46 & 6.1 & [email protected] & 1.27@xmath13 0.18 & 7:41:26.9 & + 40:05:08 & [email protected] & 2.3@xmath13 0.1 & 2.4@xmath13 0.3 + ddo 47 & 5.2 & [email protected] & & 7:42:00.0 & + 16:49:40 & [email protected] & 4.2@xmath13 0.2 & + ddo 50 & 3.4 & [email protected] & 2.65@xmath13 0.27 & 8:18:23.8 & + 70:43:18 & [email protected] & 3.3@xmath13 0.1 & 1.8@xmath13 0.2 + ddo 52 & 10.3 & [email protected] & 2.80@xmath13 1.35 & 8:28:26.3 & + 41:50:25 & [email protected] & 2.7@xmath13 0.1 & 1.2@xmath13 0.6 + ddo 53 & 3.6 & [email protected] & 0.62@xmath13 0.09 & 8:34:11.8 & + 66:10:12 & [email protected] & 2.5@xmath13 0.1 & 1.9@xmath13 0.3 + ddo 63 & 3.9 & [email protected] & 1.31@xmath13 0.10 & 9:40:56.0 & + 71:12:15 & [email protected] & 4.2@xmath13 0.1 & 2.2@xmath13 0.2 + ddo 69 & 0.8 & [email protected] & 0.27@xmath13 0.05 & 9:59:12.4 & + 30:44:37 & [email protected] & 4.0@xmath13 0.2 & 2.8@xmath13 0.5 + ddo 70 & 1.3 & [email protected] & 0.13@xmath13 0.07 & 10:00:05.4 & + 5:17:47 & [email protected] & 3.1@xmath13 0.1 & 10.3@xmath13 5.5 + ddo 75 & 1.3 & [email protected] & 0.71@xmath13 0.08 & 10:11:08.5 & -4:39:07 & [email protected] & 7.7@xmath13 0.4 & 1.9@xmath13 0.2 + ddo 87 & 7.7 & [email protected] & 0.99@xmath13 0.11 & 10:49:25.1 & + 65:30:41 & [email protected] & 3.5@xmath13 0.1 & 4.3@xmath13 0.5 + ddo 101 & 6.4 & [email protected] & 1.16@xmath13 0.11 & 11:55:36.4 & + 31:31:18 & [email protected] & 1.3@xmath13 0.1 & 1.1@xmath13 0.1 + ddo 126 & 4.9 & [email protected] & 0.60@xmath13 0.05 & 12:27:14.1 & + 37:08:01 & [email protected] & 4.0@xmath13 0.6 & 5.6@xmath13 0.5 + ddo 133 & 3.5 & [email protected] & 2.25@xmath13 0.24 & 12:32:48.9 & + 31:31:42 & [email protected] & 1.8@xmath13 0.1 & 1.0@xmath13 0.1 + ddo 154 & 3.7 & [email protected] & 0.62@xmath13 0.09 & 12:53:58.4 & + 27:07:20 & [email protected] & 5.5@xmath13 0.2 & 4.3@xmath13 0.6 + ddo 167 & 4.2 & [email protected] & 0.56@xmath13 0.11 & 13:13:20.6 & + 46:19:32 & [email protected] & 3.2@xmath13 0.2 & 1.2@xmath13 0.3 + ddo 168 & 4.3 & [email protected] & 0.72@xmath13 0.07 & 13:14:31.4 & + 45:54:12 & [email protected] & 2.7@xmath13 0.1 & 3.1@xmath13 0.3 + ddo 187 & 2.2 & [email protected] & 0.28@xmath13 0.05 & 14:15:54.9 & + 23:03:34 & [email protected] & 1.1@xmath13 0.2 & 1.5@xmath13 0.3 + ddo 210 & 0.9 & [email protected] & & 20:46:51.7 & -12:51:20 & [email protected] & 1.8@xmath13 0.1 & + ddo 216 & 1.1 & [email protected] & 1.77@xmath13 0.45 & 23:28:39.7 & + 14:44:47 & [email protected] & 1.1@xmath13 0.0 & 0.3@xmath13 0.1 + f564-v3 & 8.7 & [email protected] & 0.73@xmath13 0.40 & 9:02:55.6 & + 20:04:31 & [email protected] & 2.0@xmath13 0.3 & 1.7@xmath13 0.9 + ic 1613 & 0.7 & [email protected] & 0.71@xmath13 0.12 & 1:05:08.9 & + 2:14:15 & [email protected] & 3.3@xmath13 0.1 & 2.5@xmath13 0.4 + lgs 3 & 0.7 & [email protected] & 0.27@xmath13 0.08 & 1:03:56.4 & + 21:53:56 & [email protected] & 2.0@xmath13 0.1 & 1.2@xmath13 0.4 + m81dwa & 3.6 & [email protected] & 0.38@xmath13 0.03 & 8:23:52.5 & + 71:01:31 & [email protected] & 2.6@xmath13 0.1 & 1.9@xmath13 0.2 + ngc 1569 & 3.4 & [email protected] & 0.85@xmath13 0.24 & 4:30:58.9 & + 64:50:20 & [email protected] & 2.5@xmath13 0.1 & 1.3@xmath13 0.4 + ngc 2366 & 3.4 & [email protected] & 2.57@xmath13 0.80 & 7:29:36.2 & + 69:16:22 & [email protected] & 3.6@xmath13 0.5 & 2.6@xmath13 0.8 + ngc 3738 & 4.9 & [email protected] & 1.16@xmath13 0.20 & 11:35:47.2 & + 54:31:31 & [email protected] & 1.6@xmath13 0.1 & 1.0@xmath13 0.2 + ngc 4163 & 2.9 & [email protected] & 0.71@xmath13 0.48 & 12:12:09.2 & + 36:10:27 & [email protected] & 1.5@xmath13 0.1 & 0.7@xmath13 0.4 + ngc 4214 & 3.0 & [email protected] & 0.83@xmath13 0.14 & 12:15:30.8 & + 36:13:46 & [email protected] & 7.3@xmath13 0.1 & 6.6@xmath13 1.1 + ngc 6822 & 0.5 & [email protected] & [email protected] & 19:43:34.1 & -14:33:42 & [email protected] & 1.5@xmath13 0.1 & [email protected] + sagdig & 1.1 & [email protected] & 0.57@xmath13 0.14 & 19:30:00.8 & -17:39:56 & [email protected] & 2.0@xmath13 0.3 & 1.1@xmath13 0.3 + wlm & 1.0 & [email protected] & 0.83@xmath13 0.16 & 0:01:46.9 & -15:27:52 & [email protected] & 1.7@xmath13 0.0 & 2.5@xmath13 0.5 + haro 29 & 5.8 & [email protected] & 1.15@xmath13 0.26 & 12:26:18.3 & + 48:29:46 & [email protected] & 2.6@xmath13 0.2 & 0.7@xmath13 0.2 + haro 36 & 9.3 & [email protected] & 1.16@xmath13 0.13 & 12:46:59.4 & + 51:36:53 & [email protected] & 1.8@xmath13 0.1 & 1.5@xmath13 0.2 + mrk 178 & 3.9 & [email protected] & 0.38@xmath13 0.00 & 11:33:33.7 & + 49:13:31 & [email protected] & 7.6@xmath13 0.2 & 3.8@xmath13 0.1 + viizw 403 & 4.4 & [email protected] & 1.02@xmath13 0.29 & 11:28:01.9 & + 78:59:50 & [email protected] & 0.6@xmath13 0.1 & 0.3@xmath13 0.1 + we used @xmath0 ( 1516 ) images obtained by _ galex _ @xcite to identify knots of emission in the outer disks of each galaxy . in order to better distinguish knots from the wide - spread diffuse emission , we subtracted the stellar continuum from each @xmath0 image using the @xmath8-band image . the @xmath0 image was geometrically transformed to match the pixel scale and dimensions of the @xmath8-band image . the @xmath8-band image was scaled and gaussian - smoothed so that , as well as possible , when it was subtracted from the @xmath0 image , the diffuse emission disappeared . in order to select @xmath0 knots that we were confident belonged to the galaxy , we used @xmath14 false - color pictures produced by the _ galex _ pipeline to pick out bright , blue knots that stood out compared to the surrounding background / foreground populations . in particular we were looking at the density of comparably bright and blue objects well beyond the galaxy . the shape of the object is also a clue : is it resolved compared to a star or lumpy ? as a training exercise , we began with a few galaxies also studied by @xcite , and compared our identifications of individual regions with theirs . the distance to the furthest knot from the center of the galaxy in the plane of the galaxy , @xmath15 , was determined by comparing the location of the knot to ellipses superposed that were used for surface photometry . the major axis and minor - to - major axis ratio of the ellipses were determined for the galaxy from the @xmath8-band image . we then identified the annulus or ellipse boundary at which the @xmath0 knot fell . the @xmath15 are given in table [ tab - data ] , along with the ratios of @xmath15 to the disk scale length , @xmath16 , and to the break radius , @xmath17 . the break radius is the radius at which the exponential stellar surface brightness profile changes slope . dwarf irregular galaxies , despite having irregular and sporadic star formation , have well - behaved exponential stellar disk profiles , even out to 6 - 10@xmath16 @xcite . these azimuthally - averaged profiles , like those in spirals , often show abrupt breaks , most often dropping in brightness into the outer disk more steeply than in the interior disk ( type ii profiles ) but occasionally dropping less steeply ( type iii ) . surface brightness profiles without breaks ( type i ) are actually rare . @xmath16 and @xmath17 were taken from @xcite . two galaxies ( ddo 47 , ddo 210 ) do not have breaks in their exponential disk surface brightness profiles . the @xmath0 regions are identified on the @xmath0 images shown in figure [ fig - bw ] , along with the h@xmath4i shown as contours from the integrated h@xmath4i maps . the h@xmath4i extends much further than the outermost @xmath0 regions , which are usually located just inside the 5th or 6th contour , counting inwards from the outermost one . the h@xmath4i column density there is @xmath18 atoms cm@xmath6 , which corresponds to @xmath20 pc@xmath6 assuming a mean atomic weight of 1.36 times the hydrogen mass . the h@xmath4i maps are irregular , as expected for the dirr morphology , and the @xmath0 regions are often associated with one of these irregularities , indicating the presence of a local gas concentration . we show in [ relations ] that the average h@xmath4i density at the radius of the @xmath0 region is several times less than this local density . we have estimated ages and masses of the knots by fitting aperture photometry from the @xmath0 , @xmath21 , @xmath22 , @xmath9 , and @xmath8 images @xcite . we determined the brightness of the underlying stellar disk for subtraction from the region photometry for most of the regions using the mode of the counts in an annulus 5 pixels ( 1.35 to 5.67 ) wide and 2 pixels ( 0.54 to 4.6 ) in radius beyond the photometric aperture . since the regions are generally in the far outer parts of the disk , they are not crowded and aperture photometry is adequate . for some galaxies , such as those in which the star formation is centrally concentrated , the @xmath0 regions are too crowded with other regions for an annulus around the region to give a reliable measure of underlying galaxy . for these galaxies ( ddo 53 , ddo 101 , m81dwa , ngc 3738 , ngc 4163 , haro 29 , haro 36 , viizw 403 ) , we determined the background from nearby areas at approximately the same disk surface brightness as that where the @xmath0 region was found . no @xmath23 photometry was available from our data for the @xmath0 region in ngc 6822 . the aperture radius , colors , and e(@xmath24 ) used for reddening corrections are given in table [ tab - phot ] . e(@xmath24 ) includes foreground reddening from @xcite and 0.05 mag for internal reddening . the internal reddening is half that determined from balmer decrements in h@xmath4ii regions in a sample of 39 dirr galaxies , as described in @xcite . the expectation is that these stars are not as embedded as one would find in h@xmath4ii regions . furthermore , since most of these regions are in the outer disks of the dwarfs , extinction due to dust is expected to be minor . we use the reddening law of @xcite to produce the extinction in each optical filter . for the _ galex _ filters , we use @xmath25 and @xmath26 taken from @xcite and discussed in relation to the little things dwarfs by @xcite . to model the ages and determine masses , we compared the integrated cluster photometry with cluster evolutionary models of @xcite . these models trace the evolution of a @xmath27 m_@xmath2 cluster over time , assuming a @xcite stellar initial mass function from 0.1 to 100 m_@xmath2 . we used the models for @xmath28 metallicity , since this is typical of the little things dwarfs . see @xcite and @xcite for more details on our color fitting procedure . the ages and masses of the clusters are given in table [ tab - phot ] , values are shown as histograms in figures [ fig - age ] and [ fig - mass ] , and mass is plotted against age in figure [ fig - massvsage ] . the uncertainties in the ages and masses represent the range of ages allowed by the colors . the ages are all less than 20 myrs . young ages are expected since the regions were selected to be bright knots of @xmath0 emission . masses range from 20 m_@xmath2 to @xmath29 m_@xmath2 . there is no correlation between age or mass and distance from the center of the galaxy . eight of the galaxies in our sample were also analyzed by @xcite . working from the @xmath21 images , they identified , not just the furthest regions , but all of the discrete regions in the galaxies . the ensembles of clusters of these 8 galaxies extended to radii that were comparable to ours : 0.7 - 1.1 times the radii where we identified our furthest @xmath0 knots . their ensembles included older clusters , but the masses of our regions are within the range of the masses found by @xcite out at their furthest radius . the exceptions are ddo 70 and ddo 210 which have masses that are lower by a factor of 15 and 20 , respectively , in our study . thus , we conclude that the regions we identified in our study are not unusual compared to the collection of star clusters one finds in the outer disks of dirr galaxies in general . 0.25truein lccccccccccc cvnidwa & 0.055 & 5.7 & 20.786 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 3@xmath13 2 & 360@xmath13 250 + ddo 43 & 0.105 & 5.7 & 21.009 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 5@xmath13 1 & 2000@xmath13 490 + ddo 46 & 0.103 & 5.7 & 23.051 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 0 & & + ddo 47 & 0.073 & 7.9 & 20.711 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 3@xmath13 2 & 960@xmath13 670 + ddo 50 & 0.073 & 14.7 & 18.804 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 4@xmath13 1 & 1900@xmath13 440 + ddo 52 & 0.080 & 7.9 & 20.977 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 4@xmath13 3 & 5000@xmath13 1500 + ddo 53 & 0.075 & 5.0 & 19.402 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 5@xmath13 2 & 2000@xmath13 90 + ddo 63 & 0.062 & 34.0 & 19.212 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 16@xmath13 2 & 15000@xmath13 4700 + ddo 69 & 0.050 & 9.1 & 21.245 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 12@xmath13 5 & 30@xmath13 16 + ddo 70 & 0.063 & 5.7 & 21.486 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 4@xmath13 1 & 18@xmath13 4 + ddo 75 & 0.068 & 4.7 & 20.101 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 3@xmath13 1 & 110@xmath13 55 + ddo 87 & 0.050 & 6.8 & 21.938 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 3@xmath13 3 & 550@xmath13 16 + ddo 101 & 0.058 & 6.8 & 20.447 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 15@xmath13 3 & 10600@xmath13 3900 + ddo 126 & 0.050 & 13.6 & 20.077 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 7@xmath13 2 & 2400@xmath13 1500 + ddo 133 & 0.053 & 11.3 & 18.827 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 4@xmath13 1 & 3000@xmath13 710 + ddo 154 & 0.058 & 10.2 & 20.747 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 0 & & + ddo 167 & 0.050 & 4.5 & 21.541 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 13@xmath13 5 & 1200@xmath13 710 + ddo 168 & 0.050 & 6.8 & 19.342 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 20@xmath13 2 & 100300@xmath1319000 + ddo 187 & 0.050 & 5.7 & 20.289 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 4@xmath13 2 & 350@xmath13 90 + ddo 210 & 0.085 & 5.4 & 21.306 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 6@xmath13 1 & 16@xmath13 1 + ddo 216 & 0.073 & 5.0 & 21.297 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 8@xmath13 3 & 100@xmath13 60 + f564-v3 & 0.068 & 4.9 & 22.388 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 2@xmath13 2 & 1100@xmath13 150 + ic 1613 & 0.055 & 23.2 & 18.606 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 3@xmath13 2 & 110@xmath13 80 + lgs3 & 0.085 & 5.7 & 23.321 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 12@xmath1312 & 20@xmath13 15 + m81dwa & 0.072 & 6.8 & 21.248 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 10@xmath13 3 & 48000@xmath1324000 + ngc 1569 & 0.558 & 4.2 & 22.336 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 15@xmath1312 & 24000@xmath1318000 + ngc 2366 & 0.093 & 33.6 & 19.557 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1 & 6@xmath13 3 & 8800@xmath13 6700 + ngc 3738 & 0.050 & 4.5 & 16.175 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 8@xmath13 3 & 130000@xmath1387000 + ngc 4163 & 0.050 & 4.5 & 18.736 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 15@xmath13 2 & 11000@xmath13 1800 + ngc 4214 & 0.050 & 6.3 & 21.373 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 3@xmath13 2 & 110@xmath13 80 + ngc 6822 & 0.265 & 120.0 & 16.710 & [email protected] & [email protected] & & & & 1 & 3@xmath13 1 & + sagdig & 0.188 & 4.9 & 20.694 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 8@xmath13 3 & 290@xmath13 180 + wlm & 0.068 & 6.8 & 19.524 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 4@xmath13 1 & 60@xmath13 14 + haro 29 & 0.050 & 4.9 & 19.518 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 17@xmath13 5 & 27000@xmath13 9000 + haro 36 & 0.050 & 4.9 & 19.777 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 9@xmath13 4 & 21000@xmath13 9400 + mrk 178 & 0.050 & 6.7 & 20.271 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 2 & 11@xmath13 7 & 1300@xmath13 900 + viizw 403 & 0.073 & 4.3 & 18.209 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 3 & 3@xmath13 2 & 6100@xmath13 4300 + histograms of the ratios @xmath30 and @xmath31 are shown in figure [ fig - histrats ] . there is structure due to star formation present in almost all of the outer disks . all of the furthest @xmath0 knots in the dirr galaxies are beyond one disk scale length and 69% are beyond 2@xmath16 . a few ( ddo 70 and ngc 4214 ) are out as far as 7 - 8@xmath16 . in terms of the break in the @xmath8-band surface brightness profiles , in 57% of the dirr galaxies the furthest @xmath0 knots are at radii @xmath322@xmath17 , and so the regions are found roughly near the break . in two of the galaxies ( ddo 75 and ngc 6822 ) the regions are found around 10@xmath17 . ngc 6822 is one of the extreme dirr galaxies with a @xmath31 ratio of 10 . the @xmath0 region is located in an h@xmath4i cloud to the northwest of the galaxy center at a distance of 4.6 kpc . this region had been suggested to be a companion that is interacting with ngc 6822 @xcite , but an _ hst _ study of the stellar populations of this region and others along the h@xmath4i extent found that the star formation histories of all six of their positions are similar over the past 500 myr @xcite . @xcite argue that the companion " has an old stellar population that is like that of the extended halo of ngc 6822 . therefore , we include it here as the furthest @xmath0 region in ngc 6822 . ddo 75 is also a dirr with a @xmath31 ratio of 10 , but it also has a peculiar surface brightness profile . the @xmath8-band surface brightness profile is flat to mildly increasing in brightness from the center to 0.7 kpc , and then the surface brightness drops like a normal exponential profile to a radius of 1.3 kpc @xcite . after that , as the surface brightness becomes fainter with radius , there are superposed up and down wiggles in the surface brightness profile @xcite . therefore , @xmath16 may not be as well defined as it is in other galaxies . the two dirrs with the highest @xmath30 ratios are ddo 70 and ngc 4214 . the surface brightness profile of ddo 70 is also flat but only to a radius of 0.1 kpc and then is well - behaved after that @xcite . ngc 4214 has a steeper profile in the inner disk , to a radius of 0.8 kpc . beyond that radius , the surface brightness drops as a normal exponential . in spite of these small peculiarities , it is not obvious what is special about ddo 70 and ngc 4214 that they would have @xmath0 knots at such large @xmath16 ; there are other dwarfs that are more peculiar than these with smaller @xmath30 . for the bcds , 43% of the furthest @xmath0 regions are beyond 2@xmath16 and 80% are found relatively near the break radius , with only one further out at 3 - 4@xmath17 . statistics are poor , of course , with only five bcds . nevertheless , we expect bcds to be more centrally concentrated than normal dirrs , and their @xmath15 are roughly in agreement with that . the bcds in the little things sample have also been shown to likely be the result of an interaction or merger that has driven material towards the central regions of the galaxies @xcite . to explore a potential connection between large - scale structural peculiarities and the extended nature of star formation in dwarfs , we look at the relationship between @xmath30 and differences between optical and h@xmath4i kinematic position angles . the position angle of the optical major axis was determined from smoothed outer isophotes of the @xmath8-band images @xcite , and the h@xmath4i kinematic major axis was determined from iterative fits to the bulk velocity field of the gas @xcite . twenty - six of the little things galaxies were in the study by @xcite and all but two are included in this study . we would expect these two position angles - one describing the stellar disk and one describing the gas disk to be co - located , but over half of the little things galaxies have @xmath33pa that are larger than 20 . the large differences could represent some peculiarity of the galaxy structure . in figure [ fig - delpa ] we plot @xmath30 as a function of the difference in the position angles . we find that galaxies with large @xmath33pa are not distinguished in terms of @xmath30 compared to those with small @xmath33pa , although there is a decline in the upper envelope such that @xmath30 declines as @xmath33pa increases . -0.1truein -0.1truein -0.1truein in figure [ fig - sfr ] we plot @xmath30 as a function of the galactic integrated sfr determined from the @xmath0 luminosity @xcite : @xmath34 ( m_@xmath2yr@xmath35 kpc@xmath6 ) . the sfr is normalized to the area contained within one disk scale length . the thought here is that if a galaxy has a higher sfr , it might also have more extended star formation . in figure [ fig - sfr ] we see that , with one exception , the galaxies with the highest sfrs do also have the furthest @xmath0 knots in terms of disk scale lengths . in addition , the galaxies with lowest sfrs have among the lower @xmath30 ratios . however , the galaxies in between these extremes display a range of sfrs and a range of @xmath30 without a clear trend . to explore the relationship between the extent of the @xmath0 knots and the average gas densities at that radius , we plot @xmath30against the log of the azimuthally - averaged h@xmath4i surface density @xmath36 @xcite at the radius of @xmath15 in figure [ fig - hi ] . there is no trend of @xmath30 with @xmath36 , but most points are centered around a value of 1 m_@xmath2pc@xmath6 . fifteen galaxies in the present study have outer star formation knots within the radial range of the survey of dirr in @xcite , for which we determined gas column densities , scale heights , midplane densities , and toomre @xmath37 values as a function of radius for gas and stars . the scale heights and densities came from solutions to the vertical equilibrium of a three component fluid consisting of gas , stars , and dark matter , using observed velocity dispersions and column densities . the @xmath37 values also included the observed rotation curves to give the epicyclic frequency . figures [ fig - hisgas ] and [ fig - hisstar ] show histograms of these quantities for gas and stars , respectively , at the radii of the @xmath0 knots . comparing the two figures , the average densities and column densities for the stars are much lower than for the gas at these radii , emphasizing that the @xmath0 knots are in gas - dominated regions . @xmath37 is high for both gas and stars , but higher for stars because of the lower stellar surface densities . the scale heights at these radii are about the same for the two components . the average environment at the radius of an @xmath0 region is important for understanding how the gas first got concentrated into a cloud to begin the process of star formation . h@xmath4iexcesses near the identified regions are sometimes directly evident as deformations in the h@xmath4icontours in figure [ fig - bw ] . almost all of the @xmath0 knots occur in larger - scale @xmath0 structures too . these large - scale structures typically have a size comparable to the disk scale length . their formation would presumably involve processes that operate on the local average dynamical time , which depends on the average midplane density . in these far - outer regions , the average midplane density is typically 10 or more times lower than the gas density in the solar neighborhood ( which is about 1 @xmath38 ) as a result of both the low h@xmath4i surface density and the large gas scale height from weak gravity ( fig . [ fig - hisgas ] ) . consequently , the dynamical time over the region that the star - forming gas came from was long : @xmath39 myr for midplane gas density @xmath40 at the typical value of @xmath41 @xmath38 . the free fall time is taken to be @xmath42 for @xmath43 and @xmath44 g. this time is 20 times the average age of the knots , which is 7 myr from table 2 . unless there is a relatively dense layer of undetected molecular gas near the midplane or some triggering event like an extragalactic cloud impact , the formation of the observed knots by large - scale dynamical processes requires a relatively long time . this suggests there was a threshold density or column density for the cluster to form so the cloud could grow without early disruption from the first few stars . a threshold surface density for star formation in the outer parts of dirr galaxies is most likely not the same as that in the solar neighborhood ( e.g. , * ? ? ? * ) because the pressures and metallicities in dirrs are much lower . the threshold column density for strong self - gravity in a cloud scales as the square root of the boundary pressure , with collapse proceeding exponentially slower as @xmath45 increases @xcite . with @xmath46 lower by a factor of 10 , the threshold column density for strong gravity in our case should be lower by a factor of about 3 . that would make it @xmath47 pc@xmath6 inside the star - forming region of a cloud , if we take the solar neighborhood threshold as @xmath48 pc@xmath6 from @xcite . another threshold involves opacity and the formation of strong coolants , like co. the opacity threshold , measured in terms of a mass column density threshold , scales inversely with metallicity , and the metallicities for these dirrs are lower than solar by a factor of about 8 . thus the mass column density for an opacity threshold should be 8 times larger in these dirrs than in the solar neighborhood . if @xmath49 mag is the local threshold for co formation @xcite , then the corresponding surface density of gas would be @xmath50 pc@xmath6 for the clouds that formed our @xmath0 knots . this is what we observe for co emission in the dirr wlm , where the metallicity has this value @xcite . if both strong self - gravity and high extinction are required for star formation , then the larger of these two thresholds should apply . however , if strong self - gravity alone is required and the opacity threshold automatically follows during the ensuing collapse , then only the lower threshold is required . this cloud threshold , @xmath51 pc@xmath6 , is a factor of @xmath52 larger than the average surface density of h@xmath4i at the radius of the @xmath0 knots , which is @xmath53 pc@xmath6 from figures [ fig - hi ] and [ fig - hisgas ] . a factor of 13 increase in column density corresponds to a factor of @xmath54 contraction in two dimensions . this is about the same factor that clouds in the solar neighborhood would have to contract if the cloud threshold for star formation always scales with the square root of the ambient pressure . in this sense , star formation in outer dirr disks may not be qualitatively different from other star formation . this picture of star formation at the local dynamical rate changes if a particular event was triggered by something rapid , such as an extragalactic cloud impact . some bcds and extremely low - metallicity dwarf galaxies have single dominant star - forming regions with lower metallicities than in the rest of the galaxy @xcite . these regions could be the sites of extragalactic cloud impacts with low metallicities from nearly pristine gas . one example , kiso 5639 @xcite , has a star formation rate surface density in a @xmath55 pc region that is @xmath56 larger than the value expected from the kennicutt - schmidt relation at the likely gas surface density . the total young mass there is @xmath57 for a @xmath58 galaxy . these regions are all much larger and brighter than the @xmath0 regions studied here , which look more typical for disk galaxies , although they formed at extremely low gas densities . one of the motivations for the present study was to determine if star formation could account for the outer @xmath0 disks of dirr galaxies , as opposed to stellar scattering from the inner disk . here we compare the star formation rate inside each @xmath0 knot , obtained from the ratio of the knot mass to the age , to the total rate inside an annulus of width @xmath59 at the radius of the knot , as determined from the average kennicutt - schmidt ( ks ) relation for dirrs at the local gas surface density @xcite . if the total rate in the small knot is comparable to the total rate expected from the gas in the annulus , which is based on large - scale averages from many other observations , then one might conclude that star - forming regions like these can generally populate outer disks . if the annular rate is much larger than the local rate , then either we are missing @xmath0 regions that are too small , faint , or dispersed to detect ( e.g. , * ? ? ? * ) , or the average ks relation over - estimates the star formation rate at and below a gas surface density of @xmath11 pc@xmath6 , where our @xmath0 regions begin to disappear . such an over - estimate would imply that star formation virtually stops below @xmath11 pc@xmath6 , according to our understanding of conventional processes ( e.g. , * ? ? ? * ) . in that case , stellar scattering from the inner disk would seem to be important in filling the outer disk with stars . the outer disks have relatively smooth exponential radial profiles in @xmath8-band down to @xmath60 pc@xmath35 or lower @xcite . this is below the average stellar surface density at the radius of the @xmath0 regions , as shown in figure [ fig - hisstar ] . the left - hand panel of figure [ fig - hissfr ] shows a histogram of star formation rates in the @xmath0 knots . there is a wide range of rates , primarily because the regions vary a lot in mass and age the middle panel shows a histogram of the ratio of the theoretical star formation rate in an annulus of width @xmath59 around a knot to the rate in the knot . the right - hand panel shows the ratio of the theoretical rate integrated from the knot radius to infinity compared to the knot rate . for the theory , we use the expression @xmath61 for h@xmath4i surface density @xmath62 in @xmath63 pc@xmath6 . this implies that in the annulus , @xmath64 and for the integral @xmath65 the theoretical rate follows from the expression @xmath66 for a pure gas disk after substituting @xmath67 and @xmath68 inside the equation for free fall time given in the previous section , i.e. , @xmath69 . \label{eq : tff}\ ] ] the theoretical rate compares well with observations of 20 dirr galaxies if @xmath70 and the velocity dispersion is the observed value of @xmath71 km s@xmath35 @xcite . it also agrees with observations in the outer parts of spiral galaxies , which are gas - dominated like the dirrs @xcite . this is the steep part of the ks relation , beyond the main disks of spirals , where atoms dominate molecules , gas dominates stars , and the total consumption time approaches 100 gyrs @xcite . note that equation ( [ eq : tff ] ) gives @xmath72 myr for @xmath73 pc@xmath6 and @xmath74 km s@xmath35 , so the gas consumption time would be @xmath75 gyr with @xmath76 . for equation ( [ eq : int ] ) , we assume that the gas disk has an exponential radial profile with a scale length twice that of the @xmath8-band disk , as found for the 20 dirrs and for the outer parts of 23 blue spirals by @xcite . the middle and right - hand panels in figure [ fig - hissfr ] suggest that the annuli around the knots and the outer disks would have @xmath77 larger star formation rates than the knots have now , given the ks relation for average rates extrapolated to the local average gas surface densities around the knots . figure [ fig - hissfr1 ] illustrates this point by plotting the two theoretical star formation rates for each knot location versus the knot s star formation rate . red dots are for the annulus and blue dots are for the outer disk . for the largest knots the two rates agree well , but for the smallest knots , the star formation rates fall short of the outer - disk expectations by a factor of @xmath78 . as mentioned above , this shortfall could be because smaller , fainter and more dispersed regions can not be seen in our data , or because the sfr is actually below the extrapolated ks relation at @xmath11 pc@xmath6 . we have identified on _ galex _ images the furthest out distinct @xmath0 knot in the little things sample of dirr and bcd galaxies that we are confident belong to the galaxy . these knots are found at distances from the center of the galaxies of 1 to 8 disk scale lengths and have ages of @xmath1 myrs and masses of 20 m_@xmath2 to @xmath3 m_@xmath2 . the presence of young clusters in the outer disks of dwarf galaxies shows that dirrs do have _ in situ _ star formation taking place in their outer disks . most regions are found around an h@xmath4i surface density of 1 m_@xmath2 pc@xmath6 , making this look like a threshold for the formation of prominent ob associations . the environments of these remote star - forming regions are extreme compared to the solar neighborhood and inner milky way ( [ environ ] ) , or even compared to the inner parts of dirr galaxies . the average pressures and midplane densities are lower by a factor of @xmath78 , the disks are relatively thick compared to the radial extents @xcite , and the metallicities are @xmath79 times lower . consequently , self - gravity is weak in the interstellar medium , cloud contraction times are long , @xmath80 myr or more , and molecules are sparse . moreover , the threshold column density for strong gravity inside clouds should be @xmath81 times lower than it is locally , placing this collapse threshold firmly in the optically thin regime where atomic gas should still dominate . presumably molecules form after the collapse begins @xcite . the star formation rates in the most remote @xmath0 knots fall short by a factor of @xmath79 of the average rates obtained from the kennicutt - schmidt relation for these types of regions . only for the most massive @xmath0 knots are the rates comparable . either we are missing other discrete star - forming regions owing to faintness or rapid dispersal , or the star formation rate drops below the ks relation for outer disks starting at around @xmath11 pc@xmath6 . in the latter case , stellar scattering from the inner disk would have to fill the outer disks with stars , which go out much further than the last @xmath0 knot seen here . we are grateful to the lowell observatory research fund for funding , including a summer internship for eg . we are also grateful to john and meg menke for a donation to lowell observatory that covered part of the page charges and to ibm for covering the remainder . dah also appreciates the considerable help on installing and running python to produce figure 1 from joe llama ( university of st . andrews ) , michael mommert ( nau ) , and david trilling ( nau ) . we appreciate helpful comments by the referee . ashley , t. , simpson , c. e. , & elmegreen , b. g. 2013 , , 146 , 42 ashley , t. , elmegreen , b.g . , johnson , m. , nidever , d.l . , simpson , c.e . , pokhrel , n.r . 2014 , , 148 , 130 bellazzini , m. , beccari , g. , fraternali , f. , et al . 2014 , a&a , 566 , 44 bigiel , f. , leroy , a. , walter , f. , brinks , e. , de blok , w. j. g. , madore , b. , & thornley , m. d. 2008 , aj , 136 , 2846 bigiel , f. , leroy , a. , walter , f. , blitz , l. , brinks , e. , de blok , w.j.g . , madore , b. 2010 , aj , 140 , 1194 billett , o. h. , hunter , d. a. , & elmegreen , b. g. 2002 , , 123 , 1454 burstein , d. , & heiles , c. 1984 , , 54 , 33 cannon , j. m. , oleary , e. m. , weisz , d. r. , et al . 2012 , , 747 , 122 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 carraro , g. , villanova , s. , demarque , p. , mcswain , m. v. , piotto , g. , & bedin , l. r. 2006 , , 643 , 1151 de blok , w. j. g. , & walter , f. 2000 , , 537 , l95 de blok , w. j. g. , walter , f. , brinks , e. , trachternach , c. , oh , s .- h . , & kennicutt , r. c. , jr . 2008 , , 136 , 2648 elmegreen , b.g . 2015 , , 814 , l30 elmegreen , b.g . , & struck , c. , 2013 , , 775 , l35 elmegreen , b. g. , & hunter , d. a. 2015 , , 805 , 145 elmegreen , d.m . , elmegreen , b.g . , snchez almeida , j. , muoz - tun , c. , mendez - abreu , j. , gallagher , j.s . rafelski , m. , filho , m. , & ceverino , d. 2016 , apj , submitted glover , s.c.o . & clark , p.c . 2012 , mnras , 421 , 9 herrmann , k. a. , hunter , d. a. , & elmegreen , b. g. 2013 , , 146 , 104 hodge , p. w. , & hitchcock , j. l. 1966 , pasp , 78 , 7 hunter , d. a. , & elmegreen , b. g. 2004 , , 128 , 2170 hunter , d. a. , & elmegreen , b. g. 2006 , , 162 , 49 hunter , d. a. , elmegreen , b. g. , dupuy , t. j. , & mortonson , m. 2003 , , 126 , 1836 hunter , d. a. , elmegreen , b. g. , & ludka , b. c. 2010 , , 139 , 447 hunter , d. a. , elmegreen , b. g. , oh , s .- h . , et al . 2011 , 142 , 121 hunter , d. a. , ficut - vicas , d. , ashley , t. , et al . 2012 , , 144 , 134 kennicutt , r. c. , jr . 1989 , , 344 , 685 kennicutt r. c. , & evans n. j. , 2012 , ara&a , 50 , 531 komiyama , y. , okamura , s. , yagi , m. , et al . 2003 , , 590 , l17 krumholz , m.r . 2012 , , 759 , 9 lada , c.j . , lombardi , m. , & alves , j.f . 2010 , , 724 , 687 landolt , a. u. 1992 , , 104 , 340 leitherer , c. , schaerer , d. , goldader , j. d. , et al . 1999 , , 123 , 3 martin , d.c . , fansom , j. , schiminovich , d. , et al . 2005 , , 619 , l1 martn - navarro , i. , trujillo , i. , knapen , j.h . , bakos , j. , fliri , j. 52 2014 , mnras , 441 , 2809 melena , n. w. , elmegreen , b. g. , hunter , d. a. , & zernow , l. 2009 , , 138 , 1203 minchev , i. , famaey , b. , quillen , a. c. , di matteo , p. , combes , f. , vlajic , m. , erwin , p. , bland - hawthorn , j. 2012 , a&a , 548 , a126 nidever , d. , ashley , t. , slater , c. t. , et al . 2013 , apjl , 779 , l15 oh , s .- h . , hunter , d. a. , brinks , e. , et al . 2015 , , 149 , 180 padoan , p. , haugblle , t. , nordlund , . , 2012 , , 759 , l27 ] pellerin , a. , meyer , m. , harris , j. , & calzetti , d. 2008 , aspc , 388 , 379 radburn - smith , d. j. , roskar , r. , debattista , v. p. , et al . 2012 , , 753 , 138 roskar , r. , debattista , v. p. , quinn , t. r. , stinson , g. s. , & wadsley , j. 2008 , , 684 , l79a rubio , m. , elmegreen , b.g . , hunter , d.a . , brinks , e. , corts , j.r . , & cigan , p. 2015 , nature , 525 , 218 saha , a. , olszewski , e. w. , brondel , b. snchez almeida , j. , muoz - tun , c. elmegreen , d. , elmegreen , b. , & mendez - abreu , j. 2013 , apj , 767 , 74 snchez almeida , j. morales - luis , a.b . , muoz - tun , c. , elmegreen , d.m . , elmegreen b.g . , & mendez - abreu , j. 2014a , apj , 783 , 45 snchez almeida , j. elmegreen , b.g . , munoz - tunon , c. , elmegreen d.m . , perez - montero , e. , amorin , r. , filho , m.e . , ascasibar , y. , papaderos , p. , & vilchez , j.m . 2015 , apjl , 810l , 15 salpeter , e. e. 1955 , , 121 , 161 toomre , a. 1964 , , 139 , 1217 van den bergh , s. 1988 , pasp , 100 , 344 walter , f. , brinks , e. , de blok , w. j. g. , et al . 2008 , , 136 , 2563 wang , j. , fu , j. , aumer , m. , kauffmann , g. , jzsa , g.i.g . , serra , p. , huang , m .- l . , brinchmann , j. , van der hulst , t. , bigiel , f. 2014 mnras , 441 , 2159 wyder , t. k. , et al . 2007 , , 173 , 293 zhang , h .- x . , hunter , d. a. , elmegreen , b. g. , gao , y. , & schruba , a. 2012 , , 143 , 47

we present an updated catalog of 1300 objects in the field of m31 , including 670 likely star clusters of various types , the rest being stars or background galaxies once thought to be clusters . the coordinates in the catalog are accurate to 0.2 , and are based on images from the local group survey ( lgs , * ? ? ? * ) or from the dss . archival hst images and the lgs were inspected to confirm cluster classifications where possible , but most of the classifications are based on spectra taken of objects with the hectospec fiber positioner and spectrograph on the 6.5 m mmt . the spectra and images of young clusters are analyzed in detail in this paper ; analysis of older clusters will appear in a later paper . ages and reddenings of 140 young clusters are derived by comparing the observed spectra with model spectra . seven of these clusters also have ages derived from hst color - magnitude diagrams ( two of which we present here ) ; these agree well with the spectroscopically determined ages . combining new v band photometry with the m / l values that correspond to the derived cluster ages , we derive masses for the young clusters , finding them to have masses as great as with a median of , and a median age of 0.25 gyr . in comparison therefore , milky way open clusters have the lowest median mass , the milky way and m31 globulars the highest , and the lmc young massive clusters and the m31 young clusters are in between . the young clusters in m31 show a range of structure . most have the low concentration typical of milky way open clusters , but there are a few which have high concentrations . we expect that most of these young clusters will be disrupted in the next gyr or so , however , some of the more massive and concentrated of the young clusters will likely survive for longer . the spatial distribution of the young clusters is well correlated with the star - forming regions as mapped out by mid - ir emission . a kinematic analysis likewise confirms the spatial association of the young clusters with the star forming young disk in m31 .

0809.5283

in the milky way , there is a clear separation between known open clusters ( which have diffuse structure , generally have low masses and ages , and belong to the disk ) and globular clusters ( which have a more concentrated structure , higher masses and ages , and in many cases belong to the halo ) . when the only well - studied globular cluster system was that of the milky way , it was generally thought that this separation was because globular clusters were fundamentally different from other star clusters , perhaps because of conditions in the early universe @xcite . however , it is possible to produce this apparent bimodality from clusters formed in a single process , with the same cluster initial mass function . in this picture , cluster disruption mechanisms , which are more effective at destroying low - mass clusters in particular because of two - body relaxation @xcite , would remove almost all of the low - mass older clusters . if all clusters were born with similar cluster mass functions , then we would expect to see the occasional high - mass young cluster . in fact , we do see these in other galaxies . the `` populous blue clusters '' of the lmc @xcite have been suggested as examples of young objects which will evolve into globular clusters . m33 also has a few likely massive young clusters @xcite , and such clusters have been found in a number of normal isolated spirals @xcite . it is possible that the seeming absence of such objects in the milky way is merely an observational selection effect ; recently , there have been discoveries of heavily reddened open clusters such as westerlund 1 , which likely has mass in excess of @xmath3m@xmath4 @xcite . what of m31 s clusters ? while its clusters have been studied since the 1960s ( eg * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and it was noted even then that some of these clusters had colors indicating young populations , their nature is still not entirely clear . @xcite called them open clusters , while @xcite adopted the simple convention of calling any cluster projected on m31 s disk a `` disk cluster '' until proved otherwise by kinematics . this avoids the question of whether young clusters are fundamentally different from globular clusters in structure , formation , etc . our detailed study of m31 young clusters , incorporating kinematics , should cast some more light on these questions . it is only recently that detailed constraints on the mass , kinematics , age and structure of cluster populations in m31 have been obtained , particularly for clusters projected on the inner disk and bulge . multi - fiber spectroscopy and hst imaging have played an important role here . a number of m31 globular cluster catalogs have been created over the years , giving a very heterogeneous result , with significant contamination by both background galaxies , foreground objects and even non - clusters in m31 itself . while the work of @xcite , @xcite , @xcite , @xcite , and @xcite has gone a long way towards cleaning up the catalogs and winnowing out the non - clusters , still more work is needed for both young and old clusters . here we present a new catalog of m31 star clusters which were originally classified as globular clusters , all with updated high - quality coordinates . we have observed a large number of these clusters with the mmt and the hectospec multi - fiber system . in this paper we study more than 100 young m31 clusters in detail . in subsequent papers we will address the kinematics , ages and metallicities of the older clusters . the m31 young clusters have been studied both by authors aiming to study its open clusters ( eg * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and also by others who have aimed to study its globular clusters . for example , @xcite pointed out the existence of young clusters in m31 , estimated masses between @xmath5 and @xmath6 , and drew attention to their similarity to the populous blue clusters in the lmc . @xcite noted the existence of 8 clusters with strong balmer lines in their spectra , which they tentatively classified as young globular clusters . @xcite ( confirmed with a later sample by * ? ? ? * ) added more clusters , and commented that hst imaging ( then unavailable ) was needed to distinguish between structure typical of open and globular clusters . @xcite added more new young clusters , bringing the total to 19 , and @xcite increased the sample to 67 . in general , authors have associated these young clusters with m31 s disk , although @xcite invoke an accretion of an lmc - sized galaxy by m31 . observations are particularly challenging for clusters projected on m31 s disk : many of the early velocities had large errors , and there were issues with background subtraction . here we discuss high - quality spectroscopic measurements of kinematics and age for the young clusters , supplemented with hst imaging to delineate the structural , spatial and kinematical properties of these young clusters . we find that while they are almost all kinematically associated with m31 s young disk , and their age distribution will allow us to test suggestions that m32 has had a recent passage through m31 s disk @xcite . the young clusters have structure and masses which range all the way from the low mass milky way open clusters to higher mass , more concentrated globular clusters , although they are dominated by the lower concentration clusters . we will also discuss the likelihood that these clusters will survive . the starting point for our cluster catalog was the revised bologna catalog ( rbc ) @xcite , itself a compilation of many previous catalogs . coordinates from this catalog were used to inspect images from the local group survey ( lgs ) images @xcite , which cover a region out to 18 kpc radius on the major axis and 2.8 kpc away from the major axis , hst archival wfpc2 and acs images , and the dss for the outermost clusters . the lgs images have median seeing @xmath7 1 . we also added in some new clusters , found visually by ourselves on the lgs images and on hst images ( discussed below ) . even a casual inspection of the lgs images , particularly the i band images , reveals the presence of a vast number of uncataloged faint disk clusters , presumably similar to the galactic open clusters ( * ? ? ? * estimate 10000 such clusters from hst images ) . we have elected not to take on the enormous task of cataloging and measuring those clusters in this paper ; rather we choose to deal with the more massive clusters for which some information , however fragmentary , already exists . at a later stage in the project , the catalogs of @xcite , extracted from kpno 0.9 m images , and @xcite , from hst images became available , from which we collected objects not already in our own catalog and subjected them to the same editing as we now describe . the archival images were thus used to answer the following questions from the catalog we created from the rbc and our own additions . first , did the catalog coordinates correspond to a unique object ? in cases where the identification of the cluster on the local group survey was ambiguous or unclear , we consulted the original papers and finding charts where these were available @xcite . in some cases , there is no clear object that can be associated with the published coordinates , or the nearest object in fact already had a different i d . the large number of hectospec fibers meant that we were able to verify classifications of many low - probability candidates . the cataloging and observation parts of this program occurred in a feedback fashion , allowing some target names and/or coordinates to be changed for the succeeding observations . as a result , we had some objects whose identifications were incorrect ; to these we add an `` x '' to their original name in our tables below . second , were the existing coordinates accurate ? in general , the answer was no . our final catalog contains 1200 objects from the rbc , without considering the newer candidates of @xcite and @xcite . 830 of those required coordinate corrections larger than 0.5to place them on the fk5 system used in the dss and the lgs images . the median error in coordinates is 0.8 , with the largest error being of order 10 ; at which point the identification of the actual object becomes uncertain . similarly , there are 379 objects in the @xcite catalog found within 2of an lgs object . for these , the median error is 0.8offset , where the largest error is 1.9 . many of the discrepant velocities between us and the rbc or @xcite tabulations reported here are likely due to inaccurate coordinates used in previous spectroscopic work . the coordinates newly derived from the lgs images are accurate to 0.2 rms . third , were the targets really clusters ? the lgs v and i band images , and wfpc2 or acs images taken with non - uv filters were used to confirm the cluster nature of the objects , by visual inspection as well as the automated image classifier contained in the sextractor code . a number of cluster candidates were stellar on the lgs images ; all of these were later confirmed as stars in our spectra , from either the spectral characteristics or the velocities , in regions of m31 where there is no confusion between the local m31 velocity field and the velocity distribution of foreground galactic stars . we found that about 90 of the @xcite candidates listed as new , probable and possible ( indicating that they appeared non - stellar in their kpno 0.9 m images ) appeared stellar on the lgs or hst images . objects for which we have no new classification data are kept in the catalog , but noted as still in question in our table . the large majority of the misclassified objects are stars ( foreground galactic or m31 members ) ; more than 130 objects considered to be clusters as recently as 2007 by @xcite are in fact stars . quite a number of cataloged objects are background galaxies , and a few are either unidentifiable , or are accidental clumpings of galactic or m31 member stars . @xcite have recently stated that four clusters that were classified as disk clusters by @xcite are `` asterisms '' . they note that in their adaptive optics ( ao ) images there was no cluster visible generally there were only a few bright stars . however , for young clusters , red supergiants would dominate the light at infrared wavelengths and the hotter mainsequence stars would appear much fainter . we would need to use multi - wavelength data to classify these objects correctly . we show below that the optical spectra of those four clusters are indeed consistent with clusters of massive , main - sequence stars , and although the magnitudes , and hence masses , of these few objects were certainly overestimated , the objects will still be considered as clusters in our catalog , at least until hst images show otherwise . figure [ cohen ] shows this complication for one cluster , by comparing the high resolution , but long wavelength ao image with the lgs i band image . a star cluster is clearly visible in the i band , and even more prominent at bluer wavelengths - the object is indeed a young star cluster , though certainly not a globular and not very massive . our own hst images do reveal two cataloged clusters as asterisms : these are comprised of a small number of ob stars and late supergiants , resulting in a distinctive integrated spectrum with strong balmer and he i lines in the blue , and tio bands in the red . even if these two are real clusters , the derived masses are small enough to exclude them from a list of massive clusters . some cataloged objects have no real object even within a generous radius . in a few cases , a nearby background galaxy had also led to confusion in previous papers ( though not in the most recent version of the rbc ) , whereby an actual cluster was labeled as background . thus , while for the most part we have removed objects from the list of clusters , we have also restored a few objects to the cluster list . table [ main ] lists all objects believed to be clusters . for object names , we use the naming convention of @xcite where possible , where the name consists of a prefix with the rbc number followed by the number of the object from the next most significant catalog . objects for which we have no new information other than improved coordinates , and which have not been convincingly shown to be clusters by previous workers , are italicized . objects in the rbc which we did not observe and for which our coordinates are within 0.5of the rbc coordinates are not listed here , nor are the @xcite or @xcite cluster candidates that we did not observe . some objects that we did not observe could of course still be background even though they have non - stellar profiles , but these , few in number , are still listed here . a rough classification based on the spectra is included in this table , for objects with good quality spectra . `` young '' clusters are those with ages less than 1 gyr , `` interm '' refers to those with ages between 1 and 2 gyr , and `` old '' refers to clusters older than that . a subsequent paper will provide a detailed analysis and evidence that few if any of these `` old '' clusters have ages less than 10 gyr . `` hii '' indicates the spectrum is emission - line dominated . `` na '' appears for objects known to be clusters from an hst image , but for which we have no spectrum , or cases where the spectrum is too poor to determine the age , even in a coarse manner . the v magnitudes come from this paper , using the aperture size listed ( see [ s : phot ] ) or otherwise as indicated . column c describes what information was used to classify the target as a cluster . the possibilities are `` s '' , where our spectrum clearly indicates a star cluster , `` l '' , where the lgs image is non - stellar , and/or `` h '' , where an hst image indicates a star cluster . table [ young ] gives a list of those clusters that have ages less than 2 gyr . ( sections [ s : acs ] , [ s : ages ] and [ s : masses ] of this paper will discuss the measurement of ages and masses for these clusters . ) table [ stars ] lists objects from previous cluster catalogs that are in fact stars . many of these had also been classified as stars by previous workers . asterisms are also listed here . table [ maybestars ] gives a list of possible stars . in these cases , the local group survey imaging indicates a stellar profile , but either we have no spectrum , or the spectrum is ambiguous . note that some of the stars in tables [ stars ] and [ maybestars ] are certainly members of m31 . objects thought to be clusters in the @xcite catalog but which have stellar profiles in the lgs images are listed in [ maybestars ] , with coordinates derived from the lgs . table [ gals ] lists background galaxies . table [ nobodyhome ] lists cataloged objects where there was no obvious object within a reasonable distance of the previously published coordinates , which are listed here again . as a commentary on the difficulty experienced by all of those who have endeavored to collect true m31 star clusters ( including us ) , here is a brief summary of the contents of the most recent version of the rbc , excluding the additions of @xcite and @xcite , but including the lists compiled by other astronomers starting with edwin hubble . the rbc , restricted as just mentioned , contains 1170 entries . we here , and others ( particularly * ? ? ? * ) , have provided classifications for 991 of these . only 620 entries are actually star clusters , while 20 more could be considered clusters though the large amount of ionized gas present indicates the clusters may still be forming . 270 entries are stars , mostly foreground , and 224 entries are background galaxies or agn . at least 2 objects are chance arrangements of luminous m31 stars , together which appear as clusters from the ground . in the @xcite catalogs , there 113 , 258 and 234 `` new '' , `` probable '' and `` possible '' clusters , respectively . the lgs survey contains images of 94 , 152 and 105 members of those catalogs , respectively . of those subsets , 79 , 106 , and 129 , respectively have non - stellar profiles in the lgs , some of which were observed with hectospec . the hectospec multi - fiber positioner and spectrograph is ideally suited for this project , in that its usable field is 1 degree in diameter , and the instrument itself is mounted on the 6.5 m mmt telescope . we obtained data in observing runs in the years 2004 to 2007 , and now have high - quality spectral observations of over 400 confirmed clusters in m31 , and lower quality spectra of 50 more . we used the 270 gpm grating ( except for a small number of objects taken with a 600 gpm grating ) , which gave spectral coverage from 36509200 and a resolution of @xmath75 . the normal operating procedure with hectospec , and other multi - fiber spectrographs , is to assign a number of fibers to blank sky areas in the focal plane , and then combine those in some fashion to allow sky subtraction of the target spectra . for instance , the 4 - 5 fibers nearest on the sky to the target may be combined . these methods are satisfactory for our outer m31 fields , but not for the central areas where the local background is high relative to the cluster targets . for those fields , we alternated exposures on - target and off - target to allow local background subtraction to be performed . many of the discrepancies between our bulge cluster velocities and those of previous workers might be explained by their lack of proper background subtraction , and/or inaccurate target coordinates . we obtained exposures for 25 fields , with total exposure times varying between 1800s and 4800s . the signal / noise ratios for the 500 objects we classified as clusters have a median of 60 at 5200 and 30 at 4000 , with more than 100 clusters having a ratio at 5200 better than 100 . figure [ hectospec ] shows the locations and types of objects observed in all of these fields . the multifiber spectra were reduced in a uniform manner . for each field , or configuration , the separate exposures were debiased and flat fielded , and then compared before extraction to allow identification and elimination of cosmic rays through interpolation . spectra were then extracted , combined and wavelength calibrated . each fiber has a distinct wavelength dependence in throughput , which can be estimated using exposures of a continuum source or the twilight sky . the object spectra were thus corrected for this dependence next , followed by a correction to put all the spectra on the same exposure level . the latter correction was estimated by the strength of several night sky emission lines . sky subtraction was performed , using object - free spectra as near as possible to each target . for the targets where local m31 background was high , the method was reversed , such that only sky spectra far from the disk of m31 were used . an offset exposure for such fields , taken concurrently , was reduced in a similar way ( thus contemporaneous sky subtraction was performed for on- and off - target exposures ) , and then the off - target spectra were subtracted from the on - target . this process increased the resultant noise of course , but we deemed it essential for targets in the bulge and disk of m31 . the off - target exposures have the additional advantage of giving measurements of the unresolved light in over 800 locations over the entire disk of m31 . velocities were measured using the sao xcsao software . given the wide variety of spectra in this study , it was deemed necessary to develop new velocity templates , from the spectra themselves . the procedure was to derive an initial velocity of all spectra using library templates ( typically a k giant star ) . the spectra were shifted to zero velocity , and sorted into three different spectral types , a , f and g type spectra . the best spectra in each group were combined to make new templates , and the procedure was repeated now using the new templates . by using these templates we have assured that all the m31 targets are on the same velocity system . they are tied to an external velocity in the initial step , whose accuracy depended on the accuracy of the initial set of templates used . a good test of the internal accuracy was provided by repeat measurements of clusters . we have 386 repeat measurements ( on different nights ) for 224 clusters . the median difference in velocity for these repeats was 0.5 , with an implied median single measurement error of 11 ( smaller than our formal errors listed in the tables ) . we will present external comparisons in a subsequent paper , but note that the cluster velocities agree very well with the hi rotation curve ( see [ s : kin ] ) . velocities for the young clusters , stars and galaxies are presented in tables [ young ] , [ stars ] , [ maybestars ] , and [ gals ] . velocities for the old clusters will presented in a subsequent paper . the spectra were corrected to relative flux values , using observations of standard stars ( the mmt f/5 optics system employs an atmospheric dispersion compensator , adc ) . the flux correction has been determined to be very stable over several years . thus , observations of the same targets taken in different seasons can be combined where available . @xcite highlighted the heterogeneous quality of the m31 cluster catalogs when they claimed , using ao techniques at k , that four out of six observed young clusters were in fact asterisms . figure [ blue_spectra ] shows , from top to bottom , spectra of the four disputed clusters , the average of three genuine young clusters verified by acs images , and the average spectra of single supergiant stars ( these were verified to be stars from the lgs images , and members of m31 from their velocities ) . if the disputed clusters were in fact merely a few stars , those stars would have to be supergiant stars , whose absorption line widths would be as narrow as seen in our blue supergiant spectra . this is not the case for the four disputed clusters ( note in particular the h@xmath8 and h@xmath9 widths , narrow in the stars and wide in the other spectra ) , and we conclude that those objects are true clusters and not asterisms . to be sure , these particular clusters are not globular clusters either , and , additionally , are perhaps not massive enough to be considered young , populous clusters . high spatial resolution imaging can both check for asterisms and also explore the clusters spatial structure : is their concentration low , like typical milky way open clusters , or high , like globular clusters ? there are acs or wfpc2 images available for 25 of the clusters with ages less than 2 gyr . two of these show no evidence of an underlying cluster , but the remaining 23 are clearly not asterisms . figure [ acs_images ] shows the range of structure seen in these young clusters . while many of them show the low - concentration structure typical of milky way open clusters , a number of them , such as b374-g306 and b018-g071 , are quite centrally concentrated , like the majority of the milky way globular clusters . @xcite have measured the structure of m31 clusters with available hst imaging at the time of publication . there are 70 clusters in their sample which we have classified as old , and 7 of our young clusters . it is straightforward to compare the structure of the clusters they study with the milky way globulars , because they use the same technique as @xcite , who have produced a careful summary of the structure of the milky way globular clusters . however , it should be noted that their fitting technique ( fitting ellipses to cluster isophotes ) is not well - suited to very low - concentration clusters , and in fact one of our young clusters , b081d , is omitted from their analysis because of its low density . figure [ structure ] compares their results for old clusters from our sample with the structure of milky way globulars ( from the work of * ? ? ? it can be seen that the concentration parameter = log(@xmath10 ) for king model fits , @xcite p. 307 ] for the old clusters in m31 has a similar distribution to the milky way globulars . we note that although there are no old clusters in our sample with concentration greater than 2.2 , such clusters are definitely present in the @xcite sample so this absence is unlikely to be significant . the similarity in structure is interesting , because at first sight it would seem that the m31 clusters with high concentration would be preferentially discovered in surveys . perhaps the m31 globular cluster surveys ( which , as we have seen above , include a large number of non - globular clusters , as well as the low - concentration young clusters ) are now sufficiently thorough that they are not strongly affected by this bias . although only seven of our young m31 clusters were analyzed by @xcite , it can be seen from figure [ structure ] ( where they are marked by asterisks and five - pointed stars in the top panel ) that their concentrations cover the whole range of the older clusters in m31 and in the milky way . ( the five - pointed star represents hubble v from ngc 205 . ) our observational selection biases may over - emphasize high - concentration clusters , but it is still interesting to see that three of the young clusters have quite high concentration parameters . how does their structure compare with the milky way open clusters ? it is quite difficult to answer this question because the available samples of milky way open clusters are severely incomplete , and it is a challenging task to fit king models to the known open clusters , because cluster membership is hard to determine . the 2mass database @xcite has been used by @xcite to measure the structure of 21 open clusters . they used a cmd - fitting technique to remove contamination from disk field stars . we have also used data from @xcite , who used radial velocity to decide membership . because of the high background in all these cases , it is possible that the limiting radius " given by the authors is in fact smaller than the tidal radius , in which case the cluster concentrations would be smaller than those plotted . it can be seen in the middle panel of figure [ structure ] that all these open clusters have quite low concentrations . however , the sample of clusters with concentration measurements is quite small , and it is quite possible that there are a few open clusters in the milky way which are yet to be discovered and have high concentrations , like the two m31 young clusters . in summary , the young clusters in m31 show a range of structure . most have the low concentration typical of milky way open clusters , but there are a few which have high concentrations , like most milky way globulars . we note that any survey of m31 clusters will preferentially discover ones with high concentrations . in addition , the incompleteness of milky way open cluster samples and the difficulty of measuring cluster concentration in crowded fields means that we can not rule out the existence of such clusters in the milky way . in this section , we describe methods for determining ages from the spectra and color magnitude diagrams for the verified clusters . since the emphasis in this paper is on the younger clusters , and more specifically , on their m / l ratios , our task is first to distinguish young from old clusters , and then to obtain accurate age measurements among the younger clusters . a more refined age determination ( for clusters older than 2 gyr ) is postponed for a later paper . the methods for obtaining ages for young stellar populations from their integrated spectra are similar to those used for older stellar populations , except that instead of employing empirical stellar libraries ( e.g. , * ? ? ? * ) , modelers focussing on younger stellar populations have used synthetic spectra , partly due to a lack of empirical spectra of young stars over a range of metallicities . here we have made use of the starburst99 stellar population modeling program ( sb99 , * ? ? ? to distinguish young from old clusters , we compared our cluster spectra with two external sets of spectra , which served as population templates . one set was the sample of 41 mw globular spectra obtained by @xcite , covering the abundance range of @xmath11[fe / h]@xmath12 . these spectra have a wavelength coverage from 3500 to 6300 , a dispersion of 1 / pixel and a resolution of about 4.5 , and served as our old population templates . our young population templates were created from the sb99 program using the padova z=0.05 interior models and z=0.04 stellar atmospheres , since it seemed likely the young clusters have supersolar abundances . using a solar abundance set of isochrones in the models has the net effect of increasing the derived ages for clusters older than 0.1 gyr , in a logarithmic scaling such that a 1 gyr supersolar model and a 2.5 gyr solar abundance model have nearly the same resultant spectra . a grid of 40 high resolution spectra was created for ages from zero to 2 gyr , with a logarithmic time step of 0.07 between models . our approach then was to simplify the old populations by only using a set of mw globular cluster templates , and to simplify the young populations by restricting ourselves to a single metal - rich chemical composition . the evolution of integrated spectra is essentially logarithmic with time ; for example , the difference between a 5 and 12 gyr population is relatively small compared to the large changes which occur over the first 1 gyr . thus our simplification seems warranted , particularly since in this paper we are concerned only with identifying the younger clusters and then studying them in detail . the major issue , as seen below , in separating old from young clusters , is the potential degeneracy between very metal - poor old populations and young populations , both of which are dominated by balmer lines in their integrated spectra . both sets of template spectra were rebinned to the same resolution and dispersion as the m31 set , and then each template spectrum and m31 spectrum had a low order fit subtracted . the significance of this step was that we did not use the continuum shape to help determine the best matching template . a scaling factor was then determined between each template and all the available m31 cluster spectra , and the reduced @xmath13 calculated over the spectral ranges of 3750 - 4500 , 4750 - 5000 , and 5080 - 5360 . these ranges were chosen to exclude spectral regions where there are few lines , as well as where the mw cluster spectra have no data due to bad columns in the ccd used for that set . specifically included were the balmer lines from h@xmath8 to h@xmath14 , the mg b lines , ca ii h&k , and the he i lines at 4009 and 4026 , the last being prominent in ob stars and thus strong in the youngest clusters . the noise used in the reduced @xmath13 calculations was that calculated from the spectra themselves . a few logic decisions had to be made in analyzing the resultant list of reduced @xmath13 values . for most cases , the lowest reduced @xmath13 occurred clearly either in the sb99 set or the mw cluster set , thus allowing a particular cluster to be identified as young or old . for the young clusters , the age chosen was the average of ages where the reduced @xmath13 was within 10% of the lowest value . about 10% of the clusters were equally well fit by a mw cluster , typically a metal - poor one with [ fe / h ] @xmath15 , and a metal - rich sb99 model . nearly all of these are low s / n , and thus the poor fits were not surprising . visual inspection of the spectrum clarified 11 of these as being very young ( with very blue continuum shapes ) , and the remaining 48 we grouped as old . figure [ young_spectra ] compares the data and best - fitting models for a range of determined ages . we estimate the errors in determined ages for the young clusters to be about a factor of two , which leads to m / l uncertainties of 50% . the model m / l values for each chosen age then allowed masses to be estimated from the cluster integrated v band photometry , which is described in [ s : phot ] . to allow a comparison of young clusters to be made , we also calculated spectroscopic m / l values for the old m31 clusters from the models in @xcite , by obtaining estimates of [ fe / h ] for each cluster via fe and mg line indices and assuming an age of 12 gyr for each . the detailed analysis of the old cluster spectra will be presented in future paper . many readers may be more familiar with extracting ages from diagrams that plot a largely age - sensitive index versus a largely metallicity - sensitive index . to help demonstrate the efficacy of the @xmath13 approach we show two such diagrams . figure [ indices ] is a plot of a balmer line index versus a metal line index . we have defined m@xmath16=(fe5270+mgb)/2 and h@xmath16=(h@xmath17 h@xmath18h@xmath8)/3 , all lick indices @xcite . these indices are equivalent widths : units are . for clarity in these diagrams , a signal - to - noise ratio cutoff was made , which eliminated 20% of the clusters from being plotted . different symbols represent mw globular clusters , and four age bins for m31 clusters , which were determined by the @xmath13 method : very young ( @xmath19 0.1 gyr ) , young ( 0.1 @xmath19 age @xmath19 1 gyr ) , intermediate ( 1 @xmath19 age @xmath19 2 gyr ) , and old . clearly , there is a sequence of m31 clusters that closely matches the sequence of mw globular clusters , and a large spray of clusters that fall mostly in regions of higher balmer equivalent width . clusters with younger ages will of course have stronger balmer absorption , until the very youngest ages are reached , at which point the balmer strength declines again . the second diagram ( figure [ indices2 ] ) uses indices defined in @xcite , namely the ratio of residual light in the h@xmath9 line to the nearby fe4045 line , and the ratio of the line at 3969 which contains both h@xmath20 and caii h , to the caii k line . the indices are unitless . this diagram also shows the old cluster metallicity sequence and again distinguishes them from the young clusters , again excepting for the extremely young clusters . the shortcoming of using such diagrams is that occasionally , due to the nature of real data , one index will be bad , causing the cluster to look young or old in one diagram , and the opposite in another diagram . we are thus more confident in a fitting procedure that uses many diagnostic lines , but are gratified that in the vast majority of cases , these two diagrams verify the ages assigned by the @xmath13 method . @xcite estimated ages of many young disk clusters in m31 from wfpc2 color - magnitude diagrams ( cmd ) and isochrone fitting to the main sequence or to luminous evolved stars . four of their clusters are bright enough to be in our spectroscopic study . their ages agree quite well with ours ( table [ ages ] ) . additionally , as part of hst go proposal 10407 , we obtained acs images of several young clusters , three of which we report on here . the multidrizzle package @xcite was used to combine the 3 individual exposures taken in the f435w and f606w filters ( corresponding to b and v , respectively ) . stellar photometry was obtained using the daophot package of @xcite , modeling the spatially variable psfs for each of the combined images separately , using only stars on those images . psfs were constructed using 5 - 10 bright stars which had no pixels above a level of 20,000 counts , the point at which an ostensible non - linearity set in and the psf no longer matched those of fainter stars . aperture corrections were also measured using these stars , to determine any photometric offset between the psf photometry and the aperture magnitude within 0.5 . @xcite have provided aperture corrections from that aperture size to infinity , in all acs filters . generally , two passes of photometry were run . first , a star list was made and entered into allstar , which aside of the photometry , produces a star - subtracted image . stars missed in the first round were located in the subtracted image and added to the original list . the original frame was then measured again by allstar . the photometry was then placed on the standard johnson / kron - cousins vi system using the aperture corrections and synthetic transformations provided in @xcite . to lessen the severe problems with crowding in these clusters , only stars that fall in an annulus with radii of 15 and 50 pixels ( 0.75 and 2.5 ) are shown in the color - magnitude diagrams ( figure [ cmd ] ) . the background field shown has the same area as the cluster fields , and refers to an annulus around b049-g112 with inner radius of 60 pixels . isochrones with super solar abundances from the padova group @xcite have been placed in the diagrams to allow age determination , using a distance modulus of m31 of 24.43 and the reddenings determined above for these two clusters ( 0.25 for both ) . these cmds and that of a third we have worked on ( b367-g292 ) give ages in reasonable agreement with those from the spectroscopic analysis ( table [ ages ] ) . multicolor photometry for most of the clusters in this project is already collated in @xcite , but enough clusters are missing to warrant remeasuring all the clusters , to allow the photometry to be used with the spectroscopic m / l values to obtain masses . the lgs survey of m31 consisted of 10 separate but overlapping fields . stellar photometry from these fields using psf - fitting has been reported in @xcite , but the aperture photometry needed for resolved star clusters has not yet been reported . to limit the scope of the work , we elected to measure objects in our catalogs only in the v band . targets from our entire catalog were located on the images , and photometry for 12 separate apertures ranging from 0.7 to 16(spaced logarithmically ) was collected using daophot . growth curves from these apertures were constructed , and used in an automatic fashion to estimate the aperture which enclosed the total light of the cluster . these apertures were then inspected on the images and increased , if the clusters were in fact larger , or decreased for cases where the apertures included substantial light from objects clearly not part of the clusters . the local background was measured in annuli with inner radii 1 pixel larger than the outer radius of the aperture for the object . the apertures used are listed in table [ main ] . extraneous stars remaining in the apertures were accounted for by measuring their magnitudes separately , and subtracting their contribution to the cluster aperture magnitudes . the resultant instrumental magnitudes were then placed on the standard v system using stars from the @xcite tables which we were measured in the same way as the clusters . the color term in the v mag transformation ( described in * ? ? ? * ) was ignored , as it was smaller than the errors we report . tables [ main ] , [ stars ] , and [ maybestars ] list the results from this work . the formal errors in the standardized photometry were less than 0.03 mag , set by the uncertainties in the transformations . however , since our goal was actually cluster total magnitudes , our calculated uncertainties refer to the uncertainty in setting the proper apertures and correcting for extraneous objects within those apertures . specifically , the uncertainties were set to be equal to the difference in the magnitude of the aperture chosen as best representing the limiting radius of the cluster , and that of the next larger aperture in our logarithmic spacing of aperture sizes . in practice this means that clusters in crowded fields have larger uncertainties than those in less dense areas . the tables also list photometry from other sources for the objects that are outside of the lgs images ; we do not list the errors in such cases . comparing our aperture magnitudes with those collected from various sources and listed in the rbc , we find excellent agreement over the magnitude range of 14 to 18 , with an rms in the differences of 0.18 mag in the set of 200 objects in common ; this in spite of the fact that no effort was made to insure that the apertures used in the two data sets were the same . between v=18 and 20 , our photometry tended to be fainter by 0.2 mag and the scatter increased to 0.5 mag , some of which was likely due to differences in object identification . a further comparison was made with the 58 v magnitudes measured for m31 clusters from archival hst images in @xcite . aside of one cluster whose v magnitude appears to be a typo ( b151-g205 ) , the rms of the differences of that set with the magnitudes presented here is 0.28 , with no apparent systematics . @xcite misclassified 15 of the young clusters as old disk globular clusters ( 17% of their disk globulars ) . this was pointed out by @xcite . in some cases this was due to the low s / n of the wyffos spectrum , in others the problem was misinterpretation of the spectra . our new study of the clusters has resolved most of this confusion and changed the classification of a number of clusters from old and massive to younger , not very massive . however , we still find clusters with significant masses , above @xmath21m@xmath22 , and with ages less than 150 myr ( see table [ young ] ) . of the 10 clusters with those physical characteristics , the hst or lgs images of three confirm them as clusters similar in appearance to the populous clusters of the lmc ( these are b315-g038 , b318-g042 , and vdb0 , the latter still the most massive , young cluster known in m31 , @xcite ) . five appear more like ob associations , and thus may not survive as bound clusters ( b319-g044 , b327-g053 , b442-d033 , b106d and bh05 ) . the case for the other two ( b040-g102 , b043-g106 ) is not as clear , but their lgs images are more similar to the cases like b315-g038 than to the ob associations . figure [ mips ] shows a spitzer / mips 24@xmath23 mosaic of m31 @xcite with the positions of the young clusters overlaid . clusters younger than 0.1 gyr , between 0.1 and 0.32 gyr , and 0.32 and 2 gyr are shown in different colors . the latter two groupings divide the clusters older than 0.1 gyr into two equal parts . it can be seen that the spatial distribution of the young clusters is well correlated with the star - forming regions in m31 , with the majority associated with the 10 kpc `` ring of fire '' . the comparison of these young clusters and the warm dust emission is distinct from the comparison of the latter with the location of hii regions , as we have excluded the clusters embedded in hii regions from our sample . @xcite and @xcite use the curious appearance of the mid - ir ring - the split near the location of m32 , creating the appearance of a `` hole '' in between , and the possible offset of the ring from the nucleus - to suggest a recent encounter of m32 and m31 . both groups suggest that the split is caused by m32 s passage through the disk , and their models also produce rings offset from m31 s center , albeit not as extreme as is observed ( * ? ? ? * produce an offset of 1@xmath24 , not 6@xmath24 ) . an examination of our figure [ mips ] shows that the cluster distribution favors the outer parts of the hole , and is generally quite symmetric about m31 s nucleus . most models of an offset ring assume that the inner part of the split in the observed ring is the one which should be traced by star formation . however , this is not where most of the clusters are found . ] @xcite need a very recent interaction between m32 and the disk ( their model has the disk passage occurring @xmath25 years ago ) because the passage of m32 through the disk in their model results in a burst of star formation that propagates outward through the disk . however , we do not see any radial trends with cluster age , which might be expected with a propagating ring of star formation ( figure [ radial_age ] ) . @xcite prefer a collision about @xmath26 years ago , which triggers expanding density waves . our young cluster ages range from 0.04 gyr to 1 gyr , but most are between 10@xmath27 and 10@xmath28 years . if the ring of fire " was produced by a single event , as modeled by the above authors , we might expect the age distribution of clusters associated with it to be more peaked . the ages of the younger clusters presented in @xcite range from around @xmath29 to @xmath26 years , and there is little evidence of a peak in star formation in this age range either . these results seem to suggest that star formation has been fairly high in this region of m31 for 1 gyr or more . in summary , we see no evidence of enhancement in star formation rate or any spatial age separation , as we might expect from the m32 disk passage . do the kinematics of the young clusters bear out the disk origin suggested by their close association ( in projection ) with star forming regions in m31 s disk ? m31 s inner disk kinematics are more complex than originally supposed , due to m31 s bar @xcite . the velocities from our many sky fibers , taken both as a part of regular observing and also from entire exposures devoted to offset sky in the crowded inner regions , give us a new way of quantifying disk rotation throughout the inner regions where the young clusters are found . ( we plan to use these data in a study of bar kinematics , athanassoula et al . in preparation ) . figure [ spider_diag ] shows the disk mean velocity field . we also use kinematics of hii regions that we observed as fillers in the hectospec fields to give us an indication of the kinematics of young disk objects . these data will be published in athanassoula et al . ( in preparation ) . figure [ dist_vel ] shows the kinematics of the mean disk light , the young clusters , and hii regions , versus major axis distance x , in kpc . we have split the sample into objects which are close to the major axis ( @xmath19 1 kpc ) and those projected from @xmath30 kpc from the major axis , because the projection of a circular orbit looks different in these two cases . objects on the major axis in circular orbits have all of their velocity projected on the line of sight ; as we move further away from the major axis , less of the circular velocity is projected on the line of sight and so the tilt of the distance - velocity line becomes smaller . this can be seen clearly in figure [ dist_vel ] for all types of objects . the curious flattening of the mean velocities from absorption line spectra for major axis distances between 3 and 10 kpc is likely to be caused by the bar . it can be seen that the young clusters follow the disk mean velocity curve from absorption spectra quite well , and show an even better correlation with the kinematics of the youngest objects : hii regions and sky fibers showing emission spectra . this kinematic analysis confirms our spatial association of the young clusters with the star forming young disk in m31 . the m / l values obtained from the spectroscopic age estimates can be combined with the v band photometry to derive masses of all the observed m31 clusters , young and old . reddening values are of course also needed , and a large number of e(b v ) values were derived from photometry in @xcite . they and we consider only the total reddenings , foreground and internal to m31 . the methodology used in @xcite meant that the reddenings would only be valid for old clusters , and indeed few of the clusters we have identified here as young were included in their study , thus reddenings for those objects are needed . therefore , we elected to rederive the reddenings for all of the clusters in our study , young and old . in the case of the young clusters we compared the fluxed spectra with the sb99 model spectra of the appropriate age . as described above , the ages were obtained by matching spectral line features in the observed and model spectra , not by comparing the continua shapes . once the ages have been found in this way , differences in the continua shapes may be assumed to be due to reddening , except for a few cases where a late - type star , whether member or not , clearly dominates the redder wavelengths as evidenced by the presence of tio bands . for those cases , we use the mean reddening for the young clusters of 0.28 . for the old clusters we did not use models , but rather the sample of spectra themselves . initial values of reddenings were obtained from @xcite , and were used to deredden the spectra of those clusters with e(b@xmath31v)@xmath32 , about 190 in number ( there are about 350 old clusters in our spectroscopic sample ) . these spectra were ordered in metallicity , which was estimated from the spectral line indices as mentioned above , rebinned to a coarse grid in wavelength , and normalized to have the same intensity at the arbitrary wavelength of 5000 . interpolation formulae were developed from these spectra , via a least squares method to avoid bad spectra , for intensity as a function of both wavelength and metallicity . as a result , a cluster spectrum of arbitrary metallicity could be created , dereddened to the accuracy of the @xcite reddenings . the individual spectra in this low reddening sample were then compared with the appropriate interpolated spectra , and reddenings were adjusted as needed to bring their continua shapes closer to that of the expected template shape . the method is thus similar to methods that use the metal abundance to predict the intrinsic broad band colors , and then require the derived reddening to reproduce the observed colors . the overall goal in working with the low reddening sample was to retain the mean value of the reddenings found in @xcite , but to correct those that varied significantly . after cleaning up those reddenings , the interpolation formulae were then used to derive reddenings for the 150 clusters for which we have spectra and whose reddenings were not measured in @xcite . thus while we have not improved upon the absolute levels of the m31 old cluster reddenings , we believe we have improved the precision of the values in a relative sense , and as well have nearly doubled the number of reddenings available . about 10% of the spectra were taken during nights when the adc was not operating properly , thus we ca nt use the continuum shape to estimate reddenings . for objects whose only spectra were taken on those nights , we assume the average reddening of 0.28 . a comparison of our derived e(b@xmath31v ) values ( which range up to 1.4 mag ) and those in @xcite results in a scatter of 0.17 mag rms , which is good enough for our overall goal of comparing the m31 cluster system in bulk with that of other galaxies . interestingly , both the young cluster and old cluster groups have clusters with e(b@xmath31v)@xmath33 , though the highest measured value ( e(b@xmath31v)=1.4 ) is still found in the old cluster b037-v327 , probably a selection effect since that cluster also has the highest luminosity in all of m31 . the young cluster reddenings are listed in table [ young ] ; those of the old clusters will be presented in a subsequent paper . by using the position of blue - plume stars in the color - magnitude diagram , @xcite estimated the average reddening for young stars in m31 to be 0.13 mag , significantly lower than the mean of the clusters younger than 100 myr presented here , which may place a constraint on the accuracy of the values presented here . the mass histogram for the all of the young clusters is shown in figure [ massdist ] . we have also shown the mass distribution of milky way open clusters within 600 pc of the sun . this is based on the sample of @xcite , with mass calculations by @xcite . the kharchenko catalog is the most homogeneous and complete catalog of open clusters in the solar neighborhood currently available , and is based on a stellar catalog complete to v=11.5 . the cluster masses were estimated by counting the number of cluster members brighter than the limiting magnitude , then correcting for the stars fainter than this using a salpeter mass function and a lower mass limit of 0.15 m@xmath4 . this catalog does not include the most massive clusters in the galaxy because of its relatively small sample size ; for example , there have been recent discoveries of more distant young clusters which may have masses as high as @xmath3m@xmath4 ( e.g. @xcite ) , and we add westerlund 1 to the histogram as an example . the milky way globular and lmc young massive cluster histograms are shown in the bottom two panels ( from * ? ? ? * ) . these mass estimates are based on king model fits . obviously , m31 clusters with masses less than @xmath34m@xmath35 and ages greater than a few @xmath36 years are too faint to be part of this study , and await a future study . @xcite estimate over 10000 such clusters in the disk of m31 ; these would form the low mass tail in the mass distribution of figure [ massdist ] . nonetheless , there is a trend in cluster mass , with the milky way open clusters having the lowest median mass , the milky way and m31 globulars the highest , and the lmc young massive clusters and the m31 young clusters in between . this trend is consistent with a single cluster imf plus disruption , taking into account the small size of the volume searched for clusters in the milky way . would we expect these young m31 clusters to survive as they age , or to disrupt ? one of the main processes that leads to cluster disruption is 2-body relaxation enhanced by an external tidal field @xcite . the lower - mass clusters suffer more strongly from relaxation effects . another property of the cluster itself which will affect its survival is its density lower - density clusters will disrupt more quickly @xcite . thus we would expect that massive , concentrated clusters such as b018 and bh05 would be more likely to survive . @xcite derive an empirical expression for the disruption of clusters as a function of their mass , studying cluster populations in the solar neighborhood , the smc , m33 and m51 . @xcite point out that observational selection effects could mimic the decrease in the number of clusters with age which boutloukos et al . ascribe to cluster disruption . however , this is almost certainly not true of the solar neighborhood open clusters studied by @xcite using a similar analysis . we show the age - mass diagram for the young m31 clusters in figure [ agemass ] . while our sample is clearly very incomplete below @xmath37 years , the diagram shows some similarity to the lmc cluster age - mass diagram of @xcite in the age range we cover . unfortunately , we do not expect our catalog to be complete enough to permit an analysis using the techniques of boutloukos et al . ] environmental effects also control the tidal stripping of the cluster . for stars whose orbits are mostly confined to the disk , encounters with giant molecular clouds and spiral arms contribute to their disruption @xcite . for clusters whose orbits are not confined to the disk , bulge and disk shocking are more important @xcite . the similarity of the m31 young cluster kinematics to that of other young disk objects suggests strongly that these clusters are confined to m31 s disk plane , so giant molecular clouds should be the relevant external disruptor . @xcite show that disruption times for clusters in galaxies ranging in size from m51 to the smc , scale with molecular gas density in the expected way . s molecular gas density is highest near the ring of fire " where many of our clusters are found @xcite . this density is similar to the molecular gas density in the solar neighborhood @xcite . thus we would expect the survival due to giant molecular cloud interactions of the m31 young clusters to be similar to that of the solar neighborhood open clusters . we expect that most of these young clusters will be disrupted in the next gyr or so ( * ? ? ? * derive a disruption time of 1.3 gyr for a cluster of mass 10@xmath38 in the solar neighborhood ) . however , some of the more massive and concentrated of the young clusters will likely survive for longer . we present a new catalog of 670 m31 clusters , with accurate coordinates . in this paper we focus on the 140 clusters ( many originally classified in the literature as globular clusters ) which have ages less than 2 gyr : most have ages between @xmath37 and @xmath39 years . using high - quality mmt / hectospec spectra , excellent ground based images , and in some cases , hst images , we explore the nature of these clusters . with the exception of ngc 205 s young cluster , they have spatial and kinematical properties consistent with formation in the star - forming disk of m31 . many are located close to the 10 kpc `` ring of fire '' which shows active star formation . the age distribution of our clusters , plus that of the younger clusters of @xcite , shows no evidence for a peak in star formation there between @xmath25 and @xmath39 years ago , which we might expect if the ring was created by a recent passage of m32 through the disk , as suggested by @xcite and @xcite . we have estimated their masses using spectroscopic ages and m / l ratios , ( in some cases ) acs color - magnitude diagrams , and new photometry from the local group survey . the clusters have masses ranging from 250 to 150,000 @xmath40 . these reach to higher values than the known milky way open clusters , but it must be remembered that our sample of open clusters in the milky way is far from complete . the most massive of our young clusters overlap the mass distributions of m31 s old clusters and the milky way globulars . interestingly , although most of the young clusters show the low - concentration structure typical of the milky way open clusters , a few have the high concentrations typical of the milky way globulars and the old m31 clusters . we estimate that most of these young clusters will disrupt in @xmath30 gyr , but the massive , concentrated clusters may well survive longer . we would like to thank dan fabricant for leading the effort to design & build the hectospec fiber positioner and spectrograph , perry berlind & mike calkins for help with the observations , john roll , maureen conroy & bill joye for their many contributions to the hectospec software development project , and phil massey , pauline barmby & jay strader for comments and data tables on m31 . hlm was supported by nsf grant ast-0607518 , and would like to thank dean mclaughlin for helpful conversations . work on this project has also been supported by hst grant go10407 . baade , w. , & arp , h. 1964 , , 139 , 1027 barmby , p. et al . 2000 , aj , 119 , 727 barmby , p. , & huchra , j. p. 2001 , , 122 , 2458 barmby , p. , mclaughlin , d. e. , harris , w. e. , harris , g. l. h. , & forbes , d. a. 2007 , , 133 , 2764 battistini , p. , bonoli , f. , braccesi , a. , fusi - pecci , f. , malagnini , m. l. , & marano , b. 1980 , , 42 , 357 battistini , p. , bonoli , f. , braccesi , a. , federici , l. , fusi pecci , f. , marano , b. , & borngen , f. 1987 , , 67 , 447 battistini , p. l. , bonoli , f. , casavecchia , m. , ciotti , l. , federici , l. , & fusi - pecci , f. 1993 , , 272 , 77 beasley , m. a. et al . 2004 , aj , 128 , 1623 beaton , r. l. , et al . 2007 , , 658 , l91 beichman , c. a. , chester , t. j. , skrutskie , m. , low , f. j. , & gillett , f. 1998 , , 110 , 480 bertin , e. , & arnouts , s. 1996 , , 117 , 393 bica , e. , bonatto , c. , & blumbeqrg , r. 2006 , , 460 , 83 bica , e. , & bonatto , c. 2008 , , 384 , 1733 binney , j. and tremaine , s. , galactic dynamics , second edition , princeton university press , princeton nj block , d. l. , et al . 2006 , , 443 , 832 bonatto , c. , & bica , e. 2005 , , 437 , 483 bonatto , c. , & bica , e. 2007 , , 473 , 445 boutloukos , s. g. , & lamers , h. j. g. l. m. 2003 , , 338 , 717 burstein , d. , et al . 2004 , apj , 614 , 158 chandar , r. , bianchi , l. , & ford , h. c. 1999 , , 517 , 668 cioni , m .- l. , girardi , l. , marigo , p. , & habing , h. j. 2006 , , 448 , 77 cioni , m .- l. , girardi , l. , marigo , p. , & habing , h. j. 2006 , , 452 , 195 clark , j. s. , negueruela , i. , crowther , p. a. , & goodwin , s. p. 2005 , , 434 , 949 k. , & cameron , p. b. 2005 , , 634 , l45 cohen , j. g. , matthews , k. , & cameron , p. b. 2005 , apjl , 634 , l45 crampton , d. , cowley , a. p. , schade , d. , & chayer , p. 1985 , , 288 , 494 fall , s. m. , & rees , m. j. 1985 , , 298 , 18 freeman , k.c . 1980 , in star clusters , iau symposium 85 , edited by j. hesser ( reidel , dordrecht ) , p317 friel , e. d. 1995 , , 33 , 381 fusi pecci , f. et al . 2005 , aj , 130 , 554 galleti , s. , federici , l. , bellazzini , m. , fusi pecci , f. , & macrina , s. 2004 , , 416 , 917 galleti , s. , bellazzini , m. , federici , l. , buzzoni , a. , & fusi pecci , f. 2007 , , 471 , 127 gieles , m. , athanassoula , e. , & portegies zwart , s. f. 2007 , , 376 , 809 gieles , m. , lamers , h. j. g. l. m. , & baumgardt , h. 2008 , iau symposium , 246 , 171 gordon , k. d. , et al . 2006 , apjl , 638 , l87 de grijs , r. , & anders , p. 2006 , , 366 , 295 hodge , p.w . 1981 , in astrophysical parameters for globular clusters , iau colloquium 68 , edited by a.g.d . phillip and d.s . hayes ( schenectady , davis ) hodge , p. w. 1979 , aj , 84 , 744 hodge , p. w. , mateo , m. , lee , m. g. , & geisler , d. 1987 , pasp , 99 , 173 hubble , e. 1932 , , 76 , 44 huxor , a. p. , tanvir , n. r. , irwin , m. j. , ibata , r. , collett , j. l. , ferguson , a. m. n. , bridges , t. , & lewis , g. f. 2005 , , 360 , 1007 kharchenko , n. et al . 2005 , a&a , 438 , 1163 kennicutt , r. c. , jr . , & chu , y .- h . 1988 , aj , 95 , 720 kent , s. m. 1989 , , 97 , 1614 . kinman , t. d. 1963 , , 137 , 213 kim , s. c. , et al . 2007 , , 134 , 706 koekemoer , a. m. , fruchter , a. s. , hook , r. n. , & hack , w. 2002 , the 2002 hst calibration workshop : hubble after the installation of the acs and the nicmos cooling system , proceedings of a workshop held at the space telescope science institute , baltimore , maryland , october 17 and 18 , 2002 . edited by santiago arribas , anton koekemoer , and brad whitmore . baltimore , md : space telescope science institute , 2002 . , p.337 , 337 krienke , o. k. , & hodge , p. w. 2007 , , 119 , 7 krienke , o. k. , & hodge , p. w. 2008 , , 120 , 1 lamers , h. j. g. l. m. et al . 2005 , a&a , 441 , 117 larsen , s. s. 2002 , extragalactic star clusters , iaus 207 , 421 larsen , s. s. 2004 , the formation and evolution of massive young star clusters , 322 , 19 lee , m. g. , hwang , h. s. , kim , s. c. , park , h. s. , geisler , d. , sarajedini , a. , & harris , w. e. 2008 , , 674 , 886 mclaughlin , d. e. , & van der marel , r. p. 2005 , apjs , 161 , 304 mateo , m. 1993 , the globular cluster - galaxy connection , 48 , 387 massey , p. et al . 2006 , aj , 131 , 2478 massey , p. , olsen , k. a. g. , hodge , p. w. , jacoby , g. h. , mcneill , r. t. , smith , r. c. , & strong , s. b. 2007 , , 133 , 2393 mochejska , b. j. , kaluzny , j. , krockenberger , m. , sasselov , d. d. , & stanek , k. z. 1998 , acta astronomica , 48 , 455 morrison , h. l. , harding , p. , perrett , k. , & hurley - keller , d. 2004 , , 603 , 87 peebles , p. j. e. , & dicke , r. h. 1968 , , 154 , 891 perrett , k. m. , bridges , t. j. , hanes , d. a. , irwin , m. j. , brodie , j. p. , carter , d. , huchra , j. p. , & watson , f. g. 2002 , , 123 , 2490 puzia , t. h. , perrett , k. m. , & bridges , t. j. 2005 , , 434 , 909 santos , j. f. c. , jr . , bonatto , c. , & bica , e. 2005 , , 442 , 201 sargent , w. l. w. , kowal , c. t. , hartwick , f. d. a. , & van den bergh , s. 1977 , , 82 , 947 schiavon , r. p. , rose , j. a. , courteau , s. , & macarthur , l. a. 2005 , , 160 , 163 sirianni , m. , et al . 2005 , , 117 , 1049 spitzer , l. j. , & harm , r. 1958 , , 127 , 544 spitzer , l. j. 1958 , , 127 , 17 spitzer , l. j. , & shapiro , s. l. 1972 , , 173 , 529 stetson , p. b. 1987 , , 99 , 191 whitmore , b. c. , chandar , r. , & fall , m. 2007 , , 133 , 1067 williams , b. f. , & hodge , p. w. 2001 , , 559 , 851 williams , b. f. , & hodge , p. w. 2001 , , 548 , 190 worthey , g. 1994 , , 95 , 107 worthey , g. , & ottaviani , d. l. 1997 , , 111 , 377

we present the results of a search for faint hi 21-cm emission line wings at velocities forbidden by galactic rotation in the galactic plane using the leiden / dwingeloo hi survey data and the hi southern galactic plane survey data . these `` forbidden - velocity wings ( fvws ) '' appear as protruding excessive emission in comparison with their surroundings in limited ( ) spatial regions over velocity extent more than km s in large - scale (, ) diagrams . their high - velocities imply that there should be some dynamical phenomena associated . we have identified 87 fvws . we present their catalog , and discuss their distribution and statistical properties . we found that 85% of fvws are not coincident with known supernova remnants ( snrs ) , galaxies , or high - velocity clouds . their natures are currently unknown . we suspect that many of them are fast - moving hi shells and filaments associated with the oldest snrs that are essentially invisible except via their hi line emission . we discuss other possible origins .

0705.3684

large - scale ( @xmath3,@xmath4 ) diagrams of hi 21-cm line emission in the galactic plane usually show faint high - velocity bumps protruding from their surroundings . these faint wing "- like features are extended to the velocities well beyond the maximum or minimum velocities permitted by the galactic rotation . some of those forbidden - velocity wings ( fvws ) " are shown in fig . [ introfig ] , which is a ( @xmath3,@xmath4 ) diagram of the galactic hi emission in the 1st and the 2nd quadrants at @xmath5 . fvws are the ones that are small and smoothly extend from the main galactic hi , which distinguishes them from high - velocity clouds ( hvcs ) that have discrete peaks isolated from the galactic hi . fvws are probably the sites where kinetic energies are expelled into the interstellar medium ( ism ) by some violent events . our interest about fvws was first motivated by the statistics of snrs , namely that the number of known snrs is much less than that of the expected ( * ? ? ? * hereafter kk04 ) . the number of presently - known galactic snrs , most of which have been identified in radio continuum , is 265 according to the green s catalog @xcite , which provides a most complete and up - to - date list of identified galactic snrs . this is much less than the number ( 20 30,000 ) expected from the sn rate and the life time of snrs . therefore , most snrs are `` missing '' , and it must be mainly old snrs that are missing considering their large population . this is not surprising because old snrs are faint in radio continuum , and it is difficult to identify faint radio sources because of the confusion due to the galactic background emission and also because of observational limitations ( e.g. , * ? ? ? kk04 paid attention to the fact that an old snr consists of fast expanding hi shell and that the shell might last after the remnant becomes too faint to be visible in radio continuum . if its expansion velocity is greater than the minimum or maximum velocities permitted by the galactic rotation , then the snr shell or part of it ( maybe the caps ) could be detected as high - velocity gas , e.g. , fvws , in fig . [ introfig ] . they proposed that the fvws could be possible candidates for the missing old snrs . indeed , @xcite have carried out high resolution hi line observation toward fvw 190.2 + 1.1 , one of fvws in this paper , and detected a rapidly expanding ( @xmath6 km s@xmath2 ) hi shell . the parameters of this shell seem only consistent with those of the remnant of a sn explosion . this shell is not seen in any other wave band , suggesting that it represents the oldest type of snr , that is essentially invisible except via its hi line emission . although snr is a strong candidate of fvws , the high velocities of fvws could be also produced by stars expelling their mass into the ism , e.g. , stellar winds from massive stars , and outflows from young stellar objects or asymptotic giant branch ( agb ) stars . high- or intermediate - velocity clouds ( hvcs , ivcs ) that happen to be close to the maximum / minimum velocity of the hi gas along the line of sight would appear as fvws , too . in order to understand the nature of fvws and their relation to old snrs , therefore , a systematic study of fvws is required . we have searched fvws in the galactic plane ( @xmath7 ) using leiden / dwingeloo hi survey ( lds ) data and the hi southern galactic plane survey ( sgps ) data , and studied their statistical characteristics . some of our preliminary results were published by @xcite . in section [ secdata ] , we explain the hi surveys used for the search and the identification procedure . the catalog of fvws is presented in section 3 , where we show their statistical properties and the result of comparison with other objects , too . possibility of fvws to be snrs and some other possible origins will be discussed in section 4 . we finally summarize our results in section 5 . we used two galactic hi survey data ; the leiden / dwingeloo hi survey ( lds ) @xcite and the parkes survey , which is a part of the hi southern galactic plane survey ( sgps ) @xcite . the lds was carried out with the 25-m dwingeloo telescope ( fwhm @xmath8 ) on a half - degree grid using frequency - switching mode . the survey covered the sky north of declination @xmath9 completely on this grid . its effective velocity coverage spans local standard of rest ( lsr ) velocities from @xmath10 to @xmath11 km s@xmath2 , resolved into spectral channels of 1.03 km s@xmath2 width . stray radiation correction is applied to lsd data . the nominal brightness - temperature sensitivity of the lds is 0.07 k. the velocity resolution is not critical for the identification of fvws because fvws are extended over a quite wide velocity range . we smoothed each spectrum to a resolution of 3.09 km s@xmath2 , so that the rms noise of the final cubes used for the identification becomes @xmath12 k. the parkes survey is a part of sgps , a survey of the hi spectral line and 21-cm continuum emission in the fourth quadrant of the galactic plane . the parkes survey covers @xmath13@xmath14 and @xmath15 . the parkes data were obtained with the inner seven beams of the parkes multibeam system , a 13 beam 21-cm receiver package at prime focus on the parkes 64 m radio telescope ( fwhm@xmath16 ) , by the process of `` on - the - fly '' mapping . the final calibrated parkes data consist of 10 cubes with a cell size of @xmath17 , and a velocity resolution of 0.82 km s@xmath2 covering from @xmath18 to @xmath19 km s@xmath2 . the rms noise of the survey in brightness temperature varies from 0.13 to 0.27 k depending on the cube . the main survey area ( @xmath20 ) was re - observed to put limits on the amount of stray radiation , so that the rms noise is significantly better in this area ( @xmath21 k ) . for the coherence of this study , we convolved the parkes data using a gaussian beam to have the same angular resolution as the lds and then re - gridded the data with a grid interval of @xmath22 . the velocity channels are not smoothed ( but just regridded to a grid of 1.03 km s@xmath2 interval ) because the resulting data cubes have rms noise ( 0.020.04 k ) already lower than that of the lds data cubes . we have identified fvws by drawing large scale ( @xmath3,@xmath4 ) diagrams of each quadrant every @xmath22 in galactic latitude for galactic latitudes of @xmath7 . all identification was carried out with bare eyes . we first selected areas protruding from the galactic background hi emission at contours between 0.1 k and about 1.0 k. among them , we chose the ones with velocity extent greater than @xmath2320 km s@xmath2 compared to the neighborhood area . then , we looked into the channel images to make sure that their high - velocity features are clearly visible . we limited our selection only for those that are confined in small regions ( @xmath24 ) . the @xmath25 criterion is not based on an astrophysical consideration , but is rather empirical . however , this should be a reasonable choice for identifying missing snrs because most of them might reside in the inner galaxy and smaller than @xmath26 ( kk04 ) . in fig . [ introfig ] , for example , the features near @xmath27 , @xmath28 and 173are selected . but , the ones confined in somewhat broad longitude regions ( @xmath29 ) near @xmath30 , 40 , and 100are not selected . those features do not seem to be real but to be caused by data reduction process . in lds data , the spectra in a @xmath31 box are reduced at same time and these excess emissions are confined in @xmath31 square areas almost exactly . fvws could not be identified in the areas where groups of hvcs or intermediate - velocity clouds ( ivcs ) , or supershells make the velocity boundary complicated or where the velocity boundary is unclear or clumpy . for example , anti - center shell , and hvc complexe h and g highly contaminate the negative velocity boundaries near @xmath32 , and some velocity boundaries at high latitudes in the inner galaxy are clumpy due to many hvcs . table [ comarea ] summarizes those areas . we note that there are gaps at @xmath33 , at @xmath34 , and at @xmath35 between the two surveys . ( the exact longitude interval depends on latitude . ) cccl ( 0 , 50 ) & ( @xmath36 , @xmath37 ) & @xmath38 & + ( 10 , 30 ) & ( @xmath39 , @xmath40 ) & @xmath38 & + ( 10 , 30 ) & ( @xmath41 , @xmath42 ) & @xmath43 & + ( 90 , 150 ) & ( @xmath44 , @xmath45 ) & @xmath43 & hvc complexe g + ( 110 , 160 ) & ( @xmath46 , @xmath41 ) & @xmath43 & hvc complexe h + ( 160 , 180 ) & ( @xmath46 , @xmath47 ) & @xmath43 & anti - center shell + ( 160 , 200 ) & ( @xmath47 , @xmath42 ) & @xmath43 & anti - center shell + ( 180 , 205 ) & ( @xmath48 , @xmath49 ) & @xmath43 & anti - center shell + ( 260 , 330 ) & ( @xmath39 , @xmath40 ) & @xmath38 & + ( 330 , 360 ) & ( @xmath50 , @xmath40 ) & @xmath51 & + ( 330 , 360 ) & ( @xmath52 , @xmath37 ) & @xmath51 & + totally 87 fvws have been identified ; 73 in the northern sky and 14 in the southern sky . the parameters of the fvws are summarized in table [ catafvw ] . the entries of the catalog are the following . : ( 1 ) name of fvws based on the galactic coordinates . the central coordinates are determined in the images integrated over the given velocity extent . ( 2 ) velocity ranges where excess hi emission is seen . ( 3 ) integrated intensity of hi 21-cm line emission over the given velocity extent at the center of fvws . ( 4 ) rank indicating definiteness of protruding feature . higher rank means it is more apparent in ( @xmath3,@xmath4 ) or ( @xmath53 ) diagram ( see next paragraph ) . ( 5 ) associated celestial objects ( see section 3.3 ) . we assign rank 3 to fvws that are well - defined in both in ( @xmath3,@xmath4 ) and ( @xmath54 ) diagrams . the prototypical ones are fvw [email protected] , fvw 94.5 + 8.0 , fvw 203.0 + 6.5 , etc . they are usually located in areas with flat and clear velocity boundaries . rank 1 fvws are those with relatively unclear high - velocity features in both ( @xmath3,@xmath4 ) and ( @xmath54 ) diagrams . the prototypical ones are fvw 10.0 + 7.5 , fvw 51.5 + 3.5 , fvw [email protected] , etc . they are located in areas of complicated velocity boundaries , but still show stronger emission than their surroundings over the velocity intervals in table 2 . we assign rank 2 to the rest , i.e. , the ones that are relatively well - defined in ( @xmath3,@xmath4 ) diagram but not in ( @xmath54 ) diagram , or vice versa . the prototypical ones are fvw [email protected] , fvw 197.0 + 7.0 , fvw [email protected] . rank 3 includes 33 fvws and both of rank 2 and rank 1 have 27 fvws . we note that , for some fvws , e.g. , fvw @xmath55 , fvw @xmath56 , or fvw 196.5 + 1.5 , the classification is ambiguous so that the ranks may be considered as a general guide of protrusion level . the integrated images and position - velocity diagrams of 87 fvws are presented in fig . [ fvws ] . since identifying fvws by eyes is a subjective process , we note that the fvw catalog may be incomplete and needs a careful consideration for statistical interpretation ( see next section ) . lcrcl [email protected] & @xmath43135,@xmath43108 & 3.3@xmath57 5.3 & 2 & + fvw10.0 + 7.5 & @xmath4381 , @xmath4339 & 28.6@xmath57 4.6 & 1 & + [email protected] & @xmath4348 , @xmath4317 & 14.9@xmath57 1.6 & 2 & + [email protected] & @xmath4353 , @xmath4323 & 16.3@xmath57 2.4 & 2 & + [email protected] & @xmath4345 , @xmath4324 & 3.7@xmath57 3.1 & 2 & + + fvw27.5 + 6.5 & @xmath43101 , @xmath4375 & 10.0@xmath57 2.5 & 2 & + [email protected] & 117 , 138 & 19.6@xmath57 8.4 & 1 & + fvw39.0 + 4.0 & 89 , 131 & 35.2@xmath57 8.2 & 2 & + fvw40.0 + 0.5 & @xmath43113 , @xmath4380 & 21.9@xmath57 2.3 & 3 & + [email protected] & 79 , 119 & 53.8@xmath57 14.2 & 1 & + + [email protected] & @xmath43111 , @xmath4388 & 6.5@xmath57 0.8 & 3 & + [email protected] & 91 , 138 & 34.9@xmath57 4.6 & 3&snr , w 51 + fvw51.5 + 3.5 & 71 , 86 & 24.3@xmath57 9.7 & 1 & + [email protected] & 85 , 105 & 2.8@xmath57 0.7 & 2&snr , hc 40 + fvw69.0 + 2.5 & 35 , 89 & 26.6@xmath57 5.6 & 3&snr , ctb 80 + + [email protected] & @xmath43129,@xmath43103 & 57.4@xmath57 5.1 & 2 & + fvw75.5 + 0.5 & 30 , 60 & 29.9@xmath57 3.0 & 1 & + fvw78.0 + 2.0 & 43 , 57 & 13.0@xmath57 3.5 & 1&snr , dr 4 + fvw79.0 + 1.0 & 35 , 57 & 22.4@xmath57 5.6 & 1 & + fvw81.5 + 1.5 & 33 , 53 & 41.0@xmath57 5.2 & 1 & + + fvw83.5 + 4.0 & 30 , 52 & 29.0@xmath57 5.7 & 1 & + fvw84.5 + 0.0 & 34 , 55 & 45.9@xmath57 5.7 & 1 & + [email protected] & 20 , 57 & 54.0@xmath57 7.7 & 2 & + [email protected] & 21 , 42 & 12.0@xmath57 4.4 & 1 & + fvw88.5 + 5.0 & 30 , 88 & 24.6@xmath57 5.2 & 3&snr , hb 21 + + fvw94.5 + 8.0 & 31 , 114 & 18.5@xmath57 2.2 & 3&galaxy , cepheus 1 + fvw95.5 + 11.5 & 25 , 180 & 45.3@xmath57 2.4 & 3&galaxy , ngc 6946 + fvw95.5 + 7.0 & 26 , 65 & 7.3@xmath57 2.4 & 3 & + fvw96.0 + 4.0 & 18 , 38 & 14.3@xmath57 2.5 & 2 & + [email protected] & 28 , 48 & 7.5@xmath57 1.3 & 1 & + + [email protected] & 28 , 52 & 5.5@xmath57 1.3 & 1 & + [email protected] & 11 , 48 & 60.5@xmath57 12.2 & 2 & + [email protected] & 20 , 57 & 28.9@xmath57 4.8 & 3 & + fvw104.5 + 2.0 & 20 , 52 & 7.4@xmath57 1.4 & 3 & + fvw107.0 + 8.0 & 17 , 58 & 15.2@xmath57 3.0 & 3 & + + fvw108.0 + 7.0 & 27 , 53 & 5.4@xmath57 2.0 & 3 & + fvw108.0 + 10.0 & 29 , 51 & 3.2@xmath57 0.8 & 3 & + fvw109.5 + 2.5 & 8 , 44 & 264.8@xmath57 40.3 & 3 & + [email protected] & 18 , 36 & 11.7@xmath57 1.6 & 1 & + [email protected] & 39 , 65 & 4.7@xmath57 0.8 & 1&snr ( ? ) , cassiopeia a + + fvw116.5 + 12.0 & @xmath43196,@xmath43147 & 14.0@xmath57 1.8 & 3 & + [email protected] & 24 , 55 & 5.1@xmath57 0.9 & 3 & + [email protected] & 9 , 58 & 50.3@xmath57 5.8 & 2 & + fvw133.5 + 2.5 & 26 , 47 & 1.5@xmath57 0.6 & 2 & + fvw138.0 + 10.5 & 16 , 129 & 170.2@xmath57 12.0 & 3&galaxy , ugc 2847 + + fvw139.5 + 10.5 & 26 , 155 & 32.7@xmath57 52.5 & 3&galaxy , ugca 86 + fvw166.5 + 0.5 & 29 , 59 & 4.7@xmath57 1.0 & 2 & + fvw172.5 + 6.0 & 28 , 86 & 11.2@xmath57 1.1 & 3 & + fvw173.0 + 0.0 & 22 , 40 & 5.5@xmath57 1.8 & 2 & + fvw173.0 + 3.0 & 11 , 40 & 57.0@xmath57 23.7 & 3 & + + fvw190.2 + 1.1 & 44 , 81 & 8.4@xmath57 1.6 & 3 & + [email protected] & 54 , 82 & 5.7@xmath57 1.4 & 3 & + fvw196.5 + 1.5 & @xmath43130 , @xmath4358 & 74.8@xmath57 11.1 & 1&hvc + fvw197.0 + 7.0 & 52 , 78 & 4.4@xmath57 2.1 & 2 & + fvw201.0 + 9.5 & 64 , 90 & 9.2@xmath57 0.8 & 3 & + + [email protected] & @xmath4356 , @xmath4320 & 11.8@xmath57 4.2 & 1 & + [email protected] & @xmath4360 , @xmath4317 & 17.6@xmath57 3.9 & 1 & + fvw203.0 + 6.5 & 85 , 132 & 22.4@xmath57 1.3 & 3&hvc + fvw203.5 + 4.5 & 80 , 106 & 5.5@xmath57 2.3 & 3 & + fvw211.5 + 9.0 & 66 , 98 & 10.7@xmath57 3.8 & 1 & + + fvw212.5 + 5.5 & 78 , 115 & 9.8@xmath57 2.2 & 3 & + fvw212.5 + 3.5 & 67 , 127 & 45.1@xmath57 13.5 & 2 & + fvw215.0 + 7.0 & 84 , 115 & 11.1@xmath57 1.7 & 3&hvc + fvw215.5 + 7.5 & @xmath4343 , @xmath4320 & 5.8@xmath57 0.8 & 3 & + fvw225.5 + 6.0 & 103 , 128 & 7.2@xmath57 2.1 & 1 & + + fvw233.5 + 5.0 & 114 , 148 & 8.9@xmath57 1.6 & 1 & + [email protected] & 134 , 155 & 4.0@xmath57 1.2 & 1 & + fvw238.5 + 0.5 & 141 , 167 & 5.2@xmath57 0.9 & 3 & + fvw240.0 + 4.0 & 127 , 153 & 7.6@xmath57 2.2 & 1 & + fvw242.5 + 2.0 & 141 , 161 & 6.7@xmath57 1.7 & 1 & + + [email protected] & 135 , 167 & 19.0@xmath57 3.5 & 3 & + fvw251.0 + 12.0 & @xmath4336 , @xmath4313 & 27.9@xmath57 4.2 & 3 & + fvw252.5 + 10.5 & @xmath4350 , @xmath4330 & 6.6@xmath57 0.7 & 2 & + [email protected] & 159 , 184 & 16.8@xmath57 5.9 & 2 & + fvw261.6 + 2.3 & 158 , 178 & 13.9@xmath57 5.9 & 2 & + + fvw283.4 + 3.0 & @xmath4357 , @xmath4342 & 13.0@xmath57 5.0 & 2 & + fvw291.0 + 2.9 & @xmath4372 , @xmath4355 & 17.9@xmath57 9.8 & 2 & + fvw293.0 + 7.9 & @xmath4374 , @xmath4353 & 20.5@xmath57 7.8 & 1 & + [email protected] & 135 , 167 & 19.2@xmath57 8.4 & 1 & + [email protected] & @xmath4391 , @xmath4366 & 40.7@xmath57 13.4 & 2 & + + [email protected] & @xmath43109 , @xmath4392 & 8.9@xmath57 4.1 & 2 & + fvw312.3 + 3.6 & @xmath4399 , @xmath4380 & 42.3@xmath57 36.1 & 2 & + fvw313.4 + 0.6 & @xmath43117,@xmath43100 & 12.2@xmath57 4.8 & 2 & + fvw315.2 + 2.7 & @xmath43107 , @xmath4387 & 20.9@xmath57 8.9 & 1 & + fvw319.8 + 0.3 & @xmath43123 , @xmath4398 & 22.9@xmath57 8.1 & 3 & + + fvw324.5 + 4.2 & @xmath43132,@xmath43112 & 7.0@xmath57 4.8 & 2 & + fvw342.5 + 1.7 & 96 , 110 & 3.9@xmath57 1.9 & 3 & + fig . [ lbvifig ] presents the distribution of the fvws in ( @xmath3,@xmath58 ) plane with information about their lsr velocities and integrated intensities . fvws are marked by different symbols depending on their associations with other known objects ( see section 3.3 ) . the center of each symbol indicates its galactic coordinates and the area presents its integrated intensity . the color indicates the mean velocity with the scales shown by the color bar on the top . the hatched areas represent the highly complicated areas in table [ comarea ] . the areas filled with solid black show gaps between data , and the dotted line shows the boundary of the search in latitude . fvws are scattered on the ( @xmath3,@xmath58 ) plane . they are neither clustered in the inner galaxy nor strongly concentrated toward the galactic plane . this is more clearly seen in figs . [ distlbn ] ( left ) and ( right ) which show the distribution of fvws in galactic longitude and latitude , respectively . along the galactic longitude , there are relatively large number of fvws between @xmath59 and @xmath60 and perhaps at @xmath61 , while the number is relatively small between @xmath62 and @xmath63 and between @xmath64 and @xmath65 . the distribution of fvws in galactic latitude is broad and slightly skewed to high positive latitudes . a gaussian fitting yields @xmath66 as the center and @xmath67 as the full width at half maximum ( fwhm ) of the distribution . the fwhm in the inner galaxy is smaller than that in the outer galaxy , i.e. , @xmath68 and 17@xmath69 , respectively . a noticeable feature in fig . [ lbvifig ] is the change of the velocity of fvws with galactic longitude ( see also fig . [ distlbn ] ( left ) ) . between @xmath70 and @xmath71 , most ( 90 % ) fvws have positive velocities , while in the rest , fvws with negative velocities are dominant ( 70 % ) . in total , the fraction of fvws at positive velocities is significantly greater , 70% vs. 30% . the distribution of fvws with positive and negative velocities are also markedly different in their latitude distribution ( fig . [ distlbn ] ( right ) ) : positive - velocity fvws are concentrated toward the galactic plane although the fwhm is large , while negative - velocity fvws are roughly uniformly distributed . the above statistical properties might reflect the detectability of fvws in the first place , e.g. , how smooth and sharp the boundary of the general galactic hi emission is in ( @xmath72 ) diagram . in the 1st and 2nd quadrants , the boundaries at negative - velocities ( hereafter negative - boundaries ) , which correspond to the edge of the galaxy , are wiggly and diffuse , so that the identification of negative fvws are difficult there . also , anti - center shell ( acs ) and hvc complexes make negative - boundaries in the 2nd quadrant complicated ( table [ comarea ] ) . the boundaries at positive velocities ( hereafter positive - boundaries ) are generally smooth and sharp , although they become complicated toward the inner galaxy , particularly between @xmath73 and @xmath74 where there is not much ism and the gas motion is not circular . the above explains why most of the fvws in the 1st and 2nd quadrants have positive velocities and also why there are relatively few fvws in the innermost galaxy . in the 3rd and 4th quadrants , it is negative - boundaries which are smooth and sharp . but they are not as sharp as the positive - boundaries in the 1st and 2nd quadrants . in addition , acs makes the negative - boundaries at @xmath75@xmath76 complicated . this may explain why the negative - velocity fvws in these quadrants are not as dominant as the positive - velocity fvws in the 1st and 2nd quadrants . on the other hand , there is no reason for positive - velocity fvws at @xmath77@xmath63 to be more difficult to be identified than those at @xmath78@xmath60 because both areas have very smooth and sharp positive - boundaries . therefore , there should be intrinsically less positive - velocity fvws in this area . also there is no obvious changes in boundary shapes to cause the asymmetry and the broad extent in latitude distribution . [ colden ] shows the distribution of column density of fvws . about 90 % of fvws have values n@xmath79 @xmath80 . the median n@xmath81 is @xmath82 @xmath80 . note that this is the column density over the velocity interval in table [ catafvw ] . lllccc [email protected]&[email protected] & w51 & 30&4.1 & 1 + [email protected]&[email protected] & hc40 & 40&3.3 & 2 + fvw69.0 + 2.5&g69.0 + 2.7 & ctb 80 & 80&2.0&3 + fvw78.0 + 2.0&g78.2 + 2.1 & dr4 , gamma cygni snr & 60&1.4&2 + fvw88.5 + 5.0&g89.0 + 4.7 & hb21&120 @xmath83 90&0.8&4 + [email protected]&[email protected] & cassiopeia a , 3c461 & 5 & 3.4&5 + we compared the positions of fvws with those of snrs , nearby galaxies , hvcs , and found that 6 , 4 , and 3 fvws are coincident with those objects , respectively . for comparison with snrs , we used the green s snr catalog @xcite . we have found six fvws coincident with snrs and the results are summarized in table [ snrs ] . systematic studies to search for high - velocity hi gas have been made toward 200 snrs @xcite . they identified 25 snrs with fast - moving hi gas localized on snr areas . five snrs out of 25 are identified as fvws in this paper , i.e. , fvws in table [ snrs ] excluding [email protected] which is coincident with cas a. the other 20 snrs are not identified as fvws due to several reasons . some snrs expansion velocities are not large enough to be identified as a fvw in a large - scale ( @xmath3,@xmath4 ) diagram . some have small size so that their excess emission at high velocities are smoothed out in the data with low angular resolution . for three snrs in table [ snrs ] , high - resolution studies of the shocked hi gas have been made previously , e. g. , w51 by @xcite , ctb80 by @xcite , and dr4 by @xcite . we have detected fvw toward cas a , but , considering that cas a is only @xmath84 years old , it seems to be a chance coincidence . spectrum of cas a region is very noisy due to strong radio continuum . lrlrccc fvw94.5 + 8.0 & ( 94.37 , + 8.01)&cepheus 1 & 11.7&58&6.0&1 + fvw95.5 + 11.5 & ( 95.72 , + 11.68)&nbg 2146 , ngc 6946 & 11.2&46&5.5&2 + fvw138.0 + 10.5 & ( 138.18 , + 10.58)&nbg 329 , ugc 2847 & 16.1&32&3.9&2 + fvw139.5 + 10.5 & ( 139.77 , + 10.64)&nbg 348 , ugca 86 & 1.0&72&4.4&2 + for comparison with galaxies , we used the ` nearby galaxies catalog ' of @xcite , which includes 2367 galaxies with systemic velocities less than 3000 km s@xmath2 . three fvws are identified as galaxies according to this catalog . another fvw , fvw94.5 + 8.0 , looks similar to those 3 fvws , and turned out to be a low surface brightness galaxy named cep 1 @xcite . their parameters are summarized in table [ gals ] . fvws associated with galaxies have velocity extents wider than 80 km s@xmath2 , and are easily distinguishable from other fvws . fvw172.5 + 6.0 also show its excess emission over wide velocity extent comparable to those of galaxies . there is indeed a galaxy named ` leda 2126032 ' toward this direction according to the simbad astronomical database . the galaxy has a half - light radius of @xmath233 arcsec ( 2mass extended source catalog ) , and it s radial velocity is not known . if it is a dwarf galaxy having typical physical half - light radius of 100 pc , it would be at 6 mpc . its association with fvw172.5 + 6.0 is possible . no other fvws appear to be external galaxies . lccrcc fvw196.5 + 1.5 & ( 197.0 , + 1.5)&@xmath85 & @xmath4366.0&100&1 + fvw203.0 + 6.5&(202.8 , + 6.4)&2.5&108.4&120&2 + fvw215.0 + 7.0&(215.0 , + 7.0)&1.1&103.0&50&2 + for comparison with hvcs , we used hvc catalogs of @xcite including about 560 hvcs of all sky , and @xcite containing summary catalog based on the absorption line study for the hvcs and ivcs . in addition , we compared the catalog of fvws with those of compact isolated hvc ( chvc ) @xcite . @xcite identified about 900 chvcs using lds in the northern sky , and @xcite identified about 2000 compact and extended hvcs in southern sky using the hi parkes all - sky survey ( hipass ) , which covers lsr velocity range between @xmath86 km s@xmath2 and @xmath87 km s@xmath2 with @xmath88 spatial resolution and 26 km s@xmath2 velocity resolution . as shown in table [ hvcs ] , three fvws are identified as hvcs . although @xmath23 30% of fvws unrelated to the known sources show isolated bumps in their spectra ( see section 4.2 ) , most of them are not matched with hvcs in available hvc catalogs . we have identified 87 fvws . among them , 6 , 4 , and 3 fvws are found to be coincident with snrs , nearby galaxies , and hvcs , respectively . the rest ( 85% ) are not associated with any obvious objects that would be responsible for their large velocities . we already pointed out that fvws associated with galaxies have large , easily - distinguishable velocity extents and none of the rest fvws , perhaps except fvw172.+6.0 , seems to be associated with external galaxies . therefore , fvws are thought to be galactic objects . in this section , we discuss possible origins based on their observed properties . kk04 argued that the number of observed snrs with hi shell ( hi snrs ) is much less than that of the expected and suggested that fvws could be such missing old snrs . in fact , in large - scale @xmath89 diagrams , the morphologies of most fvws can not be discriminated from the fvws associated with known snrs . and high - resolution observation showed that at least one fvw , fvw 190.2 + 1.1 , is a rapidly ( @xmath6 km s@xmath2 ) expanding hi shell that is most likely an old snr remnant @xcite . hence , old snrs are considered to be primary candidates for fvws . the distribution of fvws , however , appears significantly different from either that of the known snrs or what we expect for old hi snrs . the known snrs are strongly concentrated toward the inner galaxy and toward the galactic plane . if we use 265 snrs in green s catalog @xcite , the longitude distribution has a maximum at @xmath90 and a fwhm of @xmath91 while the latitude distribution is almost symmetric with a fwhm of @xmath92 . the expected distribution of old snrs identifiable in hi emission had been calculated by kk04 using a simple model where it was assumed that the galaxy is composed of a uniform , axi - symmetric gaseous disk with a central hole and that only type ia supernovae produce isolated snrs which are exponentially distributed with a radial scale length of 3 kpc . according to their result , visible hi snrs are concentrated in the inner galaxy along the loci of tangent points ( @xmath93 and @xmath94 ) because the sn rate is higher in the inner galaxy and because the systemic velocities of snrs in those regions are close to either the maximum or the minimum velocities of the galactic background emission . the number of visible hi snrs is @xmath95 , 75% of which is within @xmath96 . kk04 did not calculate the latitude distribution , but if sn ia distribution follows thin old stellar disk with an exponential scale height of @xmath84 pc as in their model , the fwhm of these visible hi snrs would be @xmath97 . these results , showing that the expected distribution of visible hi snrs is quite different from that of fvws , however , does not necessarily rule out the sn origin for fvws : firstly , the number of hi snrs in the inner galaxy could be significantly less than that in the model if the interstellar space there is largely filled with a very tenuous gas instead of warm neutral medium ( 0.14 @xmath98 ) as kk04 pointed out . and , secondly , in the outer galaxy , there could be more visible hi snrs than the model because the identification is relatively easy there . also core - collapse sne may not be spatially and/or timely correlated in there , so that they result in hi snrs , too . however , the significantly larger number in the outer galaxy and their large scale height seem to indicate that not all of fvws could be old snrs . if fvws are snrs , it may be possible to see faint radio continuum emission associated with them . we have searched associated radio emission toward 38 fvws covered in the effelsburg 11-cm continuum survey @xcite ( @xmath99 , @xmath100 ) . 9 fvws are found to have noticable extended emission within @xmath101-sized circle in addition to the 6 fvws coincident with the known snrs ( fig . [ bonnfig ] ) . they are summarized in table [ contitable ] . the cases of fvw173.0 + 0.0 and fvw173.0 + 3.0 are interesting : there is an about @xmath102-size , incomplete shell - like continuum feature centered at ( @xmath103 ) @xmath104 , and hi emission of fvw173.0 + 3.0 appears to be enclosed by the filamentary continuum emission while fvw173.0 + 0.0 lies on the south of the shell - like continuum . other than these two , there is no obvious coincident shell - like continuum features . in order to prove the association of these radio features with fvws , further studies are required . ll fvw75.5 + 0.5 & extended feature , not well matched with hi + fvw79.0 + 1.0 & complicated area + fvw81.5 + 1.5 & complicated area + fvw84.5 + 0.0 & complicated area + fvw104.5 + 2.0 & hii region , not well matched with hi + fvw109.5 + 2.5 & hii region , not well matched with hi + fvw173.0 + 0.0 & 3-sized filamentary shell - like feature + fvw173.0 + 3.0 & part of 3-sized filamentary shell - like feature + fvw190.2 + 1.1 & hii region , not well matched with hi + about 30% of fvws unrelated to known snrs , galaxies , or hvcs show isolated bumps in their spectra , e.g. , [email protected] in fig . they appear as hi 21 cm wings in large - scale ( @xmath3,@xmath4 ) diagrams because they are very close to the galactic hi emission . these fvws therefore could be clouds with anomalous velocities rather than expanding shells . table [ isopeak ] lists such fvws . the fvws associated with known snrs , hvcs , or galaxies are not included , even though they show isolated bumps in their spectra . the bumps in some fvws in table [ isopeak ] , e.g. , fvw75.5 + 0.5 , fvw293.0 + 7.9 , and fvw312.3 + 3.6 , are weak ( @xmath105 k ) and not apparent in fig . [ fvws ] ( see the notes in table [ isopeak ] ) . on the other hand , there are fvws that appear to have very weak ( @xmath106 k ) , broad bumps in fig . [ fvws ] , e.g. , fvw95.5 + 7.0 and [email protected] , but they are hardly identified as isolated bumps in their spectra and not included in table [ isopeak ] . we have determined central lsr velocity , fwhm , and brightness temperature of fvws with isolated bumps by fitting their spectra into a gaussian plus a second order polynomial function . table [ isopeak ] summarizes the result . their central velocities vary from @xmath107 @xmath108 to @xmath109 km s@xmath2 . their fwhms are 7.025.5 km s@xmath2 with median value of 12.4 km s@xmath2 . their median peak brightness temperature is 1.0 k , and the median hydrogen column density is @xmath110 @xmath80 . table [ isopeak ] also lists the deviation velocity , which is defined as the excess of central velocity over the minimum or maximum velocities permitted by the galactic rotation @xcite . here , we assume a galactic model that @xcite used , i.e. , a 4 kpc thick disk with radius 26 kpc and a flat rotation curve with a value of @xmath111 km s@xmath2 and @xmath112 kpc . the deviation velocities range from @xmath113 to + 46.6 km s@xmath2 , and the median value is 31.9 km s@xmath2 . the negative @xmath114 means that the centeral velocity is within the allowed velocity range , which implies that the used disk model does not describe well the real velociy boundaries . the velocities of fvws indicate that they could have been classified either as hvcs or ivcs . hvcs are generally defined as clouds with velocities greater than 90 km s@xmath2 with respect to the local standard of rest ( lsr ) @xcite . at low galactic latitudes , however , this definition is not very useful and an alternative definition in terms of the deviation velocity could be used , i.e. , @xmath114 @xmath115 5060 km s@xmath2 @xcite . observationally ivcs are just lower velocity counterparts of hvcs , which are defined as clouds with 3040@xmath116 km s@xmath2 @xcite in terms of lsr velocity . about one third of the fvws in table [ isopeak ] have central lsr velocities @xmath117 to 170 km s@xmath2 , so that they may be classified as hvcs while the rest as ivcs . on the other hand , all of them have deviation velocities less than 50 km s@xmath2 , which puts them in the category of ivcs . an interesting feature of hvcs is that they show a systematic velocity change , i.e. , most hvcs are at positive velocities between @xmath118 and @xmath94 , while most of them are at negative velocities in the rest @xcite . this asymmetric velocity structure was explained by a model where hvcs are debris of the parental cloud of the local group , falling toward the center of the local group @xcite . the fvws in table [ isopeak ] show somewhat opposite velocity trend , i.e. , about 70% of fvws between @xmath118 and @xmath94 are at negative velocities , which is similar to the trend of the entire fvws ( 3.2 ) . this makes it difficult to consider the fvws as a low - velocity population of hvcs . it is also possible that the fvws with isolated bumps are small , fast - moving clouds unresolved in the survey . recently , small ( @xmath119 pc ) , fast - moving clouds are revealed by high - resolution observations in the disk / halo interface region of the inner galaxy , so called halo clouds @xcite , in the disk ( @xmath120 ) of the inner galaxy @xcite , and in the anti - center region @xcite . it was suggested that our galaxy may possess a population of those clouds distributed both throughout the disk and up into the halo rotating with the galactic disk , and that about one - half of the galactic hi halo could be in the form of discrete clouds @xcite . they could be clouds formed in galactic fountain flows or infalling intergalactic material in the on - going construction of the galaxy ( * ? ? ? * and references therein ) . the observational properties of the halo clouds and the clouds in the inner galactic disk and in the anti - center region are summarized by @xcite ( their table 3 ) . the properties of fvws in table . [ isopeak ] are quite comparable to those of halo clouds , i.e. , their velocity dispersion ( 12 km s@xmath2 ) and hi column densities ( 2@xmath121 @xmath80 ) are similar . the fast - moving clouds in the inner galactic disk have a factor of 2 smaller fwhm and much larger hi column densities , while the clouds in the anti - center region have comparable hi column densities but a factor of 3 smaller fwhm . but , since fvws are not resolved , this comparison should be considered as provisional until high - resolution maps of fvws are available . we should keep in mind that the isolated bump in spectrum does not necessarily mean that a fvw is not part of a snr . because the ism is not uniform , the accelarated hi shells of snrs often show bump - like features in their hi spectra ( e. g. , * ? ? ? it is also worth to note that the fast - moving small clouds usually appear related to diffuse filamentary structures @xcite , which suggests a possibility of disrupted snr shells . the nature of those cloud - like fvws needs further studies . lrrrrrl fvw10.0 + 7.5 & @xmath4343.6 & 20.3 & 16.6 & 1.5 & 50.3&1 + [email protected] & @xmath4334.1 & 12.4 & 13.8 & 0.7 & 19.9&1 + [email protected] & @xmath1229.3 & @xmath436.2 & 12.9 & 0.9 & 23.4&1 + [email protected]&@xmath43105.9 & @xmath4333.8 & 7.2 & 2.5 & 35.5&1 + fvw75.5 + 0.5 & 38.4 & 31.4 & 8.1 & 0.9 & 13.6&2 + fvw84.5 + 0.0 & 45.6 & 44.5 & 12.2 & 2.8 & 65.8&1 + [email protected] & 46.9 & 46.4 & 5.0 & 1.8 & 17.8&3 + fvw201.0 + 9.5 & 72.3 & 20.0 & 17.0 & 0.4 & 14.4&1 + fvw252.5 + 10.5 & @xmath4334.4 & 34.4 & 8.9 & 0.4 & 6.3&1 + [email protected] & 170.3 & 24.2 & 25.5 & 1.6 & 82.5&1 + fvw261.6 + 2.3 & 164.0 & 17.6 & 19.1 & 0.9 & 32.7&1 + fvw283.4 + 3.0 & @xmath4347.3 & 41.3 & 7.0 & 0.9 & 11.7&2 + fvw291.0 + 2.9 & @xmath4361.2 & 46.6 & 11.2 & 2.4 & 52.2&1 + fvw293.0 + 7.9 & @xmath4356.7 & 39.4 & 10.1 & 1.0 & 20.7&2 + [email protected] & 154.5 & 24.4 & 11.2 & 0.6 & 13.3&4 + [email protected] & @xmath4376.6 & 38.1 & 14.1 & 1.9 & 51.6&1 + [email protected]&@xmath43100.8 & 44.6 & 12.3 & 0.6 & 14.5&1 + fvw312.3 + 3.6 & @xmath4382.3 & 25.2 & 12.5 & 1.2 & 30.4&2 + fvw313.4 + 0.6&@xmath43102.6 & 42.5 & 12.4 & 1.0 & 24.2&1 + fvw324.5 + 4.2&@xmath12324.3 & 32.3 & 22.4 & 0.9 & 39.3&1 + median values & @xmath124 & 31.9 & 12.4 & 1.0 & 23.8 + another candidate for fvws is stellar winds and wind - blown shells . low- and high - mass protostars produce fast expanding winds or jets . ob stars or wolf - rayet ( wr ) stars eject strong winds forming expanding ionized or neutral hi shells . agb stars also expel their energies into the surrounding medium as slow and fast winds . here , we will discuss if they could be detectable in the hi surveys used in this work . low- and high - mass protostars lose their masses by stellar winds and jets . the wind could be neutral intrinsically @xcite or neutral hydrogen could be produced by dissociation of molecular hydrogen in shocks @xcite . neutral atomic stellar wind was detected toward young stellar object hh 7 - 11 at 350 pc from the sun @xcite . the average antenna temperature of neutral wind of the hh 7 - 11 is @xmath125 k when it is observed at the arecibo telescope . this corresponds to an antenna temperature of @xmath126 k for a 25-m or 64-m telescope . therefore , if the observed parameters of hh 7 - 11 represent typical wind parameters of low - mass protostars , fvws identified in this paper ( @xmath127 k ) can not be due to such neutral winds . neutral atomic stellar wind has been detected toward high - mass protostars , too . @xcite detected atomic stellar wind expanding at @xmath128 km s@xmath2 toward dr 21 , which contains one of the most powerful outflow sources in the galaxy . the hi mass of the wind is 24 m@xmath129 at the generally accepted distance of dr 21 ( 3 kpc ) , which is three orders of magnitude more massive than the hi wind seen in hh 7 - 11 . this would yield @xmath130k for a 25 m telescope . if we assume the observed parameters of dr 21 is typical , the neutral winds from massive protostars would not be detectable unless they are closer than @xmath131 kpc . strong stellar wind from early type ob stars or wr stars is likely to be an another possible source producing fast - moving hi gas . the mechanical luminosity of winds from ob stars range from 10@xmath132 erg s@xmath2 to @xmath133 erg s@xmath2 with a typical expansion velocity of 1,500 km s@xmath2 @xcite . this wind sweeps up ambient medium into a shell which could be observed as fvws . most observed hi shells around ob stars , however , have low expansion velocities , i.e. , @xmath134 km s@xmath2 ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? but these expansion velocities might be lower limits because their endcaps are not usually detected . according to @xcite , the radius of expanding shell is given by @xmath135 pc , where @xmath136 is ambient atomic density of h nuclei in @xmath98 , @xmath137 is the wind luminosity in @xmath138 erg s@xmath2 , and @xmath139 is the age in @xmath140 yr . from this , we may express the mass in the shell @xmath141 in terms of its expansion velocity @xmath142 ; @xmath143 @xmath144 m@xmath129 , where @xmath145 is expansion velocity in 100 km s@xmath2 . when this shell is observed with a 25-m telescope , it would yield a brightness temperature of @xmath146 , where m(t ) is in m@xmath129 , d is distance in kpc , and @xmath147 is velocity interval over which the hi shell would be detected in km s@xmath2 . since 0.1 k is a detection limit in this work , the ones at distance less than @xmath148 kpc could be detected assuming @xmath149 . if we assume that the shells of expansion velocity @xmath150 km s@xmath2 are detectable , then the distance should be less than @xmath151 kpc . thus , nearyby ( @xmath152 kpc ) wind shells with large wind luminosity ( @xmath153 erg s@xmath2 ) can be detected , assuming @xmath154 @xmath98 . we checked if o - type or wolf - rayet stars having large wind luminosities are in the region of fvws , using the galactic o star catalog of @xcite and the 7th catalog of galactic wolf - rayet stars of @xcite . about 15 % of fvws unrelated to the known sources have more than one stars in 1 -radius area . they are summarized in table [ owrstars ] . 12 and 3 fvws have nearby o stars and wr stars , respectively . all 3 fvws coincident with wr stars are coincident with o stars , too . they are at quite close distance ( @xmath155 kpc ) . thus , some of those fvws in table [ owrstars ] could be caused by winds from massive stars . however , it is worthy to note that the fwhm of the latitude distribution of early type stars ( @xmath105 ) @xcite is much smaller than that of fvws ( @xmath156 ) . lccllcc + fvw6.5 - 2.5 & 2 & ( 6.9 , @xmath432.1)&hd 165921 & o7.5v & 0.58 & 1.3 + fvw15.5 - 10.5 & 1 & ( 15.3,@xmath4310.6)&hd 175876&o6.5iii & 0.22 & 2.3 + fvw75.5 + 0.5 & 6 & ( 75.9 , + 0.8)&hd 193682 & o5 . v & 0.50 & 1.8 + fvw79.0 + 1.0 & 4 & ( 79.0 , + 1.3)&bd + 40 4179 & o8 . v & 0.30 & 1.2 + fvw83.5 + 4.0 & 1 & ( 83.8 , + 3.3)&bd + 45 3216 & o8 . & 0.77 & 1.9 + fvw104.5 + 2.0 & 1&(103.8 , + 2.6)&hd 210839 & o6 . i & 0.92 & 0.8 + fvw109.5 + 2.5 & 5&(109.6 , + 2.7)&hd 216532 & o9.5v & 0.21 & 0.9 + fvw173.0 + 0.0 & 4&(173.0 , + 0.1)&hd 35619 & o7 . v & 0.10 & 3.2 + fvw173.0 + 3.0 & 3&(173.5 , + 3.2)&hd 37737 & o9.5iii & 0.54 & 1.3 + fvw190.2 + 1.1 & 1&(190.0 , + 0.5)&hd 42088 & o6.5v & 0.50 & 1.5 + fvw313.4 + 0.6 & 3&(313.5 , + 0.2)&hd 125241 & o8.5i & 0.47 & 3.2 + fvw342.5 + 1.7 & 9&(342.8 , + 1.7)&hd 151515 & o7 . iii & 0.30 & 1.9 + + fvw75.5 + 0.5 & 3 & ( 75.7 , + 0.3 ) & sand 5&wo2 & 0.30 & 0.9 + fvw79.0 + 1.0 & 2 & ( 79.7 , + 0.7 ) & v1923 cyg , as 422 & wn7/wce+ ? & 0.77 & + fvw342.5 + 1.7 & 1&(343.2 , + 1.4 ) & hd 151932 & wn7h&0.77 & 2.0 + highly evolved cool stars in asymptotic giant branch ( agb ) eject strong stellar winds too and could be an another candidate for fvws . for example , molecular outflow at a speed of @xmath157 km s@xmath2 has been detected in agb carbon star v hya @xcite . these molecular winds may collide with the interstellar medium to produce fast - moving atomic gas . in particular , the fwhm of the latitude distribution of carbon stars is comparable to that of fvws , e.g. , @xmath158 based on the catalog of galactic carbon stars of @xcite . but the mass loss rate of agb stars is low ( @xmath159 m@xmath129 yr@xmath2 ) , so that the resulting 21-cm emission might be too faint . in summary , while some of fvws could be produced by nearby massive stars with large wind luminosity , the stellar winds or wind - blown shells from low- or high - mass protostars , or agb stars do nt seem to be capable of producing most of fvws . fvw appears as a faint extended wing - like feature in large - scale ( @xmath3,@xmath4 ) diagram . since they have excessive emission at high velocities forbidden by the galactic rotation , there should be some dynamical phenomena associated . thus , fvws are closely related to the dynamical energy sources in the galaxy and their effects on the ism . in this paper , we have identified 87 fvws in the galactic plane in use of the lds and the sgps data . fvws show double - peak distribution in longitude . they have wide fwhm of about @xmath160 in latitude and slightly skewed to positive latitude . about 70 % of fvws do have positive lsr velocities . these statistical properties appear to reflect the identifiability of fvws in the first place , e.g. , how smooth and sharp the boundary of the general galactic hi emission is in @xmath89 diagram . but there are properties that appear to be intrinsic too , e.g. , the small number of fvws at @xmath161180@xmath162 . the column densities of most fvws are less than @xmath163 @xmath80 . we compared the catalog of fvws with those of snrs , galaxies , and hvcs , and found that @xmath164 % are not coincident with those known objects . we investigated the possibility that fvws could be the old hi snrs invisible in radio continuum ( kk04 ) . the distribution of fvws is different from that of the expected hi snrs , and the fwhm of fvws is wider than that of snrs . the possibility of fvws to be the old hi snrs is not promising in this regard . instead a good fraction of fvws may be related to anomalous velocity clouds or o - type / wr stars : about 30% of fvws unrelated to the known sources have discrete hi bumps that are a bit seperated from the galactic hi . those fvws could be isolated clouds similar to the newly discovered halo clouds @xcite . about 15% of fvws unrelated to the known sources have nearby o - type stars . thus , these fvws could be due to the stellar winds from those stars . the origin of the remaining 55% of fvws unrelated to the known sources is uncertain . in a sense that the hi snrs are , at least , strong and fast enough to be detected as fvws , snrs still remain as a strong candidate of fvws . if fvws are indeed the oldest type of snr , that which is essentially invisible except via its hi line emission , then only high resolution hi observations might be able to reveal their nature . we are very grateful to jae - jun lee for his help with preparing fig . we would like to thank naomi melissa mcclure - griffiths for providing the sgps data . we would like to thank jay lockman , myung gyoon lee , and ho - seung hwang for their valuable discussions about other possibilities of fvws . we also thank to the anonymous referee for his / her comments that improved the presentation of this paper . this work was supported by the korea science and engineering foundation ( abrl 3345 - 20031017 ) . h . k. has been supported in part by the bk 21 program . alksnis , a. , balklavs , a. , dzervitis , u. , eglitis , i. , paupers , o. , & pundure , i. 2001 , baltic astronomy , vol . bally , j. , & stark , a. a. 1983 , , 266 , l61 bieging , j. h. 1990 , evolution of the interstellar medium , ed . l. blitz ( san francisco : astronomical society of the pacific ) , 137 blitz , l. , spergel , d. n. , teuben , p. j. , hartmann , d. , & burton , w. b. 1999 , , 514 , 818 boulva , f. 1999 , iappp , 76 , 84 braun , r. , & strom , r. g. 1986 , a&as , 63 , 345 brogan , c. l. , gelfand , j. d. , gaensler , b. m. , kassim , n. e. , & lazio , t. j. w. 2006 , , 639 , l25 burton , w. b. , braun , r. , walterbos , r. a.m. , & hoopes , c. g. 1999 , , 117 , 194 cappa , c. e. , & herbstmeier , u. 2000 , , 120 , 1963 cappa , c. e. , pineault , s. , arnal , e. m. , & cichowolski , s. 2002 , , 395 , 955 de heij , v. , braun , r. , & burton , w. b. 2002 , , 391 , 159 de heij , v. , braun , r. , & burton , w. b. 2002 , , 392 , 417 frst , e. , reich , w. , reich , p. , & reif , k. 1990 , a&as , 85 , 691 garmany , c. d. , conti , p. s. , & chiosi , c. 1982 , , 263 , 777 green d. a. , 2006 , ` a catalogue of galactic supernova remnants ( 2006 april version ) ' , astrophysics group , cavendish laboratory , cambridge , united kingdom hartmann , d. , & burton , w. b. 1997 , atlas of galactic neutral hydrogen , cambridge press kang , j .- h . , heiles , c. 2004 , asp conf . ser . 317 : milky way surveys : the structure and evolution of our galaxy , 317 , 38 knapp , g. r. , jorissen , a. , & young , k. 1997 , , 326 , 318 koo , b .- c . , & heiles , c. 1991 , , 382,204 koo , b .- c . , & kang , j .- h . 2004 , mnras , 349 , 983 koo , b .- c . , kang , j .- h . & mcclure - griffiths , n. m. 2004 , jounal of the korean astonomical society , 37 , 61 koo , b .- kang , j .- h . , & salter , c. j. 2006 , , 643 , l49 koo , b .- c . , & moon , d .- s . 1997 , , 475 , 194 koo , b .- c . , reach , w. t. , heiles , c. , fesen , r. a. , & shull , j. m. 1990 , , 364 , 178 lizano , s. , heiles , c. , rodriguez , l. f. , koo , b .- c . , shu , f. h. , hasegawa , t. , hayashi , s. , & mirabel , i. f. 1988 , , 328 , 763 lockman , f. j. 2002 , , 580 , l47 mcclure - griffiths , n. m. 2001 , the hi southern galactic plane survey , phd thesis milne , d. k. 1979 , australian journal of physics , 32 , 83 paladini , r. , burigana , c. , davies r. d. , maino , d. , bersanelli , m. , cappellini , b. , platania , p. , & smoot , g. 2003 , , 397 , 213 putman , m. e. et al . 2002 , , 123 , 873 reed , j. e. , hester , j. j. , fabian , a. c. , & winkler , p. f. 1995 , , 440 , 706 reich , w. , frst , e. , reich , p. , & reif , k. 1990 , a&as , 85 , 633 richter , p. , sembach , k. r. , & howk , j. c. 2003 , a&a , 405 , 1013 russell , a. p. g. , bally , j. , padman , r. , & hills , r. e. 1992 , , 387 , 219 sato , f. 1973 , pasj , 25 , 135 stanimirovi , s. , et al . 2006 , , 653 , 1210 stil , j. m. , et al . 2006 , , 637 , 366 tully , r. b. 1988 , cambridge and new york , cambridge university press , 1988 , 221 p. uyaniker , b. , landecker , t. l. , gray , a. d. , & kothes , r. 2003 , , 585 , 785 van der hucht , k. a. 2001 , new astronomy review , 45 , 135 wakker , b. p. 1991 , , 250 , 499 wakker , b. p. , & van woerden , h. 1991 , , 250 , 509 wakker , b. p. 2001 , , 136 , 463 weaver , r. , mccray , r. , castor , j. , shapiro , p. , & moore , r. 1977 , , 218 , 377

the late time accelerated expansion of the universe demands that even in local galactic - scales it is desirable to study astrophysical phenomena , particularly relativistic accretion related phenomena in massive galaxies or in galaxy mergers and the dynamics of the kiloparsecs - scale structure and beyond , in the local - galaxies in schwarzschild - de sitter ( sds ) background , rather than in schwarzschild or newtonian paradigm . owing to the complex and nonlinear character of the underlying magnetohydrodynamical equations in general relativistic ( gr ) regime , it is quite useful to have an newtonian analogous potential containing all the important gr features that allows to treat the problem in newtonian framework for study of accretion and its related processes . from the principle of conserved hamiltonian of the test particle motion , here , a three dimensional newtonian analogous potential has been obtained in spherical geometry corresponding to sds / schwarzschild anti - de sitter ( sads ) spacetime , that reproduces almost all of the gr features in its entirety with remarkable accuracy . the derived potential contains an explicit velocity dependent term of the test particle that renders an approximate relativistic modification of newtonian like potential . the complete orbital dynamics around sds geometry and the epicyclic frequency corresponding to sds metric have been extensively studied in the newtonian framework using the derived potential . applying the derived analogous potential it is found that the current accepted value of moderately influences both sonic radius as well as bondi accretion rate , especially for spherical accretion with smaller values of adiabatic constant and temperature , which might have interesting consequences on the stability of accretion disk in agns / radio galaxies .

1410.7132

the simplest and most attractive general relativistic model that explains the late time accelerated expansion of the universe [ 1 ] is the @xmath1cdm model where @xmath1 denotes the positive cosmological constant with a value of nearly @xmath2 [ 2 ] and cdm refers to cold dark matter . the @xmath1cdm model is more or less consistent with all the current cosmological observations [ 3 ] though the origin of cosmological constant still remains elusive [ 4 ] and is one of the most pressing problems in modern physics . in the presence of a repulsive cosmological constant ( positive ) ` @xmath1 ' the spacetime geometry exterior to a static spherically symmetric metric is schwarzschild - de sitter ( sds ) , which describes an isolated black hole ( bh ) in a spatially inflated universe , rather than schwarzschild metric . therefore , the cosmological constant may affect any local gravitational phenomenon like perihelion shift of the orbits of gravitationally bound systems [ 5 ] , gravitational bending of light [ 6 ] , geodetic precession [ 5 ] etc . , but the general perception is that owing to its tiny value , cosmological constant does not lead to any significant observable effects in a local gravitational phenomenon . in solar system the influence of cosmological constant is known to be maximum in the case of perihelion shift of mercury orbit where the @xmath1 contribution is about @xmath3 of the total shift [ 5 ] . however , the contribution of repulsive @xmath1 could be significant ( larger than the second order term ) even in a local gravitational phenomenon when kiloparsecs to megaparsecs - scale distances are involved , such as the gravitational bending of light by cluster of galaxies [ 7 ] . the study of particle dynamics in sds spacetime shows a significant contribution of @xmath1 when kiloparsecs - scale distance is involved [ 8,9 ] . the trajectories of both small and large magellanic clouds in the gravitational field of the milky way are affected significantly ( @xmath4 level or higher ) by @xmath1 [ 10 ] . cosmological constant also leads to non - spherical effects with non - negligible contributions on the local dynamics of clusters and superclusters [ 11 ] . some authors claimed [ 12 ] that the repulsive @xmath1 is responsible for significantly lower value ( @xmath5 ) of hubble parameter in our close neighborhood than its large - scale value ( @xmath6 ) . the cosmological constant also influences gravitational equilibrium [ 13 ] . a plausible local scenario where cosmological constant may contribute significantly is the relativistic accretion phenomena around massive bhs which involve distance - scale of the order of hundreds of parsecs or even more . this is because the asymptotic character of the bh spacetime changes significantly in the presence of cosmological constant [ 7 ] . note that the accretion of matter on to bh is believed to act as the engine for the active galactic nuclei ( agns ) . a few studies have been carried out so far to investigate the effect of @xmath1 in astrophysical jet / accretion flow paradigm [ 14 - 19 ] . it was found that the presence of cosmological constant leads to suppression of black hole evaporation [ 14 ] ; instead of evaporation black hole will accrete energy . the effect of @xmath1 on the dynamics of kiloparsecs to megaparsecs - scale astrophysical jets would be worth intriguing to investigate as suggested by stuchlk , slay and hledk [ 15 ] . their work indicated that positive @xmath1 can have strong collimation effect on jets . rezzolla et al . [ 16 ] have analyzed the effect of @xmath1 on the dynamical stability of a geometrically thick accretion disk with constant angular momentum around a sds bh . their work showed that @xmath1 introduces an outflow of matter through the outer cusp of disk causing a considerable impact on the runway instability . slobodov et al . [ 17 ] demonstrated that characteristic peaks in the iron line profile generated in the bh accretion disk in a sds background are less noticeable and distinguishable with the increase in the value of @xmath1 . in an interesting scenario , karkowski and malec [ 18 ] and mach , malec and karkowski [ 19 ] have recently explored the possible dependence of @xmath1 on relativistic bondi - type accretion flow around a nonrotating bh . they defined a quasilocal mass accretion rate and found that @xmath1 suppresses the mass accretion rate of the flow and dramatically impacts the transonic nature of the accretion flow . all the works on the effect of @xmath1 on accreting systems are carried out under some restricted conditions / situations . this is because the study of accreting bh systems involves solving general relativistic ( gr ) hydrodynamic / magnetohydrodynamic ( mhd ) equations in a strong gravitational field regime . owing to the complex and nonlinear character of the underlying equations in gr regime , analytical / quasi numerical treatment of the problem is virtually discarded . even numerical simulation is complicated by several issues such as different characteristic time scales for propagating modes of general relativity and relativistic hydrodynamics . several early works on these accretion related phenomena were based on pure newtonian gravity . a few gr effects were incorporated ad hocly . after the seminal work of paczyski and witta [ 20 ] , most of the authors treated accretion and its related processes around bhs using hydrodynamical / mhd equations in the newtonian framework by using some pnps which are essentially modification of newtonian gravitational potential developed with the objective to reproduce ( certain ) features of relativistic gravitation . this is to avoid gr gas dynamical equations , which in most occasions become inconceivable in practice in describing a complex physical system like accreting plasma . consequently , adopting pnps , one can comprehensively construct more realistic accretion flow models in simple newtonian paradigm , while the corresponding pnp would capture the essential gr effects in the vicinity of the compact objects . instead of pnps , some authors simply use parameterized post newtonian ( ppn ) expansion up to a certain order [ 21 ] . since the ppn expansion converges very slowly , the latter option is valid only for orbits at large distances , but not for particle trajectories in the vicinity of the bhs . several pnps exist in the astrophysical literature which are mostly prescribed either in an ad hoc manner or are devised employing certain explicit method [ 20,22 - 33 ] , mostly developed for schwarzschild and kerr bhs . a pnp corresponding to sds metric has also been prescribed [ 34 ] based on the method in [ 30 - 32 ] for a keplerian rotation flow which can be more appropriately used to study gas dynamical properties of accretion disk , and thereby concentrating mostly in its use in accretion tori [ 35 ] , though it has also been recently used to investigate the influence of the repulsive cosmological constant on the kinematics of small and large magellanic cloud [ 10 ] . although notionally pnps are aimed at mimicking gr geometries , however , they , in general , are devoid of their uniqueness to effectively describe the gr features in its entirety . to replicate gr features , pnps lay emphasize mostly to reproduce marginally stable and marginally bound orbits . moreover , in general , they are unable to reproduce observationally verified tests in general relativity like geodetic precession , gravitational bending of light or gravitational time delay . if the effects of cosmological constant need to be properly revealed in different astrophysical phenomena in local galactic - scales circumventing the complex gr treatment , it is desirable to have a correct newtonian analogous like potential corresponding to sds geometry that will reproduce all the salient features of the sds metric with reasonable accuracy and extensively mimic a wide spectrum of gr behavior . recently , tejeda and rosswog [ 36 ] derived a generalized effective potential for a schwarzschild bh based on a proper axiomatic procedure . their generalized potential , which has an explicit dependence on radial velocity and orbital angular velocity of test particle , reproduces exactly several relativistic features of corresponding schwarzschild geometry . later , the work has been extended for the kerr bh [ 37 ] . in the present work , we would derive a modified newtonian like potential from the conserved hamiltonian of the test particle motion [ 36 ] , which would then be an approximate relativistic potential analogue corresponding to the sds metric . we would show from a detailed investigation that the derived potential , which depends on the velocity of the test particle , reproduces almost all of the corresponding gr features with remarkable accuracy . it would be worthy enough to explore the feasibility of measurement of cosmological constant in local - scales or in galactic - scales , and further to explore its influence on the dynamics on kiloparsecs - scale structure and beyond in the local - galaxies , or the accretion related scenarios / conditions where cosmological constant may contribute significantly , in which case , the proposed modified newtonian like potential analogue to sds geometry considering it as a background metric , would be of remarkable use . the plan of the paper is the following . in the next section , we would formulate the approximate relativistic potential analogue in the newtonian framework corresponding to both sds and schwarzschild anti - de sitter ( sads ) spacetimes simultaneously , starting from the conserved hamiltonian of the system in the low energy limit of the test particle motion [ 36 ] . subsequently in , we evaluate the geodesic equations of motion and extensively analyze their solutions corresponding to sds geometry , with both our analogous potential and general relativity , for the current accepted value of @xmath0 . in we explicitly investigate the influence of repulsive @xmath1 on bondi accretion rate . finally , we end in with comments and discussion . for a general class of static spherically symmetric spacetimes of the form ( in the standard coordinates system ) @xmath7 where , @xmath8 and f(r ) is the generic metric function , the lagrangian density of a particle of mass @xmath9 is given by @xmath10 from the symmetries , one obtains two constants of motion corresponding to two ignorable coordinates @xmath11 and @xmath12 as given by @xmath13 and @xmath14 where , @xmath15 and @xmath16 are specific energy and generalized specific angular momentum of the orbiting particle , respectively . here , @xmath17 and @xmath18 represent the derivatives of ` @xmath11 ' and ` @xmath19 ' with respect to proper time @xmath20 . it needs to be mentioned that from now onwards , throughout the paper , the terms related to momentum , energy / hamiltonian , potential and frequency , all of which are in fact their specific quantities , would be addressed without the using of word ` specific ' . using eq . ( 3 ) we then can write @xmath21 using @xmath22 and substituting eqs . ( 3 ) and ( 4 ) in eq . ( 2 ) , we obtain @xmath23 - f(r ) \ , \frac{\lambda^2}{r^2 } \ , . \label{6}\end{aligned}\ ] ] by considering a locally inertial frame for a test particle motion , we write @xmath24 ( ` @xmath25 ' symbolizes ` gr - newtonian ' ) . second term in the above definition of @xmath26 is the rest mass energy of the particle which is subtracted from relativistic energy owing to low energy limit . from eqs . ( 5 ) and ( 6 ) and using eq . ( 4 ) , we get @xmath27 - f(r ) \,{\dot \omega}^2 \frac{r^2}{f(r ) } } \ , , \label{7}\end{aligned}\ ] ] where , @xmath28 is the derivative with respect to coordinate time ` @xmath11 ' . using the condition for low energy limit @xmath29 , as we prefer to , @xmath26 is given by @xmath30 \ , . \label{8}\end{aligned}\ ] ] in the asymptotic non - relativistic limit @xmath26 reduces to the newtonian mechanical energy (= hamiltonian of the motion ) . the generalized hamiltonian @xmath26 in the low energy limit should then be equivalent to the hamiltonian in newtonian regime . the hamiltonian in the newtonian regime with the generalized analogous potential in spherical polar geometry will then be equivalent to @xmath26 in eq . ( 8) . thus @xmath31 where , @xmath32 is the non - relativistic kinetic energy of the test particle . @xmath33 is the derivative with respect to coordinate time ` @xmath11 ' . @xmath34 is the analogous potential which would then be given by @xmath35}{2}- \frac{[1-f(r)}{2f(r ) } \left [ \frac{1+f(r)}{f(r ) } \dot r^{2 } + r^{2 } \ , { \dot \omega}^2 \right ] \ , . \label{10}\end{aligned}\ ] ] @xmath34 , thus , is the generalized three dimensional potential in spherical geometry in newtonian analogue , corresponding to any generalized static gr metric given in eq . ( 1 ) , with test particle motion in the low energy limit . note that , the first term on the right hand side of the potential contains the explicit information of the source . for a purely spherically symmetric gravitational mass with zero charge and without any external effects , the classical newtonian gravitational potential @xmath36 will be recovered from this term . the second term is the explicit velocity dependent and contains the information of the test particle motion , thus contributing to the modification of newtonian gravity . for sds / sads metric the metric function @xmath37 , where @xmath1 is the cosmological constant . @xmath38 represents the sds metric with spatially inflated universe , where @xmath39 represents a sads metric corresponding to negative vacuum energy density . @xmath40 . from eq . ( 10 ) we then obtain the three dimensional generalized newtonian analogous potential in spherical geometry , corresponding to sds / sads geometry in the low energy limit , which is given by @xmath41 where , subscript ` ds / ads ' denotes schwarzschild de - sitter / anti - de sitter . with @xmath42 , the above potential reduces to the potential corresponding to the simplest static schwarzschild geometry , given in [ 36 ] . @xmath43 is the mass of the bh / central object . the denominator in the second term of the potential contains the exact metric function @xmath44 of sds / sads geometry , and hence the potential @xmath45 would reproduce the exact location of the event horizon and cosmological horizon and other horizon properties , as that in full general relativity . introducing a dimensional parameter or cosmological parameter @xmath46 , the vanishing cubic polynomial @xmath44 with repulsive cosmological constant ( @xmath38 ) would give two real positive roots representing the locations of two horizons , namely the bh horizon and the cosmological horizon . the locations of these two horizons are then given by @xmath47 $ ] and @xmath48 $ ] , respectively . in the next section , we will analyze different aspects of particle dynamics in the gravitational field of sds geometry in the modified newtonian framework . in the newtonian framework , the lagrangian of a particle in the presence of the sds analogous potential @xmath49 per unit mass is given by , @xmath50 \ , + \ , \frac{gm}{r } + \frac{\lambda \ , c^2 r^2}{6 } \ , , \label{12}\end{aligned}\ ] ] where , @xmath51 . here overdots denote the derivative with respect to coordinate time ` @xmath11 ' . we then compute the conserved angular momentum and hamiltonian using @xmath49 , given by @xmath52 and @xmath53 \ , - \ , \frac{gm}{r } - \frac{\lambda \ , c^2 r^2}{6 } \ , , \label{14}\end{aligned}\ ] ] respectively . using eqs . ( 13 ) and ( 14 ) , we obtain @xmath33 that uniquely describes the test particle motion , which is given by @xmath54 which is exactly equivalent to @xmath33 in general relativity in low energy limit . replacing @xmath28 and @xmath33 in eq . ( 11 ) using eqs . ( 13 ) and ( 15 ) respectively , sds analogous potential can be written in terms of conserved hamiltonian @xmath55 and angular momentum @xmath56 , given by @xmath57 \nonumber \\ - \left(2r_s + \frac{\lambda \ , r^3}{3}\right ) \left[\frac{1}{r^2 } \left(r - r_s - \frac{\lambda \ , r^3}{6}\right ) \left(2 e_{\rm ds / ads } + \frac{2gm}{r } + \frac{\lambda \ , c^{2 } r^2}{3 } \right ) \right ] \label{16}\end{aligned}\ ] ] in fig . 1a we show the variation of @xmath49 in the form given in eq . ( 16 ) with the radial distance @xmath58 , corresponding to @xmath38 with @xmath59 in the low energy limit of test particle motion . the stated value of @xmath60 corresponds to @xmath61 for @xmath62 . unless stated explicitly , we shall use such a combination of @xmath1 and @xmath63 throughout the paper . the profile of @xmath49 clearly shows both the bh event horizon as well as the cosmological horizon . with @xmath64 , cosmological horizon extends up to @xmath65 . for a bh of @xmath66 , it gives a radius of @xmath67 . figure 1b resembles that of 1a but for semi - relativistic test particle energy ( @xmath68 ) . although the locations of bh event horizon as well as cosmological horizon remain unaltered with the increase in test particle energy , nevertheless , there is a noticeable change in the magnitude of @xmath49 at both horizon radii . it is found that @xmath56 has no effect on the nature of potential just beyond @xmath69 . these figures indicate that with the increase in the value @xmath56 , inner horizon shifts to larger radii . figures 1c and 1d show the variation of effective potential @xmath70 for the same value of @xmath1 as in figures 1a , b , corresponding to sds spacetime . the nature of the profiles resemble that in figures 1a , b , however @xmath71 attains a higher peak as compared to @xmath49 in the vicinity of inner horizon for values of @xmath72 . the figure 1c shows the variation of ` gr effective potential ' [ see eq . ( 30 ) ] for the same values of angular momentum . ( 16 ) ] with radial distance @xmath58 . the solid and long - dashed lines in figure 1a correspond to @xmath49 in low energy limit @xmath73 for sds geometry with @xmath74 with @xmath75 , respectively . the short - dashed curve represents schwarzschild spacetime ( @xmath42 ) . figure 1b is for semi - relativistic energy @xmath68 , otherwise resemble 1a . figures 1c and 1d correspond to effective potential @xmath71 , with other parameters same to those in figures 1a and 1b , respectively . figure 1e corresponds to ` gr effective potential ' for sds geometry with the same physical parameters as those in figures 1c , d , but independent of conserved energy . @xmath55 and @xmath56 are expressed respectively in units of @xmath76 and @xmath77 . ] in fig . 2a , we depict the difference in the magnitude between @xmath49 in eq . ( 16 ) and the analogous potential for schwarzschild metric in [ 36 ] @xmath78 , with radial distance @xmath58 . for semi - relativistic energy of test particle , the magnitude of @xmath49 may even exceed the corresponding schwarzschild value by @xmath79 at the outer radii near the cosmological horizon . in fig . 2b , we show the relative deviation ( in percentage ) between the @xmath49 and @xmath80 , as a function of radial distance @xmath58 . for any arbritary physical quantity @xmath81 , the relative deviation @xmath82 is defined as @xmath83 which essentially implies deviation between the analogous potential for sds and schwarzschild geometry relative to their average value . the subscript ` sw ' represents corresponding quantities in ` schwarzschild ' geometry . it is found from the fig . 2b that @xmath84 sharply increases with @xmath58 , especially , at the outer radii . at smaller distances @xmath84 is very small , and detecting such small deviations experimentally does not seem to be possible at present or in near future . at what distance @xmath84 can be said as non - negligible depends entirely on experimental accuracy . the solar system experiments presently are capable to detect any deviation from general relativity at @xmath85 level . if we demand a similar experimental accuracy , we can then say that when @xmath84 attains a value of @xmath86 ( or equivalently @xmath87 the relative deviation may be considered to be non - negligible with the corresponding distance @xmath88 . with this consideration , this then implies that beyond such radius the influence of @xmath1 can not be neglected . at @xmath89 , the relative deviation becomes substantial with @xmath90 . figure 2b also shows that @xmath84 increases rapidly up to a certain radius @xmath91 , beyond which it mostly remains constant . around this particular radius , the @xmath1 effects become too dominating , which renders @xmath80 to become negligible in compare to that of @xmath49 . we describe this radius @xmath91 as some upper bound @xmath92 , where @xmath84 attains a values of @xmath93 . for @xmath62 , in accordance with the bh mass in many agns / quasars or in massive galaxies , the lower bound in @xmath94 , gives a radius of @xmath95 kiloparsec , whereas the upper bound @xmath96 gives a radius of @xmath97 kiloparsec . thus , the region between @xmath98 and @xmath96 or a region approximately existing between few kiloparsecs to a few @xmath99 kiloparsecs would be strongly affected by cosmological constant @xmath1 , thereby , directly influencing the kiloparsecs - scale structure in massive galaxies , in the local observable universe . with a similar bh mass , the region beyond @xmath100 kiloparsec would have most significant @xmath1 effects , where @xmath84 would be greater than @xmath4 . ( @xmath59 ) and @xmath80 with @xmath58 . solid and short - dashed curves in fig . 2a show the difference between @xmath49 and @xmath80 , corresponding to @xmath73 and @xmath68 , respectively . solid curve in fig . 2b shows the variation of the relative deviation ( in percentage ) @xmath84 with @xmath58 , for both @xmath73 and @xmath101 ( which coincide ) . the two extreme vertical dashed lines , denoted as @xmath98 and @xmath96 ( in units of @xmath102 , correspond to the value of @xmath84 @xmath103 and @xmath104 , respectively ) . the vertical dashed line in the middle represents @xmath58 at which @xmath105 . @xmath55 is expressed in units of @xmath76 . note that @xmath56 has no effect on the nature of potentials when @xmath106 . ] next we obtain the equation of the orbital trajectory using eqs . ( 13 ) and ( 15 ) , as given by @xmath107 \ , . \nonumber \\ \label{17}\end{aligned}\ ] ] although eq . ( 17 ) is derived with the condition of low energy limit , yet it is exactly the same as that in full general relativity which one can easily obtain using eqs . ( 4 ) and ( 7 ) with relevant @xmath44 . to furnish a complete behavior of the particle dynamics in sds background , we obtain the equations of motion of the test particle in the presence of @xmath49 from the euler - lagrange equations in spherical geometry which are given by @xmath108 @xmath109 and @xmath110 respectively . @xmath111 and @xmath112 equations are exactly the same to that in general relativity , whereas @xmath113 equation in ( 18 ) corresponds to that in general relativity in the low energy limit . the corresponding @xmath113 equation in general relativity is given by @xmath114 it needs to be mentioned that previously a pnp has been prescribed in [ 34 ] corresponding to sds geometry based on a method adopted by [ 30 - 32 ] , which has been formulated by considering a keplerian rotation profile of the test particle motion . the form of the pnp is given by @xmath115 \left(2-r+ r^3 \frac{\lambda}{3}\right ) } \ , , \label{22}\end{aligned}\ ] ] which is derived on the premise that @xmath116 at the static radius , preserving the analogy that the gravitational potential tends to zero in asymptotically flat spacetime . this potential does not have any dependence on test particle velocity . with @xmath42 , the pnp in eq . ( 22 ) reduces to that of paczyski - witta potential corresponding to schwarzschild geometry . the corresponding gr behavior is mimicked through this pnp by intending only to reproduce the marginally stable and bound orbits for circular orbital trajectory . this is in sharp contrast with the velocity dependent potential @xmath49 , which , a priori , focused on to replicate the general relativity by resembling the geodesic equations of motion . this ensures that most of the gr features could be reproduced accurately . the equation of the orbital trajectory and the equations of motion for the corresponding pnp in eq . ( 22 ) in spherical geometry are then given by @xmath117\left(2-r+ r^3 \frac{\lambda}{3}\right ) } - \frac{\lambda^2}{r^3 } \right ] \ , , \label{23}\end{aligned}\ ] ] @xmath118 @xmath119 and @xmath120 it is seen that the trajectory equation as well as the equations of motion corresponding to the pnp in eq . ( 22 ) do not at all resemble the equivalent relations in general relativity and hence this pnp can not reproduce the features of sds spacetime accurately . notwithstanding , in the next few sections we would compare the dynamical behavior of the test particle motion obtained with @xmath49 and the pnp in eq . ( 22 ) as well as with full general relativity , over the entire spatial regime relevant for sds background . in order to compare the behavior of the particle motion in presence of @xmath49 and those in general relativity , we compute the dynamical variables for the simplest circular orbit trajectory . with the conditions for the circular orbits @xmath121 and @xmath122 , we obtain corresponding angular momentum @xmath123 , hamiltonian @xmath124 and the orbital angular velocity @xmath125 with @xmath49 using eqs . ( 13 ) , ( 15 ) and ( 18 ) , given by @xmath126 @xmath127 \ , \left(\frac{\lambda r^2}{3 } + \frac{4r_s}{r } - 2 \right ) \ , \label{28}\end{aligned}\ ] ] and @xmath128 respectively . with @xmath42 , all the above equations get reduced to that obtained in schwarzschild geometry . to compute these variables in exact sds geometry we use the corresponding ` gr effective potential ' , given by @xmath129 as usual , circular orbits occur in general relativity when @xmath130 and @xmath131 . we obtain energy @xmath15 for particle motion in circular orbit , given by @xmath132 angular momentum ( @xmath133 ) and the equivalent hamiltonian @xmath134 @xmath135 $ ] for circular orbits in general relativity then exactly resemble the corresponding values obtained with @xmath49 , given by eqs . ( 27 ) and ( 28 ) respectively . the orbital angular velocity in general relativity is then given by @xmath136 whose analytical expression is not exactly equivalent to @xmath125 in eq . note that the circular orbit corresponding to sds metric exists down to @xmath137 in similarity to that in schwarzschild geometry which represents the null hypersurface . the photon orbit is thus independent of the cosmological constant . for pnp in eq . ( 22 ) , the corresponding expressions of @xmath138 , @xmath134 and @xmath139 are given in [ 34 ] . the said pnp does not reproduce the correct photon orbit . in fig . 3 we show the appropriate comparison of the nature of potential @xmath49 in eq . ( 11 ) and that of pnp in eq . ( 22 ) , corresponding to test particle motion in circular orbit . in the very inner and outer regions near to both the bh and cosmological horizons , the behavior of these potentials differ significantly ( figures 3c , d ) . one of the major distinctive features of the pnp in eq . ( 22 ) is that it becomes zero at the static radius @xmath140 , whereas @xmath49 attains a value of @xmath141 ( fig . 3b ) . in the intermediate region , the nature of the potentials remain mostly similar . it is seen that at @xmath142 , where the effect of @xmath1 is prominent , the pnp in eq . ( 22 ) differs significantly as compared to @xmath49 . this radius approximately resembles @xmath98 in fig . 2b . and the pnp in eq . ( 22 ) with @xmath58 corresponding to circular motion trajectory for @xmath143 . the solid and short - dashed lines correspond to @xmath49 in eq . ( 11 ) and pnp in eq . ( 22 ) , respectively . figure 3a covers the entire distance , figures 3b , c focus only on the outer radii ; near to the static radius and the cosmological horizon , respectively , whereas figure 3d focuses on the inner radii near the bh event horizon . ] in fig . 4 , we furnish the variation of @xmath123 corresponding to @xmath49 which exactly coincides with that in general relativity , for the entire spatial regime . in fig . 4b , we compare the nature of @xmath123 with that corresponding to the pnp in eq . ( 22 ) at the outer radii . the figures show that for sds spacetime , there is a clear static radius at the outer radii where angular momentum abruptly falls to zero value . for bh of @xmath66 , with @xmath61 , the static radius will be located at @xmath144 . figure 4b shows that the angular momentum for circular orbit trajectory corresponding to @xmath49 as well as for the pnp in eq . ( 22 ) behaves quite similarly at the outer radii , and coincides near the static radius . however , due to the profound effect of @xmath1 on the outer radii , the nature of @xmath123 at the outer radii differs significantly from that corresponding to @xmath80 , especially beyond @xmath145 . in the inner radii , where the effect of @xmath1 is negligible , @xmath123 show appreciable deviation from the angular momentum profile corresponding to the pnp in eq . ( 22 ) . for a comparison of the profiles of different pnps in the inner radii , the readers should see [ 36 ] . resembling fig . 2b , in fig . 5 , we show the relative deviation @xmath146 in angular momentum as a function of radial distance @xmath58 , showing approximately similar behavior as that in fig . for this profile ( fig . 5 ) , we found that the value of @xmath147 and the value of @xmath148 , corresponding to the value of @xmath84 @xmath149 and @xmath93 , respectively . at @xmath96 , which is also the static radius , the curve gets truncated . at @xmath150 , where the effect of @xmath1 is significant , large relative deviation ( @xmath151 ) in angular momentum profile is noticed . for circular orbit trajectory . the short - dashed curve in fig . 4a corresponds to @xmath49 with @xmath59 that coincides with the corresponding gr results . the solid curves in figures 4a and 4b are due to @xmath80 . the long - dashed line in fig . 4b corresponds to @xmath49 which coincides with the angular momentum profile corresponding to the pnp given in eq . ( 22 ) . ] and @xmath96 , correspond to value of @xmath84 @xmath103 and @xmath93 , respectively . the curve is truncated at @xmath96 which also corresponds to the static radius . the vertical dashed line in the middle represents @xmath58 at which @xmath84 @xmath4 . ] figure 6 show the variation of conserved hamiltonian @xmath124 with @xmath58 , which is compared with the hamiltonian corresponding to the pnp in eq . ( 22 ) and the hamiltonian corresponding to @xmath80 . it needs to be noted that @xmath124 corresponding to @xmath49 exactly resembles the corresponding expression in general relativity . in the outer region of the test particle motion in circular orbit , where the effect of repulsive @xmath1 is prominent ( figures 6b , c ) , there is a marked difference between the profile of conserved hamiltonian corresponding to @xmath49 and that corresponding to the pnp in eq . ( 22 ) , especially at @xmath152 . this radius @xmath153 approximately resembles the location @xmath98 , the lower bound in @xmath58 , at which the relative deviation @xmath146 in the hamiltonian attains the value of @xmath103 ( see fig . in fact , the value of @xmath98 in this case is @xmath154 . around static radius @xmath155 , the hamiltonian corresponding to the pnp in eq . ( 22 ) attains zero value ( fig . 6b ) , unlike the case of @xmath124 . in the vicinity of cosmological horizon @xmath156 , the hamiltonian corresponding to the pnp in eq . ( 22 ) diverges , showing a contrasting behavior with respect to the profile of @xmath124 . @xmath124 after attaining a certain negative value becomes zero at a large outer radius , which is the outermost bound orbit ( fig . 6c ) . due to the substantial effect of @xmath1 , @xmath124 deviates significantly from that corresponding to @xmath80 at @xmath157 ( fig . 6b ) , which precisely resembles the radius at which @xmath84 attains the value of @xmath4 ( fig . resembling fig . 2b , here too , @xmath96 is located at that radius where @xmath84 attains a value of @xmath93 . the value of @xmath96 is then @xmath158 . in the inner radii of the test particle in the circular orbit , where the effect of @xmath1 is negligible , @xmath124 shows appreciable deviation from the hamiltonian profile corresponding to the pnp in eq . we do not show the corresponding profiles separately in this region , as already been furnished in [ 36 ] . for test particle in circular orbit corresponding to @xmath59 . the solid and long - dashed lines in fig . 6a are for the hamiltonian corresponding to @xmath80 and @xmath49 , respectively . the profile of @xmath124 corresponding to @xmath49 coincides with that in general relativity . the solid , long - dashed and short - dashed lines in figures 6b , c are for the hamiltonian corresponding to @xmath80 , @xmath49 and the pnp in eq . ( 22 ) , respectively . figures 6b , c focus on the outer radii ; near to the static radius and the cosmological horizon , respectively . 6d resembles fig . 2b and fig . 5 , however , showing the the relative deviation in hamiltonian ( in percentage ) , as a function of radial distance @xmath58 . ] in fig . 7a , instead of the profile of orbital angular velocity @xmath125 , we depict the profile of corresponding angular frequency @xmath159 in @xmath58 , which truncates at the static radius at @xmath160 . at outer radii where the effect of @xmath1 is quite substantial , the value of angular frequency of the particle in circular orbit is negligible . owing to which , @xmath161 , its corresponding gr counterpart , angular frequency corresponding to @xmath80 and the angular frequency corresponding to the pnp in eq . ( 22 ) , all coincide . hence , we do not show the angular frequency profile of the test particle motion in the circular orbit separately , at the outer radii , as well the corresponding relative deviation profile . however , in the inner radii , where the effect of @xmath1 is negligible , @xmath161 deviates from the corresponding gr result by a maximum of @xmath162 . moreover , the angular frequency profile of the particle in the circular orbit corresponding to the pnp in eq . ( 22 ) shows a huge deviation from the profile of @xmath161 , as well as from the corresponding gr result , with an error margin of @xmath163 . the corresponding orbital angular velocity profiles in the inner radii have already been shown in [ 36 ] . for the test particle in circular orbit . solid , long - dashed and short - dashed curves in fig . 7a depict angular frequency profiles corresponding to @xmath49 with @xmath143 , the gr counterpart , and the pnp in eq . ( 22 ) respectively . note that @xmath161 coincides with the angular frequency profile corresponding to @xmath80 , both in the inner as well as at the outer radii . figure 7b depicts the variation of epicyclic frequency with @xmath58 , corresponding to @xmath49 ] as discussed in the previous subsection about the importance of the repulsive cosmological constant on the dynamical behavior of the particle orbits and their stability at the outer radii near the static radius , it is necessary for us to investigate the perturbative effects on the orbital dynamics , which would indeed be affected by the positive cosmological constant around the cosmological horizon . the perturbative effects would plausibly have substantial influence on the accretion flow stability in the outer regions in sds background , in the expanding universe . we follow our analysis by computing the epicyclic frequency for small perturbation of the particle orbit in circular trajectory . it should be reminded that till date there is no exclusive analytical relation of epicyclic frequency in full general relativity corresponding to sds spacetime , unlike that in the case of schwarzschild and kerr bhs [ 38 ] . henceforth , we evaluate the relation for epicyclic frequency using our correct sds analogous potential @xmath49 . we restrict ourselves in the equatorial plane of test particle orbit in a circular trajectory . we then perturb @xmath58 and @xmath19 and their derivatives accordingly : @xmath164 and @xmath165 respectively . using eqs . ( 18 ) , ( 19 ) and ( 20 ) , the linearized perturbed equations of motion are then obtained as @xmath166 \right ] \ , + 2 { \dot \phi } \ , { \vert}_c \ , ( r - 3r_s ) \delta { \dot{\phi}}\ , , \label{33}\end{aligned}\ ] ] @xmath167 and @xmath168 respectively . here , @xmath169 is the orbital angular velocity of the particle motion in the circular orbit in equatorial plane . we concentrate ourselves in computing only the radial epicyclic frequency as it is more significant . using eqs . ( 33 ) and ( 34 ) , with the following relations of perturbed quantities @xmath170 and @xmath171 for harmonic oscillations , where @xmath172 is the radial epicyclic frequency and @xmath173 and @xmath174 are amplitudes ( see [ 38 ] ) , we eventually obtain the expression for radial epicyclic frequency @xmath172 corresponding to @xmath49 after rigorous algebra , given by @xmath175 \right ] } \ , . \label{36}\end{aligned}\ ] ] the relation in eq . ( 36 ) reduces to that in schwarzschild geometry with @xmath42 whose value lies within a maximum error of just @xmath176 , as compared to the gr result . although the radial epicyclic frequency @xmath172 has been derived with sds analogous potential , we anticipate that this relation could also be accurately used in full gr framework , as the effect of @xmath1 can only be realized at large outer radii from the central object , where the relativistic effect will itself be diminishing . as no other effective expressions for epicyclic frequency corresponding to sds geometry is known , in fig . 7b , we only show the variation of epicyclic frequency corresponding to @xmath49 with @xmath58 . resembling @xmath161 , the value of @xmath172 at the outer radii is negligible . consequently , the profile of epicyclic frequency corresponding to @xmath49 coincides with that corresponding to @xmath80 , both in the inner as well as at the outer radii , and hence the corresponding relative deviation would also be negligible . in the inner radii , where the effect of @xmath1 is negligible , the pnp in eq . ( 22 ) reduces to paczyski - witta potential . the epicyclic frequency profile corresponding to this potential shows a huge deviation from the profile corresponding to @xmath49 by over an error margin of @xmath163 ( see [ 36 ] ) . before analyzing the dynamics of orbital trajectory , we display the profile of the radial velocity @xmath177 ( see fig . 8) of the test particle falling radially towards the central bh in the sds background . the expression of @xmath177 , here , solely depends on the energy @xmath55 of the test particle motion . in figures 8a , b we show the profiles of @xmath177 corresponding to @xmath49 , @xmath80 , and the pnp in eq . ( 22 ) , for the test particle with @xmath73 and @xmath68 , respectively . expression for @xmath177 corresponding to @xmath49 coincides with the gr expression . in figures 8c , d , we depict the detailed nature of the corresponding profiles at the outer radii which is our region of interest , where the effect of @xmath1 is significant . the figures show that @xmath177 profiles corresponding to the pnp in eq . ( 22 ) distinctly differ ; rather show opposite behavior from @xmath177 corresponding to @xmath49 at the outer radii , at @xmath178 . for semi - relativistic / relativistic test particle energy , the radial velocity corresponding to the pnp in eq . ( 22 ) even surpasses velocity of light at that radii , thus showing physically inconsistent behavior ( figures 8b , d ) . in figures 8e , f , we show the profiles of relative deviation @xmath179 in @xmath177 , corresponding to @xmath73 and @xmath68 , respectively . @xmath98 and @xmath96 corresponding to @xmath177 for @xmath73 are located at @xmath180 and @xmath181 , at which @xmath84 attains the value of @xmath103 and @xmath93 , respectively . however , the relative deviation corresponding to semi - relativistic / relativistic test particle energy ( fig . 8f ) show marked difference as compared to test particle motion in the low energy limit , unlike that in fig . for the semi - relativistic / relativistic energy , @xmath177 corresponding to @xmath49 deviates from that corresponding to @xmath80 at much outer radii . @xmath98 and @xmath96 thus move further outward attaining the value of @xmath182 and @xmath183 , where @xmath84 is @xmath103 and @xmath93 , respectively . in fig . 8f , the curve gets truncated at @xmath96 which is also the cosmological horizon . the dynamics of orbital trajectory can be obtained from eq . ( 17 ) , which is identical to that obtained in general relativity , implying that the nature of trajectory of particle orbit obtained with @xmath49 would be similar to that in general relativity . in fig . 9 we show the profile of @xmath184 with @xmath58 corresponding to @xmath59 . the profiles indicate that for semi - relativistic / relativistic test particle energy , @xmath184 corresponding to @xmath49 deviate from @xmath184 corresponding to @xmath80 at much outer radii ( figures 9b , d ) , as compared to test particle energy in the low energy limit ( figures 9a , c ) . @xmath184 profiles corresponding to pnp in eq . ( 22 ) also show different behavior , from that corresponding to @xmath49 . corresponding to @xmath49 with @xmath59 , @xmath80 and pnp in eq . ( 22 ) , respectively , for test particle with @xmath73 . figure 8b is similar to that in 8a , however , for @xmath68 . in figures 8c , d which resemble figures 8a , b , the detailed nature of the corresponding profiles at the outer radii is depicted . in figures 8e , f we show the relative deviation ( in percentage ) in @xmath177 , corresponding to @xmath73 and @xmath68 , respectively . the vertical lines in the figures are same to those in fig . @xmath55 is expressed in units of @xmath76 . ] in radial direction @xmath58 corresponding to @xmath59 . the solid , long - dashed and short - dashed curves in all the figures correspond to @xmath49 , pnp in eq . ( 22 ) , and @xmath80 . figures 9a and 9c are for @xmath73 corresponding to angular momentum @xmath185 , respectively . figures 9b and 9d resemble figures 9a , c , however for @xmath68 . @xmath55 and @xmath56 are expressed in units of @xmath76 and @xmath77 , respectively . ] with @xmath59 . in fig . 10a , the solid line is due to @xmath80 and the dotted line for @xmath49 . the particle trajectory coincides with that in general relativity . figure 10b shows the profile of perihelion advancement @xmath186 as a function of eccentricity corresponding to @xmath49 for @xmath187 . the solid and long - dashed curves correspond to @xmath80 and @xmath49 respectively . ] in fig . 10a , the elliptical - like trajectory of particle orbit in the @xmath188 plane due to @xmath49 , along with that due to @xmath80 , obtained from the equations of motion are shown . the particle starts with a tangential initial velocity of @xmath189 from an apoapsis @xmath187 attaining a constant eccentricity of magnitude @xmath190 . we obtained the plot of elliptical trajectory using cartesian transformation adopting the method of euler - cromer algorithm which preserves energy conservation . the distinct precessional effect can be seen around @xmath191 . this is precisely the radius at which the @xmath84 corresponding to the potential ( fig . 2b ) becomes greater than @xmath4 . in fig . 10b , we show the apsidal precession or the perihelion advancement @xmath186 [ given by eq . ( 37 ) ] as a function of eccentricity which is identical to the gr results . this also guarantees that the photon orbit trajectory obtained with @xmath49 would resemble that in general relativity , which is independent of the cosmological constant . @xmath192 where @xmath193 is the usual half orbital period of the test particle . @xmath194 and @xmath195 are periapsis and apoapsis of the orbit respectively . we can obtain the last stable circular orbit using potential @xmath49 with the condition @xmath196 or equivalently @xmath197 which is exactly similar to that obtained in full general relativity . thus , @xmath49 would accurately reproduce the last stable circular orbit in general relativity . also the last bound circular orbit is obtained from eq . ( 28 ) with @xmath198 , and by virtue , is similar to that in full general relativity . the equivalent relation is @xmath199 for cosmological parameter @xmath200 , the familiar circular orbit stability limit and the last bound circular orbit for the schwarzschild metric is recovered . unlike the case in the schwarzschild geometry , circular orbits here become unstable at both inner and outer radii . equation ( 38 ) gives two real positive roots . one of the root is the last stable circular orbit in the inner radii resembling the usual schwarzschild case . the other root is the maximum possible circular stable orbit @xmath201 at the outer radii which is located near to the static radius . the location of @xmath201 is @xmath202 , whereas the corresponding static radius is located at @xmath160 . resembling the case for stable circular orbit , eq . ( 39 ) gives two real positive roots , corresponding to the marginally bound circular orbit in the inner radii and the maximum possible bound circular orbit @xmath203 at the outer radii . inner marginally bound circular orbit is the usual limit corresponding to schwarzschild geometry . pnp in eq . ( 22 ) , too , accurately reproduces both marginally stable and bound orbits corresponding to sds geometry . however , it needs to be noted that the test particle in sads background with negative @xmath1 does not yield any outermost stable or outermost bound circular orbit . although we have studied the entire test particle dynamics with @xmath204 for bh of @xmath66 , however , in fig . 11 , we will depict the variation of @xmath201 and @xmath203 with @xmath60 , and the corresponding values of the dynamical variables for the test particle motion in circular orbit , in order to see the effect of @xmath1 on the stability of the orbits . with the decrease in value of @xmath60 , both @xmath201 and @xmath203 move further outwards . for all values of @xmath60 , @xmath203 is located at much outer radii as compared to that of @xmath201 . we show the variation of @xmath205 and @xmath161 along @xmath201 corresponding to various values of @xmath60 in figures 11b , c , d , respectively , which are identical to the gr results , as well as corresponding to the pnp in eq . ( 22 ) . on the other hand , the corresponding dynamical variable do not furnish any real values along @xmath203 . on the contrary , the values of the dynamical variables corresponding to @xmath49 in the innermost stable and innermost bound orbits where the effect of @xmath1 is negligible , differ considerably from that corresponding to the pnp in eq . ( 22 ) , as discussed in [ 36 ] . , obtained with @xmath206 . the solid and dashed lines in fig . 11a show the variation of @xmath201 and @xmath203 with @xmath1 , respectively . figures 11b , c , d show the variation of hamiltonian , angular momentum and angular frequency along @xmath201 with corresponding values of @xmath1 , obtained with @xmath49 . the y - axis in fig . 11a is in logarithmic - scale . ] for supermassive bhs ( smbhs ) in agns or quasars with @xmath207 , the current accepted value of @xmath1 yields a maximum possible radius of the outermost stable circular orbit in the range of [ @xmath208 kiloparsec . this also implies that the outer edge of the largest keplerian accretion disk in agns / quasars will never exceed length - scale of kiloparsecs . the instability of circular orbits at large outer radii can have serious astrophysical consequences . in the next section , we analyze the influence of @xmath1 on a simplistic astrophysical phenomenon , bondi accretion rate . in the last section , we have exemplified that the test particle dynamics corresponding to @xmath49 are mostly identical to the corresponding gr results . this is due to the fact that geodesic equations obtained from the lagrangian corresponding to @xmath49 are identical to the gr results , in the low energy limit . the derived potential exactly reproduces the gr results like horizon properties , orbits including apsidal precession , angular momentum , energy , radial velocity of the test particle and also the temporal properties . in the inner radii , where the effect of @xmath1 is negligible , the profile of @xmath49 and particle dynamics corresponding to it essentially coincide with those of @xmath80 however , as seen from the analysis in the earlier subsections , the effect of @xmath1 starts to become prominent in the outer region from an approximate radius @xmath98 , at which @xmath84 corresponding to most of the quantities attain a value of @xmath103 . with @xmath143 , for most of the quantities , the corresponding values of @xmath98 is found out to be approximately in the range of @xmath209 . this is approximately the region from where @xmath49 and most of its corresponding dynamical quantities starts to deviate effectively , from that corresponding to @xmath80 . for @xmath62 , this gives a radius of @xmath210 kiloparsec . this implies that in massive galaxies , beyond a scale of few kiloparsecs , the influence of repulsive cosmological constant may not be neglected . nevertheless , the most significant effect of @xmath1 occurs from around a radius of @xmath211 , where @xmath84 corresponding to most of the dynamical quantities attain a value of @xmath151 . with bh of @xmath212 , this gives a radius of @xmath213 kiloparsec . this is approximately the region where @xmath49 and its corresponding dynamical quantities largely deviate from that corresponding to schwarzschild case . this is also the radius where strong orbital precessional effect in the sds background can be seen ( fig . on the other hand , the results obtained with @xmath49 differ significantly from the corresponding results of the pnp in eq . ( 22 ) , both in the inner radii as well as at the outer radii . in the inner radii @xmath214 , where the effect of @xmath1 is negligible , the pnp in eq . ( 22 ) reduces to that of paczyski - witta potential , and the particle dynamics corresponding to this pnp deviates largely from that obtained with @xmath49 , or in general relativity . the details of which have been elucidated in [ 36 ] . however , in the outer region where the effect of @xmath1 is predominant , especially around the static radius and beyond @xmath215 , the pnp in eq . ( 22 ) as well as the profile of conserved hamiltonian show prominent deviations from that of @xmath49 or that in general relativity , for test particle motion in circular orbit ; sometimes showing contrary behavior near to the cosmological horizon ( fig . in fact , the hamiltonian corresponding to the pnp in eq . ( 22 ) starts deviating from that corresponding to @xmath49 or gr result , from an approximate location of @xmath98 . moreover , the pnp in eq . ( 22 ) gives opposite trend for the test particle radial velocity ( figures . 8c , d ) , as compared to @xmath49 or in general relativity , at @xmath178 . however , in general , for circular motion , the velocity dependent part of @xmath49 vanishes and it gives results close to that predicted by eq . therefore , major differences are expected for non - circular motions and in such situations the pnp proposed in [ 34 ] does not yield consistent results with the gr ( with @xmath1 ) and hence with potential @xmath49 . thus the pnp in eq . ( 22 ) does not reproduce gr features with modest accuracy , unlike that with @xmath49 . this occurs owing to the fact that @xmath49 has been derived based on the prerequisite , that , any analogous potential of corresponding gr geometry should reproduce identical or nearly - identical geodesic equations of motion . however , the only gr notion that the pnp in eq . ( 22 ) tries to mimic is to reproduce last stable circular orbits and marginally bound circular orbits ( both innermost and outermost ) . bondi accretion is a spherically symmetric steady gaseous accretion onto a compact star [ 39 ] . although bondi accretion flow is more hypothetical , yet it has serious astrophysical relevance . it is generally accepted that low luminous agns or low excitation radio galaxies ( lergs ) are presumably powered by radiatively inefficient hot mode accretion , accreting gaseous plasma from hot x - ray emitting phase of interstellar medium ( ism ) or intergalactic medium ( igm ) [ 40,41 ] . consequently the accretion paradigm is quasi spherical in nature , accreting at near ` bondi accretion rate ' from ism / igm . in the expanding universe , bondi accretion rate itself would be plausibly influenced by repulsive @xmath1 , thus would explicitly govern the accretion dynamics [ 19 ] . dynamically bondi solution is transonic in nature with two critical radii or sonic radii ( @xmath216 ) at which the radial velocity @xmath177 exactly equals to sound speed @xmath217 in the limits @xmath218 and @xmath219 respectively . the limit @xmath218 in astrophysical sense represents the location of the ambient medium ( ism / igm ) . as usual , the flow is considered to be locally adiabatic , with @xmath220 and @xmath221 . @xmath222 is the gas pressure , @xmath223 is the density of the accreting gas and @xmath224 the adiabatic index . the mass accretion rate here is then defined by the relation @xmath225 without going into the detailed physics of bondi flow which has been extensively discussed in literature ( for e.g. [ 39,42 ] ) , we simply furnish the relations at critical point @xmath216 , obtained from eq . ( 40 ) and radial momentum conservation equation . using the condition at critical radius , i.e. , at @xmath226 , @xmath227 and using @xmath49 , instead of usual newtonian gravity , the relations at critical point @xmath216 are then given by @xmath228 and @xmath229 respectively . solving these two equations we obtain @xmath216 and corresponding @xmath230 in terms of @xmath231 , where @xmath231 is the sound speed at @xmath218 . using this , the density at @xmath216 is given by @xmath232 , where , @xmath233 is the density at @xmath218 . the transonic accretion rate is then given by @xmath234 @xmath235 can then be expressed in terms of @xmath231 and @xmath233 , which is defined as ` bondi accretion rate ' @xmath236 in sds background with repulsive cosmological constant in presence of @xmath49 . assuming the accreting plasma to be of purely ionized hydrogen , we define @xmath237 and @xmath238 , where @xmath239 and @xmath240 are the temperature and the number density of the accreting gas at @xmath218 . @xmath241 and @xmath242 are the usual proton mass and boltzmann constant , respectively . with these relations @xmath243 in presence of @xmath49 can then be explicitly expressed in terms of @xmath244 for newtonian gravitational potential the usual relation of @xmath245 , where @xmath246 . @xmath239 and @xmath233 in astrophysical sense are the temperature and density of the ambient medium , like hot x - ray emitting phase of ism / igm . in fig . 12 , we study the variation of @xmath243 and sonic radius @xmath216 with @xmath239 and @xmath224 for @xmath0 . it is found that the current accepted value of cosmological constant does not have much impact on the sonic radius as well as the bondi accretion rate for large values of ambient temperature . however , it is found that there is a moderate increase in the value of @xmath243 as compared to the usual value of @xmath243 with @xmath42 , for an ambient temperature @xmath247 , and for smaller values in @xmath224 ( @xmath248 ) . the sonic radius of the flow , too , shift little inwards in sds back ground as that compared to @xmath42 , corresponding to @xmath247 and @xmath249 . this impact of @xmath1 on bondi accretion rate might have interesting consequences on the advective accretion dynamics , especially in low luminous agns or in lergs . obtained with @xmath49 , for @xmath0 . figure 12a shows the variation of the ratio of sonic radius in the sds background to the sonic radius with @xmath42 , in temperature . figure 12b is similar to 12a , but for the ratio of bondi accretion rate in sds background to bondi accretion rate with @xmath42 . solid , long - dashed and dotted curves in the figures 12a , b are for adiabatic constant @xmath250 , respectively . bh mass is chosen to be @xmath66 . ] the influence of positive cosmological constant is important in understanding the kinematics of the kiloparsecs - scale regions and beyond , in the local - galaxies . any observational inference on the possible dependence of the local - scale structure of the galaxies on @xmath1 would be very intriguing , for which , newtonian analogous potential would be of remarkable use . here , we have obtained modified newtonian analogous potential corresponding to sds / sads geometry which reproduces almost all of the gr features with remarkable accuracy . we have investigated the dynamical behavior of the test particle motion in sds background for a typical bh mass @xmath251 in accordance with the bh mass in many agns / quasars . for such massive black hole system , the non - negligible effects of cosmological constant @xmath1 start to occur at radius @xmath252 where the relative deviation between the quantities of sds and schwarzschild geometries attain a value of @xmath103 . the significant effects ( at the level of @xmath253 and above over those due to pure schwarzschild geometry ) of @xmath1 occur at @xmath254 and consequently the local astrophysical kinematics in many massive agns / quasars would be strongly influenced by @xmath1 beyond a distance of few kiloparsecs . ` low excitation radio galaxies ' , which are often massive elliptical galaxies [ 43 ] are powered by the gaseous accretion directly from the x - ray emitting phase of ism or either from the hot x - ray halos surrounding the galaxy or from the hot phase of the igm [ 40,41 ] , in which case the accretion flow region may well exceed parsecs - scale , extending even to kiloparsecs - scale . quasar accretion disk may be susceptible to gravitational instability if the disk is self - gravitating and massive , which may trigger massive star formation in the outer regions extending beyond parsecs - scale [ 44 ] . effects of @xmath1 in the outer regions of these kiloparsecs - scale accretion disk , especially on their dynamical instabilities , would then have interesting consequences in the expanding universe . as remarked by [ 15 ] that positive cosmological constant could also influence the accretion process onto primordial bhs in the early epoch of the expanding universe and can also induce strong collimation effects in astrophysical jets . over kiloparsecs to megaparsecs - scale jets are , in fact , observed [ 45,46 ] in quasars and radio galaxies whose dynamics and structure would likely to be influenced by the effect of repulsive @xmath1 in the expanding universe . seemingly , those are not the only plausible ways in which a positive cosmological constant can , in principle , influence the astrophysical processes on the kiloparsecs - scale regions and beyond , in the local - galaxies in sds background . recently [ 47,48 ] , it has been suggested that nuclear spirals can be one of the mechanisms to power central smbhs by transferring gas from galactic - scales ( over kiloparsec - scales ) to the central parsec region . in such circumstances @xmath1 can influence the feeding process where the mass inflow to nuclear smbhs occur from kiloparsec - scales , and thus controlling the mass accretion rate of the accreting bh . another interesting paradigm to explore the possible effects of @xmath1 , could be in the area of agn feedback , where outflows , jets and radiation from the nuclear region of the active galaxies harboring smbhs interact with the ism and beyond , with length - scales ranging over several kiloparsecs to even megaparsecs . the feedback might considerably effect the star formation rates as well as can curtail accretion onto host smbh , and thus may influence the formation and evolution of galaxies [ 49,50 ] . there would be significant effects of @xmath1 at these length - scales in the expanding universe . to have a glimpse on how repulsive @xmath1 could influence the realistic astrophysical phenomena as well as to examine the effectiveness of the potential derived here , as a toy model , we investigated the impact of @xmath1 on bondi accretion rate , which approximately governs the accretion dynamics plausibly in low luminous agns or in lergs . using sds analogous potential , it is found that the current accepted value of cosmological constant impact the sonic radius as well as the bondi accretion rate moderately , for an ambient temperature @xmath247 and for smaller values in @xmath224 ( @xmath248 ) . thus , it appears that the said value of repulsive cosmological constant might influence the bondi accretion phenomenon , with measurable effect in the local galactic - scales . this might have interesting consequences , especially on the dynamics of advective accretion flow ; the detailed study of which would be pursued in the near future . note that , [ 19 ] also analyzed the influence of @xmath1 on bondi accretion rate , however with significantly large values of @xmath1 . interestingly the authors of [ 19 ] adopted unrealistic high value of @xmath1 , however because for small @xmath1 their adopted technique becomes too difficult under pure gr framework . the next goal is to test the sds analogous potential in regard to the realistic astrophysical processes in local galactic - scales in order to explore the influence of @xmath1 on kiloparsecs - scale structure and beyond , in the local - galaxies . the authors are thankful to an anonymous reviewer for insightful comments and suggestions that help us to improve the manuscript . here we furnish the cartesian transformation of the acceleration terms of the test particle motion using potential @xmath49 in sds back ground , which are useful to depict the elliptical like trajectory of the particle orbit . the acceleration of the test particle in @xmath255 direction , where @xmath256 is given by s. weinberg , rev . phys . * 61 * , 1 ( 1989 ) ; p. j. e. peebles , b. ratra , rev . mod . phys . * 75 * , 559 ( 2003 ) ; m ishak , a. peel , and m. a. troxel , phys . letts . * 111 * , 251302 ( 2013 ) ; m. ishak , j. richardson , d. garred , d. whittington , a. nwankwo , r. sussman , phys . d , * 78 * , 123531 ( 2008 ) ; f. bauer , j. sol , h. tefani , mod . . letts . * a26 * , 2559 ( 2011 ) ; s. finazzi , s. liberat , and l. sindoni , phys . rev . 108 * , 071101 ( 2012 )

this survey paper discusses some uniqueness questions for symplectic forms on compact manifolds without boundary .

1211.2940

let @xmath0 be an oriented @xmath1-manifold . a symplectic form on @xmath0 is a closed @xmath2-form @xmath3 whose top exterior power @xmath4 is a volume form . fundamental questions in symplectic topology are which derham cohomology classes @xmath5 are represented by symplectic forms ( the existence problem ) and whether any two symplectic forms on @xmath0 that represent the same cohomology class are related by a suitable equivalence relation ( the uniqueness problem ) . necessary conditions for the existence of symplectic forms are the existence of an almost complex structure and , in the closed case , the existence of a cohomology class @xmath5 with @xmath6 . whether or not these conditions are also sufficient , when @xmath0 is a closed manifold , is completely open in dimensions @xmath7 . in dimension four additional necessary conditions for existence arise from seiberg witten theory . in the present survey paper the focus is mainly on the uniqueness problem . section [ sec : symp ] gives precise formulations of some relevant questions , section [ sec : ex ] discusses what is known about these question in some examples ( without any claim of providing a complete picture ) , and section [ sec : d ] discusses some conclusions and conjectures as well as the donaldson geometric flow approach to the uniqueness problem for hyperkhler surfaces . [ para : equiv]let @xmath0 be a manifold without boundary . consider the following statements for two symplectic forms @xmath8 , @xmath9 on @xmath0 . ( a ) : : @xmath8 and @xmath9 are connected by a path of cohomologous symplectic forms . ( b ) : : @xmath8 and @xmath9 are connected by a path of symplectic forms . ( c ) : : @xmath8 and @xmath9 are connected by a path of nondegenerate @xmath2-forms . ( d ) : : @xmath8 and @xmath9 have the same first chern class in @xmath10 . ( a ) : : there exists a diffeomorphism @xmath11 of @xmath0 such that @xmath12 . ( b ) : : there exists a diffeomorphism @xmath11 of @xmath0 such that @xmath8 and @xmath13 are connected by a path of symplectic forms . ( c ) : : there exists a diffeomorphism @xmath11 of @xmath0 such that @xmath8 and @xmath13 are connected by a path of nondegenerate @xmath2-forms . ( d ) : : there exists a diffeomorphism @xmath11 of @xmath0 such that @xmath14 . these statements define equivalence relations . two symplectic forms @xmath8 and @xmath9 are called * isotopic * if they are related by ( a ) ( i.e. a path of cohomologous symplectic forms ) , they are called * homotopic * if they are related by ( b ) ( i.e. a path of symplectic forms ) , they are called * diffeomorphic * if they are related by ( a ) ( i.e. a diffeomorphism ) , and they are called * deformation equivalent * if they are related by ( b ) ( i.e. a diffeomorphism , followed by a path of symplectic forms ) . for closed manifolds these equivalence relations are related as follows @xmath15 in particular , by moser isotopy , assertion ( a ) holds if and only if there exists a diffeomorphism @xmath16 isotopic to the identity such that @xmath17 . thus ( a ) implies ( a ) . for ( a ) , ( b ) , ( c ) it may also be interesting to restrict to diffeomorphisms that act as the identity on homology ( see karshon kessler pinsonnault @xcite ) . for ( a ) this would mean that equivalent symplectic forms represent the same cohomology class and for ( b ) that they have the same first chern class and the same gromov witten invariants . [ para : questions]the uniqueness problem in symplectic topology is the problem of understanding these equivalence relations on the space of symplectic forms . here are some questions that may serve as a guideline . fix a derham homology class @xmath18 and denote the space of symplectic forms on @xmath0 representing the class @xmath19 by @xmath20=a\right\}.\ ] ] this is an open set in the frchet space of all closed @xmath2-forms on @xmath0 representing the class @xmath19 . the following three questions correspond to the equivalence relations ( a ) , ( a ) , ( b ) . in all three questions assume that @xmath21 is nonempty ( respectively that @xmath0 admits a symplectic form ) . * question 1 . * _ is @xmath21 connected ? _ * question 2 . * _ are any two symplectic forms in @xmath21 diffeomorphic ? _ * question 3 . * _ are any two symplectic forms on @xmath0 deformation equivalent ? _ section [ sec : ex ] discusses what is known about these questions in some examples . section [ sec : d ] includes a brief discussion of the donaldson geometric flow approach to question 1 in dimension four ( see @xcite ) . [ para : ac]denote the set of almost complex structures on @xmath0 that are compatible with a nondegenerate @xmath2-form @xmath22 by @xmath23 this space is nonempty and contractible . in fact the space of nondegenerate @xmath2-forms on @xmath0 is homotopy equivalent to the space of almost complex structures on @xmath0 . ( the homotopy equivalence is a smooth map @xmath24 , depending on the choice of a background metric , which assigns to every nondegenerate @xmath2-form @xmath22 an almost complex structure @xmath25 ; see @xcite . ) thus every nondegenerate @xmath2-form @xmath22 on @xmath0 determines a cohomology class @xmath26 called the * first chern class of @xmath22*. consider the set @xmath27 if this set contains more than one element , the answer to question 1 is negative . moreover , define @xmath28 * further questions : * what is the set @xmath29 ? is @xmath30 ? does the diffeomorphism group act transitively on @xmath31 ? [ para : homotopyj]let @xmath0 be a compact connected oriented smooth four - manifold without boundary . denote by @xmath32 the euler characteristic and by @xmath33 the signature of @xmath0 . by a theorem of wu an integral cohomology class @xmath34 is the first chern class of an almost complex structure on @xmath0 , compatible with the orientation , if and only if it is an integral lift of the second stiefel whitney class and @xmath35 ( the necessity of equation is the hirzebruch signature theorem . ) given such a class @xmath36 , denote by @xmath37 the space of almost complex structures on @xmath0 with first chern class @xmath36 and compatible with the orientation of @xmath0 . if @xmath0 is simply connected then @xmath38 has precisely two connected components for every @xmath39 . in general , there is a bijection @xmath40 for every @xmath39 . ( this was pointed out to me a long time ago by tom mrowka , but i do nt know a reference . ) the first factor is the set @xmath41 it characterizes the isomorphism classes of spin@xmath42 structures on @xmath43 with first chern class @xmath36 ( see [ para : sw ] and lawson michelsohn ( * ? ? ? * appendix d ) ) . the second factor characterizes the set of homotopy classes of almost complex structures on @xmath0 whose canonical spin@xmath42 structures are isomorphic to a given spin@xmath42 structure with first chern class @xmath36 . ( this can be proved with a standard pontryagin manifold type construction as in milnor @xcite . ) [ para : existence]a natural question is under which conditions a given cohomology class @xmath5 can be represented by a symplectic form . this is the fundamental existence problem in symplectic topology . the obvious necessary conditions are the existence of an almost complex structure on @xmath0 and , in the case of a closed manifold , the condition @xmath44 . in the open case a theorem of gromov @xcite asserts that the existence of an almost complex structure is also sufficient . in the closed case there are counterexamples in dimension four , based on seiberg witten theory ( see taubes @xcite ) . the simplest example is @xmath45 which does not admit a symplectic form because its seiberg witten invariants vanish . in higher dimensions the existence problem is completely open . [ para : diffeo]it is also interesting to investigate the topology of the space @xmath21 of symplectic forms in a given cohomology class beyond the question of connectedness . this is closely related to the topology of certain diffeomorphism groups . let @xmath46 be a compact symplectic manifold without boundary and denote @xmath47\in h^2(m;{{\mathbb r}})$ ] . consider the diffeomorphism groups @xmath48 thus @xmath49 denotes the group of symplectomorphisms of @xmath0 that are _ smoothly _ isotopic to the identity . care must be taken to distinguish it from the subgroup of all symplectomorphisms of @xmath0 that are _ symplectically _ isotopic to the identity ( i.e. the identity component of @xmath50 ) . denote by @xmath51 the connected component of @xmath52 in @xmath21 . ( here @xmath53 denotes the equivalence relation ( a ) in [ para : equiv ] . ) by moser isotopy the group @xmath54 acts transitively on @xmath55 . hence the map @xmath56 induces a homeomorphism @xmath57 thus the topology of @xmath55 is closely related to the topology of the symplectomorphism group . for example , if there exists a symplectomorphism of @xmath46 that is smoothly , but not symplectically , isotopic to the identity then @xmath49 is not connected and hence @xmath55 has a nontrivial fundamental group ( see example [ ex : seidel ] ) . [ para : sw ] many results about symplectic structures in dimension four rely on taubes seiberg witten theory . here is some relevant background . let @xmath0 be a closed oriented smooth four - manifold . denote the set of equivalence classes of spin@xmath42 structures on @xmath0 by @xmath58 . the tensor product with complex line bundles defines a free and transitive action @xmath59 . each @xmath60 has a first chern class @xmath61 and @xmath62 . * every nondegenerate @xmath2-form @xmath63 , compatible with the orientation , determines a canonical equivalence class of spin@xmath42 structures @xmath64 , associated to the @xmath22-compatible almost complex structures . it depends only on the homotopy class of @xmath22 and has the first chern class @xmath65 . * every nondegenerate @xmath2-form @xmath63 , compatible with the orientation , determines a _ homological orientation of @xmath0 _ , i.e. an orientation @xmath66 of the real vector space @xmath67 ( see donaldson @xcite ) . if @xmath68 then the _ positive cone _ @xmath69 has two connected components . if @xmath22 is a symplectic form , compatible with the orientation , let @xmath70 be the connected component containing @xmath71 $ ] . the seiberg witten invariants of @xmath0 take the form of a map @xmath72 this map depends on the choice of a homological orientation @xmath73 of @xmath0 ( see 3 ) . changing the homological orientation reverses the sign of the invariant . when @xmath74 the seiberg witten invariant also depends on the choice of a connected component @xmath75 of @xmath76 ( see 4 ) . changing the connected component is governed by the wall crossing formula of li liu @xcite . let @xmath22 be a symplectic form , compatible with the orientation , and @xmath77 $ ] . when @xmath78 , taubes proved in @xcite that @xmath79 @xmath80 @xmath81 for all @xmath82 . in @xcite taubes proved that every cohomology class @xmath83 with @xmath84 is poincar dual to a @xmath22-symplectic submanifold of @xmath0 . these results continue to hold when @xmath74 , with @xmath85 replaced by @xmath86 ( see @xcite ) . an immediate corollary of taubes theorems is the uniqueness of the first chern class for cohomologous symplectic forms in dimension four . * corollary a. * _ let @xmath0 be a closed smooth four - manifold . two cohomologous symplectic forms on @xmath0 have equivalent spin@xmath42 structures and hence have the same first chern class . _ let @xmath87 be cohomologous symplectic forms on @xmath0 and @xmath77=[\rho']$ ] . choose the orientation such that @xmath88 and choose @xmath83 such that @xmath89 . assume first that @xmath78 . by , @xmath90 and so @xmath84 . hence @xmath91 by . interchanging the roles of @xmath22 and @xmath92 gives @xmath93 , hence @xmath94 , and hence @xmath95 by . this shows that @xmath96 and @xmath97 . the proof for @xmath74 is verbatim the same . include @xmath98 , respectively @xmath99 , as an argument of @xmath100 and use the fact that @xmath101 , because @xmath71=[\rho']$ ] . the next corollary is a special case of a result by conolly - l - ono @xcite . it strengthens corollary a in the simply connected case . * corollary b. * _ let @xmath0 be a simply connected closed smooth four - manifold . _ * ( i ) * _ two symplectic forms on @xmath0 with the same first chern class and inducing the same orientation on @xmath0 are homotopic as nondegenerate @xmath2-forms . _ * ( ii ) * _ two cohomologous symplectic forms on @xmath0 are homotopic as nondegenerate @xmath2-forms . _ ( following conolly l ono @xcite . ) assume @xmath74 . ( the case @xmath78 is easier . ) fix an orientation of @xmath0 and an integral lift @xmath34 of the second stiefel whitney class with @xmath102 . let @xmath103 be the set of nondegenerate @xmath2-forms on @xmath0 with first chern class @xmath36 and compatible with the orientation . then @xmath103 has precisely two connected components , by . let @xmath104 be symplectic forms . then @xmath105 and so @xmath106 , since @xmath0 is simply connected . by , @xmath107 the wall crossing formula in li liu @xcite asserts , in the simply connected case , that @xmath108 whenever @xmath109 . since @xmath110 is even , it follows that @xmath111 . hence @xmath112 , and hence @xmath113 . in ( * prop . 3.25 ) and ( * ? ? ? * lemma 6.4 ) donaldson proved that there is a free involution on @xmath114 , which reverses the homological orientation . hence any two symplectic forms in @xmath103 belong to the same connected component . this proves ( i ) . assertion ( ii ) follows from ( i ) and corollary a. [ ex : open]let @xmath0 be a connected noncompact smooth manifold that admits an almost complex structure . then the h - principle rules . namely , a theorem of gromov asserts that , for every derham cohomology class @xmath5 , the inclusion of the space @xmath21 of all symplectic forms representing the class @xmath19 into the space of all nondegenerate @xmath2-forms is a homotopy equivalence . ( see @xcite , @xcite , ( * ? ? ? * theorem 10.2.2 ) , ( * ? ? ? * theorem 7.34 ) . ) this implies the existence statement in [ para : existence ] and shows that the uniqueness problem reduces to topological obstruction theory . for euclidean space @xmath115 the answer to question 3 is positive and the answers to questions 1 and 2 are negative . namely , there are two homotopy classes of symplectic forms on @xmath116 , one for each orientation , and in @xcite gromov constructed a symplectic form @xmath52 on @xmath116 ( for @xmath117 ) such that @xmath118 is not symplectomorphic to any open subset of @xmath119 with the standard symplectic form ( see also @xcite and ( * ? ? ? * example 13.8 ) ) . by contrast , if @xmath52 is a symplectic form on @xmath120 that agrees with the standard symplectic form @xmath8 at infinity , then @xmath121 is symplectomorphic to @xmath122 by another theorem of gromov ( see @xcite and ( * ? ? ? * theorem 9.4.2 ) ) . [ ex : riemann]let @xmath0 be a closed orientable two - manifold . then the space @xmath21 is nonempty and convex for every nonzero cohomology class @xmath5 . since @xmath0 admits an orientation reversing diffeomorphism , questions 1 , 2 , and 3 have positive answers . in higher dimensions the uniqueness problem for symplectic forms on closed manifolds does not reduce to topological obstruction theory . there is often a dramatic difference between the space of nondegenerate two - forms and the space of symplectic forms , as the following examples show . [ ex : mcduff]this example is due to mcduff @xcite ( see also ( * ? ? ? * theorem 9.7.4 ) ) . here the set @xmath21 is disconnected . however , by an arbitrarily small perturbation of the cohomology class @xmath19 the two known distinct connected components of @xmath21 merge to a single connected component . mcduff s paper @xcite also contains a variant of this construction in dimension eight . consider the manifold @xmath123 identify the @xmath2-torus with the product of two circles . for @xmath124 and @xmath125 let @xmath126 be the rotation about the axis through @xmath127 by the angle @xmath128 . consider the diffeomorphism @xmath129 defined by @xmath130 it acts as the identity on cohomology . let @xmath3 be the product symplectic form , where both @xmath131 factors have the same area , and denote @xmath132=[\psi^*{{\omega}}]\in h^2(m;{{\mathbb r}}).\ ] ] mcduff s theorem asserts that @xmath52 and @xmath133 can be joined by a path of symplectic forms , but not by a path of cohomologous symplectic forms . the proof that @xmath52 and @xmath133 are not isotopic can not be based on the gromov witten invariants because these are invariant under deformation of symplectic forms ( equivalence relation ( b ) in [ para : equiv ] ) . mcduff s proof does involve the moduli space of holomorphic spheres . the relevant evaluation maps represent the same homology class but are not homotopic ; they have different hopf invariants . the argument breaks down for symplectic forms where the @xmath131 factors have different areas , because in that case the relevant moduli spaces are no longer compact . * conclusion . * _ for the six - manifold @xmath134 question 1 has a negative answer and questions 2 and 3 are open problems . _ [ ex : cp2]a theorem of taubes @xcite asserts that any two symplectic forms on @xmath135 with the same volume are diffeomorphic . the proof uses a theorem of gromov @xcite which requires , as an additional hypothesis , the existence of a symplectically embedded two - sphere ( see also ( * ? ? ? * theorem 9.4.1 ) ) . the existence of the required symplectic sphere follows from taubes _ `` seiberg witten equals gromov '' _ theorem @xcite . combining taubes theorem with moser isotopy one finds that , for every cohomology class @xmath5 with @xmath136 , there is a bijection @xmath137 where @xmath52 is the fubini study form representing the class @xmath19 and @xmath138 another theorem of gromov @xcite asserts that @xmath50 retracts onto the isometry group @xmath139 of @xmath140 ( see also ( * ? ? ? * theorem 9.5.3 ) ) . * conclusion . * _ questions 2 and 3 have positive answers . moreover , @xmath21 is connected if and only if every diffeomorphism of @xmath140 that acts as the identity on homology is isotopic to the identity_. [ ex : s2s2]the discussion of example [ ex : cp2 ] carries over to @xmath141 as follows . first , every class @xmath5 with @xmath142 is represented by a symplectic form . second , theorems of gromov @xcite and mcduff @xcite assert that every symplectic form for which the homology classes @xmath143 $ ] and @xmath144 $ ] ( with either orientation ) are represented by symplectically embedded spheres is diffeomorphic to a standard form ( see ( * ? ? ? * theorem 9.4.7 ) ) . third , taubes theorem @xcite establishes the existence of the required symplectic spheres . fourth , a theorem of gromov @xcite asserts that the group of symplectomorphisms that preserve @xmath145 retracts onto the isometry group @xmath146 when @xmath145 have the same area ( see ( * ? ? ? * theorem 9.5.1 ) ) . fifth , a theorem of abreu and mcduff @xcite asserts that the symplectomorphism group is connected when @xmath145 have different areas . thus @xmath147 and @xmath50 has two connected components when @xmath148 , and is connected otherwise . * conclusion . * _ questions 2 and 3 have positive answers . moreover , @xmath21 is connected if and only if every diffeomorphism of @xmath149 that acts as the identity on homology is isotopic to the identity_. [ ex : ruled]let @xmath0 be an orientable smooth four - manifold that admits the structure of a fibration over a closed orientable surface @xmath150 of positive genus with fibers diffeomorphic to the @xmath2-sphere : @xmath151 fix an orientation of @xmath0 and an orientation of the fibers . let @xmath152 be the homology class of the fiber . call a symplectic form @xmath52 on @xmath0 * compatible with the fibration * if it restricts to a symplectic form on each fiber . call it * compatible with the orientations * if its cohomology class @xmath19 satisfies @xmath153 here are some basic facts . * by seiberg witten theory @xcite a cohomology class @xmath5 is represented by a symplectic form if and only if @xmath136 and @xmath154 . any such cohomology class is uniquely determined by the numbers @xmath155 and @xmath156 . * 2 . * a theorem of mcduff @xcite ( see also ( * ? ? ? * theorem 9.4.1 ) ) shows that every symplectic form on @xmath0 that admits a symplectically embedded two - sphere in the homology class @xmath157 or @xmath158 is diffeomorphic to one that is compatible with the fibration . the existence of the required symplectic sphere follows from taubes seiberg witten theory @xcite . * @xmath0 admits an orientation preserving diffeomorphism that preserves the fibration and reverses the orientation of the fiber , and an orientation reversing diffeomorphisms that preserves the fibration and the orientation of the fiber . thus every symplectic form on @xmath0 is diffeomorphic to one that is compatible with the fibration and orientations . * 4 . * a theorem of lalonde mcduff @xcite asserts that any two symplectic forms @xmath159 on @xmath0 that are compatible with the fibration and orientations can be joined by a path of symplectic forms . they also proved that the path can be chosen in the same cohomology class when @xmath160=[{{\omega}}_1]$ ] . thus @xmath161=[{{\omega}}_1]\ ] ] here ( a ) and ( b ) denote the equivalence relations in [ para : equiv ] . * conclusion . * _ on a ruled surface any two symplectic forms are deformation equivalent , and diffeomorphic if they represent the same cohomology class . in the latter case they are homotopic if and only if they are isotopic . _ [ ex : cp21]let @xmath162 let @xmath163 be the homology class of the line in @xmath140 and @xmath164 be the class of the exceptional divisor , both with their complex orientations . they have self - intersection numbers @xmath165 and @xmath166 . then @xmath0 admits an orientation reversing diffeomorphism that interchanges @xmath167 and @xmath168 , an orientation preserving diffeomorphism that preserves @xmath167 and reverses @xmath168 , and an orientation preserving diffeomorphism that reverses @xmath167 and preserves @xmath168 . by taubes theorem @xcite @xmath167 or @xmath169 , and @xmath168 or @xmath170 , is represented by a symplectically embedded sphere for every symplectic form on @xmath0 . hence a class @xmath5 is represented by a symplectic form if and only if @xmath171 by the theorems of mcduff @xcite and taubes @xcite any two symplectic forms representing the same cohomology class are diffeomorphic , as in example [ ex : ruled ] . thus @xmath147 . by every diffeomorphism preserving @xmath19 acts as the identity on homology . moreover , a theorem of abreu and mcduff @xcite asserts that @xmath50 is connected . * conclusion . * _ questions 2 and 3 have positive answers . moreover , @xmath21 is connected if and only if every diffeomorphism of @xmath0 that acts as the identity on homology is smoothly isotopic to the identity . _ [ ex : cp22]consider the oriented four - manifold @xmath172 let @xmath163 be the homology class of the line with self - intersection number @xmath165 and @xmath173 be the homology classes of the exceptional divisors with self - intersection numbers @xmath174 . then , by with @xmath175 , @xmath176 the set @xmath177 of all classes @xmath164 satisfying @xmath166 is given by @xmath178 a theorem of li liu @xcite asserts the following for every @xmath5 . * ( i ) * _ @xmath30 . _ * ( ii ) * _ the diffeomorphism group acts transitively on @xmath31 . _ * ( iii ) * _ @xmath179 if and only if @xmath88 and @xmath180 for every @xmath181 . _ * ( iv ) * _ @xmath182 . _ * ( v ) * _ if @xmath159 are symplectic forms on @xmath0 , then there exists a diffeomorphism @xmath11 of @xmath0 such that @xmath8 and @xmath13 can be joined by a path of symplectic forms . _ assuming seiberg witten theory , assertion ( vi ) below is lemma 3.11 in karshon pinsonnault @xcite and ( vii ) is theorem 1.1 in mcduff @xcite . * ( vi ) * _ two symplectic forms @xmath159 on @xmath0 have the same first chern class if and only if there exists a diffeomorphism @xmath11 of @xmath0 , inducing the identity on homology , such that @xmath8 and @xmath13 can be joined by a path of symplectic forms . _ * ( vii ) * _ two symplectic forms @xmath159 on @xmath0 represent the same cohomology class if and only if there exists a diffeomorphism @xmath11 of @xmath0 , inducing the identity on homology , such that @xmath17 . _ in particular , questions 2 and 3 have positive answers . the proofs rely on the following observations . * for every @xmath39 there exist precisely three classes @xmath183 such that @xmath184 . the numbering can be chosen such that @xmath185 each class @xmath186 is represented by an embedded sphere . for the standard first chern class @xmath187 these can be chosen as the classes @xmath188 , @xmath189 , @xmath190 . for general elements @xmath39 this follows from the fact that , by the results of wall @xcite , the diffeomorphism group acts transitively on @xmath191 . liu @xcite for details . ) * 2 . * by taubes seiberg witten theory @xcite the homology classes @xmath192 are represented by symplectically embedded spheres for every symplectic form on @xmath0 with first chern class @xmath36 . a cohomology class @xmath5 can be represented by a symplectic form @xmath52 with first chern class @xmath193 if and only if for every @xmath164 @xmath194 by 2 . condition is necessary for the existence of @xmath52 . the converse follows from the nakai moishezon criterion . ( see demazure @xcite and also mcduff polterovich @xcite for a direct symplectic proof . ) let @xmath5 such that @xmath88 . close examination shows that there exists a cohomology class @xmath39 such that condition holds if and only if @xmath195 for _ all _ @xmath181 . liu @xcite for details . ) assertions ( i ) , ( ii ) , ( iii ) follow from 1.-4 . and ( iv ) follows from corollary a. . ] * 5 . * to understand ( iv ) geometrically , it is convenient to slightly change the point of view . fix a cohomology class @xmath39 and define @xmath196\in h^2(m;{{\mathbb r}})\,|\,\rho\in{{\omega}}^2(m),\ , d\rho=0,\,\rho\wedge\rho>0,\,c_1(\rho)=c\right\}. \end{split}\ ] ] then @xmath197 ( see 3 . these cones define a chamber structure on @xmath198 . ( see figure [ fig : cp22 ] , where the standard basis of @xmath199 is denoted by @xmath200 and the symplectic cohomology classes satisfy the normalization conditions @xmath201 , @xmath202 , @xmath203 . ) the chamber structure is determined by straight lines connecting pairs of rational points on the unit circle . in this notation ( iv ) asserts that the chambers are disjoint . * * here is a sketch of a proof of ( v ) and ( vi ) , which was worked out in a discussion with yael karshon . construct a model for the symplectic blowup @xmath204 as follows . choose distinct points @xmath205 and disjoint neighborhoods @xmath206 of @xmath207 , respectively , equipped with holomorphic coordinate charts @xmath208 defined on the @xmath209 ball in @xmath210 , centered at the origin . then use the standard complex blowup construction at @xmath211 and @xmath212 to define @xmath0 as a complex manifold . there is a holomorphic projection @xmath213 with singular values @xmath207 . their preimages @xmath214 and @xmath215 are the exceptional divisors that , with their complex orientations , represent the classes @xmath216 . the projection @xmath217 is a holomorphic diffeomorphism from @xmath218 to @xmath219 and the holomorphic sphere @xmath220 represents the class @xmath167 . to define a symplectic form on @xmath0 , let @xmath221 denote the fubini study form on @xmath140 such that @xmath222 has area @xmath217 . choose an exact perturbation @xmath8 of @xmath221 such that @xmath223 is standard on @xmath224 for @xmath225 . ( warning : @xmath8 can not be equal to @xmath221 . ) now use the construction in ( * ? ? ? * lemma 7.15 ) to obtain a symplectic form @xmath226 on @xmath0 such that @xmath227 on @xmath228 and the exceptional spheres @xmath229 are symplectic submanifolds of small areas . let @xmath52 be any symplectic form on @xmath0 with first chern class @xmath230 and choose @xmath192 as in part 1 above such that holds . denote @xmath231 , e^c_i\rangle},\qquad i=1,2,3,\ ] ] and assume @xmath232 , l^c\rangle}=\pi$ ] . by taubes witten theory @xcite , the classes @xmath233 are represented by pairwise disjoint symplectically embedded spheres @xmath234 . hence there exist disjoint weinstein neighborhoods @xmath235 of @xmath236 in which the symplectic form is standard ( see ( * ? ? ? * theorem 3.30 ) ) . the symplectic blowdown construction replaces @xmath236 by closed four - balls of radii @xmath237 with @xmath238 . this results in a symplectic manifold @xmath239 , equipped with symplectic embeddings @xmath240 of closed four - balls with images @xmath241 , and a symplectomorphism @xmath242 thus @xmath243 descends to a symplectically embedded sphere @xmath244 , disjoint from @xmath245 , with area @xmath217 and self - intersection number @xmath246 . the homology class of @xmath247 generates @xmath248 . hence a theorem of gromov @xcite and mcduff @xcite ( see also ( * ? ? ? * theorem 9.4.1 ) ) asserts that there exists a symplectomorphism @xmath249 . consider the embeddings @xmath250 composing @xmath251 with a suitable hamiltonian isotopy we may assume without loss of generality that @xmath252 and that @xmath253 agrees with @xmath254 near the origin . thus @xmath255 is a neighborhood of @xmath256 and the complement of their union is symplectomorphic to a subset of @xmath0 under @xmath257 . the formula in the proof of ( * ? ? ? * lemma 7.15 ) gives rise to a diffeomorphism @xmath258 such that @xmath259 , @xmath260 and @xmath261 . the diffeomorphism @xmath262 satisfies @xmath263 the symplectic form @xmath264 agrees with @xmath226 on @xmath265 . it can be joined to @xmath226 by a path of symplectic forms via a scaling argument . if @xmath266 one can choose @xmath267 and then @xmath262 induces the identity on homology . this proves ( v ) and ( vi ) . the proof of ( vii ) hinges on the fact that holds for the two point blowup of @xmath140 and symplectic forms with the standard first chern class ( i.e. two cohomologous symplectic forms with first chern class @xmath268 are homotopic if and only if they are isotopic ) . this was proved by mcduff in @xcite . thus ( vii ) follows from ( vi ) and corollary a. [ ex : delpezzo]the discussion in example [ ex : cp22 ] carries over almost verbatim to all del pezzo surfaces @xmath269 * 1 . * the characterization of the symplectic cone @xmath197 in remains unchanged and the number of exceptional classes is @xmath270 ( for @xmath271 these are the 27 lines on a cubic in @xmath272 . ) thus condition is still necessary and sufficient for the existence of a symplectic form that represents the class @xmath19 and has first chern class @xmath36 . that the condition is necessary follows again from taubes seiberg witten theory @xcite . that the condition is also sufficient for the standard class @xmath273 is a classical result in khler geometry . that the condition is sufficient in general , was proved by li liu @xcite , by reducing the assertion to the standard chamber via the results of wall @xcite . * assertion continues to hold for all blowups of @xmath140 , i.e. two cohomologous symplectic forms are homotopic if and only if they are isotopic . this was first proved by mcduff @xcite for @xmath274 , then by biran @xcite for @xmath275 , then by mcduff @xcite for all @xmath276 . the proofs ( for @xmath277 ) are based on the inflation techniques of lalonde mcduff @xcite . * 3 . * with this understood it follows by the same arguments as in example [ ex : cp22 ] , parts 6 and 7 , that assertions ( v ) , ( vi ) , ( vii ) continue to hold for all del pezzo surfaces . liu @xcite and karshon kessler pinsonnault @xcite . ) * conclusion . * _ assertions ( i)-(vii ) in [ ex : cp22 ] remain valid for all del pezzo surfaces . thus questions 2 and 3 have positive answers and @xmath30 . two cohomologous symplectic forms are homotopic if and only if they are isotopic . _ [ ex : elliptic]consider the closed oriented smooth four - manifolds @xmath278 , @xmath279 . * 1 . * when @xmath279 it is still true that a class @xmath5 is represented by a symplectic form with first chern class @xmath280 if and only if @xmath281 and holds . thus , in the notation of , the symplectic cone is given by @xmath282 . for @xmath283 this no longer follows from khler geometry . it is an existence theorem in symplectic topology by li liu @xcite . an earlier theorem by biran @xcite asserts that @xmath284 . the proofs are based on taubes seiberg witten theory and the inflation techniques developed by lalonde and mcduff @xcite . the nagata conjecture asserts that the khler cone for the standard complex structure should agree with the symplectic cone . ( see biran @xcite for a symplectic approach to this question . ) consider the * elliptic surface * @xmath285 this manifold admits the structure of a holomorphic lefschetz fibration over @xmath222 with elliptic curves as regular fibers and twelve singular fibers . ( choose two cubics in general position and blow up their nine points of intersection . ) in the standard basis @xmath286 of @xmath199 the homology class of the fiber is @xmath287 . its poincar dual @xmath288 is the first chern class of the standard complex structure . * 3 . * for @xmath279 the diffeomorphism group no longer acts transitively on @xmath191 , however , it acts transitively on @xmath289 . for @xmath290 the theory of wall @xcite is still applicable and shows that , for every @xmath39 , there is a unique odd integer @xmath291 such that @xmath36 is diffeomorphic to @xmath292 . by taubes results , @xmath36 is the first chern class of a symplectic form if and only if @xmath293 . for @xmath294 , li liu proved in ( * ? ? ? * theorem 1 ) that @xmath295 in ( * ? ? ? * theorem d ) they showed that any two symplectic forms with the standard first chern class are deformation equivalent . assuming , the proof of ( v)-(vii ) for all @xmath276 follows the same line of argument as in example [ ex : cp22 ] , parts 6 and 7 , in the case @xmath296 . ( see karshon kessler pinsonnault @xcite . ) there is a chamber structure on @xmath198 as in examples [ ex : cp22 ] and [ ex : delpezzo ] . however , the set @xmath297 is now infinite for each @xmath298 . for more details see biran @xcite and li - liu @xcite . * conclusion . * _ assertions ( ii)-(vii ) in [ ex : cp22 ] remain valid for @xmath279 . thus questions 2 and 3 have positive answers . however , @xmath289 . two cohomologous symplectic forms are homotopic if and only if they are isotopic . _ [ ex : d ] * 1 . * this example is due to smith @xcite . consider the four - manifolds @xmath299 where @xmath300 is a degree five hypersurface . then @xmath301 admits the structure of a lefschetz fibration with generic fiber @xmath157 a torus and 60 singular fibers . both manifolds are simply connected , not spin , and have the betti numbers @xmath302 hence they are homeomorphic and hence @xmath303 and @xmath304 are diffeomorphic . however , the first chern class of @xmath301 with its standard khler structure is divisible by three ( it is @xmath305 ) , while the first chern class of @xmath306 is primitive . these divisibility properties carry over to the product . hence there exist symplectic forms @xmath307 and @xmath308 on diffeomorphic manifolds such that @xmath309 for every diffeomorphism @xmath310 . this gives rise to two symplectic forms on the same manifold that are not even related by the weakest equivalence relation ( d ) in [ para : equiv ] . hence they are not deformation equivalent . * in @xcite smith constructed a simply connected four - manifold @xmath301 that admits two symplectic forms @xmath8 , @xmath9 such that @xmath311 is divisible by three and @xmath312 is primitive . his four - manifold is obtained by forming a fiber connected sum of @xmath313 with five copies of @xmath314 . as in 1 . above the first chern classes of @xmath8 and @xmath9 are not diffeomorphic , and hence @xmath8 and @xmath9 are not deformation equivalent . in fact , for each integer @xmath315 he constructed a simply connected four - manifold @xmath316 with at least @xmath315 different deformation equivalence classes of symplectic forms , distinguished by the divisibility properties of their first chern classes . taking the product @xmath317 , one obtains a @xmath318-manifold with @xmath315 pairwise deformation inequivalent symplectic forms . an earlier example of a symplectic four - manifold with two deformation inequivalent symplectic forms was constructed by mcmullen taubes @xcite . in their example the symplectic forms are distinguished by their seiberg witten invariants ( see also @xcite ) . in @xcite vidussi constructed homotopy k3 s with arbitrarily many deformation inequivalent symplectic forms . * 4 . * in all these examples it is not clear whether two deformation inequivalent symplectic forms can be found in the same cohomology class . [ ex : seidel]in his phd thesis seidel constructed symplectomorphisms on many symplectic four - manifolds @xmath46 that are smoothly , but not symplectically , isotopic to the identity . he proved the following two theorems ( see @xcite and ( * ? ? ? * cor 2.9 & thm 0.5 ) ) . * theorem a. * _ let @xmath0 be a closed symplectic four - manifold with betti numbers @xmath319 assume @xmath0 is minimal ( i.e. it does not contain a symplectically embedded two - sphere with self - intersection number minus one ) . _ if @xmath320 is a lagrangian sphere , then the square @xmath321 of the dehn seidel twist determined by @xmath322 is smoothly , but not symplectically , isotopic to the identity . * theorem b. * _ let @xmath0 be an algebraic surface and a complete intersection , not diffeomorphic to @xmath140 or @xmath323 . then there is a symplectomorphism @xmath16 that is smoothly , but not symplectically , isotopic to the identity . _ here are some remarks . for a much more detailed and wide ranging discussion of examples and ramifications of these theorems see seidel @xcite . the assumptions of theorem a allow for examples of symplectic four - manifolds that do not admit khler structures ( see for example gompf mrowka @xcite and fintushel stern @xcite ) . the assumptions of theorem b allow for examples that are not minimal , such as the six - fold blowup of the projective plane ( cubics in @xmath272 ) . the symplectomorphism @xmath324 in theorem a is smoothly isotopic to the identity by an isotopy localized near @xmath322 . seidel computed the floer cohomology group @xmath325 with its module structure over the quantum cohomology ring , via his exact sequence @xcite . as a result he was able to show that , under the assumptions of minimality and @xmath326 , the floer cohomology group @xmath325 is not isomorphic to the floer homology group @xmath327 hence @xmath328 and @xmath329 are not hamiltonian isotopic . when @xmath330 it then follows that they are not symplectically isotopic . * the assumption @xmath326 can not be removed in theorem a. an example is @xmath331 with its monotone symplectic structures and with @xmath322 equal to the anti - diagonal . in this example @xmath324 _ is _ symplectically isotopic to the identity , by gromov s theorem in @xcite ( see example [ ex : s2s2 ] ) . the assumption of minimality can not be removed in theorem a. examples are the blowups @xmath332 with @xmath333 . on these manifolds there exist symplectic forms ( monotone in the cases @xmath334 ) and lagrangian spheres such that the squares of the dehn seidel twists _ are _ symplectically isotopic to the identity ( see ( * ? ? ? * examples 1.10 and 1.12 ) ) . in contrast , for @xmath335 , the square of a dehn seidel twist in the @xmath276-fold blowup of the projective plane with its monotone symplectic form is _ never _ symplectically isotopic to the identity ( see ( * ? ? ? * example 2.10 ) ) . seidel showed that , by an arbitrarily small perturbation of the cohomology class of @xmath52 , the square of the dehn seidel twist deforms to a symplectomorphism that _ is _ symplectically isotopic to the identity . if @xmath46 is a symplectic four - manifold satisfying the assumptions of theorem a or theorem b then the space @xmath55 of all symplectic forms @xmath22 on @xmath0 that are isotopic to @xmath52 is not simply connected ( see [ para : diffeo ] ) . however , the nontrivial loops in @xmath55 that arise from dehn seidel twists _ are _ contractible in the space of nondegenerate @xmath2-forms . * 7 . * in @xcite kronheimer used seiberg witten theory to prove the existence of symplectic four - manifolds @xmath46 such that @xmath55 is not simply connected . in fact , he developed a method for constructing , for each integer @xmath336 , symplectic four - manifolds @xmath46 such that @xmath337 . [ ex : k3]as a smooth manifold the k3-surface can be constructed as the fiber connected sum @xmath338 on the k3-surface with its standard orientation , every cohomology class @xmath339 with @xmath88 is represented by a symplectic form ( with an associated hyperkhler structure ) . by taubes results @xcite , every symplectic form on the k3-surface with the standard orientation has first chern class zero ( see [ para : sw ] ) . hence it follows from corollary b in section [ sec : symp ] that _ any two symplectic forms on the k3-surface , compatible with the standard orientation , are homotopic as nondegenerate @xmath2-forms . _ [ ex : torus]every cohomology class @xmath340 with @xmath142 is represented by a symplectic form and , by taubes result in @xcite , every symplectic form on @xmath313 has first chern class zero ( see [ para : sw ] ) . there are infinitely many homotopy classes of nondegenerate @xmath2-forms with first chern class zero and compatible with a fixed orientation . the set of such homotopy classes is in bijective correspondence to the set @xmath341 ( see [ para : homotopyj ] ) . the conolly - l - ono argument in @xcite removes only half of these classes as candidates for containing a symplectic form ( see corollary b in section [ sec : symp ] ) . _ it remains an open question whether any two symplectic forms on @xmath313 , that induce the same orientation , are homotopic as nondegenerate @xmath2-forms . _ [ ex : torus1]a recent theorem by latschev mcduff schlenk @xcite asserts that a closed four - ball admits a symplectic embedding into the four - torus with any constant coefficient symplectic form if and only if the volume of the four - ball is smaller than the volume of the four - torus . this settles the existence problem for the one point blowup of the four - torus @xmath342 namely , let @xmath164 be the homology class of the exceptional divisor . then a cohomology class @xmath5 can be represented by a symplectic form if and only if @xmath343 thus the symplectic cone of @xmath0 is strictly bigger than the khler cone . the uniqueness problem for symplectic forms on @xmath0 is still far from understood . [ ex : generaltype]let @xmath344 be a minimal khler surface of general type with first chern class @xmath230 such that @xmath345 < 0.\ ] ] then @xmath346 are the only seiberg witten basic classes . hence , by taubes results in @xcite , every symplectic form @xmath22 on @xmath0 satisfies @xmath347<0,\qquad c_1(\rho)=\pm c.\ ] ] thus any two symplectic forms , compatible with the orientation , have the same first chern class up to sign . _ it is an open question whether every cohomology class @xmath19 with @xmath88 and @xmath348 can be represented by a symplectic form with first chern class @xmath36 _ ( as conjectured by tianjun li ) . here are some of the conclusions that can be drawn from the examples in section [ sec : ex ] about the equivalence relations in [ para : equiv ] . * many symplectic manifolds ( with nonzero first chern classes ) admit anti - symplectic involutions . thus two symplectic forms related by ( a ) in [ para : equiv ] ( a diffeomorphism ) need not be related by ( d ) ( same first chern class ) . * example [ ex : mcduff ] shows that two cohomologous and diffeomorphic symplectic forms on a closed manifold @xmath0 that can be joined by a path of symplectic forms need not be isotopic . thus @xmath349=[{{\omega}}_1],\quad { { \omega}}_0\stackrel{(\mbox{\tiny{a}})}{\sim}{{\omega}}_1,\quad { { \omega}}_0\stackrel{(\mbox{\tiny{b}})}{\sim}{{\omega}}_1 \qquad{{\hspace{7pt}\not\hspace{-7pt}\implies}}\qquad { { \omega}}_0\stackrel{(\mbox{\tiny{a}})}{\sim}{{\omega}}_1.\ ] ] however , for ruled surfaces and all blowups of the projective plane it is known that @xmath160=[{{\omega}}_1]$ ] and ( b ) imply ( a ) ( see examples [ ex : cp2]-[ex : elliptic ] ) . _ it is an open question whether two symplectic forms on a closed manifold that can be joined by a path of nondegenerate @xmath2-forms can always be joined by a path of symplectic forms . _ in other words , it is an open question whether the equivalence relations ( b ) and ( c ) agree . for open manifolds they agree . * 4 . * for all closed simply connected smooth four - manifolds with nonzero euler characteristic , the equivalence relation ( c ) ( homotopic as nondegenerate @xmath2-forms ) agrees with ( d ) ( same first chern class ) , by corollary b in section [ sec : symp ] . * 5 . * for all blowups of @xmath140 the relations ( c ) and ( d ) agree with a stronger form of ( b ) ( a diffeomorphism acting as the identity on homology , followed by a path of symplectic forms ) ; see [ ex : cp22]-[ex : elliptic ] . the symplectic forms on these manifolds have only one equivalence class with respect to ( b ) , ( c ) , ( d ) . * _ in dimensions @xmath350 it is an open question whether two cohomologous symplectic forms on a closed manifold always have the same first chern class . _ for closed oriented smooth four - manifolds it follows from seiberg witten theory that @xmath160=[{{\omega}}_1]$ ] implies @xmath351 ( corollary a in section [ sec : symp ] ) . * _ it is an entirely different question whether any two symplectic forms , compatible with the orientation , have diffeomorphic first chern classes up to sign . _ examples with positive answers include all symplectic four - manifolds with @xmath74 ( see li liu @xcite ) and examples [ ex : k3]-[ex : generaltype ] . for negative answers in dimension four and higher see example [ ex : d ] . * example [ ex : d ] , part 2 , shows that in any dimension @xmath352 there exist closed manifolds with pairs of symplectic forms whose first chern classes are not diffeomorphic ( even up to sign ) ; hence they are not related by ( d ) . [ con : don46 ] let @xmath33 be a symplectic form on the @xmath2-sphere . let @xmath353 and @xmath354 be closed symplectic four - manifolds . then @xmath301 and @xmath306 are diffeomorphic if and only if the symplectic six - manifolds @xmath355 and @xmath356 are deformation equivalent , i.e. there exists a diffeomorphism @xmath357 such that @xmath358 and @xmath359 can be joined by a path of symplectic forms on @xmath360 . the examples of smith @xcite in [ ex : d ] show that the @xmath2-sphere in conjecture [ con : don46 ] can not be replaced by the @xmath2-torus . the conjecture is nontrivial in either direction . when @xmath301 and @xmath306 _ are not _ diffeomorphic but @xmath360 and @xmath361 _ are _ diffeomorphic , the conjecture suggests that the two symplectic structures on these six - manifolds should still _ remember _ the differences in the smooth structures on @xmath301 and @xmath306 . at the time of writing the only known methods for distinguishing smooth structures on four - manifolds are the donaldson invariants and the seiberg witten invariants . by taubes seiberg witten theory @xcite the seiberg witten invariants can also be interpreted as symplectic invariants , and it is conceivable that they give rise to a method for distinguishing the symplectic structures on the products with the two - sphere . an example where @xmath301 and @xmath306 _ are not _ diffeomorphic , @xmath360 and @xmath361 _ are _ diffeomorphic , and the symplectic forms on @xmath360 and @xmath361 can be distinguished by their gromov witten invariants was found by ruan @xcite ( see also ( * ? ? ? * example 9.7.1 ) ) . more examples along these lines were found by ruan tian @xcite and ionel parker @xcite . [ con : torus ] let @xmath0 be a closed hyperkhler surface ( i.e. a four - torus or a k3 surface ) and let @xmath5 be a cohomology class such that @xmath88 . then the space @xmath21 of symplectic forms on @xmath0 representing the class @xmath19 is connected . the uniqueness problem for the four - torus is a longstanding open question in symplectic topology , which goes back at least to the early eighties . in @xcite donaldson proposed a remarkable geometric approach to the uniqueness question for symplectic forms on hyperkhler surfaces which i explain next . the starting point of donaldson s approach is to view the space @xmath362 of diffeomorphisms from a symplectic four - manifold @xmath363 to a hyperkhler surface @xmath364 , that pull back the cohomology class @xmath19 of @xmath9 to that of @xmath33 , as an infinite dimensional hyperkhler manifold . the group of symplectomorphisms of @xmath363 acts on @xmath362 by hyperkhler isometries and the action of the subgroup of hamiltonian symplectomorphisms is generated by a hyperkhler moment map . the negative gradient flow of the square of the hyperkhler moment map is a parabolic type equation for a path of diffeomorphisms @xmath365 depending on a real parameter @xmath366 . it has the form @xmath367 where @xmath368 . equation is preserved by the symplectomorphism group of @xmath363 . one can elmininate the action of the reparametrization group by pushing forward the sympectic form @xmath33 under the diffeomorphisms @xmath369 . the resulting path @xmath370 satisfies the equation @xmath371 this is the * donaldson geometric flow*. it is the negative gradient flow of the energy functional @xmath372 defined by @xmath373 the inner products on the tangent spaces @xmath374 are associated to the norms @xmath375 the gradient flow of the energy functional on @xmath21 with respect to the inner products determined by is meaningful for any closed symplectic four - manifold @xmath46 , equipped with a background riemannian metric . it is convenient to choose the riemannian metric on @xmath0 to be compatible with @xmath52 . then @xmath52 is the unique global minimum of @xmath177 on @xmath21 and the hessian of @xmath177 is positive definite at @xmath52 . in @xcite donaldson also observed that the energy @xmath376 controls the @xmath377-norm of @xmath22 and he proved , in the hyperkhler case , that the hessian at any other critical point ( if it exists ) can not be positive definite . the study of equation poses deep and challenging analytical problems . already regularity and short time existence are nontrivial . they are the subject of as yet unpublished work by robin krom @xcite , which is based on ideas of donaldson in @xcite . the hope is that , in the hyperkhler setting , one can establish long time existence and convergence for equation and use this to prove that the space @xmath21 of symplectic forms in a fixed cohomology class is connected . this hope is backed up by the fact that an analogous geometric flow approach in dimension two leads to the parabolic equation @xmath378 ( see @xcite ) . in this equation the proof of long time existence is based on the observation that the time derivative is nonpositive at each local maximum of @xmath379 and is nonnegative at each local minimum of @xmath379 . f. lalonde , d. mcduff , @xmath382-curves and the classification of rational and ruled symplectic @xmath381-manifolds . in : _ symplectic and contact geometry _ ed . c. thomas , camb . press 1996 , pp 342 .

we analyse the results of recent measurements of nonthermal emission from individual supernova remnants ( snrs ) and their correspondence to the nonlinear kinetic theory of cosmic ray ( cr ) acceleration in snrs . it is shown that the theory fits these data in a satisfactory way and provides the strong evidences for the efficient cr production in snrs accompanied by significant magnetic field amplification . magnetic field amplification leads to considerable increase of cr maximum energy so that the spectrum of crs accelerated in snrs is consistent with the requirements for the formation of galactic cr spectrum up to the energy ev . galactic cosmic rays ; supernova remnants ; shock acceleration ; nonthermal emission : radioemission , x - rays , gamma - rays

0809.0738

the main reason why supernova remnants ( snrs ) are considered as a cosmic ray ( cr ) source is a simple argument about the energy required to sustain the galactic cosmic ray ( gcr ) population against loss by escape , nuclear interactions and ionization energy loss . the mechanical energy input to the galaxy from each supernova ( sn ) is about @xmath1 erg so that with a rate of about one every 30 years the total mechanical power input from supernovae is of the order @xmath2 erg / s ( e.g. * ? ? ? thus supernovae have enough power to drive the gcr acceleration if there exists a mechanism for channeling about 10% of the mechanical energy into relativistic particles . the high velocity ejecta produced in the supernova explosion interacts with the ambient medium to produce a system of strong shocks . the shocks in turn can pick up a few particles from the plasma flowing into the shock fronts and accelerate them to high energies . the only theory of particle acceleration which at present is sufficiently well developed and specific to allow quantitative model calculations is diffusive acceleration applied to the strong shocks associated with snrs ( e.g. see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * for review ) . considerable efforts have been made during the last years to empirically confirm the theoretical expectation that the main part of gcrs indeed originates in snrs.theoretically progress in the solution of this problem has been due to the development of the kinetic nonlinear theory of diffusive shock acceleration @xcite . the theory includes all the most relevant physical factors , essential for snr evolution and cr acceleration , and it is able to make quantitative predictions of the expected properties of crs produced in snrs and their nonthermal radiation . detail information about high - energy cr population in snrs can be obtained from the observations of the nonthermal emission produced by accelerated crs in snrs . the electron cr component is evident in a wide wave length range by the radiation that they produce in snrs , from radio to @xmath3-ray emission , whereas in the case of the nuclear cr component a @xmath3-ray detection is the only possibility to find it . if this nuclear component is strongly enhanced inside snrs then through inelastic nuclear collisions , leading to pion production and subsequent decay , @xmath3-rays will be produced at the detectable level . high - energy @xmath3-rays can also be produced by cr electrons due to the inverse compton ( ic ) scattering of the background photons . therefore it is not so obvious which cr component ( nuclear or electron ) produces tev emission , detected from cassiopeia a ( cas a ) @xcite , rxj1713.7 - 3946 @xcite and vela jr @xcite . for this determination the snr magnetic field value plays the crucial role . the relative role of electrons in @xmath3-ray production is low if snr magnetic field is strongly amplified . in addition , if the magnetic fields in snrs are considerably larger than the typical interstellar value , then snrs are able to generate a power - law spectrum of accelerated crs at least up to the knee energy @xcite . in fact , there is strong evidence that the actual magnetic field in snrs is indeed intrinsically enhanced : the analysis of the nonthermal emission of young snrs show that all of them have amplified fields @xcite . required strength of the magnetic field , that is significantly higher than the interstellar medium ( ism ) field , has to be attributed to nonlinear field amplification at the sn shock by cr acceleration itself , in fact by the dominant nuclear cr component @xcite . the application of kinetic theory to individual snrs ( see * ? ? ? * for a review ) has demonstrated its power in explaining the observed snr properties and in predicting new effects like the extent of magnetic field amplification , leading to the concentration of the highest - energy electrons in a very thin shell just behind the shock . the theory also predicted the enhancement of the high - energy galactic @xmath3-ray background radiation due to the contribution of crs confined in parent snrs , that was partly confirmed by the measured background tev - emission . in order to perform the detail comparison of theoretical expectation with the experiment one needs sufficient number of individual snrs with known values of relevant physical parameters such as the age , distance , ism density . unfortunately the number of such snrs are very limited , especially thous which are seen in all wavelength from radio to @xmath3-ray . it is why every new experimentally established property of nonthermal emission of snr represents considerable interest for theoretical analysis . therefore we start this review with a brief consideration of three snrs for which essentially new information about the efficiency of cr production was recently obtained . then we shall consider the correspondence of the interior magnetic field values extracted from the analysis of the experimental data with theoretical expectation . finally , we discuss the maximum energy of crs which can be achieved during their acceleration in snrs , taking into account realistic values of magnetic field in snrs . experimental and theoretical study of young galactic snrs , tycho s snr , rx j1713.7 - 3946 and sn 1987a , performed recently provides very useful information about cr acceleration by sn shocks . therefore below we briefly analyse these results . the kinetic theory has been applied in detail to tycho s snr , the result of a type ia sn explosion in a ( roughly ) uniform interstellar medium ( ism ) , in order to compare the results with existing data , using a stellar ejecta mass @xmath4 , distance @xmath5 kpc , and ism number density @xmath6 h - atoms @xmath7 @xcite . for these parameters a total hydrodynamic explosion energy @xmath8 erg was derived to fit the observed size @xmath9 and expansion speed @xmath10 . a rather high downstream magnetic field strength @xmath11 @xmath12 g is needed to reproduce the observed steep and concave radio spectrum ( see also * ? ? ? * ) and to ensure a smooth cutoff of the synchrotron emission in the x - ray region . overall , very good consistency of the predictions of the nonlinear theory with the existing observational data was achieved . using chandra x - ray observations @xcite have recently estimated the ratio between the radius @xmath13 of the contact discontinuity ( cd ) , separating the swept - up ism and the ejecta material , and the radius @xmath14 of the forward shock . the large mean value @xmath15 of this ratio was interpreted as evidence for efficient cr acceleration , which makes the medium between those two discontinuities more compressible . this result was analysed within already performed kinetic nonlinear approach @xcite . note , that the increase of the ratio @xmath16 due to the sn shock modification by efficiently accelerated crs was predicted for the case of kepler s snr by @xcite . fig.[f1 ] and fig.[f2 ] show the calculations of shock and cd related quantities which were part of the earlier considerations @xcite , together with the azimuthally averaged experimental data available at the time . the calculated shock as well as cd radii and speeds are shown as a function of time for the two different cases of interior magnetic field strengths @xmath17 @xmath12 g and @xmath18 @xmath12 g required by the value of the synchrotron frequency , where the spectrum hardens relative to the low - energy radio spectrum . according to fig.[f1]a tycho s snr is nearing the adiabatic phase . the softness of the observed low - energy radio spectrum relative to a test particle spectrum required a proton injection rate @xmath19 . this implies a significant nonlinear modification of the shock at the current age of @xmath20 yrs . a larger magnetic field lowers the alfvnic mach number and therefore leads to a decrease of the shock compression ratio , as seen in fig.[f1]b . the result is a total compression ratio @xmath21 and a subshock compression ratio @xmath22 for @xmath17 @xmath12 g . in turn @xmath23 , @xmath24 , for @xmath18 @xmath12 g . therefore , including cr acceleration at the outer blast wave , the calculated value of the ratio @xmath16 for @xmath18 @xmath12 g is slightly lower than for @xmath17 @xmath12 g . at the current epoch we have @xmath25 which is somewhat lower than the value @xmath15 inferred from the chandra observations . projecting a highly structured shell onto the plane of the sky tends to favor protruding parts of the shell . therefore the average radius measured in projection is an overestimate of the true average radius . analyzing the amount of bias from the projection for the shock and cd radii @xcite found a corrected `` true '' value @xmath15 which is lower than their measured `` projected average '' value @xmath26 . one needs to make such a purely geometrical correction independently of the nature of the factors which produce the shell structures , if one measures the size of the quasispherical structured shell in projection instead of dealing with a 2-dimensional shell crossection . in the latter case there would be no need for such a correction . irrespective of this need to correct the observations of the cd position , if one starts from a spherically symmetric calculation of the cd radius as we do , one has to take into account that the actual cd is subject to the rayleigh - taylor ( r - t ) instability ( e.g. * ? ? ? * ; * ? ? ? * ) , and thus a correction is needed to compare such a 1-d calculation with observations of the cd . in the nonlinear regime the instability leads to effective mixing of the ejecta and swept - up ism material with `` fingers '' of the ejecta on top of this mixing region . the latter has a radius larger than @xmath27 . the minimum correction to @xmath27 is the radial extent @xmath28 of the mixing region above @xmath27 . we estimate @xmath29 , where @xmath30 is the longest finger size according to the numerical modeling of @xcite ( albeit without particle acceleration ) . this leads to a rough estimate of the corrected cd radius @xmath31 . as it is seen in fig.2 , the comparison of the corrected values @xmath32 , according to the earlier calculations @xcite , with this experimentally estimated value @xmath15 shows quite good agreement . the agreement between theoretical solutions for effective particle acceleration with the measurements of the shock and discontinuity radii has to be considered as new evidence for strong nonthermal effects in tycho s snr . new northern hemisphere tev detectors is expected to detect this source at tev - energies in , predominantly , hadronic @xmath3-rays at an energy flux level above @xmath33 . as a corollary the detection of a tev signal is not only important by itself , but it is also crucial for the correct determination of all other key snr parameters . rx j1713 - 3946 is a shell - type supernova remnant ( snr ) , located in the galactic plane , that was discovered in x - rays with rosat @xcite . further study of this snr with the asca satellite by @xcite and later by @xcite have shown that the observable x - ray emission is entirely non - thermal , and this property was confirmed in later xmm - newton observations @xcite . the radio emission of this snr is weak : only part of the shell could be detected in radio synchrotron emission up to now , with a poorly known spectral form @xcite . rx j1713 - 3946 was also detected in very high energy @xmath3-rays with the cangaroo @xcite and h.e.s.s . @xcite telescopes . especially the latter observations show a clear shell structure at tev energies which correlates well with the asca contours . the difficulty for the theoretical description is the fact that several key parameters of this source are either not known or poorly constrained . this already concerns the distance and age of the object . it was demonstrated that consistent description of this object is achieved @xcite following present consensus which puts the distance at 1 kpc , the age to about 1600 years and that the primary explosion must have been a type ii / ib sn event with a massive progenitor star whose mass loss in the main sequence phase created a hot wind bubble in a high - density environment . the solution for the overall remnant dynamics then yields the value for the expansion velocity of the outer shock , given the total mechanical energy @xmath34 erg released in the explosion . to obtain a consistent solution for the broadband nonthermal emission the injection rate @xmath19 and interior magnetic field @xmath35 g are needed . the calculated overall broadband spectral energy distribution is displayed in fig.[f3 ] , together with the experimental data from atca at radio wavelengths , as estimated for the full remnant by @xcite , the x - ray data from asca , and the most recently measured tev gamma - ray spectrum from h.e.s.s . the overall fit is impressive , noting that the choice of a few key parameters like @xmath36 , @xmath37 , @xmath38 in the theory allows a spectrum determination over more than 19 decades . note , that calculated gamma - ray stectrum ( berezhko and vlk , 2006 ) agrees with the new h.e.s.s . data ( aharonian et al . , 2007 ) even better than with earlier h.e.s.s . measurements ( aharonian et al . , 2006a ) . the properties of small scale structures of snr rx j1713.7 - 3946 seen in x - rays by @xcite , and in particular by @xcite , provide even stronger evidence that the magnetic field inside the snr is indeed considerably amplified . in order to find out if the filamentary structure found by @xcite is indeed consistent with the efficient cr production projected radial profile @xmath39 calculated for the x - ray energy @xmath40 kev is shown in fig.[f4 ] . the theory predicts the peak of the emission just behind the shock front with projected thickness @xmath41 that corresponds to an angular width @xmath42 . this width is significantly thinner than the observed width of @xmath43 . however , one should take into account the smoothing procedure with which the data were obtained / presented . if we apply a point spread function of width @xmath44 , corresponding to the smoothed xmm data , to the calculated profile , we obtain the broadened profile , which is shown in fig.[f4 ] by the dashed curve . it has a width @xmath45 , or @xmath46 , and is consistent with the observational profile of @xcite . in addition the projected radial profile ( dash - dotted line ) which corresponds to the test particle limit is presented in fig.[f4 ] . in this case the proton injection rate was taken artificially so small @xmath47 , that accelerated crs do not produce any significant nonlinear effect , like shock modification and/or magnetic field amplification . therefore interior magnetic field value @xmath48 @xmath12 g is much lower in this case . due to this fact synchrotron losses are not relevant on account of the low magnetic field , the peak of the emission is broader by a factor of five , inconsistent with the xmm observations . the projected radial @xmath3-ray emission profile , calculated for the energy @xmath49 tev , is presented in fig.[f5 ] . as a result of the large radial gradient of the gas density and the cr distribution inside the snr , the theoretically predicted radial profile of the tev - emission is concentrated within a thin shell of width @xmath50 . since the h.e.s.s . instrument has a finite angular resolution we present in fig.[f4 ] also the modified radial profile smoothed with the gaussian point spread function with @xmath51 , that corresponds to the angular resolution @xmath52 . as shown by fig.[f5 ] the smoothed radial profile of the tev - emission is much broader and what is most interesting it is characterised by a maximum to minimum intensity ratio @xmath53 . such a ratio is consistent with the h.e.s.s . measurement @xcite which obviously gives only a lower limit to the sharpness of the @xmath3-ray profile . note that the broad radial profile of the tev - emission with @xmath54 measured by the h.e.s.s . instrument is indirect evidence that the actual radial profile is significantly sharper , with a higher ratio @xmath55 . we conclude that the present observational knowledge of snr rx j1713.7 - 3946 can be interpreted by a source which ultimately converts about 10% of the mechanical explosion energy into nuclear crs and that the observed high energy @xmath3-ray emission of snr rx j1713.7 - 3946 is of hadronic origin . note , that @xcite have obtained the same type of conclusion suggesting a rather different @xmath3-ray spectrum formation scenario , compared with @xcite . supernova 1987a occurred in the large magellanic cloud . it has been extensively studied in all wavelengths from radio to @xmath3-ray . the initial short outburst of radio emission @xcite is attributed to the synchrotron emission of electrons accelerated by the sn shock propagated in the free wind of presupernova star , which was the blue supergiant ( bsg ) @xcite . after about 3 years radio emission was detected again @xcite as well as the monotonically increased x - ray emission @xcite . this second increase of emission is attributed to the entrance of the outer sn shock into the thermalized bsg wind and then in the h ii region occupied by much more dense matter , consists of the swept up wind of red supergiant ( rsg ) progenitor star @xcite . the advantage of sn 1987a for theoretical consideration is that the key parameter values are very well determined : stellar ejecta mass @xmath56 , distance @xmath57 kpc , hydrodynamic explosion energy @xmath58 erg ( e.g. * ? ? ? during an initial period the shell material has a broad distribution in velocity @xmath59 . the fastest part of this ejecta distribution is described by a power law @xmath60 with @xmath61 . the kinetic nonlinear model for cr acceleration in snrs has been in detail applied to sn 1987a , in order to compare its results with observational properties @xcite . a rather high downstream magnetic field strength @xmath62 mg is needed to reproduce the observed steep radio spectrum @xcite . the required strength of the magnetic field have to be attributed to nonlinear field amplification at the sn shock by cr acceleration itself . since for all the thoroughly studied young snrs , the ratio of magnetic field energy density @xmath63 in the upstream region of the shock precursor to the cr pressure @xmath64 is about @xmath65 @xcite and cr pressure in young snrs has a typical value @xmath66 , the upstream magnetic field @xmath67 was taken in the form @xmath68 calculated shock radius @xmath9 and speed @xmath10 shown in fig.[f6]a as a function of time are in satisfactory agreement with the values obtained on the basis of radio and x - ray measurements . note that radio data compared with x - ray data gives larger shock size at any given time @xmath69 and calculated radius @xmath70 goes between these two sets of data . to fit the spectral shape of the observed radio emission a proton injection rate @xmath71 is needed . this leads to a significant nonlinear modification of the shock : as it is seen in fig.[f6]b total shock compression ratio @xmath72 is essentially larger and a subshock compression ratio @xmath73 is lower than classical value 4 . since during the last 15 yrs the shock speed and the gas density are almost constant cr energy content grows roughly linearly with time ( see fig.[f6]c ) it gives the natural explanation of the linear increase of the radio emission detected during this evolutionary period . strongly modified sn shock generates cr spectrum @xmath74 , which is very steep at momenta @xmath75 , with index @xmath76 . cr electrons with such a spectrum produces synchrotron radioemission spectrum @xmath77 with spectral index @xmath78 , that very well corresponds to the experiment , as it is seen in fig.[f7 ] , where synchrotron energy spectra @xmath79 , calculated for five subsequent epoch together with the experimental data are presented . note that cr spectrum has a concave shape : it becomes flatter at higher momenta @xmath80 . as a consequence synchrotron spectrum @xmath81 is also concave as it is clearly seen in fig.[f7 ] at @xmath82 hz . radio data reveal this property in good consistency with theoretical prediction . strong downstream magnetic field @xmath83 mg , that corresponds to the upstream field @xmath84 mg ( see fig.[f7 ] ) , provides synchrotron cooling of electrons with momenta @xmath85 that in turn makes synchrotron spectrum at high frequencies @xmath86 hz very steep ( see fig.[f7 ] ) . concave shape of electrons continuously produced at the shock front together with their synchrotron cooling lead to a formation of two peaks in synchrotron energy spectrum @xmath79 . the first one at @xmath87 hz corresponds to cr electron momentum @xmath88 above which synchrotron energy looses are relevant , whereas the second peak at @xmath89 hz corresponds to the maximum momentum @xmath90 of accelerated electrons . under this condition calculated synchrotron flux at frequency @xmath91 hz , which corresponds to the photon energy @xmath92 kev , is below the measured flux at the epochs @xmath93 d. since the contribution of the nonthermal radiation in the observed x - ray emission of sn 1987a is not very well known ( e.g. * ? ? ? * ) , the observed x - ray flux has to be considered as the upper limit for the expected nonthermal emission . at early epoch @xmath94 d however the calculated flux exceeds the measured one ( see the curve , corresponding @xmath95 d in fig.[f7 ] ) . this can be considered as indication , that the actual magnetic field @xmath96 is few times larger then given by the eq.(1 ) . at lower magnetic field @xmath97 mg synchrotron losses of high energy cr electrons are considerably smaller compared with the previous case . due to this fact synchrotron spectra considerably exceeds at any given epoch the measured chandra flux @xcite . therefore one can conclude , that the actual interior magnetic field strength is not lower than @xmath98 mg . we see that quite a reasonable consistency of kinetic nonlinear theory with most of the observational data of sn 1987a is achieved , that gives the evidence of the efficient cr production leading to a strong shock modification and strong magnetic field amplification . calculated @xmath3-ray integral energy flux at all energies is dominated by the @xmath99-decay component @xcite . at the current epoch the expected @xmath3-ray energy flux at tev - energies is about @xmath100 erg/(@xmath101s ) and during the next four years it is expected to grow by a factor of two . the detection of high energy @xmath3-rays from sn 1987a , which could be done by instruments like h.e.s.s . , would provide very clear evidence for the efficient acceleration of nuclear crs . detailed consideration performed within a frame of nonlinear kinetic theory demonstrates that the cr production efficiency in young snrs is consistent with requirements for the galactic cr sources . recent experimental and theoretical studies of tycho s snr , rx j1713.7 - 3946 and sn 1987a provides new very strong evidence for that . consideration performed by @xcite led him to the opposite conclusion : that tev @xmath3-ray emission from the mostly fairly young snrs ( sn 1006 , cas a and tycho s ) is unexpectedly low . note , however , that this consideration is based on the simplified estimates , which does not take into account such an important aspect as the necessity of renormalization of cr production efficiency , predicted in spherical approach . we have argued before that ion injection is quite efficient at the quasiparallel portions of the shock surface and it is expected to be strongly suppressed at the quasiperpendicular part of the shock ( see * ? ? ? * for details ) . therefore one should renormalize the results for the nucleonic spectrum , calculated within the spherically symmetric model . the lack of symmetry in the actual snr can be approximately taken into account by a renormalization factor @xmath102 , roughly @xmath103 to @xmath104 , which diminishes the nucleonic cr production efficiency , calculated in the spherical model , and all effects associated with it . for sn 1006 which appears to approximate this ideal case , efficient cr production is expected to arise within two polar regions , where the sn shock is quasi - parallel . the calculated size of the efficient cr production regions which amounts to about 20% of the shock surface corresponds very well to the observed sizes of the bright x - ray synchrotron emission regions @xcite . this implies a renormalization factor of 0.2 . in addition , the distance to the object @xmath105 kpc used by @xcite for the case of sn 1006 is considerably lower than the most reliable determination of the distance @xmath106 kpc to sn 1006 @xcite . due to the above two factors the estimate of tev @xmath3-ray flux from sn 1006 , made by @xcite , should be reduced by a factor of 20 , so that it will be in agreement with h.e.s.s . upper limit . note , that the study of sn 1006 on the basis of kinetic nonlinear theory demonstrated that the lack of a tev signal that follows from the non - detection by the h.e.s.s . instrument does not invalidate the theoretical picture which gives a consistent description of the nonthermal emission characteristics ; it rather implies a constraint on the ambient gas density @xmath107 @xmath7 @xcite . using xmm - newton x - ray observations @xcite recently confirmed that the ism density is indeed low , @xmath107 @xmath7 . the example of sn 1006 illustrates that real difficulty in reliable prediction of expected @xmath3-ray emission from snrs is not a theoretical one , but rather due to poorly known values of relevant astronomical parameters , like distance to the object and ambient ism density . > from theoretical consideration @xcite it becomes clear that in the case of sn 1006 and in other similar cases @xmath3-ray measurements will at the same time give a reliable estimate of the ambient ism density . similar situation is for the two other cases of cas a and tycho s snr , considered by @xcite : renormalization of his estimates , made in spherical approach , and the adoption the most reliable astronomical parameters provide the value of the expected @xmath3-ray flux consistent with the existing measurements ( or upper limits ) ( see * ? ? ? all young snrs ( including three above cases ) to which nonlinear kinetic acceleration theory has been successfully applied up to now exhibit the strongly amplified magnetic field as result of very efficient acceleration of nuclear crs @xcite . magnetic field extracted from the spectral properties of the overall synchrotron emission and from the observed filamentary structure of the nonthermal x - ray emission agree very well and their value can be expressed in the form @xmath108 where @xmath64 is cr pressure at the shock front , @xmath109 is the upstream field , which is already amplified and therefore considerably exceeds magnetic field value @xmath110 @xmath12 g in the ambient ism . let us consider the consistency of such a field strength with the plasma physics . several attempts have been undertaken to model highly nonlinear turbulent magnetic field , created by accelerated crs upstream of the strong shock @xcite . two aspects of this process are of significant importance : the amount of magnetic field amplification and the extent of of gas heating due to the turbulence dissipation . unfortunately up to now it is not clear in which proportions cr working on the gas in the precursor region is divided between gas heating and magnetic field amplification . therefore we restrict ourselves here by a simple estimates in order to demonstrate that considerable magnetic field amplification @xmath109 , consistent with eq.(2 ) is indeed possible upstream of the strong shock . there are two different mechanisms operating upstream of the fast super - alfvnic shock and leading to the amplification of the chaotic magnetic field . the first one is related with the resonant interaction of crs with the alfvn waves , which due to the cr anisotropy triggers an instability , followed by the intense alfvn wave excitation @xcite . the other one , recently proposed by @xcite , is non - resonant amplification of mhd perturbations due to lorentz force , created by cr current . integral effect of resonant alfvn wave excitation can be written in the form @xcite @xmath111 if we suggest that the cr working @xmath112 goes on the excitation of the alfvn waves . here @xmath113 is magnetic component of the alfvn wave energy density , @xmath114 is alfvnic speed , @xmath115 is the plasma flow speed directed along the @xmath116 axis to the shock situated at the point @xmath117 . for the purpose of estimates we suggest quasistationary condition and neglect the gas speed variation @xmath118 in the upstream region @xmath119 due to the shock modification by cr backreaction , taking @xmath120 . assuming that the far upstream field @xmath121 coincides with the ism field @xmath122 and performing the integration of the above equation we have magnetic field energy density at the shock front @xmath123 where @xmath124 is alfvnic shock mach number , @xmath125 is the energy density of the far upstream field . since the strong shock provides efficient cr acceleration with @xmath126 the above expression shows that alfvn wave amplitude grows to a large value @xmath127 in the case of strong shock with mach number @xmath128 . resonant magnetic field amplification should be supplemented by two effects : non - resonant instability and the alfvn speed variation @xmath129 across the precursor due to the increase of the mean effective field @xmath130 . in this case ( see * ? ? ? * for details ) instead of ( 4 ) we have @xmath131 where @xmath132 is the mean field amplified non - resonantly . here @xmath133 , @xmath134 is the maximum cr momentum . substituting @xmath135 in the expression ( 5 ) we have @xmath136 since for all known young galactic snrs the shock speed is about @xmath137 , this expression shows , that existing mechanisms is able to amplify magnetic field within the precursor to the level , which considerably exceeds the required value ( 2 ) . therefore we can suggest that the main amount of the energy given by the expression ( 5 ) in fact goes on the gas heating due to dissipation of mhd perturbations and only small fraction of it , given by ( 2 ) , indeed goes on the magnetic field amplification . numerical simulation of considered process performed by @xcite confirmed , that the gas thermal energy an order of magnitude larger than the energy content of amplified magnetic field . note that the process of magnetic field amplification is not included in kinetic nonlinear theory ( e.g. * ? ? ? * ; * ? ? ? it simply postulates the existence of already amplified far upstream field @xmath96 , which value is consistent with experimental requirements , given by eq.(2 ) . the consistency of such a picture is provided by the fact , that the spectrum of crs produced by strong modified shock is very hard so that crs with highest energies have a largest contribution in their energy content . these the most energetic crs produce field amplification on their spacial scale that is the precursor size . therefore crs with lower energies already see the amplified field @xmath138 . note also , that kinetic nonlinear model contains in the explicit form the suggestion , that the whole cr working goes on the gas heating presumably due to some kind of dissipation mechanism . since according to the above estimates , the magnetic field energy which can be potentially achieved in the precursor region is considerably larger than required energy , given by eq.(2 ) , it can be considered as indirect evidence for the existence of such dissipation processes , which efficiently converts the energy of magnetic field perturbations into the gas thermal energy . the existence of nonadiabatic gas heating within the precursor plays an important role for the shock structure and dynamics . it restricts the shock compression ratio by the value @xcite @xmath139 if the whole cr working @xmath140 goes on the gas heating . for the appropriate upstream magnetic field values , consistent with eq.(2 ) , it provides moderate shock modification with @xmath141 . in the opposite case , when within the precursor there are no any additional gas heating except adiabatic one , the expected shock compression ratio is @xcite @xmath142 where @xmath143 is the sonic mach number , @xmath144 is the far upstream sound speed . for the typical ism parameters this expression leads to much larger shock compression @xmath145 . in the case when cr working in the upstream region goes on the gas heating , the expected gas thermal energy just ahead of the subshock is @xmath146 appropriate magnetic field value @xmath96 given by eq.(2 ) leads to a typical shock compression ratio @xmath147 , that according to eq.(8 ) corresponds to @xmath148 , that in turn according to eq.(10 ) provides gas thermal energy @xmath149 . this value @xmath150 very well corresponds to what is expected from the expression ( 7 ) , if indeed main amount of magnetic field perturbation energy is dissipated . we therefore conclude that kinetic nonlinear theory provides the consistent description at least at shock speeds @xmath151 km / s , that correspond to the most active phase of the snr evolution that is the beginning of the sedov phase . the most important consequence of magnetic field amplification in snrs is the substantial increase of the maximum energy of crs @xmath152 , accelerated by sn shocks . let us briefly analyse what value of @xmath153 can be expected taking into account the amplification of snrs magnetic field . according to theoretical consideration @xcite and the analysis of the observational data @xcite cr diffusion in the amplified magnetic field is close to the bohm limit . therefore cutoff momentum of crs accelerated at any given snr evolutionary stage can be expressed in the form @xcite @xmath154,\ ] ] where @xmath155 is cr diffusion coefficient , corresponding to bohm limit ; @xmath156 is the charge number of cr nuclei . the value of parameter @xmath157 depends on the shock expansion law @xmath70 and on the downstream gas flow radial profile @xmath158 . note that the main factors which determines the value of cr cutoff momentum @xmath159 are : i ) finite shock size ; ii ) shock deceleration and iii ) cr adiabatic cooling in the expended downstream region @xmath160 . these factors are more relevant compared with the time factor , studied by @xcite , that in turn prevents the complete shock modification due to cr backreaction @xcite . first consider the most simple case of uniform ism , that is typical for type ia sne . in such a case the main part of cr spectrum formed during the snr evolution is produced at the sedov phase . since the cr cutoff momentum at the sedov phase @xmath161 decreases with time @xmath69 the most energetic crs in the power law part of this spectrum are produced at the very end of the free expansion phase . therefore the maximum cr momentum @xmath162 depends on physical parameters as @xmath163 , where @xmath164 , @xmath165 and @xmath166 are mean ejecta speed , sweep up radius and sweep up time respectively . taking into account the value of upstream magnetic field ( 2 ) and assuming @xmath167 , we have @xcite @xmath168 it has a strong dependence on the sn parameters @xmath169 and @xmath170 , but is only weakly dependent upon the ism density @xmath171 . maximum momentum of protons , given by eq.(12 ) for @xmath172 erg , @xmath4 and @xmath173 @xmath7 is @xmath174 . this required condition to reproduce the spectrum of gcrs up to the knee energy is fulfilled due to the magnetic field amplification . for cr iron nuclei @xmath175 that provides the possibility for the formation of gcr spectrum up to the energy @xmath0 ev inside snrs . since during the early free expansion snr evolutionary phase the sn shock speed @xmath176 km / s is considerably higher than on the subsequent sedov phase ( @xmath177 km / s ) , one could expect the production of crs with energies @xmath178 essentially higher than produced on sedov phase @xcite . let us briefly consider this possibility , which was studied by @xcite and by @xcite . during the free expansion phase the sn shock evolution is determined by the velocity distribution of the fastest part of the ejecta which is described by a power law @xmath179 with @xmath180 to 12 @xcite . the ejecta kinetic energy contained in the fraction with speed @xmath59 within the interval @xmath181 is @xmath182 . the cr spectrum created by the strongly modified shock is quite hard : @xmath183 , @xmath184 . for the sake of a rough estimate we assume that all the above ejecta energy goes into crs with maximum momenta @xmath185 , where the shock speed is roughly the ejecta speed @xmath186 . this gives @xmath187 in the ultrarelativistic case . taking into account that @xmath188 , we obtain @xmath189 since in the case of uniform ism @xmath190 we have @xmath191 if we suggest for the simplicity @xmath192 . even in the case of type ia sne with a typical value @xmath180 eq.(15 ) gives @xmath193 @xcite . it means that the spectrum of crs with momenta @xmath178 accelerated during the free expansion phase is so steep , that it hardly plays any role for the formation of the gcr spectrum . in the case of type ii / ib sne , which are more numerous in our galaxy @xcite , during the initial snr evolutionary phase sn shock propagates through the progenitor star wind . therefore in this case ambient gas density at the shock front varies as @xmath194 that gives @xmath195 . even at the lowest value @xmath196 the resulting cr spectrum is very steep , @xmath197 . we therefore conclude , that crs accelerated during the free expansion phase do not play a role for gcr spectrum , except the case if the actual value of the parameter @xmath198 is as small as @xmath199 for type ia sne and @xmath200 for type ii / ib sne or the csm has more complicated structure than considered above . in fig.8 we present gcr energy spectra , expected in the galaxy , assuming that snrs are the main gcr source . compared with our previous calculations @xcite much higher magnetic field in snr is used , which varies during the snr evolution according to eq.(2 ) . to obtain these spectra @xmath201 the corresponding source spectra @xmath202 , produced in snr during its evolution , were calculated for a typical set of relevant parameters ( @xmath172 erg , @xmath4 , @xmath173 @xmath7 ) , taking into account their steepening due to the gcr escape from the galaxy with a time scale @xmath203 , where @xmath204 is the rigidity , @xmath205 is the kinetic energy of cr particle @xcite . for a comparison with the experiment we represent in fig.8 the data of only three experiments , atic-2 @xcite , jacee @xcite and kascade @xcite , which on our view are representative in the corresponding energy ranges . one can see , that the theory in a satisfactory way fits the existing data up to the energy @xmath206 ev . the essential exception is the measured in the recent atic-2 balloon experiment @xcite helium spectrum , which opposite to the theoretical expectation is noticeably flatter than the protons spectrum . note that the spectral shape of protons and helium measured in the previous atic-1 experiment at energies @xmath207 gev @xcite are very similar . similar form of the spectra for different gcr nuclei at energies @xmath208 gev if observed could be considered as the indirect confirmation of gcr origin in snrs . therefore the confirmation of atic results for protons and helium by other experiments with comparable or even better precision is very much needed . another difficulty for the theory is the fact that to get a consistency with the observed gcr spectra the value @xmath209 is needed , which is a bit beyond the experimentally determined interval @xmath210 @xcite . in addition , strong energy dependence of the gcr escape time @xmath211 makes it difficult to interpret the observed cr anisotropy , which is very low , less than one percent at energies @xmath212 ev @xcite . here @xmath213 is total energy of cr particle . total flux of gcrs escaping from the galaxy @xmath214 has energy dependence like cr source spectrum @xmath215 with @xmath216 . therefore gcr anisotropy outside the region where cr sources are situated ( galactic disk volume ) increases with energy @xmath217 , like inverse residence time @xmath218 . if @xmath12 is as big as 0.6 the expected anisotropy at @xmath219 ev exceeds the level of 100% @xcite . since our observing position is not outside the source layer , we should see a smaller gcr anisotropy @xcite . in addition one can also suggest that in the region , occupied by cr sources , gcr diffusion coefficient is higher than in outer part of gcr confinement volume , due to enhanced level of turbulence produced by gcrs streaming away from the galaxy . in such a case the gcr anisotropy in the source layer could be sufficiently small and consistent with the observations . note that the theory predicts concave spectra at energies @xmath220 gev for individual cr species and there are some experimental evidences that the actual gcr spectra are indeed concave , eventhough one needs more precise measurements at @xmath221 gev in order to make more strict conclusion . at the same time calculated all particle spectrum has almost pure power law form up to the knee energy ( see fig.8 ) . according to fig.8 the knee in the observed all particles gcr spectrum has to be attributed to the maximum energy of protons , produced in snrs . the steepening of the all particle gcr spectrum above the knee energy @xmath222 ev is a result of progressive depression of contribution of light cr nuclei with the energy increase . such a scenario is confirmed by kascade experiment which shows relatively sharp cutoff spectra of gcr species at energies @xmath223 ev @xcite . at energy @xmath224 ev gcr spectrum is dominated by iron nuclei contribution . according to the analysis , performed by @xcite the contribution of galactic cr sources into the observed gcr spectrum dominates up to the energy @xmath225 ev above which it progressively goes down so that at energy @xmath226 ev the contribution of extragalactic sources becomes dominant . at so - called second knee energy @xmath227 ev these two sources provide roughly equal contributions into gcr spectrum . as it is seen from fig.8 , calculations provide a good fit of the observed gcr up to the required energy @xmath228 ev , that is a strong argument for snrs to be considered as the main galactic cr source . essentially different scenario of the formation of gcr spectrum and the first knee in particular was proposed and discussed by erlykin and wolfandale in a series of papers ( see * ? ? ? * and references there ) . based on the analysis of measurements performed by ground based installations they claim the existence of complex feature of gcr spectrum in the knee energy region . this leads them to a conclusion that observed gcr spectrum consists of smooth background spectrum ( without the knee ) , produced by many cr sources , and superimposed single source ( presumably nearby snr ) contribution , that forms a bump on the smooth background spectrum usually interpreted as a ( first ) knee . the contribution of the single nearby snr , leading to some peculiarity of gcr spectrum at the detectable level in the knee region , is indeed quite possible @xcite . however , the existing measurements of cr composition demonstrate , that the knee is indeed peculiar point in gcr spectrum , where not only the shape of gcr spectrum but also their chemical composition undergoes a noticeable change , that is consistent with cr production in typical snrs , as described above . detailed consideration performed within a frame of nonlinear kinetic theory demonstrates that the cr production efficiency in young snrs is consistent with requirements for the galactic cr sources . recent experimental and theoretical studies of tycho s snr , rx j1713.7 - 3946 and sn 1987a provides new very strong evidence for that . synchrotron radiation from young snrs provides evidence that efficient cr acceleration followed by the significant shock modification and magnetic field amplification takes place there . magnetic field amplification leads to the considerable increase of maximum energy of crs , accelerated in snrs . calculations performed within kinetic nonlinear theory demonstrate that the expected gcr spectrum , produced in snrs , in satisfactory way fits the existing gcr data up to the energy @xmath225 ev . the first knee in the observed all particles gcr spectrum is attributed to the maximum energy of protons , produced in snrs . the steepening of the all particle gcr spectrum above the knee energy @xmath222 ev is a result of progressive depression of contribution of light cr nuclei with the energy increase . such a scenario is confirmed by kascade experiment which shows relatively sharp cutoff spectra of gcr species at energies @xmath223 ev @xcite . the nature of the whole gcr spectrum , which does not require any significant contribution of galactic cr sources other then snrs , can be the following . gcr spectrum up to the energy @xmath229 ev is dominated by the contribution of galactic sources , whereas at @xmath230 ev gcrs are predominantly extragalactic . such a picture is consistent with the analysis , presented here , and with the analysis of ultra high energy cr properties @xcite as well . detail experimental study of gcr spectrum and composition could find whether indeed the transition from galactic to extragalactic component takes place in energy range @xmath231 ev . due to high interior magnetic field in young snrs the @xmath99-decay @xmath3-rays generated by the nuclear cr component as a rule dominates over @xmath3-rays , generated by electron cr component , and calculated @xmath3-ray flux fits existing data . new generation of ground - based @xmath3-ray detectors cangaroo iii , h.e.s.s . , magic , veritas will presumably measure significantly more accurate @xmath3-ray spectra from already detected snrs , like cas a , rxj1713.7 and tycho s snr , as well as from a number of new sources . in this regard considerable interest as a potential @xmath3-ray sources represent tycho s snr , sn 1987a and kepler s snr @xcite eventhough the expected @xmath3-ray fluxes from these objects only slightly above the sensitivity of modern tev @xmath3-ray instruments . together with data , which soon will be obtained by space @xmath3-ray observatory glast , they will reproduce the full @xmath3-ray spectra of some number of individual snrs , as well as the galactic background spectrum . although the existing data together with their theoretical interpretation presents a consistent picture of gcr origin , these new measurements will allow for detailed study of snrs and the cr production inside them . this work has been supported by the russian foundation for rasic research ( grants 07 - 02 - 00221 , 06 - 02 - 96008 , 05 - 02 - 16412 ) . the author thanks the organizers of the cospar meeting e1.4 `` new high - energy results on supernova remnants and pulsar wind nebula '' for the invitation to present this paper and for the financial support , prof . m.i.panasyuk and dr.a.d.panov for their kind presentation of atic-2 data and prof . h.j.vlk for many useful discussions . acero , a. , ballet , b. inner density of sn 1006 . 36th cospar sci . assembly , beijing , abstracts p.1598 , 2006 . aharonian , f. a. , akhperjanian , a. , barrio , j. , et al . evidence for tev gamma ray emission from cassiopeia a. astron . 370 , 112 - 120 , 2001a . aharonian , f.a . , akhperjanian , a.g.,bazer - bachi , a.r . high - energy particle acceleration in the shell supernova remnant . nature 432 , 75 - 77 , 2004 . aharonian , f.a . , akhperjanian , a.g.,bazer - bachi , a.r . et al . a detailed spectral and morphological study of the gamma - ray supernova remnant rx j1713.7 - 3946 with hess , astron . 449 , 223 - 242 , 2006a . aharonian , f.a . , akhperjanian , a.g.,bazer - bachi , a.r . first detection of a vhe gamma - ray spectral maximum from a cosmic source : hess discovery of the vela x nebula , astron . 448 , l43-l47 , 2006b . aharonian , f.a . , akhperjanian , a.g.,bazer - bachi , a.r . primary particle acceleration above 100 tev in the shell - type supernova remnant rxj1713.7 - 3946 , astron . 464 , 235 - 243 , 2007 . ahn , h. s. , seo , e. s. , adams , j. h. et al . the energy spectra of protons and helium measured with the atic experiment . 37 , 1950 - 1954 , 2006 . aloisio , r. , berezinsky , v. , blasi . a dip in the uhecr spectrum and the transition from galactic to extragalactic cosmic rays . arxiv : astro - ph/0608219 , 2006 . amato , e. , blasi , p. non - linear particle acceleration at non - relativistic shock waves in the presence of self - generated turbulence . mon . not . r. astron . 371 , 1251 - 1258 , 2006 . antoni , t. , apel , w.d . , badea , a.f . kascade measurements of energy spectra for elemental groups of cosmic rays : results and open problems . 24 , 1 - 25 , 2005 . bell , a.r . the acceleration of cosmic rays in shock fronts.i . mon . not . 182 , 147 - 156 , 1978 . bell , a. r. , lucek , s.g . cosmic ray acceleration to very high energy through the non - linear amplification by cosmic rays of the seed magnetic field . mon . not . 321 , 433 - 438 , 2001 . bell , a.r . turbulent amplification of magnetic field and diffusive shock acceleration of cosmic rays . mon . not . 353 , 550 - 558 , 2004 . berezhko , e.g. , krymsky , g.f . acceleration of cosmic rays by shock waves . soviet phys .- uspekhi 31 , 27 - 51 , 1988 . berezhko , e.g. , yelshin , v.k . , ksenofontov , l.t . cosmic ray acceleration in supernova remnants . jetph 82 , 1 - 21 , 1996 . berezhko , e.g. maximum energy of cosmic rays accelerated by supernova shocks . astropart . , 5 , 367 - 378 , 1996 . berezhko , e.g. , vlk , h.j . kinetic theory of cosmic rays and gamma rays in supernova remnants . i. uniform interstellar medium . astropart . phys . 7 , 183 - 202 , 1997 . berezhko , e.g. , ksenofontov , l.t . composition of cosmic rays accelerated in supernova remnants . jetph 89 , 391 - 403 , 1999 . berezhko , e.g. , ellison , d.c . a simple model of nonlinear diffusive shock acceleration . astrophys . j. 526 , 385 - 399 , 1999 . berezhko , e. g. , ksenofontov , l. t. cosmic rays , radio and gamma - ray emission from the remnant of supernova 1987a . astronomy letters 26 , 639 - 656 , 2000 . berezhko , e.g. , vlk , h.j . kinetic theory of cosmic ray and gamma - ray production in supernova remnants expanding into wind bubbles . 357 , 283 - 300 , 2000a . berezhko , e.g. , vlk , h.j . theory of synchrotron emission from supernova remnants . 427 , 525 - 536 , 2004 . berezhko , e.g. gamma - ray astronomy and cosmic ray origin problem . adv . space res . , 35 , 1031 - 1040 , 2005 berezhko , e.g. , vlk , h.j . theory of cosmic ray production in the supernova remnant rx j1713.7 - 3946 . 451 , 981 - 990 , 2006 . berezhko , e.g. , ksenofontov , l. t. , vlk , h.j . gamma - ray emission expected from kepler s supernova remnant . 452 , 217 - 221 , 2006 . berezhko , e. g. , ksenofontov , l. t. magnetic field in supernova remnant sn 1987a . j. 650 , l59-l62 , 2006 . berezhko , e. g. , vlk , h. j. spectrum of cosmic rays , produced in supernova remnants . j. 661 , l175 - 178 , 2007 . berezinsky , v.s . , bulanov , s.v . , ginzburg , v.l . , et al . 1990 , astrophysics of cosmic rays , north - holland publ . blandford , r. , eichler , d. particle acceleration in astrophysical shocks : a theory of cosmic ray origin . 154 , 1 - 75 , 1987 . cassam - chena , g. , decourchelle , a. , ballet , j. , et al . xmm - newton observations of the supernova remnant rx j1713.7 - 3946 and its central source observations of snr rx j1713.7 - 3946 . 427 , 199 - 216 , 2004 . chevalier , r.a . self - similar solutions for the interaction of stellar ejecta with an external medium . 258 , 790 - 797 , 1982 . chevalier , r. a. , fransson , c. circumstellar matter and the nature of the sn1987a progenitor star . nature , 328 , 44 - 45 , 1987 . chevalier , r.a . , blondin , j.m . , emmering , r.t . hydrodynamic instabilities in supernova remnants - self - similar driven waves . j. 392 , 118 - 130 , 1992 . chevalier , r. a. , dwarkadas , v. v. the presupernova h ii region around sn 1987a . j. 452 , l45-l48 , 1995 . decourchelle , a. , ellison , d.c . , ballet , j. thermal x - ray emission and cosmic ray production in young supernova remnants . j. 543 , l57-l60 , 2000 . drury , l. oc . , vlk , h. j. , berezhko , e. g.existence and interpretation of smooth cosmic - ray dominated shock structures in supernova remnants . 299 , 222 - 224 , 1995 . drury , lo.c . an introduction to the theory of diffusive shock acceleration of energetic particles in tenuous plasmas . 46 , 973 - 1027 , 1983 . enomoto , r. , tanimori , t. , naito , t. , et al . the acceleration of cosmic - ray protons in the supernova remnant rx j1713.7 - 3946 nature 416 , 823 - 826 , 2002 . erlykin , a.d . , wolfendale , a.w . the origin of pev cosmic rays . 356 , l63-l65 , 2000 . gaensler , b. m. , manchester , r. n. , staveley - smith , l. et al . the asymmetric radio remnant of sn 1987a . j. 479 , 845 - 860 , 1997 gorenstein , p. , hughes , j. p. , tucker , w. h. astrophys . j. 420 , l25-l28 , 1994 . hasinger , g. , ashenbach , b. , trmper , j. the x - ray lightcurve of sn 1987a . 312 , l9-l12 , 1996 . hillas , a. m. can diffusive shock acceleration in supernova remnant account for high - energy galactic cosmic rays ? . g : nucl . part . 31 , r95-r131 , 2005 . hiraga , j.s . , uchiyama , y. , takahshi , t. , aharonian , f.a . spectral properties of nonthermal x - ray emission from the shell - type snr rx j1713.7 3946 as revealed by xmm - newton astron . 431 , 953 - 961 , 2005 . jones , e.m . , smith , b.w . & straka , w.c . formation of supernova remnants - the pre - blast - wave phase . j. 249 , 185 - 194 , 1981 . jones , f.c . , ellison , d.c . the plasma physics of shock acceleration . space sci . rev . 58 , 259 - 346 , 1991 . koyama , k. , kinagasa k. , matsuzaki , k. , et al . discovery of non - thermal x - rays from the northwest shell of the new snr rx j1713.7 - 3946 : the second sn 1006 ? pasj 49 , l7-l11 , 1997 . ksenofontov , l. t. , berezhko , e. g. and vlk , h. j. dependence of the gamma - ray emission from sn 1006 on the astronomical parameters . 443 , 973 - 980 , 2005 . lagage , p. o. , cesarsky , c. j. the maximum energy of cosmic rays accelerated by supernova shocks . 125 , 249 - 257 , 1983 . lazendic , j. s. , slane , p.o . , gaensler , b.m . , et al . a high - resolution study of nonthermal radio and x - ray emission from supernova remnant g347.3 - 0.5 . j. 602 , 271 - 285 , 2004 . lucek , s.g . , bell , a.r . non - linear amplification of a magnetic field driven by cosmic ray streaming . mon . not . 314 , 65 - 74 , 2000 . malkov , m.a . , drury , l.,oc . nonlinear theory of diffusive acceleration of particles by shock waves . 64 , 429 - 481 , 2001 . malkov , m. a. , diamond , p.h . , sagdeev , r.z . on the gamma - ray spectra radiated by protons accelerated in supernova remnant shocks near molecular clouds : the case of supernova remnant rx j1713.7 - 3946 . j. 624 , l37-l40 , 2005 . manchester , r. n. , gaensler , b. m. , wheaton , v. c. , et al . evolution of the radio remnant of sn 1987a : 1990 - 2001 . pasa 19 , 207 - 221 , 2002 . manchester , r. n. , gaensler , b. m. supernova 1987a in the large magellanic cloud . iau circ . 7757 , 2001 . mccray , r. supernova 1987a revisited . annual rev 31 , 175 - 216 , 1993 . mckenzie , j.f . , vlk , j.h . non - linear theory of cosmic ray shocks including selfgenerated alfvn waves . 116 , 191 - 200 , 1982 . michael , e. , zhekov , s. , mccray , r. , et al . the x - ray spectrum of supernova remnant 1987a astrophys . j. 574 , 166 - 178 , 2002 . muraishi , h. , tanimori , t. , yanagita , s. et al . evidence for tev gamma - ray emission from the shell type snr rx j1713.7 - 3946 . 354 , l57-l61 , 2000 . panov , a. d. , adams , j. h. , ahn , h. s. et al . the results of atic-2 experiment for elemental spectra of cosmic rays , @xmath232arxiv : astro - ph/0612377@xmath2332006 . parizot , e. , marcowith , a. , ballet , j. , gallant , y. a. observational constraints on energetic particle diffusion in young supernovae remnants : amplified magnetic field and maximum energy . 453 , 387 - 395 , 2006 . park , s. zhekov , s. a. , burrows , d. n. et al . a chandra view of the morphological and spectral evolution of supernova remnant 1987a astrophys . j. 610 , 275 - 284 , 2004 . pelletier , g. , lemoine , m. , marcowith , a. turbulence and particle acceleration in collisionless supernovae remnant shocks . i - anisotropic spectra solutions . 453 , 181 - 191 , 2006 . pfeffermann , e. , aschenbach , b. 1996 , in roentgenstrahlung from the universe , ed . zimmermann , j. trmper , & h. yorke ( mpe rep . 263 ; garching : mpe ) , 267 . ptuskin , v.s . , zirakashvili , v.n . on the spectrum of high - energy cosmic rays produced by supernova remnants in the presence of strong cosmic - ray streaming instability and wave dissipation . 429 , 755 - 765 , 2005 . reynolds , s.p . , ellison , d.c . electron acceleration in tycho s and kepler s supernova remnants - spectral evidence of fermi shock acceleration . j. 399 , l75-l78 , 1992 . reynolds , p.t . , akerlof , c.w . , cawley , m.f . survey of candidate gamma - ray sources at tev energies using a high - resolution cerenkov imaging system - 1988 - 1991 . j. 404 , 206 - 218 , 1993 . slane , p. , gaensler , b.m . , dame , t.m . nonthermal x - ray emission from the shell - type supernova remnant g347.3 - 0.5 . j. 525 , 357 - 367 , 1999 . staveley - smith , l. manchester , r. n. , kesteven , m. j. et al . birth of a radio supernova remnant in supernova 1987a . nature 355 , 147 - 149 , 1992 . tammann , g.a . , lfler , w. , schrder , a. the galactic supernova rate . 92 , 487 - 493 , 1994 . tan , s.m . , gull , s.f . the expansion of tycho s supernova remnant as determined by a new algorithm for comparing data . mon . not . 216 , 949 - 970 , 1985 . takahashi , y. , asakimori , k. , burnett , t.h . et al . elemental abundance of high energy cosmic rays nuclear phys . b ( proc . suppl . ) 60 , 83 - 92 , 1998 . turtle , a. j. , campbell - wilson , d. , bunton , j. d. et al . a prompt radio burst from supernova 1987a in the large magellanic cloud . nature 327 , 38 - 40 , 1987 . vladimirov , a. , ellison , d. c. , bykov , a. nonlinear diffusive shock acceleration with magnetic field amplification . j. 652 , 1246 - 1258 , 2006 . vlk , h.j . , berezhko , e.g. , ksenofontov , l.t . , rowell , g.p . the high energy gamma - ray emission expected from tycho s supernova remnant . 396 , 649 - 656 , 2002 . vlk , h.j . , berezhko , e.g. , ksenofontov , l.t . variation of cosmic ray injection across supernova shocks . 409 , 563 - 571 , 2003 . vlk , h.j . , berezhko , e.g. , ksenofontov , l.t . magnetic field amplification in tycho and other shell type supernova remnants . 433 , 229 - 240 , 2005 . vlk , h.j . , berezhko , e.g. , ksenofontov , l.t . new evidence for strong nonthermal effects in tycho s snr . accepted in astrophys . space sci . uchiyama , y. , aharonian , f.a . & takahshi , t. fine - structure in the nonthermal x - ray emission of snr rx j1713.7 - 3946 revealed by chandra . 400 , 567 - 574 , 2003 . wang , c .- y . , chevalier , r.a . instabilities and clumping in type ia supernova remnants . j. 549 , 1119 - 1134 , 2001 . warren , j.s . , hughes , j.p . badenes , c. et al . cosmic - ray acceleration at the forward shock in tycho s supernova remnant : evidence from chandra x - ray observations . j. 634,376 - 389 , 2005 . winkler , p. f. , gupta , g. and long , k. s. the sn 1006 remnant : optical proper motions , deep imaging , distance , and brightness at maximum . j. 585 , 324 - 335 , 2003 . , contact discontinuity radius @xmath13 , shock speed @xmath234 , and contact discontinuity speed @xmath235 , for tychos snr as functions of time , including particle acceleration ; ( b ) total shock ( @xmath236 ) and subshock ( @xmath237 ) compression ratios @xcite . the _ dotted vertical line _ marks the current epoch . the _ solid and dashed lines _ correspond to the internal magnetic field strength @xmath17 @xmath12 g and ( @xmath18 @xmath12 g ) , respectively . the observed mean size and speed of the shock , as determined by radio measurements @xcite , are shown as well.,width=283 ] of the radii of the contact discontinuity and the forward shock of tycho s snr as a function of time @xcite . solid and dashed lines correspond to the same two cases as in fig.[f1 ] . the experimental point is taken from @xcite.,width=283 ] ev corresponds to @xmath99-decay @xmath3-ray emission , whereas the dashed and dash - dotted curves indicate the inverse compton ( ic ) and nonthermal bremsstrahlung ( nb ) emissions , respectively . , width=510 ] kev from rx j1713.7 - 3946 @xcite . the solid line corresponds to the high - injection model ; the dashed line again represents the above profile , but smoothed to the resolution of the xmm - newton data used in @xcite ; the dash - dotted line corresponds to the test - particle limit . , width=283 ] -ray emission for @xmath3-ray energies @xmath49 tev @xcite . the calculated radial profile is given by the solid line ; the dashed line represents the calculated profile smoothed with a gaussian point spread function of the width @xmath238 . for purposes of presentation both profiles are normalised to their peak values in this figure . , width=283 ] , shock speed @xmath234 , gas density @xmath239 and upstream magnetic field @xmath96 at the current shock position ; ( b ) total shock ( @xmath236 ) and subshock ( @xmath237 ) compression ratios ; ( c ) kinetic energy of ejecta @xmath240 and accelerated cr energy content @xmath241 as functions of time for sn 1987a @xcite . the observed radius of the sn shock , as determined by radio @xcite and x - ray measurements @xcite , are shown by circles and stars respectively . the scaling values are @xmath242 cm and @xmath243 km / s . the _ dotted vertical line _ marks the current epoch . [ f6],width=245 ]

the first generation of stars ( commonly known as population iii ) are expected to form in low - mass protogalaxies in which molecular hydrogen is the dominant coolant . radiation from these stars will rapidly build up an extragalactic ultraviolet background capable of photodissociating , and it is widely believed that this background will suppress further star formation in low - mass systems . however , star formation will also produce an extragalactic x - ray background . this x - ray background , by increasing the fractional ionization of protogalactic gas , promotes formation and reduces the effectiveness of ultraviolet feedback . in this paper , we examine which of these backgrounds has the dominant effect . using a simple model for the growth of the uv and x - ray backgrounds , together with a detailed one - dimensional model of protogalactic chemical evolution , we examine the effects of the x - ray backgrounds produced by a number of likely source models . we show that in several cases , the resulting x - ray background is strong enough to offset uv photodissociation in large-cooled protogalaxies . on the other hand , small protogalaxies ( those with virial temperatures ) remain dominated by the uv background in all of the models we examine . we also briefly investigate the effects of the x - ray background upon the thermal and chemical evolution of the diffuse igm . cosmology : theory galaxies : formation molecular processes radiative transfer

astro-ph0205308

in cosmological models based on cold dark matter ( cdm ) , the first stars are believed to form within small protogalaxies , with virial temperatures @xmath2 @xcite . cooling within these protogalaxies is dominated by molecular hydrogen , @xmath0 , which forms via the gas - phase reactions @xmath3 and @xmath4 even in the absence of dust . although the fractional abundance of @xmath0 that forms in this way is small , it is sufficient to allow for effective cooling and the formation of stars @xcite . as soon as massive stars form , however , they immediately begin to photoionize and photodissociate this @xmath0 . photoionization requires photons with energies greater than @xmath5 , which are strongly absorbed by neutral hydrogen , and is only of importance within regions . photodissociation , by contrast , occurs through the absorption of photons in the lyman - werner band system @xcite , with energies in the range @xmath6 @xmath7 . these photons are not strongly absorbed by neutral hydrogen and can readily escape into the intergalactic medium @xcite . initially , many of these photons will be absorbed by intergalactic @xmath0 , but its abundance is small and it is rapidly photodissociated ( see section [ ther_chem ] below ) . consequently , the onset of star formation is soon followed by the appearance of an ultraviolet background radiation field . this ultraviolet background acts to suppress further star formation by photodissociating @xmath0 within newly - forming protogalaxies . the effects of this background have been studied by a number of authors @xcite . in particular , @xcite study the coupled problem of the evolution of the ultraviolet background and its feedback on the global star formation rate using a simple galaxy formation model based on the press - schechter formalism @xcite . they find that cooling ( and hence star formation ) within small protogalaxies is completely suppressed prior to cosmological reionization . taken at face value , their results suggest that star formation within small , @xmath0-cooled protogalaxies is a transient phenomenon , with little impact on later stages of galaxy formation . these conclusions , however , rest on the assumption that the only free electrons present in the protogalactic gas come from the small residual fraction remaining after cosmological recombination . this is important , as free electrons ( and protons ) catalyze @xmath0 formation , as we can see from equations [ h2f1 ] to [ h2f4 ] . if the free electron abundance were significantly higher , then the @xmath0 formation rate would also be higher , offsetting the effects of photodissociation . at the very least , this would delay the suppression of star formation , and in principle could entirely negate it . it is therefore important to determine whether there is any way in which an enhanced level of ionization could be produced . one possible source is the residual ionization that would remain after the recombination of regions produced by an earlier generation of stars . this has recently been studied by @xcite , who find that it can be an effective source of @xmath0 and can dramatically reduce the effectiveness of photodissociation feedback . however , their conclusions are still somewhat uncertain , both because their simulations are under - resolved ( see figures 57 in * ? ? ? * ) and because they neglect the effects of supernovae , which would act to disperse the gas and significantly lengthen the recombination timescale , thereby delaying the formation of @xmath0 . an alternative possibility is ionization by a high redshift x - ray background . at x - ray energies , the optical depth of the intergalactic medium ( igm ) is small and any x - ray sources present will naturally generate an x - ray background . moreover , x - rays can penetrate to large depths within newly - formed protogalaxies , allowing them to raise the fractional ionization throughout the gas . the potential importance of such a background was first highlighted by @xcite , with high redshift quasars suggested as a possible source . using a very simple model for a quasar - produced x - ray background , they showed that if quasars contribute more than 10% of the uv background , then the ionization produced by the associated x - ray background is sufficient to negate the effects of uv photodissociation . indeed , they found evidence that such a background could actually promote cooling within the dense gas in the centres of protogalaxies . this scenario has not yet been firmly ruled out , but observational evidence suggests that quasars are unlikely to be present in sufficient number at high redshift @xcite . quasars , however , are not the only potential source of x - rays . star formation also leads to the production of x - rays , primarily through the formation of massive x - ray binaries @xcite , but also through bremsstrahlung and inverse compton emission from supernova remnants . these sources generate only a small fraction of the present - day x - ray background @xcite , but may become dominant at high redshifts . in this paper , we examine the effects of the x - rays produced by these sources , both on the cooling of gas within virialized protogalaxies and also on the chemical and thermal evolution of the igm . the outline of the paper is as follows . in section [ back ] , we discuss the sources responsible for producing the uv and x - ray backgrounds , and show how the build - up of these backgrounds can be computed . in section [ meth ] , we outline the method used to study the effect of this radiation on the primordial gas , and in section [ res ] apply it for a number of different x - ray source models . we present our conclusions in section [ conc ] . for an observer at redshift @xmath8 , we can write the mean specific intensity of the radiation background at an observed frequency @xmath9 as @xcite @xmath10 where @xmath11 , @xmath12 is the proper space - averaged volume emissivity , @xmath13 is the optical depth at frequency @xmath9 due to material along the line of sight from redshift @xmath8 to @xmath14 and @xmath15 is the cosmological line element . to solve this equation , we need to know how the emissivity and opacity evolve with redshift . for simplicity , we write the space - averaged emissivity in terms of the global star formation rate ( sfr ) as @xmath16 where @xmath17 is the luminosity density ( i.e. the luminosity per unit frequency ) per unit star formation rate ( in solar masses per year ) , and @xmath18 is the global star formation rate , with units @xmath19 . in principle , @xmath17 may be a complicated function of frequency and redshift . in practice , however , lyman - werner band emission is dominated by massive , short - lived ob stars and declines rapidly once star formation comes to an end ( see , for example , the instantaneous starburst models of * ? ? ? * in which the lyman - werner flux declines by an order of magnitude within 4@xmath20 ) . similarly , x - ray emission is dominated by short - lived sources such as massive x - ray binaries and supernova remnants which are end products of the same massive stars . both kinds of emission are therefore strongly correlated with the star formation rate ; to a first approximation , we can assume that they are directly proportional to it , and that any redshift dependence of @xmath17 can be neglected . with this simplification , determining the emissivity breaks down into two independent problems : determining the global star formation rate as a function of redshift , and determining the luminosity density as a function of the star formation rate . although we have observational constraints on the star formation rate up to @xmath21 , we have no direct constraints ( and few indirect ones ) at higher redshift . consequently , any model of high redshift star formation must inevitably be highly theoretical . moreover , this lack of constraints motivates us to choose as simple a model as possible ; more complicated ( and realistic ) models can always be considered once our observational knowledge improves . a good example of this kind of simple model is the one used by @xcite ; we adopt the same model here . we assume that star formation proceeds primarily through starbursts , of duration @xmath22 years , that are triggered when galaxies form . during the starburst , the star formation rate is assumed to be constant . the global star formation rate in this model is given by @xmath23 where @xmath24 is the star formation efficiency , @xmath25 is the baryon fraction ( ie the ratio of baryons to dark matter ) , and where @xmath26 is the cosmological density of matter in newly - formed galaxies ( with units of @xmath27 ) . we assume that the value of @xmath25 in the protogalaxies is the same as in the igm , or in other words that @xmath28 . for the cosmological model adopted in section [ res ] , this corresponds to @xmath29 . we further assume that the rate of change of @xmath30 is approximately the same as the rate of change of @xmath31 , the total fraction of matter in halos with virial temperatures greater than a critical temperature @xmath32 , or in other words that @xmath33 where @xmath34 is the cosmological matter density . here , @xmath32 represents the minimum virial temperature required for efficient cooling ; to a first approximation , halos with @xmath35 are unable to cool , while those with @xmath36 cool rapidly and can form stars . various different definitions of @xmath32 are in use in the literature ( see , for example * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) ; we discuss our particular choice in a later section . making this approximation is equivalent to assuming that the growth in @xmath31 is dominated by the formation of new halos with @xmath37 ( either by monolithic collapse or by the merger of smaller objects ) rather than by the accretion of matter by existing halos with @xmath36 . this is justified at high redshift when such halos are rare and @xmath31 is dominated by objects near @xmath32 , but becomes less accurate at lower redshifts . for this reason ( and others , to be discussed later ) we do not attempt to simulate the evolution of the background below @xmath38 . to solve equation [ diff_rho ] we need to know the value of @xmath32 . in general , this will depend both on redshift ( see , eg * ? ? ? * ) and on the intensities of the uv and x - ray backgrounds . indeed , understanding the evolution of @xmath32 with redshift is one of the main goals of this paper . our procedure for determining @xmath32 is discussed at length in section [ meth ] ; for now , we assume that it is known . in this case , we can calculate @xmath31 using the press - schechter formalism @xcite : @xmath39,\ ] ] where @xmath40 is the critical density threshold for collapse , @xmath41 is the square root of the variance of the cosmological density field , as smoothed on a mass scale @xmath42 , and where @xmath43 is the mass of a protogalaxy with virial temperature @xmath32 . although both @xmath44 and @xmath41 depend upon the choice of cosmological model , their behaviour is well - known , and the problem of determining @xmath45 reduces to the relatively simple one of relating @xmath43 to @xmath32 . to do this , we need to know the protogalactic density profile . hydrodynamical simulations suggest that it is approximately isothermal @xcite , but representing it as a singular isothermal sphere is physically unrealistic due to the latter s infinite central density . accordingly , we follow @xcite and represent it as a truncated isothermal sphere @xcite . with this choice , we find that @xmath46 to determine @xmath47 , we must integrate equation [ diff_rho ] over a redshift interval @xmath48 , corresponding to the duration of the starburst ; hence , @xmath49 this simple model has a number of shortcomings . for instance , it assumes that the star formation efficiency @xmath24 and starburst duration @xmath22 are both constant , independent of redshift or galaxy mass . it also assumes that each galaxy forms its stars in a single starburst and thus ignores the effects of continuous star formation and of subsequent , merger - triggered starbursts . . ] nevertheless , it has the virtue of simplicity , and is a good point from which to start our examination of the effects of the x - ray background . the ultraviolet flux of a star - forming galaxy is dominated by emission from young , massive o and b - type stars . these are short - lived , with the most massive having lifetimes of only a few myr , and thus the ultraviolet luminosity density is closely correlated with the star formation rate . its value depends upon the spectral properties of the newly - formed stellar population , and thus on their initial mass function ( imf ) , metallicity and age . in a recent paper , @xcite presents values for the photon flux in the lyman - werner bands calculated for a number of different metal - free stellar populations . if we assume that the spectrum within the bands is flat ( a reasonable approximation ) , then we can convert this photon flux into a luminosity density . for a salpeter imf with minimum mass @xmath50 and maximum mass @xmath51 ( model a in * ) , we find that @xmath52 reducing @xmath53 to the more conventional value of @xmath54 reduces this luminosity to @xmath55 as we form a greater number of low mass stars that do not contribute significantly to the lyman - werner flux . both of these results assume that the dissociative flux has stabilized at its equilibrium value and is therefore proportional to the star formation rate . this equilibrium is typically established after only 2 3 myr , so this is generally a good approximation , even for starbursts of relatively short duration . the above figures are appropriate so long as we are dealing with stars formed out of _ entirely _ metal - free gas . such stars are somewhat unusual , however , as the absence of carbon means that they are unable to generate energy via the cno cycle , which otherwise would dominate energy production in stars of mass @xmath56 . as a result , metal - free stars are hotter than their metal - enriched counterparts @xcite and have harder spectra . a surprising consequence of this fact is that a low - metallicity stellar population will produce a larger dissociative flux than a metal - free population the lower effective temperatures move the peak in the thermal emission from the most massive stars closer to the lyman - werner band , causing the ionizing flux to fall but the dissociative flux to rise . we can use the data presented in @xcite for a stellar population with @xmath57 to examine the difference that this effect makes to the lyman - werner flux . using the same imf as in equation [ flux_dis ] , we find that @xmath58 thus , raising the metallicity increases the lyman - werner flux , but only by about 6070% ; as we will see in section [ res_nox ] , this has little effect on the evolution of @xmath32 . clearly , there are many possible models other than those considered here . indeed , there is growing evidence that the imf of population iii stars is strongly biased towards high masses @xcite . however , this remains uncertain , and in this paper we have chosen to err on the side of caution and assume that the high - redshift imf is similar to that at the present day . in a recent study , @xcite collate data on a number of local starburst galaxies and compare the 2 10 @xmath59 x - ray fluxes measured by asca with the 8 1000 @xmath60 infrared fluxes measured by iras . they find that a clear correlation exists , with the total x - ray and infrared fluxes related by @xmath61 similar correlations have previously been reported by @xcite and @xcite for x - rays in the 0.5 4.5 @xmath59 and 2 30 @xmath59 energy bands respectively . theoretically , we would expect such a correlation , with both the x - ray and infrared emission tracing the underlying star formation rate . for the x - rays , this occurs because the emission is dominated by massive x - ray binaries ( mxrbs ) : binary systems consisting of a massive ob star accreting onto a compact companion ( a neutron star or black hole ) . x - ray emission from such systems generally switches on a few million years after the formation of the compact object , and the lifetime of the emitting phase is short , typically of the order of ( 2 5@xmath62 @xcite . these short timescales tie the emission closely to the underlying star formation rate @xcite . the far - infrared flux , on the other hand , tracks star formation far more directly , being dominated by emission from dust heated by short - lived , massive stars . to use this observed correlation to determine the x - ray luminosity of a star - forming galaxy as a function of its star formation rate , we use the result from the starburst models of @xcite that @xmath63 together with equation [ x_to_ir ] to write the x - ray luminosity as @xcite @xmath64 we then assume that this relationship between x - ray luminosity and star formation rate remains valid as we move to higher redshifts . note that the same need not be true for the relationship between x - ray luminosity and infrared luminosity , or infrared luminosity and star formation rate : although they serve to establish the correlation between x - ray luminosity and star formation rate at @xmath65 , the assumption that this correlation remains valid at higher redshift does not imply that these other correlations also remain valid . indeed , we would expect the infrared luminosity of dust - free protogalaxies to be very much lower than would be predicted by equation [ ir ] . evidence that the correlation between x - ray luminosity and star formation rate does indeed remain valid at high redshift is provided by the recent stacking analysis of individually undetected lyman break galaxies in the _ chandra _ deep field - north @xcite . this analysis finds that the average rest frame luminosity of the lyman break galaxies in the 2 8 @xmath66 energy band is @xmath67 . assuming a typical star formation rate of @xmath68 for these galaxies @xcite , this corresponds to an x - ray luminosity of @xmath69 consistent with the value derived above . is measured in a slightly different energy band from that of equation [ lx ] , so the agreement between the two values is not quite as good as may at first appear . nevertheless , the necessary correction is small , and the values agree to within 50% . ] although far from conclusive , this result suggests that we can extrapolate the locally observed correlation to at least as far as @xmath70 . comparing our determination of @xmath71 with a recent calculation by @xcite , we find a difference of a factor of ten in our results . some of this disagreement is due to the difference in x - ray energy bands considered ( @xmath72 @xmath73 in @xcite , compared to @xmath74 @xmath75 here ) , but some must surely be due to intrinsic scatter in the observational data , suggesting that equation [ lx ] should properly be regarded as an order of magnitude estimate of the true x - ray luminosity . given equation [ lx ] for the x - ray luminosity as a function of the star formation rate , we calculate the x - ray luminosity density by assuming a template spectrum of power - law form @xmath76 where @xmath77 , and requiring that @xmath78 where @xmath79 and @xmath80 . @xcite find that a weighted average of the galaxies in their sample gives a value for the spectral index of @xmath81 . adopting this value and solving for @xmath82 , we find that @xmath83 altering @xmath84 changes @xmath82 , but never by more than 50% for values consistent with the @xcite measurement . clearly , individual galaxies may have spectra that differ markedly from this simple template , but it should be a reasonable approximation when averaging over a large number of galaxies . the above model is simple , and empirically motivated , but does assume that the x - ray emission of high - redshift star - forming galaxies is very similar to that of starbursts observed locally . this is a reasonable assumption in the absence of evidence to the contrary , and is probably valid as long as massive x - ray binaries continue to dominate the galactic x - ray emission . however , it is quite possible that at high redshift some other type of source will come to dominate the emission , particularly if the number of binary systems is small , as is suggested by recent simulations @xcite . it is therefore prudent to consider the effects of other potential sources of x - rays . the obvious candidates are supernova remnants ( snr ) ; next to x - ray binaries , they are the most significant galactic sources @xcite . they can emit x - rays through a variety of different emission mechanisms , but at high redshift the most significant will be thermal bremsstrahlung emission and non - thermal inverse compton emission . thermal bremsstrahlung is produced by the hot gas within the snr . detailed modeling properly requires a hydrodynamical treatment ( see , for example , * ? ? ? * ) , but for our purposes a simple parameterization suffices . if we assume that the hot gas has a single characteristic temperature @xmath85 , then we can write the luminosity density per unit star formation rate as @xmath86 where @xmath87 and where @xmath88 is constant . moreover , we can write the total x - ray luminosity as @xmath89 where @xmath90 is the typical supernova explosion energy ( in units of @xmath91 ) , @xmath92 is the fraction of this energy radiated as bremsstrahlung , and where @xmath93 is the number of supernovae that explode per solar mass of stars formed . the value of @xmath93 depends on the imf ; for the standard salpeter imf adopted previously , @xmath94 . finally , since @xmath95 we can fix the value of @xmath88 ; we find that @xmath96 for typical supernova parameters ( @xmath97 and an ambient density @xmath98 ) , @xcite find that a fraction @xmath99 of the explosion energy is radiated , at a characteristic temperature @xmath100 . on the other hand , the higher mean density at high redshift , together with the comparative weakness of outflows from low metallicity stars @xcite suggest that the typical ambient density may be very much higher . in particular , if it is as high as @xmath101 , then a supernova remnant will radiate its energy extremely rapidly , before the ejecta have time to thermalize @xcite . in this case , the fraction of energy radiated as bremsstrahlung is very much higher ( @xmath102 ) , as is the characteristic temperature ( @xmath103 ) . we examine both of these models in section [ res ] , with the understanding that the true picture lies somewhere in between . in addition to this thermal emission , supernova remnants also produce non - thermal x - rays . these are generated as the relativistic electrons produced by the snr gradually lose energy through synchrotron radiation , non - thermal bremsstrahlung and the inverse compton scattering of photons from the cosmic microwave background . at high redshift , the latter is likely to dominate @xcite . the spectrum of the resulting emission depends upon the energy spectrum of the relativistic electrons , but at the energies of interest is well represented by a power law : @xmath104 . the intensity of the emission depends upon the fraction of the supernova energy transferred to the electrons ; this is not well constrained , with estimates ranging from 0.1% to 10% . accordingly , we model the emission as @xmath105 where @xmath87 , @xmath106 is the fraction of energy deposited in the electrons , and where we have assumed that @xmath97 and @xmath107 as in the thermal bremsstrahlung case . this expression assumes a high - energy cutoff for the x - ray spectrum at @xmath73 but is only logarithmically dependent on the value of this cutoff . in section [ res ] , we examine results for models with @xmath108 and @xmath109 , which bracket the range of plausible values . much more detail on high redshift inverse compton emission , including a discussion of potential observational tests , is given in @xcite . the opacity @xmath13 can be separated into two distinct components absorption by dust and gas within the emitting protogalaxy , which we term intrinsic absorption , and absorption by gas along the line of sight through the igm . intrinsic absorption is difficult to model with any degree of accuracy as it depends upon a number of variables the size and shape of the galaxy , its ionization state , the position of the sources within it , the dust content etc . rather than attempt to model these in detail a significant undertaking in itself we instead adopt a highly approximate representation . we assume that the emitted x - rays are attenuated by absorption by a neutral hydrogen column density of @xmath110 plus an associated neutral helium column density @xmath111 . this absorption is assumed to be the same for all sources . these values are chosen because they are representative of the column densities of the protogalaxies studied in this paper ( which presumably contain the bulk of the x - ray sources ) . reducing @xmath112 ( as would be appropriate if much of the surrounding were photoionized and/or dispersed by the progenitors of the x - ray sources ) has little or no effect on @xmath32 , as gas in the igm and within the protogalaxy itself quickly come to dominate the total absorption . increasing @xmath112 , on the other hand , has more significant effects : an order of magnitude increase in @xmath112 produces similar results to an order of magnitude decrease in the strength of the x - ray background , which , as we will see in section [ res ] , is generally sufficient to render x - ray feedback ineffective . consequently , we will overestimate the effect of the background if the bulk of the x - ray sources reside within massive galaxies . turning to the lyman - werner bands , we note that intrinsic absorption will generally be negligible within small protogalaxies , as their @xmath0 content is rapidly photodissociated @xcite . moreover , we also assume that the effects of dust absorption are negligible . in galaxies of primordial composition , this is obviously true ; in metal - poor galaxies , it should also be a good approximation , as very large column densities are required for significant dust obscuration ( for instance , @xmath113 for @xmath114 gas , if the dust - to - gas ratio is similar to that in the milky way ) . again , these assumptions break down if the majority of sources are to be found in massive , metal - rich galaxies , but we expect such galaxies to be extremely rare at the redshifts of interest in this paper . compared to intrinsic absorption , the effects of absorption due to the igm are much simpler to treat , particularly if we can assume that the bulk of the gas remains at an approximately uniform density . this assumption proves reasonable at high redshift for photons with mean free paths much greater than the typical clumping scale , as is the case for both lyman - werner band photons and x - rays . our treatment of igm absorption is discussed in the following sections . the continuum opacity of metal - free gas is very small @xcite and for our purposes can be neglected . consequently , the only significant sources of opacity encountered by lyman - werner photons are absorption by the lyman series lines of neutral hydrogen , and by the lyman - werner lines of molecular hydrogen . the lyman series lines have the effect of absorbing any lyman - werner photons of the same frequency , and reprocessing them to lyman-@xmath84 photons plus associated softer photons . as lyman-@xmath84 lies outside of the lyman - werner band , the net effect is to block from view any sources at redshifts higher than some maximum , @xmath115 , given by @xmath116 where @xmath117 is the frequency of the appropriate lyman series line and @xmath9 and @xmath8 are the observed frequency and redshift . clearly , the size of @xmath115 depends upon the distance between @xmath9 and @xmath117 , and thus more sources are seen at frequencies that are a long way from a line . as a result , the spectrum develops a characteristic ` sawtooth ' shape ( see figure 1 in @xcite ) , with the effect becoming more pronounced as one nears the lyman limit . absorption by molecular hydrogen is rather more complicated , due to the large number of lyman - werner lines that contribute to the opacity . if we approximate the lines as delta functions then an individual line produces an opacity @xmath118 where @xmath119 and @xmath120 are the oscillator strength and wavelength of the transition , @xmath121 is the number density of @xmath0 molecules in the level giving rise to the line at @xmath122 , the redshift of absorption , and @xmath123 is the hubble constant at @xmath122 . the value of @xmath122 is given by @xmath124 where @xmath125 . if we assume that all of the photons that are absorbed in the lines are permanently removed from the lyman - werner band , then the total opacity @xmath13 is simply given by the sum over all lines with @xmath126 : @xmath127 this sum potentially involves a very large number of lines , but can be greatly simplified by assuming that all of the @xmath0 is to be found in its ortho or para ground state ; at the redshifts of interest , the population of excited states will be negligible . in deriving this expression , we have assumed that every absorption permanently removes a lyman - werner band photon . this is not entirely correct . on average , only 15% of absorptions are followed by photodissociation of the @xmath0 molecule @xcite ; the rest of the time , the molecule decays back to a bound state , emitting a photon . in their treatment of this problem , @xcite assumed that the excited @xmath0 molecule would always decay directly back into the original state , and would thus emit a photon of the same energy as the one initially absorbed . in fact , this is not correct ( t. abel , private communication ) ; most decays occur initially to highly excited vibrational states , producing photons redwards of the lyman - werner bands . only a small fraction of decays ( about 5% ) take place directly into the original state , while a slightly larger fraction ( about 15% ) produce photons that lie elsewhere in the lyman - werner band system @xcite . we do not include the effect of these photons , however ; an accurate treatment would be quite complicated and is almost certainly unnecessary as we shall see in section [ res ] , @xmath0 in the igm is rapidly destroyed by the growing lyman - werner background and is completely negligible by the time that negative feedback begins . at x - ray energies , the opacity of the intergalactic gas is dominated by the ionization of neutral hydrogen and helium ; prior to recombination , the @xmath128 abundance is small and can be neglected . the x - ray opacity can thus be written as @xmath129 \frac{{\,{\rm d}l}}{{\,{\rm d}z } } { \,{\rm d}z},\ ] ] where @xmath130 and where @xmath131 and @xmath132 are the absorption cross - sections of neutral hydrogen and helium respectively , with @xmath133 and @xmath134 being the corresponding number densities . as long as the fractional ionization of the igm remains small ( i.e. a few percent or less ) , the ratio between @xmath135 to @xmath136 can be accurately approximated by its primordial value @xmath137 where @xmath138 is the helium mass fraction , and we can write equation [ tauigm ] purely in terms of @xmath136 as @xmath139 n_{{\rm{h } } } \frac{{\,{\rm d}l}}{{\,{\rm d}z } } { \,{\rm d}z}.\ ] ] this integral is readily computable by means of numerical integration . in the previous section , we showed that , given a simple star formation model , it is relatively easy to calculate the evolution of the lyman - werner and x - ray backgrounds . two of the parameters of our star formation model the star formation efficiency @xmath24 and starburst lifetime @xmath22 we treat as free parameters ( although they can be constrained to some extent see , e.g. @xcite ) . the remaining parameter , @xmath32 , the temperature at which efficient cooling becomes possible , is determined by the strength of the backgrounds themselves . this clearly presents us with a problem : the evolution of @xmath32 is coupled to that of the backgrounds , and to know one we must first know the other . fortunately , this difficulty is easily avoided . we know that at high redshift the number of protogalaxies , and hence the star formation rate , must be very small . consequently , there must be some redshift above which the external radiation field will become too weak to affect galaxy formation . the precise redshift at which this occurs is model dependent , but for the models examined in this paper we typically find that radiative feedback is negligible above @xmath140 . by choosing an initial redshift @xmath141 , therefore , we can be sure that in our initial simulation the background radiation will have no effect . given this starting point , we next proceed incrementally to lower redshifts via the following procedure : 1 . given @xmath142 , we calculate the background radiation field at @xmath143 , assuming that @xmath144 . 2 . using the computed background , we simulate the chemical and thermal evolution of a protogalaxy with @xmath145 ; the details of this simulation are outlined in sections [ comp ] to [ halt ] below . the main aim of this simulation is to determine whether the protogalactic gas can cool efficiently . if the protogalactic gas cools , then our assumed value of @xmath32 is correct ; we store this result , and return to step one to proceed to the next redshift . if the gas fails to cool , we continue to step four . we increment our assumed value of @xmath146 by a small amount @xmath147 , and recalculate the background radiation field . we assume that @xmath32 varies linearly over @xmath148 . given the new background , we return to step two . provided that @xmath148 and @xmath147 are both small , the error in @xmath146 will also be small ; this is particularly the case once emission from larger protogalaxies ( which cool via lyman-@xmath84 radiation ) begins to dominate the background . this approach reduces the coupled problem to the simpler one of determining whether a protogalaxy with virial temperature @xmath149 and formation redshift @xmath150 will cool when exposed to a particular background radiation field . to answer this question , we need to be able to model the thermal and chemical evolution of the protogalaxy . our approach to this problem is outlined in the following sections . ideally , we would like to use a high - resolution hydrodynamical simulation to follow the thermal and chemical evolution of the protogalactic gas ( see , e.g. * ? ? ? unfortunately , including the effects of radiative transfer , particularly of photons in the lyman - werner bands , into such a simulation is not currently feasible . we are thus forced to approximate . in choosing an appropriate approximation , we are also motivated by the desire to minimize the computational requirements of our simulations , so that we can explore the effects of a variety of different source models . we make three main approximations : 1 . we assume spherical symmetry . this is a reasonable approximation for the first generation of protogalaxies , but clearly is incorrect in detail ( see , e.g. figure 2 of * ? ? ? we assume that the protogalactic gas is _ static _ , at least on the timescale of the simulation . this allows us to ignore the hydrodynamical evolution of the gas , and also substantially simplifies the treatment of radiative transfer . this assumption clearly breaks down once the gas begins to cool strongly and loses its pressure support , but as we are only interested in the evolution up to this point , this is not a significant problem . we assume that all of the @xmath0 molecules remain in the rotational and vibrational ground state , in either ortho or para form . this simplification allows us to evolve the chemistry and radiative transfer on the timescale on which the total @xmath0 abundance changes ( typically @xmath151 @xmath152 ) rather than that on which the individual level populations change ( @xmath153 @xmath154 ) . it also simplifies our treatment of the radiative transfer . we discuss this approximation in more detail in section [ pd ] . together , these approximations allow us to solve for the chemical and thermal evolution of a model protogalaxy in a matter of minutes on a fast desktop computer . this allows us to study the redshift evolution of @xmath32 at high resolution in both temperature and redshift and for a number of different x - ray source models . however , this approach has an obvious drawback we can not be sure that our approximations give a fair representation of the real protogalaxy . of particular concern is the neglect of the hydrodynamic evolution of the gas , and the consequent error in the density profile . this is potentially significant because the @xmath0 cooling rate , along with many of the chemical reaction rates , scales as the square of the density . small errors in the density can thus lead to larger ones in the computed temperature . these concerns are mitigated to some extent , however , by the close agreement between the results of detailed numerical simulations and previous semi - analytic treatments . for instance , the values of @xmath32 obtained from the smoothed - particle hydrodynamics simulations of @xcite agree well with the results of @xcite , despite the highly approximate uniform density profile adopted by the latter group . similarly , the results of @xcite , obtained with a three - dimensional adaptive mesh hydrodynamical code broadly agree with those of @xcite , who use a static model similar to that presented here . together , these results suggest that @xmath32 is insensitive to the precise details of the density profile , but clearly this remains an area of concern . our computational method can be broken down into three main stages initialization of the density profile and the chemical abundances , computation of the thermal and chemical evolution of the gas , and termination of the simulation at a suitable point . these are described below . the protogalactic density profile is modeled as a truncated isothermal sphere , with central overdensity @xmath155 and truncation radius @xmath156 the virial temperature and redshift of formation of the protogalaxy completely specify its density profile . we subdivide this profile into @xmath157 spherical shells of uniform thickness and compute the mean density within each shell . we have run a number of test simulations with different values of @xmath157 , and find that setting @xmath158 provides sufficient spatial resolution to accurately determine @xmath32 . we assume that the initial chemical composition of the protogalactic gas is the same as that of the intergalactic medium . at our initial redshift @xmath159 , we take this from @xcite . at lower redshifts , the chemical evolution of the igm is influenced by the lyman - werner and x - ray backgrounds . we therefore calculate the intergalactic abundances explicitly , using the chemical model outlined in section [ chem ] , by solving the chemical rate equations : @xmath160 where @xmath161 and @xmath162 are source and sink terms for @xmath163 . at the same time we also solve for the temperature of the intergalactic gas @xcite @xmath164 where @xmath165 is the hubble constant , @xmath166 is the total particle number density and where @xmath167 and @xmath168 are the net heating and cooling rates ( see section [ cool ] ) . we solve this set of ordinary differential equations with the stifbs integrator of @xcite . as the size of the required timestep is generally much smaller than the redshift interval @xmath148 that separates our individual determinations of @xmath32 , we compute intermediate values by interpolation and from these determine the strength of the radiation background and hence the photochemical rates . although our main aim in following this chemistry is to determine the correct initial abundances for our simulations of protogalactic evolution , the results are of interest in their own right and are presented and discussed in section [ ther_chem ] . our treatment of the chemistry of the igm does not include the effects of the ionizing photons from stars ( and/or quasars ) that are ultimately responsible for the reionization of the intergalactic gas . this is justified at early epochs , as these photons are confined within small regions surrounding the luminous sources , but this simplification restricts the validity of our results to the period prior to cosmological reionization . the post - reionization epoch , and the effect of reionization on galaxy formation , have received extensive study elsewhere ( see * ? ? ? * and references therein ) . to simulate the chemical evolution of the protogalactic gas , we adopt a chemical model consisting of thirty reactions between nine species : @xmath169 , @xmath170 , @xmath171 , @xmath172 , @xmath128 , @xmath173 , @xmath0 , @xmath174 and free electrons . the reactions included in the model are summarized in table [ chemtab ] , together with the source(s ) of the data used . this model is based in large part on that of @xcite , but has been modified to improve its accuracy when applied to optically thick gas . aside from a number of updates to the reaction coefficients in the light of new data , the main differences are as follows : 1 . we include the contribution to the hydrogen ionization rate arising from the ionizing photons produced by @xmath128 recombination , in line with the discussion in chapter 2 of @xcite . although commonly a small correction to the total rate , this can become significant when x - ray photoionization dominates . 2 . to enable us to accurately determine the @xmath128 abundance , we find that we need to include the effects of charge transfer between @xmath128 and @xmath169 ( reaction 20 ) , as this can be comparable to the recombination rate when the fractional ionization is small . for completeness we also include the inverse reaction ( no . 21 ) , although this is unimportant at @xmath175 . we include the contribution to the ionization rates of hydrogen and helium arising from secondary ionization by energetic photoelectrons , based on the recent calculations of @xcite . the contribution of secondary ionization to the other photoionization rates is small and can be neglected . 4 . we do not include the photodissociation of @xmath0 by photons above the lyman limit ( reaction 28 in @xcite ) , as in optically thick gas this will be negligible compared to the effects of @xmath0 photoionization ( reaction 26 ) . on the other hand , we do include the effects of dissociative photoionization ( reaction 27 ) , which becomes significant for photon energies greater than 30@xmath176 . more information about all of these points , and the chemical model generally , can be found in @xcite . .a list of the reactions included in our chemical model of protogalactic gas . values for the rate coefficients ( or radiative cross - sections where appropriate ) are given in @xcite . references are to the primary source(s ) of the data whenever possible ; in many cases , we have also used analytical fits to this data from @xcite , @xcite or @xcite . [ cols= " < , < , < " , ] given the initial temperature and chemical abundances , plus the set of chemical reaction rates , actually solving for the thermal and chemical evolution is relatively easy . as in the igm case , we simply solve the coupled set of chemical rate equations @xmath160 @xmath177 using the stifbs integrator of @xcite . at the start of each timestep ( hereafter time @xmath178 ) , we compute the photochemical rates as outlined in sections [ rtmethod ] . we then use stifbs to solve for the new chemical abundances and new temperature at the end of the timestep ( time @xmath179 ) , repeating this for each shell in turn . we next store these values , return to time @xmath178 , and recalculate them using the same procedure , but with _ two _ timesteps of length @xmath180 . we recalculate the photochemical rates at the intermediate point . we next test for convergence by comparing our two sets of results . if any of the chemical abundances or temperatures of any of the shells differ by more than 0.1% , then we reject the results and begin again from time @xmath178 with a smaller timestep . otherwise , we check to see whether we need to halt the simulation , using the criteria discussed below , and , if we do not , we begin the computations for a new timestep starting from @xmath179 . one final approximation that we find useful in practice is to fix the @xmath170 and @xmath174 abundances at their equilibrium values . this allows the integrator to take much larger timesteps than would otherwise be possible , but introduces very little error into the computed @xmath0 abundances . using the method outlined in the preceding sections , we compute the chemical and thermal evolution of the protogalaxy until one of two conditions is met : either the protogalactic gas begins to cool strongly , or we exceed a preset time limit , @xmath181 . to assess whether gas cooling is ` strong ' enough requires an objective cooling criterion . a number of different possibilities have been suggested in the literature @xcite . in our simulations , we adopt the criterion used by @xcite : we require that the elapsed time exceeds the cooling time , as computed at the edge of the protogalactic core , at a distance @xmath182 from the centre of the protogalaxy . the advantage of this choice is that it avoids giving us a false positive result in cases where @xmath183 drops briefly below @xmath184 at early times , but remains so for a time @xmath185 . as an additional sanity check , we also require that the final temperature be smaller than the initial temperature . if the protogalactic gas does not cool strongly , then the simulation will terminate when it reaches @xmath181 . this pre - set time limit is required on purely practical grounds , to prevent simulations in which the gas does not cool from running for excessive amounts of time , but also has a physical justification . in our simulations , we treat protogalaxies as isolated objects , uninfluenced by external events . in reality , they are part of a dynamically evolving mass distribution , and the majority will only survive for a limited time before merging with other protogalaxies of a similar or larger size . it is possible to use the press - schechter formalism to calculate the distribution of survival times as a function of mass @xcite , but for rare objects the mean survival time is typically of the order of the hubble time and thus for simplicity we set @xmath186 . in the following sections , we present results from a number of simulations that examine the effects of the x - ray backgrounds produced by the various source models discussed in sections [ def_xray ] and [ alt_xray ] . unless otherwise noted , all of these simulations assume the same parameters for the star formation model : a standard salpeter imf , with @xmath187 and @xmath188 , a star formation efficiency @xmath189 and a starburst lifetime @xmath190 . additionally , all of the simulations use the same cosmological model , the @xmath168cdm concordance model of @xcite , which has parameters @xmath191 . before examining the effect of an x - ray background on protogalactic evolution , we first briefly study the evolution of @xmath32 in its absence . as well as allowing us to determine the sensitivity of our results to variations in the uv source model , this also provides us with a necessary baseline against which to compare our other results . in figure [ baseline ] , we plot the evolution of @xmath32 with redshift for the three different star formation models discussed in section [ uvdensity ] . all three models assume a salpeter imf , with maximum stellar mass @xmath51 , as well as our standard star formation efficiency , starburst lifetime and cosmological model , described previously . our basic model assumes a metal - free stellar population , with a minimum stellar mass of @xmath187 ; the corresponding results are given by the dotted line in figure [ baseline ] . the dashed line illustrates the effect of increasing the minimum mass to @xmath50 ; the solid line assumes the same @xmath53 , together with a metallicity @xmath57 . figure [ baseline ] demonstrates that although the strength of the lyman - werner background increases by almost a factor of five as we move from our basic model to the metal - enriched model , this has less effect on @xmath32 than we might expect : the difference in @xmath32 between the three models is never more than 50% , and the qualitative details of its evolution are very similar in all three models . this suggests that the uncertainty introduced by our lack of knowledge of the properties of the primordial stellar population need not be unduly limiting . however , it is also clear that some uncertainty remains , and this will place a lower limit on the magnitude of any effect that we can reliably claim to detect , as small variations in @xmath32 due to the x - ray background will be swamped the by the error resulting from the uncertainty in the lyman - werner background . in the work that follows , we take as our baseline the results of our basic , metal - free , low @xmath53 model ; as figure [ baseline ] demonstrates , this minimizes the strength of the lyman - werner background , and thus will tend to maximize the effectiveness of the x - ray background . in figure [ base_xray ] , we plot the evolution of @xmath32 in the presence of the x - ray background generated by the massive x - ray binary model described in section [ def_xray ] . for the purposes of comparison , we also plot the results of our basic x - ray free model . initially , the evolution of @xmath32 is the same in both models , implying that the x - ray background has little or no effect on the gas . at a redshift @xmath192 and critical temperature @xmath193 , however , the models begin to diverge significantly . in the x - ray free model , the critical temperature continues to increase rapidly until it reaches its maximum value of @xmath194 . in the x - ray binary model , by contrast , the rate of increase of @xmath32 is significantly slower , and it fails to reach its maximum value by the end of the simulation at @xmath195 . because at lower redshift we expect the effects of ionizing radiation from both stellar sources and quasars to become increasingly important and thus our results to become unreliable . clearly if reionization occurs at @xmath196 , the same is true for some portion of the results plotted here . ] it is clear from figure [ base_xray ] that the presence of the x - ray background significantly affects the evolution of gas in the larger of the @xmath0-cooled protogalaxies . in small protogalaxies , on the other hand , the negative feedback caused by the lyman - werner background remains as strong as ever . a simple way to judge the importance of this effect is to examine the difference it makes to the fraction of gas in the universe that can collapse and cool . as we saw in section [ sfr ] , we can use the press - schechter formalism to express the cool gas fraction as @xmath39.\ ] ] using the relationship between @xmath43 and @xmath32 derived in that section , it is straightforward to calculate the evolution of @xmath45 in both the x - ray binary and x - ray free models . to better highlight the difference between the two models , we plot in figure [ fcoll ] the ratio of the cooled gas fraction in the x - ray binary model to that in the x - ray free model . at high redshift , the evolution of @xmath32 is the same in both models , and thus the ratio is one . below @xmath197 , the behaviour of the models begins to diverge , but this has little effect on the ratio until the models begin to diverge sharply at @xmath198 . thereafter , the ratio rises sharply to a peak at @xmath199 , and subsequently declines as the growth in the cooled mass fraction becomes dominated by the formation of larger protogalaxies that cool via lyman-@xmath84 emission . figure [ fcoll ] shows us that by ignoring the effect of the x - ray background we underestimate @xmath45 by at most a factor of two . in fact , the difference is likely to be even smaller : we have assumed that all of the gas in a protogalaxy with @xmath200 will cool , but this need not be the case if an x - ray background is present , as x - ray heating will prevent gas from cooling in the low - density outer layers of the protogalaxy . whether this will affect the star formation rate within individual protogalaxies is not clear , but in any case the difference in the globally - averaged star formation rate is unlikely to be greater than a factor of two . note , however , that this extra star formation occurs entirely in low - mass systems , from which ionizing photons @xcite and supernova ejecta @xcite can more readily escape . doubling the star formation rate may thus significantly increase the feedback of primordial star formation on the igm . finally , given the uncertainties in the data underlying our simple massive x - ray binary model ( see section [ def_xray ] ) , it is of interest to investigate the sensitivity of our results to changes in the strength of the x - ray background . accordingly , we have run additional simulations in which the strength of the x - ray sources was increased or decreased by a factor of ten . the results are plotted in figure [ alt_norm ] , together with the results of our basic x - ray binary model and of the x - ray free model . unsurprisingly , increasing or decreasing the strength of the x - ray background alters the evolution of @xmath32 . decreasing it by an order of magnitude increases @xmath32 to the point where its evolution is little different from that in the x - ray free model . increasing it by an order of magnitude , on the other hand , systematically lowers @xmath32 , although never by more than a factor of two . as we discussed in section [ alt_xray ] , it is possible that massive x - ray binaries are much less abundant at high redshift than at the present day . if so , then the x - ray emission of star - forming galaxies will be dominated by supernova remnants ( snr ) . these will generate x - rays through two main emission mechanisms : thermal bremsstrahlung from hot gas , and inverse compton scattering of the cmb by relativistic electrons . we consider the effects of these mechanisms separately . we examined two possible variants of the thermal bremsstrahlung model . in one , we assumed that the characteristics of the emission are broadly the same as those observed locally , with a fraction @xmath99 of the supernova energy being radiated as x - rays with a characteristic temperature @xmath201 @xcite . in the other model , we assumed that all high - redshift supernovae explode in extremely dense surroundings , producing x - ray bright , ultra - compact remnants with @xmath202 and @xmath203 @xcite . realistic models should lie somewhere between these two extremes . however , we found that in neither of these cases does the x - ray background affect the evolution of @xmath32 : at our level of temperature resolution , the results are identical to those obtained for the x - ray free model . the obvious conclusion is that the background produced by bremsstrahlung is simply too weak to be effective . our inverse compton model fares somewhat better . in this model , the strength of the x - ray background is proportional to the mean fraction @xmath106 of the supernova explosion energy that is transferred to relativistic electrons within the remnant . this value is not known accurately and so we considered two possible cases , one with @xmath108 and another with @xmath109 , these being conservative lower and upper bounds on the true value . in the former case , we again saw no significant effect on the evolution of @xmath32 . in the latter case , on the other hand , we saw results very similar to those obtained for the x - ray binary model , as illustrated in figure [ sn_ic ] . in principle , therefore , inverse compton emission from supernova remnants could be as important an x - ray source as emission from x - ray binaries . ultimately , however , its importance depends upon the value of @xmath106 , and no firm conclusions are possible until this value is better constrained . in this context , the possible observational tests suggested by @xcite could prove extremely valuable . as well as determining the evolution of @xmath32 , our simulations also allow us to study the thermal and chemical evolution of the diffuse intergalactic medium , as outlined in section [ model_pg ] . as a simple example , we plot in figure [ h2_in_igm ] the evolution of the fractional @xmath0 abundance in the igm for the x - ray free model ( solid line ) and the x - ray binary model ( dashed line ) . in both cases , the fractional abundance rapidly decreases from its primordial value due to photodissociation by the ultraviolet background , reaching @xmath204 by @xmath205 . subsequently , its decline slows , in part because the rate of increase in the strength of the ultraviolet background also slows . below @xmath206 , the behaviour of the two models diverges . in the x - ray free model , @xmath207 continues to decline until the end of the simulation . in the x - ray binary model , on the other hand , the increasing ionization of the igm boosts the @xmath0 formation rate to the point where it overtakes the photodissociation rate and the @xmath0 abundance , after reaching a minimum at @xmath208 , begins to climb . nevertheless , it remains extremely small at the end of the simulation , readily justifying our assertion in section [ lwabs ] that absorption by intergalactic @xmath0 does not play a significant role in determining the strength of the lyman - werner background . in figure [ igm_ion ] , we plot the evolution of the fractional ionization of the igm . in the x - ray free model , this remains approximately constant over the lifetime of the simulation , as the recombination timescale is significantly longer than the hubble time . in the x - ray binary model , on the other hand , photoionization by the growing x - ray background eventually overcomes the very small recombination rate and drives the fractional ionization upwards , increasing it by just over an order of magnitude by the end of the simulation . even so , it remains small , demonstrating that the x - ray background does not contribute significantly to cosmological reionization . the effect on the fractional ionization of helium ( the ratio of @xmath128 to @xmath172 ) is rather more striking . the post - recombination @xmath128 abundance is extremely small @xcite , and in the x - ray free model remains at this low level throughout the simulation . in the x - ray binary model , on the other hand , it increases dramatically over the course of the simulation , reaching @xmath209 by @xmath38 . this is illustrated in figure [ igm_heion ] . finally , in figure [ igm_heat ] we plot the evolution with redshift of the temperature of the igm . in the x - ray free model , adiabatic cooling dominates the thermal evolution , and the temperature falls off approximately as @xmath210 . in the x - ray binary model , on the other hand , photo - electric heating begins to heat the igm strongly at @xmath211 , driving the temperature up to @xmath212 by the end of the simulation . thus , although the x - ray background does not contribute significantly to the reionization of the igm , it does produce substantial reheating prior to reionization . moreover , given the large mean free path of the x - ray photons , this reheating occurs almost uniformly throughout the igm , rather than being localized to the vicinity of star - forming galaxies . one consequence of this reheating is that the formation of very small - scale structure will be suppressed , as the increased temperature of the igm leads to an increased jeans mass . this is unlikely to affect the global star formation rate , however , as star formation within these small structures would in any case be strongly suppressed by the ultraviolet background . nevertheless , it will reduce the mean clumping factor of the igm below the level that we would otherwise predict , which may in turn speed up reionization @xcite . reheating also affects the visibility of the igm in the redshifted @xmath213 line of neutral hydrogen . @xcite show that scattered lyman-@xmath84 emission from high redshift galaxies efficiently couples the spin temperature of the hyperfine levels to the kinetic temperature of the gas . if the gas temperature is smaller than the cmb temperature , this results in @xmath213 line absorption ; if it is greater , then it results in emission . absorption is easier to detect than emission @xcite , but the heating produced by the x - ray background implies that absorption occurs only at @xmath214 , and that concentrating on detecting @xmath213 emission may be the more viable strategy . the results of the previous section allow us to assess the impact of the high redshift x - ray background that is produced by star - forming galaxies . if we assume that the x - ray emission of these galaxies is similar to that observed locally , and that the same correlation between x - ray luminosity and star formation rate applies , then we find that the background produced is strong enough to partially offset the effects of uv photodissociation in large ( @xmath215 ) , @xmath0-cooled protogalaxies . however , local emission is dominated by massive x - ray binaries , which may not form in large numbers at high redshift . therefore , we have also explored the effect of an x - ray background produced by emission from supernova remnants . if this emission is dominated by inverse compton scattering and if the fraction of the supernova energy transferred to the relativistic electrons powering this emission is large , then the resulting background has very similar effects to one produced by x - ray binaries . on the other hand , if the fraction of energy transferred is small , then the background has little or no effect . in addition to inverse compton emission , we have also examined the effect of thermal bremsstrahlung emission from hot gas in the remnants , and find that even if all supernovae were to form x - ray bright , ultra - compact remnants , the resulting x - ray background would still be too weak to significantly affect protogalactic evolution . finally , none of these models produces an x - ray background that is strong enough to balance uv photodissociation in small protogalaxies , with virial temperatures @xmath216 . in these protogalaxies , negative feedback always dominates . how significant are these results ? one simple way to assess this is to study the evolution of the mass fraction of cooled gas , which represents the total amount of matter available to form stars . comparing its evolution in the x - ray binary model to that in the absence of an x - ray background , we find that it is increased by approximately a factor of two . given our star formation model , this corresponds to an increase in the global star formation rate by the same amount . however , this is small compared to the order of magnitude increase that would result if we were simply to ignore the effect of the uv background ( see figure 7 of * ? ? ? * ) . in reality , the difference between the two models may be greater than this because the additional star formation takes place entirely in low - mass systems , from which ionizing photons and supernova - produced metals can readily escape @xcite . doubling the star formation rate may therefore have more than double the impact on the intergalactic medium . however , to properly assess the ultimate importance of this effect requires more detailed modelling , beyond the scope of this paper . ultimately , understanding the history of star formation in the small protogalaxies studied in this paper remains important even if they do not contribute to the reionization or enrichment of the igm to any great degree . this is simply because , in a hierarchical universe , these protogalaxies are the building blocks from which larger galaxies form and therefore set the initial conditions for later stages of galaxy formation . in particular , very little metal enrichment of the primordial gas is required in order to allow the cno cycle to operate and population ii ( rather than population iii ) stars to form , and yet this can have a profound effect on the predicted spectral energy distribution of an early stellar population . although the main purpose of our study was to examine the effects of the x - ray background on the thermal and chemical evolution of gas within protogalaxies , our approach also allows us to examine the effects of the background on the diffuse igm . our main results are three - fold : 1 . we confirm the rapid destruction of @xmath0 in the intergalactic medium noted by @xcite , but also show that when an x - ray background is present the @xmath0 abundance does not continue to decline indefinitely , but eventually stabilizes and may even begin to increase . however , it never becomes large enough to significantly affect the lyman - werner background . we show that although photoionization by the x - ray background significantly increases the fractional ionization of the igm ( and in particular the fractional ionization of helium ) , the bulk of the gas remains mostly neutral , demonstrating that the contribution of the x - ray background to cosmological reionization is small . we find that the x - ray background will also heat the intergalactic gas , raising its temperature to @xmath217 by @xmath38 ( compared to @xmath218 in the x - ray free model ) . this will suppress the formation of structure on the smallest scales by increasing the jeans mass . it is unlikely to affect the global star formation rate , since @xmath219 , but may speed up the process of reionization by reducing the mean clumping factor of the igm . a number of other authors have studied the effects of radiative feedback on the formation of @xmath0-cooled protogalaxies @xcite . most of these studies have concentrated solely on the effects of the lyman - werner background , generally finding that it suppresses cooling ( and hence star formation ) by @xmath19830 , prior to cosmological reionization . our x - ray free simulations support this conclusion . in particular , comparison with the results of @xcite , who use a very similar method but with a different implementation of radiative transfer , gas chemistry and cooling , shows good agreement . rather less work has been done on the effects of the high redshift x - ray background . @xcite examined the formation of molecular hydrogen in a constant density primordial cloud illuminated by a power - law uv spectrum extending into the hard x - rays , and found that in some cases , the background could enhance @xmath0 formation . this lead @xcite to examine the effects of a quasar - produced x - ray background , using a very simple model in which the x - ray flux is a fixed fraction of the lyman - werner band flux , modulated by absorption by a fixed column density of @xmath169 and @xmath172 . they found that an x - ray to uv flux ratio of 10% was enough to overcome negative feedback , and that a higher flux ratio could potentially produce positive feedback . recently , @xcite have studied the effects of similar model backgrounds using an adaptive - mesh hydrodynamics code . they also find that x - rays reduce the effectiveness of negative feedback , but that the latter still dominates . they do not find evidence for the positive feedback predicted by @xcite . however , their simulations do not include the effects of @xmath0 self - shielding , and thus potentially underestimate the amount of @xmath0 that forms ( although see * ? ? ? * for a different view ) . there are several significant differences between our work and these previous investigations . firstly , we do not assume a fixed spectrum or intensity for the x - ray background ; rather , we specify the properties of the x - ray _ sources _ and subsequently compute the build - up of the background in a self - consistent fashion . moreover , we consider source models where the x - ray emission is proportional to the star formation rate , as is observed to be the case for star - forming galaxies at low redshift ; the relationship between star formation rate and x - ray emission in quasar - based models is far less clear . as a result , we generally consider x - ray backgrounds signficantly fainter than those studied in the papers cited above . this makes direct comparison of our results difficult . however , we note that @xcite find x - ray feedback to be ineffective below @xmath220 , regardless of the strength of the x - ray background ; at the redshifts they consider , this corresponds to a virial temperature @xmath221 , and thus agrees well with the similar result obtained in this paper . in closing , we note that a number of uncertainties still remain in our treatment of this problem . some of these the high redshift star formation rate or the appropriate population iii initial mass function , for instance we simply do not know at the present time . fortunately , changes to our assumed values can be readily incorporated in the framework laid out in this paper . other uncertainties arise from our method of simulation ; in particular , from our assumption of a static density profile for the protogalactic gas . we hope to address these issues in future work . the authors would like to acknowledge detailed comments by the anonymous referee which helped to significantly improve this paper . scog would also like to acknowledge useful discussions on various aspects of this work with tom abel , omar almaini , rennan barkana , marie machacek and si peng oh . financial support for much of this work was provided by a pparc studentship . additional support was provided by nsf grant ast99 - 85392 .

the combination of ground - based astrometric compilation catalogues , such as the fk5 or the gc , with the results of the esa astrometric satellite hipparcos produces for many thousands of stars proper motions which are significantly more accurate than the proper motions derived from the hipparcos observations alone . in paper i ( wielen et al . 1999 , a&a 347 , 1046 ) we have presented a method of combination for single stars ( si mode ) . the present paper ii derives a combination method which is appropriate for an ensemble of ` apparently single - stars ' which contains undetected astrometric binaries . in this case the quasi - instantaneously measured hipparcos proper motions and positions are affected by ` cosmic errors ' , caused by the orbital motions of the photo - centers of the undetected binaries with respect to their center - of - mass . in contrast , the ground - based data are ` mean values ' obtained from a long period of observation . we derive a linear ` long - term prediction ' ( ltp mode ) for epochs far from the hipparcos epoch , and a linear ` short - term prediction ' ( stp mode ) for epochs close to . the most accurate prediction for a position at an arbitrary epoch is provided by a smooth , non - linear transition from the stp solution to the ltp solution . we present an example for the application of our method , and we discuss the error budget of our method for the fk6 ( a combination of the fk5 with hipparcos ) and for the combination catalogue gc+hip . for the basic fundamental stars , the accuracy of the fk6 proper motions in the ltp mode is better than that of the hipparcos proper motions ( taking here the cosmic errors into account ) by a factor of more than 4 .

astro-ph0007262

in paper i ( wielen et al . 1999b ) , we have shown that the combination of the data of the hipparcos astrometric satellite ( esa 1997 ) with ground - based results ( such as the fk5 ) is able to provide for many stars individual proper motions which are significantly more accurate than the hipparcos proper motions themselves . the method has been already successfully applied in the construction of the fk6 , the sixth catalogue of fundamental stars ( part i of the fk6 : wielen et al . 1999d , part iii of the fk6 : wielen et al . 2000a ) . the method of combination presented in paper i is , however , strictly valid for single stars only . in the fk6 , we call this procedure therefore the ` single - star mode ' . in reality most of the stars are members of binaries or of multiple systems . if the duplicity of an individual object is already definitely known , either from ground - based investigations or from hipparcos observations , then the method of combination has to be changed properly in order to obtain meaningful results . we call such procedures ` special solutions ' . paper iii of this series of papers will discuss the special solutions for visual binaries and other types of double stars . the present paper ( paper ii ) describes an appropriate method of the combination of hipparcos results with ground - based observations for ` apparently single stars ' . even if we have removed from such a sample of apparently single stars all the objects with known duplicity , there remains beside the truly single - stars a large number of hitherto undetected astrometric binaries . the measured photo - center of an unresolved astrometric binary moves on the sky on a wavy curve , in contrast to the linear motion of single stars ( fig . 1 ) . in such a case , an ` instantaneously ' measured proper motion deviates from a ` mean ' proper motion , averaged over a long interval of time . the proper motions provided by hipparcos ( esa 1997 ) are essentially such instantaneously measured quantities , since they are derived from positional measurements spread over about three years only . in contrast , the proper motions given in the fk5 ( fricke et al . 1988 , 1991 ) are long - term averages over up to about 200 years . we have called the difference between the instantaneous proper - motion and the mean one the ` cosmic error ' of the instantaneous proper motion ( wielen 1995a , b , 1997 , wielen et al . 1997 ) . in some cases , the individual cosmic error is so large that the duplicity of an apparently single star is strongly indicated by this fact . we have called such objects ` @xmath2 binaries ' ( wielen et al . 1999a ) . in most cases , however , the cosmic error in the hipparcos proper motion of a hitherto undetected astrometric binary is not _ individually _ significant , but can be shown to exist only _ statistically _ in a larger sample of apparently single stars . nevertheless , an appropriate combination method should not neglect the statistical consequences of these cosmic errors . our comparison of hipparcos proper motions with ground - based results has shown that , at least for brighter stars , the average cosmic error in a hipparcos proper motion is often larger than the hipparcos measuring error , typically by a factor of three ( wielen 1995a , b , wielen et al . 1997 , 1998 , 1999c ) . wielen ( 1997 , henceforth called paper wpsa ) has developed a coherent scheme of ` statistical astrometry ' for handling the effects of cosmic errors in high - precision astrometry . in the following sections , we shall apply the concepts of statistical astrometry to the problem of combining the hipparcos results with ground - based measurements for a sample of apparently single stars . the main results will be solutions which we call the ` long - term prediction mode ( ltp ) ' and the ` short - term prediction mode ( stp ) ' . our results have already been applied for the ltp and stp solutions given in part i and part iii of the fk6 ( wielen et al . 1999d , 2000a ) . it is clear that in principle the most desirable solution for our problem would be to treat each star individually and fully correctly . this would mean ( 1 ) for truly single stars : to use the ` single - star mode ( si ) ' described in paper i , and ( 2 ) for binaries : to apply individual orbital corrections , as e.g. done for polaris by wielen et al . . however , since for ` apparently single stars ' the true nature ( single or double ) of the individual objects is unknown , we _ have _ to rely on statistical methods in order to handle such a sample of stars properly . the treatment of a sample of apparently single stars by our statistical procedures gives _ on average _ the best astrometric prediction , and it provides the most realistic error budget for such a sample . in this sense , our statistical treatment is certainly much more appropriate than to ignore the binary nature of a large fraction of a sample of apparently single stars altogether . in paper i , we have already pointed out that the older observations carry a high weight in the combination of ground - based measurements with hipparcos data , and that therefore the gc ( boss et al . 1937 ) should be also considered here because of its high number of rather well - measured stars . we call the result of the combination of the gc with hipparcos the combination catalogue gc+hip . with respect to the two catalogues which should be combined into a new one , we follow closely the situation described in sect . 2 of paper i. we are using also as far as possible the nomenclature of that section . we assume that two astrometric compilation catalogues are available , identified by the indices 1 ( e.g. for the fk5 ) and 2 ( e.g. for hipparcos ) . for the combined catalogue ( e.g. the fk6 ) , we use the index c , usually now supplemented by an additional subindex which identifies the special mode of the solution ( e.g. ltp or stp ) . each of the two basic catalogues ( @xmath3 ) provides for the stars a position @xmath4 at a central epoch @xmath5 and a proper motion @xmath6 for two angular coordinates ( e.g. right ascension @xmath7 and declination @xmath8 ) . the mean measuring errors of @xmath9 and @xmath6 are denoted by @xmath10 and @xmath11 . usually one of the catalogues ( e.g. @xmath12 ) has first to be reduced to the astrometrical system of the other catalogue ( e.g. the fk5 to the hipparcos system ) . in this case , @xmath13 and @xmath14 denote the already systematically corrected quantities , and their mean errors @xmath15 and @xmath16 include the uncertainty of the systematic corrections . for the determination of the systematic differences between two catalogues , we use methods developed for the construction of the fk5 ( bien et al . 1978 ) . with respect to the principles of statistical astrometry , we make now the important additional assumptions that catalogue 1 gives ` mean ' quantities for @xmath17 and @xmath18 , averaged over a long interval of time , while catalogue 2 provides ` instantaneously ' measured values of @xmath17 and @xmath18 . if we apply our scheme to a combination of the fk5 with hipparcos , both assumptions are fulfilled to a high degree of approximation : the fk5 is based on ground - based observations spread over about two centuries , while the hipparcos results are obtained from measurements made during a short interval of time , about three years only . in the terminology of statistical astrometry , our assumptions mean that we suppose that the fk5 is free from cosmic errors . the cosmic errors in the hipparcos positions and proper motions are denoted by @xmath19 and @xmath20 . the correlation functions @xmath21 , @xmath22 , and @xmath23 are explained in wpsa ( especially sect . 3 ) . numerical values for @xmath24 and @xmath25 as functions of the parallax @xmath26 will be provided in sect . 5 . in the combination of two astrometric catalogues of which at least one contains instantaneously measured data ( affected by cosmic errors ) , the predicted position @xmath27 at an arbitrary epoch @xmath28 should not be anymore a linear function of time . according to the principles of statistical astrometry , the ` best ' prediction @xmath27 for the true instantaneous position @xmath29 is given by the non - linear expression ( wpsa , eq . ( 82 ) , with a slight change in nomenclature ) : @xmath30 the quantities @xmath31 and @xmath32 are functions of time @xmath28 to be determined . for computational reasons , we often use @xmath33 instead of @xmath32 . the functions @xmath31 and @xmath32 are determined from the condition that the mean error @xmath34 of the predicted positions @xmath35 should be a minimum for every value of @xmath28 , averaged over the ensemble : @xmath36 the operator @xmath37 means , as in wpsa ( 1997 ) , the average of the quantity @xmath38 over the ensemble of similar stars . for an individual star , @xmath37 can be interpreted as the ` expectation value ' of the quantity @xmath38 for this star . in this sense , we consider the ensemble averages as statistical predictions for the mean behaviour of a typical individual member of the ensemble . inserting eq . ( 1 ) into eq . ( 3 ) , and using the scheme of statistical astrometry , we obtain for @xmath39 : @xmath40 the quantities @xmath41 and @xmath42 are the true mean position and the true mean proper motion of the star ( in the sense of statistical astrometry ) . @xmath43 is a linear function of @xmath28 , e.g. @xmath44 in deriving eq . ( 4 ) we have used the following relations : according to our assumption that catalogue 1 provides mean quantities , @xmath45 and @xmath14 differ from @xmath46 and @xmath47 by their measuring errors only : @xmath48 the quantities @xmath29 and @xmath49 are the true instantaneous position and proper motion of the star at epoch @xmath28 . since catalogue 2 is assumed to provide instantaneous quantities , @xmath50 and @xmath51 differ from @xmath52 and @xmath53 also by their measuring errors only : @xmath54 the correlation functions @xmath55 , and @xmath56 ( see wpsa ) occur because of : @xmath57 @xmath58 @xmath59 @xmath60 we have further to remember @xmath61 the products of the other terms occuring in the second line of eq . ( 4 ) all vanish , because the measuring errors are not correlated with the cosmic errors , for example @xmath62 and because the measuring errors of @xmath4 and @xmath63 are not correlated at the central epoch @xmath5 ( by definition of the central epoch ) , for example : @xmath64 and finally because the measuring errors of @xmath13 and @xmath14 are not correlated with those of @xmath65 and @xmath66 , for example @xmath67 the conditions for the minimum of @xmath39 with respect to @xmath68 and @xmath69 are @xmath70 and @xmath71 carrying out these procedures and using the final part of eq . ( 4 ) for @xmath39 , we obtain the following two equations of condition for the functions @xmath31 and @xmath32 which give the ` best ' values of @xmath72 : @xmath73 % ( 20 ) & & \gamma ( t ) \varepsilon^2_{\mu , 1 } ( t_2 - t_1 ) \nonumber\\ \lefteqn { + \ , \beta ( t ) ( t - t_2 ) ( \eta ( 0 ) + \varepsilon^2_{\mu , 1 } + \varepsilon^2_{\mu , 2 } ) } \nonumber\\ & & = \zeta ( t - t_2 ) + \varepsilon^2_{\mu , 1 } ( t - t_1 ) \,\ , . % ( 21)\end{aligned}\ ] ] the eqs . ( 20 ) and ( 21 ) correspond to the eqs . ( 86 ) and ( 87 ) in wpsa . solving the eqs . ( 20 ) and ( 21 ) for the unknowns @xmath31 and @xmath32 we obtain : @xmath74 @xmath75 with the auxiliary quantity @xmath76 : @xmath77 for deriving the functions @xmath31 and @xmath32 , we have up to now implicitely assumed that all the stars in the ensemble have the same measuring errors . this is , of course , not strictly true in reality . however , the ensemble averages are actually neccessary for handling the cosmic errors only , but not for treating the measuring errors . hence we shall use all the equations derived above by inserting the individual measuring errors if we treat individual stars of the ensemble . formally we may imagine to handle subsamples of stars in which the stars have the overall behaviour with respect to the cosmic errors , but in which the common measuring errors are equal to those of the individual star under consideration . it is a more severe problem that we do not make use of our knowledge of how large the individual cosmic errors in @xmath17 and @xmath18 are for a given individual object ( e.g. @xmath78 ) . as discussed in wpsa , this would require ` conditioned ' correlation functions . since this information is presently not available , we are treating here the consequences of the cosmic errors on the level of ensemble averages only . inserting these results for @xmath31 and @xmath32 from eqs . ( 22 ) and ( 23 ) into eqs . ( 1 ) and ( 4 ) , we derive the prediction @xmath27 for the instantaneous position @xmath29 of the star and the mean error @xmath72 of this prediction . the prediction @xmath35 is a non - linear function of @xmath28 , because the correlation functions @xmath79 and @xmath80 , which occur in the formulae for @xmath31 and @xmath81 , are non - linear functions . a typical run of @xmath82 is shown in fig . 2 . in order to illustrate the properties of our prediction @xmath27 , we consider in the following subsections 2.3 and 2.4 two limiting cases in which we either neglect the measuring errors or the cosmic errors . if we neglect all the measuring errors and set @xmath83 , then we obtain for @xmath31 and @xmath32 @xmath84 these results were already derived and discussed in wpsa ( sect . 4.2.4 and fig . 10 ) . the corresponding prediction @xmath85 , shown in fig . 2 , passes through the point @xmath50 with the slope @xmath51 . therefore , the measured instantaneous position and proper motion at epoch @xmath86 are exactly reproduced by @xmath85 . for @xmath87 , the prediction @xmath85 approaches asymptotically the mean position @xmath88 of the star . the uncertainty @xmath89 of the prediction is given by @xmath90 we see that @xmath91 is zero at the epoch @xmath86 and approaches @xmath92 for @xmath87 , since in this limit @xmath27 is equal to @xmath41 , and @xmath41 differs from the instantaneous position @xmath29 by the cosmic error @xmath93 on average . if we neglect the cosmic errors and set @xmath94 , then we obtain @xmath95 % ( 28 ) \lefteqn { b_{\rm{nce } } ( t ) = \beta_{\rm{nce } } ( t ) ( t - t_2 ) = \big ( - \varepsilon^2_{x , 1 } \varepsilon^2_{\mu , 1 } ( t_2 - t_1 ) } \nonumber\\ & & + \ , ( \varepsilon^2_{x , 1 } + \varepsilon^2_{x , 2 } ) \varepsilon^2_{\mu , 1 } ( t - t_1)\big ) / n_{\rm{nce } } \,\ , , % ( 29)\end{aligned}\ ] ] with @xmath96 equation ( 29 ) illuminates one of the advantages of introducing @xmath97 as a substitute for @xmath32 . while @xmath98 , and in general @xmath99 for non - zero measuring errors , remains finite , the quantity @xmath100 , and in general @xmath69 , tends towards infinity for @xmath101 . only in some degenerated cases , such as @xmath102 or @xmath103 or @xmath104 , the quantity @xmath100 , and in general @xmath69 , remains finite for @xmath101 . if we insert @xmath105 and @xmath106 into eq . ( 1 ) , the corresponding prediction @xmath107 for the position of the star at an epoch @xmath28 is now a strictly linear function of @xmath28 , since @xmath105 and @xmath106 are linear in @xmath28 . in order to facilitate the understanding of the behaviour of the prediction @xmath107 , we rewrite @xmath108 by using the auxiliary quantities @xmath109 . if we insert eqs . ( 28 ) and ( 29 ) into eq . ( 1 ) , we obtain after some lengthy algebra : @xmath110 with the auxiliary quantities @xmath111 % ( 32 ) x_{\rm{c , nce } } ( t_{\rm{c , nce } } ) & = & \frac{w_{x , 1 } x_1 ( t_1 ) + w_{x , 2 } x_2 ( t_2 ) } { w_{x , 1 } + w_{x , 2 } } \,\ , , \\[1ex ] % ( 33 ) \mu_{\rm{c , nce } } & = & \frac{w_{\mu , 1 } \mu_1 + w_{\mu , 2 } \mu_2 + w_{\mu , 0 } \mu_0 } { w_{\mu , 1 } + w_{\mu , 2 } + w_{\mu , 0 } } \,\ , , \\[1ex ] % ( 34 ) \mu_0 & = & ( x_2 ( t_2 ) - x_1 ( t_1 ) ) / ( t_2 - t_1 ) \,\ , . % ( 35)\end{aligned}\ ] ] the weights @xmath112 are given by @xmath113 % ( 36 ) w_{x , 2 } & = & \frac{1}{\varepsilon^2_{x , 2 } } \,\ , , \\[1ex ] % ( 37 ) w_{\mu , 1 } & = & \frac{1}{\varepsilon^2_{\mu , 1 } } \,\ , , \\[1ex ] % ( 38 ) w_{\mu , 2 } & = & \frac{1}{\varepsilon^2_{\mu , 2 } } \,\ , , \\[0.5ex ] % ( 39 ) w_{\mu , 0 } & = & \frac{1}{\varepsilon^2_{\mu , 0 } } = \frac{(t_2 - t_1)^2}{\varepsilon^2_{x , 1 } + \varepsilon^2_{x , 2 } } \,\ , . % ( 40)\end{aligned}\ ] ] inserting @xmath105 and @xmath106 into eq . ( 4 ) , we obtain the mean error @xmath114 of @xmath108 . using the form of @xmath108 as given in eq . ( 31 ) and the auxiliary quantities with the index c , we find @xmath115 with @xmath116 @xmath117 a comparison of the eqs . ( 31)(42 ) with the analytic version of the single - star - mode solution of paper i ( sect . 2 , especially eqs . ( 19 ) , ( 23 ) , ( 30)(32 ) ) shows that the prediction @xmath107 and its mean error @xmath114 are strictly identical with the single - star - mode solution @xmath118 and its mean error @xmath119 . this result is very pleasing , since it proves the internal consistency of our scheme : in the limit of vanishing cosmic errors , the prediction @xmath27 according to eq . ( 1 ) is asymptotically approaching the single - star mode solution @xmath118 . this result is not apriori self - evident , since our definitions for the ` best ' solution for predicting @xmath29 differ at least formally in paper i and in this paper ii ( i.e. , eqs . ( 5 ) and ( 35 ) of paper i versus eq . ( 3 ) of paper ii ) . the general solution of our combination problem is given in sect . 2.2 . for all epochs the solution @xmath27 is a non - linear function of @xmath28 , and requires the knowledge of the correlation functions @xmath120 and @xmath23 as functions of the epoch difference @xmath121 . at present , we do not have a well - established knowledge about the run of @xmath120 and @xmath23 . only the cosmic errors @xmath20 and @xmath92 can be empirically determined from the comparison of the fk5 with hipparcos , assuming that the fk5 is giving ` mean ' quantities . even if we would know the run of @xmath120 and @xmath23 as a function of @xmath121 , the non - linearity of @xmath27 would demand a table of @xmath27 for a sequence of epochs , e.g. for each year , if the user should not have the burden to do the full calculation himself by running a program . we propose the following solution : the general solution for @xmath27 allows rather easily to obtain two limiting solutions for @xmath122 and for @xmath123 . we call the solution for @xmath124 the ` long - term prediction ( ltp ) ' , and the solution for @xmath123 the ` short - term prediction ( stp ) ' around the epoch @xmath86 . both the ltp and stp solutions are linear in @xmath28 . they can be therefore given in the usual astrometric style , i.e. as a position at a central epoch and a proper motion , together with their mean errors . the details on the ltp and stp solutions @xmath125 and @xmath126 are given in the sects . 3 and 4 . the general solution @xmath27 is a smooth transition from short - term prediction @xmath126 ( for epochs around @xmath86 ) to the long - term prediction @xmath125 for epochs @xmath28 far away from @xmath86 . in sect . 5 , we will discuss a convenient ( but only approximately valid ) method to carry out this transition , if we know the run of @xmath23 . this gives at least a rough indication on the process of transition as a function of the epoch difference @xmath121 , even if @xmath23 is not well - established . if , in the future , @xmath23 should be better determined , then our method would allow rather conveniently the ( approximate ) determination of @xmath27 also for epochs inbetween the validity ranges of @xmath126 and @xmath125 . we consider in this section the limiting case of the general solution @xmath27 for @xmath127 . this ` long - term prediction ' @xmath125 is valid for epochs not too close to the epoch @xmath86 of the instantaneous catalogue 2 ( i.e. here the hipparcos catalogue with @xmath128 ) . we assume that the epoch difference @xmath129 is so large that the correlation functions @xmath120 and @xmath23 both vanish . setting @xmath130 and @xmath131 , we obtain from the general eqs . ( 22 ) and ( 23 ) for the ltp solution @xmath132\right ) \nonumber\\ & & \hspace{-0.40 cm } + \varepsilon^2_{\mu , 1 } \left[\varepsilon^2_{\mu , 2 } + \eta ( 0)\right ] \left(t_2 - t_1\right ) \left(t - t_1\right)\big ) / n \,\ , , \\[1ex ] % ( 44 ) b_{\rm{ltp } } ( t ) & = & \beta_{\rm{ltp } } ( t ) \left(t - t_2\right ) \nonumber\\ & = & \big(- \varepsilon^2_{x , 1 } \varepsilon^2_{\mu , 1 } \left(t_2 - t_1\right ) \nonumber\\ & & \hspace{-0.40 cm } + \left(\varepsilon^2_{x , 1 } + \left[\varepsilon^2_{x , 2 } + \xi ( 0)\right]\right ) \varepsilon^2_{\mu , 1 } \left(t - t_1\right)\big ) / n \,\ , , % ( 45)\end{aligned}\ ] ] where @xmath76 is still given by eq . ( 24 ) . if we compare the eqs . ( 43 ) , ( 44 ) , and ( 24 ) for @xmath133 , @xmath134 , and @xmath135 with the corresponding eqs . ( 28)(30 ) for @xmath136 , @xmath137 , and @xmath138 , we find that they are identical if we replace in the equations for the nce solution the quantity @xmath139 by @xmath140 and @xmath141 by @xmath142 this is very plausible , since the instantaneously measured quantities @xmath143 and @xmath51 are affected by the cosmic errors @xmath93 and @xmath144 . if we add these cosmic errors to the corresponding measuring errors @xmath145 and @xmath146 , we obtain ` apparent ' measuring errors @xmath147 and @xmath148 . since the cosmic errors are not correlated with the measuring errors , the summation has to be done quadratically . using this finding we obtain for the long - term prediction @xmath149 with + @xmath150 % ( 49 ) x_{\rm{ltp } } ( t_{\rm{ltp } } ) & = & \frac{w_{x , 1}\ , x_1 ( t_1 ) + w_{x , 2 , \rm{ltp } } \ , x_2 ( t_2 ) } { w_{x , 1 } + w_{x , 2 , \rm{ltp } } } \,\ , , \\[0.5ex ] % ( 50 ) \mu_{\rm{ltp } } & = & \frac{w_{\mu , 1 } \ , \mu_1 + w_{\mu , 2 , \rm{ltp } } \ , \mu_2 + w_{\mu , 0 , \rm{ltp } } \ , \mu_0 } { w_{\mu , 1 } + w_{\mu , 2 , \rm{ltp } } + w_{\mu , 0 , \rm{ltp } } } \ , . % ( 51)\end{aligned}\ ] ] the weights @xmath112 are given by eqs . ( 36 ) and ( 38 ) , and by @xmath151 % ( 52 ) w_{\mu , 2 , \rm{ltp } } & = & \frac{1 } { \varepsilon^2_{\mu , 2 } + c^2_\mu } \,\ , , \\[0.5ex ] % ( 53 ) w_{\mu , 0 , \rm{ltp } } & = & \frac{(t_2 - t_1)^2 } { \varepsilon^2_{x , 1 } + \varepsilon^2_{x , 2 } + c^2_x } \,\ , . % ( 54)\end{aligned}\ ] ] the long - term prediction @xmath125 has in fact two conceptionally different properties : ( 1 ) as described above , it is the limit of the general prediction @xmath27 for the true _ instantaneous _ position @xmath29 for @xmath152 . ( 2 ) on the other hand , @xmath125 is for all epochs @xmath28 the best prediction for the _ mean _ position @xmath41 of the object . this means especially that @xmath153 is the best estimate of the center - of - mass velocity of the object . the two different concepts produce two different error estimates @xmath154 for @xmath125 . if we consider @xmath155 as the prediction for the _ mean _ position @xmath41 then the mean error @xmath156 is given by : @xmath157 with @xmath158 @xmath159 if we consider @xmath155 as the prediction for the _ instantaneous _ position @xmath29 for large values of @xmath160 , then the uncertainty @xmath161 of @xmath125 is given by @xmath162 similarly , the mean error @xmath163 of the predicted _ instantaneous _ proper motion for large @xmath121 is given by @xmath164 the equations above describe what we have called the ` analytic ' approach in paper i. for the ` numerical ' approach , we can take over for the ltp solution the formulae given in paper i for the single - star mode ( si ) with the following changes : ( 1 ) we should not redetermine the parallax @xmath26 of the star . the determination of @xmath26 by hipparcos requires instantaneous values of @xmath17 and @xmath18 , not the ` mean ' ltp values . formally , we set all correlations between @xmath26 and the other quantities ( position , proper motion ) equal to zero . ( 2 ) in the variance - covariance matrix * d * we replace in the diagonal line the hipparcos values of @xmath165 by @xmath166 and @xmath167 by @xmath168 , @xmath169 by @xmath170 , and @xmath171 by @xmath172 , respectively . no changes are made in the non - diagonal elements ( except for decoupling the parallax as described above ) , i.e. all the covariances remain in the ltp as they were in the si mode , e.g. @xmath173 in this section , we consider the other limiting case of the general solution @xmath27 , namely the limit for @xmath174 . this ` short - term prediction ' @xmath126 is valid for epochs close to @xmath86 ( in the case of using hipparcos : @xmath175 ) . for the case @xmath123 , we use for the correlation function @xmath120 and @xmath23 taylor series in @xmath121 , and keep only terms which are linear in @xmath121 . from the eqs . ( 54 ) and ( 55 ) of wpsa , we obtain for small values of @xmath129 @xmath176 in eq . ( 63 ) , we have made use of the differential relation ( 28 ) of wpsa . although hipparcos values are already averaged over about 3 years of observation , the use of eq . ( 63 ) is justified according to numerical investigations carried out by m. biermann ( 1996 , unpublished , see also wpsa , sect . 3.4 ) . inserting eqs . ( 62 ) and ( 63 ) into the eqs . ( 22 ) and ( 23 ) , we find for the short - term prediction @xmath177 % ( 64 ) \lefteqn { b_{\rm{stp } } ( t ) = \beta_{\rm{stp } } ( t ) ( t - t_2 ) } \nonumber\\ & & = \bigg(\varepsilon^2_{\mu , 1 } \varepsilon^2_{x , 2 } ( t_2 - t_1 ) \nonumber\\ & & \hspace{0.5 cm } + \big(\eta ( 0 ) \big(\xi ( 0 ) + \varepsilon^2_{x , 1 } + \varepsilon^2_{x , 2 } + \varepsilon^2_{\mu , 1 } ( t_2 - t_1)^2 \big ) \nonumber\\ & & \hspace{1 cm } + \ , \varepsilon^2_{\mu , 1 } \big ( \xi ( 0 ) + \varepsilon^2_{x , 1 } + \varepsilon^2_{x , 2 } \big ) \big ) ( t - t_2)\bigg ) / n \,\ , , % ( 65)\end{aligned}\ ] ] where @xmath76 is still given by eq . ( 24 ) . inserting @xmath178 and @xmath179 into eq . ( 1 ) , we obtain @xmath180 which is a linear function of the epoch @xmath28 . we do not use here a ` central ' epoch @xmath181 , since our taylor series for @xmath126 is centered around @xmath182 . the proper motion @xmath183 can be derived as the sum of the coefficients in front of @xmath28 in the eqs . ( 1 ) , ( 63 ) , and ( 64 ) : @xmath184 after some algebra , @xmath183 can be rewritten as @xmath185 the ` combined ' mean proper motion @xmath186 is derived from @xmath14 and @xmath187 by @xmath188 @xmath189 , and @xmath190 are given by the corresponding equations in sect . 2.4 , and @xmath191 by eq . the weight @xmath192 of @xmath186 in the stp solution is given by @xmath193 with @xmath194 the meaning of eqs . ( 68)-(71 ) is the following : @xmath183 is the weighted mean of @xmath66 and the combined mean proper motion @xmath186 , where the cosmic error @xmath144 in @xmath186 has to be taken into account in @xmath195 . the combined mean proper motion @xmath186 is itself a weighted mean of @xmath14 and @xmath187 , where for @xmath187 the cosmic error @xmath93 in @xmath50 has to be included into @xmath191 . the exact expression for @xmath196 is given by @xmath197 \lefteqn { - \ , \varepsilon^2_{x , 2 } \big ( \big(\eta ( 0 ) + \varepsilon^2_{\mu , 1 } + \varepsilon^2_{\mu , 2 } ) ( x_2 ( t_2 ) - x_1 ( t_1 ) - \mu_1 ( t_2 - t_1 ) ) } \nonumber\\[0.5ex ] \lefteqn { \hspace{+ 2.50 cm } - \ , \varepsilon^2_{\mu , 1 } ( t_2 - t_1 ) ( \mu_2 ( t_2 ) - \mu_1 ) \big ) / n \,\ , . } % ( 72)\end{aligned}\ ] ] for all practical purposes we can , however , neglect the last term in eq . ( 72 ) , which is proportional to @xmath139 . the measuring accuracy @xmath145 for the hipparcos positions @xmath50 at @xmath128 is so high , relative to the other measuring errors and cosmic errors , that this term is usually of the order of 0.01 mas only . we use instead of eq . ( 72 ) the very good approximation @xmath198 the mean errors of @xmath199 and of @xmath183 are given by @xmath200 @xmath201 the full uncertainty @xmath202 of @xmath126 is given by @xmath203 in eq . ( 76 ) , we have neglected the very small correlation between @xmath204 and @xmath183 which is caused by the use of @xmath143 in deriving @xmath187 which in turn enters into @xmath183 . the last term in eq . ( 76 ) is the statistical uncertainty of a prediction based on instantaneous data at @xmath86 for small epoch differences @xmath121 ( see eq . ( 57 ) of wpsa ) . @xmath205 is the fourth derivative of the correlation function @xmath120 with respect to @xmath121 at @xmath206 . in the numerical approach for deriving the short - term prediction , we modify the procedure for the single - star mode presented in paper i more strongly than for the ltp solution . as ` observations ' * b * we use first * b*@xmath207 , as in paper i , the parallax @xmath208 inclusive . the corresponding part * d*@xmath209 of the variance - covariance matrix * d * remains also unchanged . the second part of * b * , which we now call * b*@xmath210 is given by the @xmath211 and @xmath8 components of the combined mean proper motion @xmath186 . the quantity @xmath186 is obtained in a preparatory step from eq . ( 69 ) , and its mean error from eq . ( 70 ) as @xmath212 . each component of @xmath186 , i.e. of * b*@xmath210 , is considered not to be correlated with any other component of * b*. the numerical approach for stp produces values for @xmath213 and @xmath183 ( in @xmath211 and @xmath8 ) , a new parallax @xmath214 , and the corresponding variance - covariance matrix . in presenting the results in printed form , we use again central epochs @xmath181 ( different for @xmath211 and @xmath8 ) , at which @xmath213 and @xmath183 are uncorrelated . however , @xmath181 differs usually only very slightly from the individual central epochs of the basic hipparcos data . as discussed in sect . 2.4 and illustrated in fig . 2 , the general solution @xmath27 is a smooth transition from the short - term prediction @xmath126 for epochs @xmath28 close to @xmath86 to the long - term prediction @xmath125 for @xmath87 . we are now asking for the ` transition function ' @xmath215 which describes this transition in @xmath17 : @xmath216 formally we can always solve eq . ( 77 ) for @xmath215 , using our former results for @xmath217 and @xmath218 . however , the resulting transition function @xmath215 is then very complicated and depends unfortunately not only on the correlation functions , but also explicitely on the measured values of @xmath219 , and on their measuring errors . such a result is not very suitable for a practical application . there does exist , however , an approximate treatment for the transition function @xmath215 which gives a very simple and easily applicable result , and which nevertheless describes the transition quantitatively rather accurately . the basic idea is the observation that in real applications the mean position @xmath220 enters into the final result @xmath27 mainly through the proper motion @xmath187 . this is caused by the small error of the hipparcos position @xmath50 with respect to the error of the mean ( fk5 or gc ) position @xmath220 . only in cases of a large cosmic error @xmath24 in @xmath50 , our approximation becomes less accurate . we therefore consider the transition function @xmath215 for the limiting case in which @xmath15 tends towards infinity while @xmath221 remains finite ( equal to its actual value ) . the latter can be enforced by setting @xmath222 . this means that we let go @xmath223 and @xmath224 in such a way that @xmath221 remains constant . ccc + @xmath160 & @xmath225 & @xmath226 + @xmath227 $ ] & + + 0 & 1.000 & 1.000 + 1 & 0.978 & 0.934 + 2 & 0.915 & 0.752 + 3 & 0.819 & 0.491 + 4 & 0.700 & 0.202 + 5 & 0.573 & 0.065 + 6 & 0.449 & 0.270 + 7 & 0.336 & 0.397 + 8 & 0.241 & 0.445 + 9 & 0.165 & 0.430 + 10 & 0.108 & 0.373 + 11 & 0.068 & 0.297 + 12 & 0.041 & 0.219 + 13 & 0.023 & 0.152 + 14 & 0.013 & 0.099 + 15 & 0.007 & 0.060 + 20 & 0.0001 & 0.002 + if we use this special case as an approximation , we derive after some lengthy algebra the following rather simple expression for the transition function : @xmath228 a similar function has already been derived as eq . ( 70 ) in sect . 4.2.4 of wpsa . the function @xmath229 has the welcome property that it does neither depend on the measured values of @xmath17 and @xmath18 nor on the mean errors of a given star . these values are fully absorbed in the individual solutions @xmath126 and @xmath125 . hence @xmath215 is the same function for all the objects . we should remark here that the apparently more complicated form of the transition function according to eq . ( 75 ) of wpsa is caused by the fact that we have used in wpsa a slightly different definition of the long and short - term prediction , namely @xmath230 instead of @xmath218 and @xmath231 instead of @xmath125 . the function @xmath215 has the desired properties in the limits @xmath123 or @xmath232 . for @xmath182 , we have @xmath233 , because of @xmath234 for small values of @xmath121 . for @xmath235 vanishes and hence @xmath236 . an example for the full run of @xmath237 , from 1 to 0 , is illustrated by the full curve shown in fig . 8 of wpsa . table 1 gives @xmath238 for a few values of @xmath239 , using the example given in sect . 3.6 of wpsa for the correlation function @xmath56 . at @xmath240 5 - 6 years , the general solution @xmath27 is about half - way between @xmath126 and @xmath125 . from both the fig . 8 of wpsa and table 1 , we get an indication for the range of applicability of the short and long - term prediction . the short - term prediction @xmath126 has a rather limited range of applicability , namely a few years around @xmath86 only . the long - term prediction @xmath125 is a good approximation for the general solution @xmath27 for epoch differences @xmath239 which are larger than about 10 years . hence the transition from stp to ltp is rather rapid , at least in our example . we should remark here that ( by chance ) the transition function @xmath241 given by eq . ( 78 ) is even strictly valid ( exact ) for the example used for the correlation functions in sect . 3.6 of wpsa and adopted in table 1 . for more general runs of the correlation functions , @xmath242 can become quite large at epochs around the ` crossing time ' at which @xmath243 , if @xmath244 and @xmath245 are significantly different from @xmath246 at that time . the most accurate prediction @xmath247 for the instantaneous proper motion @xmath49 at an arbitrary epoch @xmath28 is formally given by @xmath248 the transition function @xmath249 for the proper motion @xmath18 is , similar to @xmath229 , in general a rather complicated function , which depends on the measured values of @xmath250 and @xmath187 . if we adopt the same approximation as used in sect . 5.1 for obtaining eq . ( 78 ) for @xmath229 , namely @xmath251 while keeping @xmath221 constant , we derive for @xmath249 also a very simple expression : @xmath252 this transition function @xmath249 for @xmath49 has the same proper limiting values as @xmath215 : for @xmath253 , we have @xmath254 and hence @xmath255 , and for @xmath256 we obtain @xmath257 and @xmath258 . in table 1 we list @xmath259 for a few values of @xmath260 , using eq . ( 80 ) and the simple example for @xmath22 given in sect . 3.6 of wpsa . the transition functions @xmath215 and @xmath249 according to eqs . ( 78 ) and ( 80 ) have another nice property : if we use the example given in sect . 3.6 of wpsa , then the function @xmath229 depends only on @xmath261 , but not on the individual cosmic error @xmath262 , since @xmath263 occurs also as a factor in @xmath264 and cancels out in eq . similarly , the transition function @xmath249 is a ` scaled ' function and depends on @xmath261 only , but not on @xmath265 . for deriving the solutions in the ltp and stp mode , we need to know the cosmic errors @xmath20 and @xmath92 . the cosmic errors depend strongly on the distance @xmath266 of the star from the sun , or equivalently on the stellar parallax @xmath26 . for 1202 ` apparently single stars ' from the basic fk5 , we have obtained the functions @xmath25 and @xmath24 empirically by using groups of stars in various distance intervals . the data can be represented by the following fit functions : @xmath267 % ( 81 ) c_x ( p ) & = \ , ( \xi ( 0 , p))^{1/2 } & = \,\ , c_3\,c_\mu ( p)\,\ , , % ( 82)\end{aligned}\ ] ] with @xmath268 these versions of @xmath25 and @xmath24 have been used for the fk6 ( wielen et al . 1999d , 2000a ) . a table for @xmath144 and @xmath93 as a function of @xmath26 or @xmath266 , based on eqs . ( 81)-(85 ) , is given on page 12 of wielen et al . ( 1999d ) . in the future we hope to determine also the dependence of the cosmic errors on the absolute magnitude ( or mass ) of the stars for a given parallax . a comparison of the results for the cosmic errors based on the fk5 stars with those derived from the ( on average fainter ) gc stars seems to indicate a rather weak dependence on the brightness of the stars ( wielen et al . 1998 ) . lrrrrrrrrrrrr + & & + quantity & @xmath211 & & @xmath8 & & @xmath211 & & @xmath8 & + + input data : + @xmath269 or @xmath270 & 7.83 & @xmath271 12.52 & & + 96.95 & @xmath271 15.30 & & 2.64 & @xmath271 49.92 & & + 236.68 & @xmath271 32.88 & + @xmath272 or @xmath273 & + 0.49 & @xmath271 0.40 & & 1.20 & @xmath271 0.38 & & + 1.69 & @xmath271 1.24 & & 3.98 & @xmath271 1.86 & + @xmath274 or @xmath275 & 1947.84 & & & 1929.73 & & & 1892.60 & & & 1890.30 & & + @xmath276 & 0.00 & @xmath271 0.77 & & 0.00 & @xmath271 0.54 & & 0.00 & @xmath271 0.77 & & 0.00 & @xmath271 0.54 & + @xmath277 & 0.00 & @xmath271 1.01 & & 0.00 & @xmath271 0.77 & & 0.00 & @xmath271 1.01 & & 0.00 & @xmath271 0.77 & + @xmath278 & 1991.26 & & & 1991.51 & & & 1991.26 & & & 1991.51 & & + @xmath208 & 49.48 & @xmath271 0.99 & & & & & 49.48 & @xmath271 0.99 & & & & + @xmath93 & 17.28 & & & 17.28 & & & 17.28 & & & 17.28 & & + @xmath144 & 2.91 & & & 2.91 & & & 2.91 & & & 2.91 & & + + @xmath279 & + 0.18 & @xmath271 0.29 & & 1.57 & @xmath271 0.25 & & + 0.03 & @xmath271 0.51 & & 2.34 & @xmath271 0.32 & + + + @xmath280 & 0.18 & @xmath271 0.76 & & + 0.14 & @xmath271 0.54 & & 0.23 & @xmath271 0.77 & & + 0.08 & @xmath271 0.54 & + @xmath281 & + 0.23 & @xmath271 0.23 & & 1.34 & @xmath271 0.20 & & + 0.04 & @xmath271 0.42 & & 2.00 & @xmath271 0.29 & + @xmath282 & 1991.12 & & & 1991.47 & & & 1991.26 & & & 1991.51 & & + @xmath283 & 0.57 & @xmath271 0.95 & & & & & 0.98 & @xmath271 0.95 & & & & + + + @xmath284 & 5.14 & @xmath271 10.14 & & + 54.37 & @xmath271 11.46 & & 0.30 & @xmath271 16.34 & & + 51.26 & @xmath271 15.30 & + @xmath285 & + 0.36 & @xmath271 0.31 & & 1.38 & @xmath271 0.27 & & + 0.28 & @xmath271 0.49 & & 2.37 & @xmath271 0.36 & + @xmath286 & 1962.77 & & & 1956.76 & & & 1980.68 & & & 1969.40 & & + + + @xmath287 & 0.01 & @xmath271 0.77 & & 0.00 & @xmath271 0.54 & & 0.02 & @xmath271 0.77 & & 0.00 & @xmath271 0.54 & + @xmath288 & 0.00 & @xmath271 0.95 & & 0.08 & @xmath271 0.74 & & 0.03 & @xmath271 0.95 & & 0.15 & @xmath271 0.74 & + @xmath289 & 1991.26 & & & 1991.52 & & & 1991.26 & & & 1991.52 & & + @xmath290 & 0.03 & @xmath271 0.99 & & & & & 0.06 & @xmath271 0.99 & & & & + + units : mas , mas / year , or years in order to illustrate our combination method for all the three modes ( si , ltp , stp ) , we give in table 2 the results for one individual star . as in paper i ( sect . 5 ) , we give the positions @xmath29 and the proper motions @xmath49 always relative to the hipparcos solution as @xmath292 and @xmath293 , in order to save printing space and to make the comparison of the results easier . the ground - based data are always reduced to the hipparcos system . the results of the single - star mode presented here in paper ii differ somewhat from those of paper i , because we now adopt slightly improved systematic differences fk5-hip and gc - hip . for a valid comparison of the results of the three different modes it is necessary to use exactly the same basic input data . table 2 shows that the short - term prediction is usually quite close to the hipparcos solution . on the other hand , the long - term prediction differs most strongly from the hipparcos solution , since the hipparcos values are entering with a lower weight into the ltp mode than in the si mode , because of the cosmic errors in the hipparcos data . the central epoch @xmath294 is the only one which is usually significantly earlier than @xmath0 , and the mean error of the central position @xmath295 is typically only slightly smaller than the cosmic error @xmath24 . in table 3 we present the error budget of proper motions in the three different modes ( si , ltp , stp ) for two samples of basic fk5 stars . the mean errors @xmath296 given in table 3 refer to one ` mean ' coordinate component . it is obtained as an rms average over @xmath297 and @xmath298 , and over all the stars in the corresponding sample . the error budget for the 1535 basic fk5 stars is slightly fictious , since this sample contains double stars for which the fk6 provides in reality ` special ' solutions instead of the ` direct ' combination solutions discussed in this paper . nevertheless , the results for this sample provide a valid indication for the overall accuracy of our combination method in the direct modes si , ltp , and stp . the sample of 1202 basic fk5 stars contains ` apparently single objects ' only . most of these stars ( 878 objects ) have direct solutions in the fk6 . the error budget for these 878 basic fk5 stars in part i of the fk6 is given in wielen et al . ( 1999d ) . the error budgets for 3272 additional fundamental stars with direct solutions in the three modes are presented in part iii of the fk6 ( wielen et al . . table 4 gives the error budget for the combination of the gc ( boss et al . 1937 ) with hipparcos . the sample of gc stars are the ` full sample ' of 29717 gc stars observed by hipparcos , and the ` subsample ' of 11737 gc stars with linear hipparcos standard solutions . from table 4 it is obvious that the original proper motions @xmath299 of the gc do not contribute significantly to the gc+hip on average . however , for some brighter and well - observed stars , the accuracy of @xmath299 is much better than the rms value of @xmath300 seen in table 4 . furthermore , the proper motion @xmath301 , derived from the central positions of the gc and the hipparcos catalogue , has usually a quite reasonable accuracy for most of the gc stars . the typical gain in accuracy in the proper motions derived in the long - term prediction mode , relative to the original hipparcos proper motions , is a factor of 4.6 for the basic fundamental stars in the fk6=fk5+hip , and a factor of 1.8 for the 11773 gc stars in the gc+hip . this improvement in the ltp mode with respect to hipparcos is a consequence of the cosmic errors in the instantaneously measured hipparcos proper motions . in contrast to the ltp mode , the short - term predictions ( stp mode ) do not differ so much from the hipparcos solutions . however , in most cases we are more interested in the long - term averaged proper motion ( ltp ) or in the single - mode result ( si ) , where the gain in accuracy is quite significant . the long - term predictions and the short - term predictions derived in the former sections are _ statistically _ the best astrometric solution for a sample of ` apparently single - stars ' . however , an unsatisfactory property of these statistically valid solutions is the fact that we are not able to make proper use of the available information on the _ individual _ behaviour of the stars . for example , we use the overall cosmic errors @xmath25 and @xmath24 , no matter whether the object is a @xmath2 binary or a single - star candidate ( wielen et al . 1999a ) . in principle , for each individual star , one would have to use ` conditioned correlation functions ' which are based on the individually observed differences between the instantaneous measurements and the mean data ( e.g. on @xmath302 ) . unfortunately , the conditioned correlation functions , i.e. ` conditioned cosmic errors ' in particular , are not available at present . we should also point out that presently the cosmic errors @xmath93 in position are much more uncertain than the cosmic errors @xmath144 in proper motion . while the typical values of @xmath144 are larger than the measuring errors @xmath303 , @xmath221 , and @xmath304 , the typical values of @xmath93 are nearly lost in the measuring errors of the ground - based data . furthermore , our knowledge about the actual form of the correlation functions @xmath120 , @xmath22 , @xmath23 for @xmath305 is presently still quite rudimentary . the simple example for the correlation functions presented in sect . 3.6 of wpsa is mainly given for illustrating the general behaviour of these functions . the application of this example to real data should be done very cautiously . hence the real transition from the stp solution to the ltp solution ( discussed in sect . 5 ) is quantitatively not well - determined . which of the three solutions offered ( si , ltp , stp ) should be used in real applications ? for single - star candidates , the single - star mode should be adopted , although some of the single - star candidates may nevertheless be binaries . if the user is handling a sample of ` apparently single stars ' ( which usually contains single - star candidates , @xmath2 binaries , and intermediate cases ) , then the ltp or stp solutions are recommended , depending on the corresponding epoch difference @xmath306 ( with @xmath307 ) . the problems discussed above are especially severe for objects detected as @xmath2 binaries ( wielen et al . 1999a ) . for @xmath2 binaries the difference between the instantaneous proper motion ( @xmath308 ) and the mean one ( e.g. , @xmath309 or @xmath187 ) is sometimes much larger than the cosmic error @xmath144 expected on average . in such a case the weight @xmath310 ( eq . ( 53 ) ) of the hipparcos proper motion ( @xmath311 ) is higher than appropriate for this individual object . the derived mean proper motion @xmath312 is then biased towards the hipparcos proper motion , and the derived mean error @xmath313 is too small . in the case of extreme @xmath2 binaries , it is better to adopt a properly weighted mean of @xmath14 ( e.g. @xmath309 ) and @xmath187 as a prediction for the mean proper motion @xmath47 . the main problem in such a procedure is the unknown individual value of the cosmic error in the hipparcos position , which is higher than the value of @xmath93 expected on average . this value enters into the weight of @xmath187 , and hence into the predicted value of @xmath47 . we shall discuss this problem in more detail in a subsequent paper . in any case , the ltp solutions ( and , of course , the si solutions ) for @xmath2 binaries are inherently the least accurate ones among the class of direct solutions , because of the disturbing double - star nature of these objects in this paper ii , we have derived and discussed an appropriate method to combine a ground - based astrometric catalogue ( such as the fk5 or gc ) with the hipparcos catalogue , taking cosmic errors ( due to undetected binaries ) in the quasi - instantaneously measured data into account . the method leads to long - term predictions ( ltp mode ) and to short - term predictions ( stp mode ) , which are the limiting cases of the general solution . the general solution is a smooth transition from the stp to the ltp mode . the case of single stars with no cosmic errors was already treated in paper i ( si mode ) . in a subsequent paper , we shall present ` special solutions ' for known double stars . wielen r. , schwan h. , dettbarn c. , jahrei h. , lenhardt h. , 1997 , in : hipparcos venice 97 , presentation of the hipparcos and tycho catalogues and first astrophysical results of the hipparcos space astrometry mission , eds . battrick b. , perryman m.a.c . , bernacca p.l . , esa sp-402 , p. 727 wielen r. , schwan h. , dettbarn c. , jahrei h. , lenhardt h. , 1998 , in : the message of the angles astrometry from 1798 to 1998 , proceedings of the international spring meeting of the astronomische gesellschaft , held in gotha , 11 - 15 may 1998 , eds . brosche p. , dick w.r . , schwarz o. , wielen r. , acta historica astronomiae , vol . 3 , verlag harri deutsch , thun and frankfurt am main , p. 123 wielen r. , schwan h. , dettbarn c. , jahrei h. , lenhardt h. , 1999c , in : modern astrometry and astrodynamics , proceedings of the international conference honouring heinrich eichhorn , held at vienna observatory , austria , 25 - 26 may 1998 , eds . dvorak r. , haupt h.f . , wodnar k. , verlag der sterreichischen akademie der wissenschaften , wien , p. 161 lcccc + + + + sample of stars : & & + [ 0.7ex ] + & rms aver . & median & rms aver . & median + + + hipparcos & & & & + random & 0.82 & 0.63 & 0.68 & 0.61 + fk5 + random & 0.76 & 0.64 & 0.77 & 0.67 + system & 0.28 & 0.25 & 0.28 & 0.25 + total & 0.81 & 0.70 & 0.83 & 0.72 + @xmath187 + random & 0.53 & 0.43 & 0.54 & 0.45 + system & 0.24 & 0.23 & 0.25 & 0.23 + total & 0.58 & 0.49 & 0.59 & 0.51 + fk6 = fk5+hip & 0.35 & 0.33 & 0.35 & 0.34 + ratio of hipparcos & 2.3 & 1.9 & 1.9 & 1.8 + to fk6 errors + + + hipparcos & & & & + random & 0.82 & 0.63 & 0.68 & 0.61 + cosmic ( in @xmath18 ) & 2.15 & 2.05 & 2.13 & 2.04 + total & 2.30 & 2.14 & 2.24 & 2.13 + fk5 + random & 0.76 & 0.64 & 0.77 & 0.67 + system & 0.28 & 0.25 & 0.28 & 0.25 + total & 0.81 & 0.70 & 0.83 & 0.72 + @xmath187 + random & 0.53 & 0.43 & 0.54 & 0.45 + system & 0.24 & 0.23 & 0.25 & 0.23 + cosmic ( due to @xmath17 ) & 0.32 & 0.30 & 0.32 & 0.30 + total & 0.66 & 0.57 & 0.68 & 0.59 + fk6 = fk5+hip & 0.49 & 0.44 & 0.49 & 0.45 + ratio of hipparcos & 4.7 & 4.9 & 4.6 & 4.7 + to fk6 errors + + + hipparcos & & & & + random & 0.82 & 0.63 & 0.68 & 0.61 + fk5 + random & 0.76 & 0.64 & 0.77 & 0.67 + system & 0.28 & 0.25 & 0.28 & 0.25 + cosmic ( in @xmath18 ) & 2.15 & 2.05 & 2.13 & 2.04 + total & 2.30 & 2.16 & 2.28 & 2.16 + @xmath187 + random & 0.53 & 0.43 & 0.54 & 0.45 + system & 0.24 & 0.23 & 0.25 & 0.23 + cosmic ( due to @xmath17 ) & 0.32 & 0.30 & 0.32 & 0.30 + cosmic ( in @xmath18 ) & 2.15 & 2.05 & 2.13 & 2.04 + total & 2.25 & 2.13 & 2.23 & 2.12 + fk6 = fk5+hip & 0.68 & 0.59 & 0.62 & 0.57 + ratio of hipparcos & 1.2 & 1.1 & 1.1 & 1.1 + to fk6 errors + lcccc + + + + sample of stars : & & + [ 0.7ex ] + & rms aver . & median & rms aver . & median + + + hipparcos & & & & + random & 1.47 & 0.73 & 0.75 & 0.69 + gc + random & 10.57 & 9.38 & 8.59 & 7.55 + system & 0.57 & 0.49 & 0.56 & 0.48 + total & 10.59 & 9.39 & 8.61 & 7.57 + @xmath187 + random & 1.78 & 1.74 & 1.42 & 1.44 + system & 0.18 & 0.12 & 0.13 & 0.12 + total & 1.79 & 1.75 & 1.43 & 1.45 + gc+hip & 0.72 & 0.63 & 0.62 & 0.58 + ratio of hipparcos & 2.0 & 1.2 & 1.2 & 1.2 + to gc+hip errors + + + hipparcos & & & & + random & 1.47 & 0.73 & 0.75 & 0.69 + cosmic ( in @xmath18 ) & 1.75 & 1.60 & 1.80 & 1.66 + total & 2.29 & 1.76 & 1.95 & 1.80 + gc + random & 10.57 & 9.38 & 8.59 & 7.55 + system & 0.57 & 0.49 & 0.56 & 0.48 + total & 10.59 & 9.39 & 8.61 & 7.57 + @xmath187 + random & 1.78 & 1.74 & 1.42 & 1.44 + system & 0.18 & 0.12 & 0.13 & 0.12 + cosmic ( due to @xmath17 ) & 0.12 & 0.10 & 0.12 & 0.11 + total & 1.79 & 1.75 & 1.43 & 1.45 + gc+hip & 1.17 & 1.15 & 1.06 & 1.05 + ratio of hipparcos & 2.0 & 1.5 & 1.8 & 1.7 + to gc+hip errors + + + hipparcos & & & & + random & 1.47 & 0.73 & 0.75 & 0.69 + gc + random & 10.57 & 9.38 & 8.59 & 7.55 + system & 0.57 & 0.49 & 0.56 & 0.48 + cosmic ( in @xmath18 ) & 1.75 & 1.60 & 1.80 & 1.66 + total & 10.73 & 9.53 & 8.79 & 7.75 + @xmath187 + random & 1.78 & 1.74 & 1.42 & 1.44 + system & 0.18 & 0.12 & 0.13 & 0.12 + cosmic ( due to @xmath17 ) & 0.12 & 0.10 & 0.12 & 0.11 + cosmic ( in @xmath18 ) & 1.75 & 1.60 & 1.80 & 1.66 + total & 2.51 & 2.37 & 2.30 & 2.20 + gc+hip & 0.82 & 0.69 & 0.69 & 0.65 + ratio of hipparcos & 1.8 & 1.1 & 1.1 & 1.1 + to gc+hip errors +

we shall give dynamics to our spacetime manifold by first identifying the local affine symmetry as the characterizing symmetry for our geometryla felix klein , this symmetry is imposed on us by the law of inertia and the law of causality . we then prescribe 16 gauge vector bosons to this symmetryla yang and mills . the locally affine symmetric yang - mills lagrangian in the presence of a background world metric , and the corresponding equations of motion , are respectively constructed and derived . spontaneous breaking of the local affine symmetry to the local lorentz symmetry is achieved by classical solutions to the equations of motion . in these classical solutions , the 16 gauge vector bosons are shown to select the schwarzschild metric as one among the admissible background world metrics . classical gravity is thus be expressed by a spontaneously broken erlangen program . we shall also show that this yang - mills gauge theory of gravity can give an explanation of the form of the galactic rotation curves , of the amount of intergalactic gravitational lensing , and of the accelerating expansion of the universe .

1312.4528

the four fundamental interactions in physics are described by two different disciplines . the gravitational interaction follows the curved spacetime approach of the general theory of relativity laid down by einstein in which the dynamical variable is the metric tensor , while the electroweak and strong interactions follow the local gauge vector boson approach pioneered by yang and mills in which the dynamical variables are the vector bosons . both disciplines give spectacular success in terms of experiments and physical observations , despite of the fact that they look very different . there are , by now , many research works done in trying to put these two disciplines into one single footing . some people try to visualize gauge vector boson interactions as geometrical manifestations in a higher dimensional manifold with our spacetime as a four dimensional sub - manifold @xcite . other people try to consider the geometrical gravitation theory in the form of a local gauge theory @xcite . all of these ideas are met with difficulties in one way or the other . and in the past decades many new discoveries in astronomy might imply that the general theory of relativity is not sophisticated enough to explain the new phenomena . for example , stellar objects at the spiral arms of galaxies are orbiting at faster speeds than that can be explained by the keplerian motions . to overcome this difficulty , people assume that some extra matter , not visible to us , is giving an extra pull on these stellar objects . extra light deflections , as observed in the intergalactic gravitational lensing , are also ascribed to the existence of this extra matter . this is known as the @xmath1 @xmath2 problem . an equally well known fact is that the universe is accelerating in its expansion . this is in contrast to our expectation that the universe should be decelerating , unless some extra energy is kept pumping into the universe . this is the so called the @xmath1 @xmath3 problem in this article , we shall give a new approach to use the yang - mills method to construct a local gauge theory of gravity , and shall show that this theory can successfully solve the @xmath1 @xmath2 and @xmath1 @xmath3 problems . here we shall consider the local gauge theory approach of gravitation , albeit in a new context . a non - minkowskian world metric for our spacetime is always regarded as what makes it curved . however , it is very difficult to regard the world metric ( or more precisely its corresponding vierbein fields ) as gauge fields because of the peculiar way it appears in the action that determines physics . in the following we shall assign the world metric of our spacetime with a much limited role . we shall assume that the only function of the world metric @xmath4 is to give us world distance ( and hence world volume element @xmath5 ) , and will have no dynamical terms ( terms that contain spatial or temporal derivatives of @xmath4 ) in the action . this assumption frees us from taking the global minkowskian metric as the de facto world metric and sets the notion that no particular metric is a priori world metric for physics . in this sense , the world metric for our spacetime serves just as an arbitrary background of measuring clock and stick in our discussion of physics . with a given world metric @xmath6 at the point @xmath7 with world coordinates @xmath8 , a set of vierbein fields @xmath9 will follow . these vierbein fields are defined in a locally flat patch that is assumed to be equipped with an arbitrary but given local minkowskian frame whose coordinates and metric are @xmath10 and @xmath11 respectively . the differentials of these two coordinate systems define the vierbein fields as @xmath12 , and hence the world metric and the vierbein fields will then be related by @xmath13 here , and in the following , the latin indices will signify the minkowskian components while the greek indices will mean the world ones . on a locally flat patch around a point of our spacetime is where we do our physics . even though we have already had a local minkowskian system @xmath10 and @xmath11 on that patch , we may still have the freedom to re - label the points on that patch with different local coordinate systems , for example , by rotating and stretching these local minkowskian coordinate axes . these rotations and stretches are reflections of the relative physical situations experienced by local observers . for examples , a relative rotation of two local minkowskian frames is always regared as a relative lorentzian motion between the two local observers . the form of the admissible local coordinate transformations depends on what are the physical principles that , we hope , to remain invariant under these coordinate changes . here , we believe that the law of inertia should remain intact under these expected coordinate transformations . this means that the concept of straight lines should be preserved , as an object moving in straight line in one coordinate system should remain moving in straight line in another coordinate system . also light should propagate in straight lines in whatever coordinate system we are using . causality is also a very important concept in physics , and hence the order of points and the ratio of segment lengths in a straight line should not change with a change in coordinate system . and of course , the concept of parallelness should also be preserved because two parallel moving objects , as well as parallel light rays , should remain parallel under a change of coordinate system . those transformations which maintain collinearity , order of points and invariant segment ratios in straight lines , and parallelness are , in fact , the affine transformations of euler @xcite . affine transformations are sometimes grouped together as dilations , rotations , shears and reflections . we shall call collectively transformations that are not rotations as strains . here we want to emphasize that our choice of the affine transformations as our admissible local coordinate transformations comes from physics . it comes from our belief that these admissible transformations should leave the above said physical principles invariant . and if we are going to call such a chosen local coordinate system as a chosen local geometric setting , then we can say that physics is assumed to be invariant under a change of local geometric setting . in physical language , what his means , is that physics looks the same to all observers having the proposed relative physical situations . and this can be regarded as an extension of the principle of relativity of the special theory of relativity to a general spacetime manifold . these transformations form a lie group , called the local affine group . it was felix klein who first suggested of classifying geometries by their underlying symmetry groups , starting with the projective geometry ( our affine geometry is a restriction of the projective geometry ) . such a mathematical program is called the erlangen program @xcite . since matter , which are world objects , are described by local fields with the reference to a local coordinate system . these local fields could have structures that depend on the geometric setting chosen at that point . for example , if we want to describe the physics of an electron , it may be convenient to choose the local affine rotations as our local lorentz transformations . then an affine rotated setting will give a set of lorentz transformed fields . as we believe that the relative differences of the local fields of the same world object at two space - time points arising from different geometric settings are physically meaningless , we have to find some way to counteract such variations . similar to what have been done by yang and mills @xcite , we shall introduce a set of vector bosons to do these counteractions . note that the introduction of vector bosons , as suggested by yang and mills , were originally used to facilitate the local identifications of internal quantum numbers for quantum systems . here we extend their ideas to the local identifications of geometric settings in our spacetime . at any point of our spacetime , these vector bosons can be transformed away locally by a suitable choice of the coordinate system at that point . a more familiar way of saying this is that these vector fields are locally equivalent to a transformation of the coordinate system . one can then draw the strong analogy between the above statement and the principle of equivalence of einstein which states that `` there is a complete physical equivalence of a gravitational field and a corresponding acceleration of the reference system '' . the principle of equivalence is the founding principle taken by einstein to build his general theory of relativity , and the gauge principle stated in the above will be shown , as given below , to give a theory of vector gravity . in the following , these vector boson fields are regarded as dynamical variables . their dynamics will be fabricated so as to ensure that physics be invariant under local affine transformations . and this will be done , again by following yang and mills , by first constructing the lagrangian that is locally affine symmetric . for a four dimensional patch , these affine transformations can be carried out by @xmath15 invertible real matrices , either actively or passively . all these @xmath15 invertible real matrices form a lie group called the real general linear group of dimension four and is designated as @xmath14 . hence the @xmath14 will be synonymous with our affine group . the @xmath14 has two sets of generators . the six anti - symmetric generators @xmath16 generate the rotations while the ten symmetric generators @xmath17 generate the strains . they satisfy the following commutation relations @xcite @xmath18 & = & -i \ { \eta^{ac}j^{bd } - \eta^{ad}j^{bc } - \eta^{bc}j^{ad } + \eta^{bd}j^{ac } \ } ; \nonumber \\ \left[j^{ab } , t^{cd}\right ] & = & -i \ { \eta^{ac}t^{bd } + \eta^{ad}t^{bc } - \eta^{bc}t^{ad } - \eta^{bd}t^{ac } \ } ; \\ \left[t^{ab } , t^{cd}\right ] & = & i \ { \eta^{ac}j^{bd } + \eta^{ad}j^{bc } + \eta^{bc}j^{ad } + \eta^{bd}j^{ac } \}. \nonumber \end{aligned}\ ] ] these generators , when combined together as @xmath19 , and with the indices lowered by @xmath11 , give a compact commutation relation of the form of @xmath20 = i \delta^{b}_c m_a^{\ d } - i \delta^{d}_a m_c^{\ b}\end{aligned}\ ] ] we know that the @xmath14 with the defining lie algebra given in eq . [ eq : commu_gen ] has no presupposition of the existence of the minkowskian metric on the locally flat patch of our spacetime . we are introducing @xmath14 into physics in our way because we want to emphasize that there are some very fundamental laws , namely , the law of inertia and the law of causality , working together to impose the @xmath14 symmetry onto our spacetime . `` invariance dictates interaction '' , said c. n. yang . this doctrine seems working well in the electroweak and the strong interactions . we propose in this article that this doctrine could apply to gravity too . the yang - mills gauge potentials for the @xmath14 are @xmath21 note that there are totally sixteen gauge bosons @xmath22 appearing in our theory . the antisymmetric parts of @xmath22 go with the generators @xmath23 while the symmetric parts go with the generators @xmath24 . these sixteen gauge bosons are world vector fields . the yang - mills field strength tensor @xmath25 is @xmath26 \nonumber \\ & = & ( { \partial_{\mu}a^{m}_{\ n\nu } - \partial_{\nu}a^{m}_{\ n\mu } + a^{m}_{\ p\mu}a^{p}_{\ n\nu}- a^{m}_{\ p\nu}a^{p}_{\ n\mu } } ) m_{m}^{\ n } \nonumber \\ & \equiv & f^{m}_{\ n\mu\nu}m_{m}^{\ n}. \end{aligned}\ ] ] the yang - mills lagrangian , which is invariant under the local @xmath14 transformations , is @xmath27 it is interesting to see how we can evaluate the trace of the products of the generators of @xmath14 . for @xmath14 , there exists a relation between the trace of the products and the product of the traces of the generators and the bilinear killing form of the special linear group @xmath28 ( as denoted by a tilde ) , namely @xmath29 where @xmath30 @xmath31 is the bilinear killing form and @xmath32 is the adjoint representation of @xmath33 which is traceless . the calculation of the trace of the products of the adjoint representation is direct , and it is @xmath34 the calculation of @xmath35 is easy , too . the generators of @xmath14 differs from the generators of @xmath28 by a dilatation part of @xmath36 , which is @xmath37 . in other words , @xmath38 , and hence we have @xmath39 combing eq . [ eq : gen_trace ] , eq . [ eq : killing_form ] , eq . [ eq : gen_trace_adj ] and eq . [ eq : gen_trace_2 ] gives @xmath40 this is independent of the kind of representation chosen for @xmath14 . note that the result given in eq . [ eq : gen_trace_3 ] comes not only from the structure constants of eq . [ eq : commu_gen ] , but also comes from a judicious choice of the dilation part of the @xmath14 generators . and hence the affine symmetric yang - mills action @xmath41 , in the presence of the background world metric @xmath4 , will be @xmath42 = \kappa \int \sqrt{-g } d^4x g^{\mu\mu ' } g^{\nu\nu ' } ( \delta_{a}^{d}\delta_{c}^{b } ) f^a_{\ b\mu\nu } f^c_{\ d\mu'\nu'},\ ] ] @xmath43 is a dimensionless coupling constant of the theory . of course , the total action will also contain a piece coming from the matter fields which are supposed to couple gauge invariantly to the gauge fields @xcite , and thus making the total action @xmath44 as @xmath45 even though the world metric serves just as a background measuring clock and stick for our spacetime , it does contribute to the feynman amplitudes in calculating physical processes because @xmath46,\ ] ] where @xmath47 $ ] is the @xmath14 invariant measure over the fields . in the following , we shall study the @xmath48 physics implied by eq . [ eq : amp ] . and this can be done by looking at the extrema of the action . the role played by @xmath4 in classical physics would be clear if we were able to integrate out the @xmath14 gauge field and then solve for the equation of motion containing only the metric . such similar programs were indeed being tried many times before and were never made to work . here we want to suggest a novel way to attack this classical problem by solving the classical yang - mills equation in a background world metric . solving a classical yang - mills gauge equation in a background world metric is by no means easy . fortunately , we are able to show that the @xmath14 gauge theory is identical to a geometric theory quadratic in the riemannian tensor albeit that the metric and connections are now independent variables . and thanks to the works of many people , this geometric theory was shown to contain the schwarzschild metric and some other world metric as solutions . in the following , we are trying to show how this can happen . a particular choice of the background world metric , and a particular set of the affine fields and matter fields , that together extremize the total action , will give us the physics that we are observing in the classical world . in other words , any choice of the world metric is allowed as background metric , but only those metrics that satisfy the extremal conditions are what we are experiencing classically . these extremal conditions are @xmath49 where @xmath50 denotes the yang - mills gauge covariant differentiation . the @xmath51 and @xmath52 are respectively the metric energy - momentum tensor and gauge current tensor of the source matter . this is the set of equations that we are proposing to describe the spatial and temporal evolution of the classical @xmath14 yang - mills and the matter fields , in our spacetime which has a background world metric @xmath4 . it turns out that not all world metrics can sustain a classical @xmath14 yang - mills field , only some selected ones can do . putting eq . [ eq : var_action ] in words : the solved @xmath53 from the second equation ( which is the yang - mills equation ) will be functionals of @xmath4 . and when we plug the solved @xmath53 into the first equation , it will become an equation for @xmath4 . and from this first equation we shall select the world metrics for our classical world . up to this point , we have not used any sophisticated concepts in geometry such as the connection , the parallelism and the curvature . the only thing we have used that may have something to do with geometry is that our spacetime should have a metric , telling us how to measure distance and volume . so the geometry of our spacetime is not riemannian , and is not even affine ; it is just metrical . from our point of view , a metric is fundamental and is needed if we want to construct an action from some local fields . the concepts of connection , parallism and curvature are , however , not fundamental . if we can show that the schwarzschild metric and some other metric can follow from our eq . [ eq : var_action ] , then we can claim that the description of gravity needs no sophisticated geometric ideas other than that of a metric . before we embark on further discussions , we should clarify the role played by our dimensionless parameter @xmath43 and should also clarify the way how the newtonian gravitational constant @xmath54 makes its appearance in gravitation phenomena . @xmath43 appears because there is an arbitrariness in fixing the relative scale between the matter and gauge parts of the lagrangian . the value of @xmath43 should have no significance in physics , and can be absorbed into the definition of the matter field . what have significance on physics are the integration constants that accompany the solutions to the equation of motion . because @xmath6 is dimensionless , @xmath7 must be scaled by some integration constant @xmath55 which has the dimension of length . it is a different @xmath55 that gives a different strength in gravitational interaction . the fact that @xmath55 is proportional to the inertial mass @xmath56 of the gravitational source , @xmath57 , is inferred by comparing the weak field limit of the solution with the newtonian gravitational potential . this situation also happens in general relativity . if we have the experience in solving eq . [ eq : var_action ] in its gauge form , we would have the above claim that classical gravity comes from the yang - mills gauge theory of @xmath14 done . just as we have said before , solving the @xmath14 yang - mills equations in a background world metric is by no means easy . one way that could help us in solving eq . [ eq : var_action ] is to cast this equation into a form that we are acquainted with . from the yang - mills fields @xmath53 and the vierbein fields @xmath58 we can construct some new fields called the @xmath59 fields by @xcite @xmath60 and then substitute the sixty - four variables @xmath22 by the new sixty - four variables @xmath61 , by plugging @xmath22 into @xmath62 in eq . [ eq : f_munu ] . note that @xmath61 is defined by @xmath22 and has , so far , nothing to do with connections . miraculously , the yang - mills field strength tensor can be re - expressed in the @xmath59 fields in a very simple way as @xcite @xmath63 where we have used @xmath64 to stand for @xmath65 . note that our @xmath64 , though looks like the riemann curvature tensor , is , in fact , a derived quantity coming from the yang - mills field tensor . plugging the result in eq . [ eq : ym_f_munu ] into the yang - mills action in eq . [ eq : ym_action ] , the yang - mills action will then look like @xmath66 = { \rm s_{ym}}\left[g , \gamma , \partial\gamma \right ] = \kappa \int \sqrt{-g } d^4x g^{\mu\mu ' } g^{\nu\nu ' } ( r^{\lambda}_{\ \sigma\mu\nu}r^{\sigma}_{\ \lambda\mu'\nu ' } ) .\ ] ] note that there is only one term quadratic in the riemann curvature appearing in the action . when the yang - mills action @xmath41 is expressed in terms of the new fields , the yang - mills equation can , of course , be obtained by extremizing @xmath67 $ ] with respect to @xmath61 . hence we have arrived at the very important fact that the yang - mills equation for the affine symmetry group can be written as @xmath68}{\delta \gamma^{\rho}_{\tau\mu}}\right \vert_{g } + \left . \frac{\delta \rm s_{matter}}{\delta \gamma^{\rho}_{\tau\mu}}\right \vert_{g } = 0.\ ] ] variation with the @xmath4 will give us another equation , which we shall display as @xmath69 note that eq . [ eq : tot_act_var ] is in fact a combined result of the first equation and the second equation of eq . [ eq : var_action ] , because the action can depend on @xmath4 through its dependence on @xmath53 when @xmath61 is held fixed . one interesting thing that is worth noting is that if we are going to integrate over all the distinctive @xmath61 ( apart from a diffeomorphism of the entire spacetime ) , we are in fact factoring out the group volume of @xmath14 as was shown explicitly in eq . [ eq : ym_pot_new ] . the legitimate measure over the fields will then be @xmath70 $ ] . now eq . [ eq : ym_pot_new ] looks like the famous geometric relation between the affine connections and the spin connections under the tetrad postulate , if we regard @xmath61 as the affine connections and @xmath53 as the spin connections . and . [ eq : ym_action_new ] , eq . [ eq : ym_act_var ] and eq . [ eq : tot_act_var ] , together , look like a geometric theory with a geometric lagrangian and a set of geometric equations of motion that we have encountered frequently in talking about gravity . hence it is natural for us to put all these in the following geometric jargons : that a parallel connection @xmath61 is introduced into the spacetime , that a riemann curvature tensor is constructed , that a gravitational lagrangian is formed out of the product of the riemannian curvature tensor , that we are trying to obtain the gravity equations by varying the connections and the metric independently @xmath0la palatini @xcite , and that gravity is a kind of metric - affine theory @xcite . from the above discussions , it is now clear that what all these things that were done in the past , were done in pieces by following the doctrines laid down by klein , einstein , yang and mills and feynman , either knowingly or unknowingly . this form of the gravitational action of eq . [ eq : ym_action_new ] has presented itself many times in the history of the development of the geometric theory of gravity , but is , in fact , carrying very different information at each one of the presentations . the point of focus is on the relation between the metric and the connections . in its very early version , as proposed by hermann weyl @xcite , the connections that appear in the theory are nothing but the christoffel symbols which are of first derivatives in the metric . this will result into a theory in which the metric is the only dynamical variable , and the variation with respect to the metric will give an equation of motion of higher order derivatives . it is well known that such a theory will possess runaway solutions . later yang @xcite , also regarded the connections as the christoffel symbols at the start , but varied the connections instead in order to get the equation of motion . the final result is , again , an equation of higher derivatives in the metric . stephenson @xcite , put the anti - symmetric parts of the connections equal to zero , and regarded the symmetric parts of the connections and the metric as independent variables . and he obtained two equations of motion by varying both the metric and symmetric parts of the connections independently . on the other hand , some people identify the symmetric parts of the connections as the christoffel symbols and regard the anti - symmetric parts and the metric as independent variables . those people working on the poincare gauge theory of gravity are taking this point of view @xcite . for us , the full connections , both the symmetric and the anti - symmetric parts , as well as the metric are independent variables . in this theory of gravity , the connections are just the transformed @xmath14 yang - mills vector potentials @xmath71 , and can be taken as being independent of @xmath4 . here we feel obligatory for us to re - assert the reason that we are making excursion into the land of geometry is because we want to make use of some of the results known to the people working in the geometric theory of gravity . and of course , we also want to know how our proposed theory looks like in geometrical languages . the different choices of the content coded in the riemann curvature tensor give different stories for physics . for example , for the weyl theory , the metric is the only dynamical variable , and hence the action will contain kinetic terms that have derivatives that are of orders higher than two . and when we look for the possible propagation modes in the theory , which can be obtained by looking at the inverse of the kinetic term , we will find that there will be propagators having the wrong signs , which will correspond to unphysical states called the ghosts or tachyons , and will end up into an unstable theory with the so called ostrogradski instability @xcite . for us , the metric is a non - dynamical background field @xcite , and the only dynamical variables are the connections which obey an equation second order in space and time derivatives . the propagating modes are the 16 vector bosons and nothing else . and hence our theory will contain no ghost and no ostrogradski instability . we are not having the pathologies that are affecting quadratic curvature theories in which the metric is dynamical and is compatible with the connections . let us now concentrate ourselves on the vacuum " solutions of eq . [ eq : ym_action_new ] , eq . [ eq : ym_act_var ] and eq . [ eq : tot_act_var ] . by vacuum " here we shall mean the case where all the matter fields are absent , except possibly at the source point . however , the @xmath14 gauge vector fields may not necessarily be vanishing . a trivial solution with a global minkowskian world metric and vanishing @xmath14 gauge potentials can be inferred immediately from the equations . solving the `` vacuum '' @xmath14 yang - mills equation in the presence of a background metric will be prohibitively difficult , even in the spherically symmetric cases ( see appendix ) . however , we are fortunate enough to have shown that the @xmath14 yang - mills gauge theory in a background metric is equivalent to the metric - affine theory of gravity with a lagrangian quadratic in the riemann curvatures , namely eq . [ eq : ym_action_new ] . in this geometric language , we can then invoke the concept of metric - affine compatibility and of the concept of torsion . it is under the requirements of compatibility and torsionless that we are trying to see if we can find a subclass of solutions to eq . [ eq : ym_action_new ] , eq . [ eq : ym_act_var ] and eq . [ eq : tot_act_var ] . in the following , we shall search for solutions @xmath72 the ansatz that the world metric @xmath4 and the connections @xmath73 are compatible with each other . we shall call this ansatz the compatibilty ansatz ( ca ) . and we consider only solutions that are torsionless . then the vacuum " version for eq . [ eq : ym_action_new ] , eq . [ eq : ym_act_var ] and eq . [ eq : tot_act_var ] can be written as @xmath74_{{\rm ca \ + \ torsionless } } = 0 , \nonumber \\ \end{aligned}\ ] ] which means @xmath75 , and @xmath76_{{\rm ca \ + \ torsionless } } \nonumber \\ & = & r^{\lambda \ \theta}_{\ \sigma \ \rho}_{\ \lambda } -\frac{1}{4}g^{\theta\tau}r^{\lambda \ \xi\rho}_{\ \sigma}r^{\sigma}_{\ \lambda\xi\rho } = 0.\end{aligned}\ ] ] because the solutions that we are searching are metric compatible and torsionless , then the solution @xmath73 will become the levi - civita connection for @xmath4 ( this is guaranteed by the fundamental theorem of riemannian geometry , and note that the substitution of @xmath73 by the levi - civita connection is done only after the variation ) . a proper decomposition of the curvature tensor and the proper use of the bianchi identities will convert eq . [ eq : ca_sky ] and eq . [ eq : ca_step ] into the stephenson - kilmister - yang equation @xcite and the algebraic stephenson equation @xcite , respectively as , @xmath77 @xmath78 here we have quoted the results given in ref . @xcite , with @xmath79 as the traceless weyl conformal curvature . @xmath80 is the covariant derivative with respect to the levi - civita connections . obviously the above two equations are satisfied simultaneously by the vanishing of the ricci curvature tensor , and hence satisfied by the schwarzschild metric . note that we did nt solve eq . [ eq : var_action ] directly , but routed ourselves into the domain of the theory of quadratic gravity and borrow the results from her . our statement that the schwarzschild metric is induced by a configuration of the 16 vector bosons is thus verified . it is also worth mentioning that there exists another metric ( the thompson - pirani - pavelle - ni , tppn , metric @xcite ) , different from the schwarzschild metric , that is also a simultaneous torsionless solution to eq . [ eq : sky ] and eq . [ eq : step ] . this new metric , together with the schwarzschild metric , suggest the existence of more than one gravitational copies of matter in nature . the schwarzschild and tppn metric are displayed in the following , respectively , @xmath81 @xmath82 where @xmath83 and @xmath84 are the integration constants for the solutions . in these two solutions , the metrics happen to be compatible with their respective connections . but we have to bear in our mind that we have not assumed metric compatibility , as a priori , in our formulation of the theory . the existence of some solutions which are metric compatible does not mean that the theory is a higher derivative theory . the tppn solution was first discovered as a solution to the vacuum eq . [ eq : sky ] , with symmetric connections and with no reference to eq . [ eq : step ] . it was later shown by baekler , yasskin , ni , and fairchild @xcite , and by hsu and yeung @xcite that this solution satisfies vacuum eq . [ eq : sky ] for the full connections . it is also easy to see that these metrics will also satisfy the vacuum eq . [ eq : step ] @xcite . in fact , it was shown @xcite that the schwarzschild and the tppn solutions are the only two possible simultaneous solutions to vacuum eq . [ eq : step ] and vacuum eq . [ eq : sky ] for spherical symmetric situation under the compatibility ansatz . the most straight forward way to show the validity of the solutions given in eq . [ eq : s_metric ] and eq . [ eq : tppn_metric ] is to cast vacuum eq . [ eq : step ] and vacuum eq . [ eq : sky ] in their spherical symmetric forms . in the appendix , we will give vacuum eq . [ eq : step ] and vacuum eq . [ eq : sky ] in their spherical symmetric forms , and we can see that both solutions satisfy the equations . a very thorny problem facing physicists when they try to develop a local affine symmetric gauge theory for gravity is the inexistence of finite dimensional spinor representations for the gauge group @xmath14 @xcite . the observed finite dimensional spinor fields with definite spins and masses will invalidate our claim that we are having @xmath14 symmetries in our laboratories . this dilemma is resolved here in the following way : in spite of the fact that the lagrangian of the theory is having the full symmetry of the affine group , the solutions of the equations of motion ( for example , the classical solutions that include the schwarzschild metric and the tppn metric and the accelerating cosmic metric as given below ) which describe our classical gravitational phenomena , retain far less symmetries . what are the residual symmetries that are left unbroken ? this can be answered by noting that the classical @xmath61 are now the levi - civita connections , which are compatible with the background metric . this compatibility will require that the observed yang - mills vector fields @xmath22 be anti - symmetric in their @xmath85 , @xmath86 indices . for this case , the ten symmetric generators @xmath17 of the @xmath14 will not be used . the remaining six anti - symmetric generators @xmath16 are just the generators of the local lorentz group . the geometric picture for the compatibility ansatz is thus that we are going to identify the defining minkowskian frame for the vierbein fields and its rotations as our admissible geometric settings in the discussions of classical gravity physics . hence what is left unbroken by our solutions is the local lorentz symmetry . the changes in the gauge potentials are compensated by the changes in the vierbein fields when the minkowskian axes are rotated , so as to leave our metrics and levi - civita connections unchanged . that explains why we are now seeing particles of definite spins and masses in our laboratories . in summary , classical gravity is expressed by a spontaneously broken erlangen program @xcite , breaking down the local @xmath14 symmetry to the local lorentz symmetry . some may wonder if it is legitimate to use @xmath14 for a local gauge theory became of its non - compactness . the reader is referred to the appendix of ref . @xcite for a discussion on the positive - definiteness of @xmath14 . though the schwarzschild metric has already found a lot of applications in the studies of various gravitational phenomena , the tppn metric was dismissed by its discoverers soon after its discovery because it failed to reproduce the classical tests that are so successfully predicted by the schwarzschild metric . and this leads , subsequently by many people , to the conclusion that the yang - mills gauge theory is not a viable physical theory . here we want to show that a suitable interpretation of the above two metrics can save this situation . and will reproduce nicely both the galactic rotation curves and the amounts of intergalactic lensing , if nature follows our following postulate . we postulate that if nature is going to make use of this yang - mills gauge theory of gravity , matter will then be endowed with either one of these two metrics . matter endowed with the first metric ( @xmath87 , metric given in eq . [ eq : s_metric ] ) will be called the regular matter , and matter endowed with the second metric ( @xmath88 , metric given in eq . [ eq : tppn_metric ] ) will be called the primed matter . both the regular matter and the primed matter were produced during the creation of the universe , though in different amounts , probably because of the difference in the requirements of energy in producing them . there might possibly be other particles endowed with some yet undiscovered metric solutions . these particles are assumed to be too heavy to be produced in an amount large enough to be noticeable in the present day astronomical observations . in other words , we will suffice ourselves by considering the physics played by these two gravitational copies of matter . the immediate question is : how do we know that there exist solution to eq . [ eq : s_metric ] and eq . [ eq : tppn_metric ] that describe the co - existence of the regular matter @xmath56 and primed matter @xmath89 as the source of the gravitational field ? the gravitational field equations are highly nonlinear , a rigorous solution for a source with both regular matter @xmath56 and primed matter @xmath89 is hard to find , but we can infer the existence of such a solution by forming a linear superposition of @xmath90 this linearly superposed metric is asymptotically minkowskian . as we are going to explain in the appendix , the fact that @xmath83 and @xmath91 are integration constants for the solutions will imply that they do not appear in the equations that embody the solutions . in the limit of @xmath92 ( or @xmath93 ) , @xmath94 will be getting closer and closer to be a solution . the deviations of @xmath94 from a solution will be proportional to @xmath95 or its higher orders as @xmath92 ( or to @xmath96 or its higher orders as @xmath93 ) . note that @xmath97 signifies the significance of the schwarzshild component in @xmath94 while @xmath98 signifies the significance of the tppn metric in @xmath94 . more details on this can be found in the appendix . the next question is : how does a test particle , which is either a regular matter of mass @xmath85 or a primed matter of mass @xmath99 reacts to the gravitational field which is produced by a mixture of @xmath56 and @xmath89 ? the energy - momentum tensor @xmath100 is a functional of the metric . it is natural to assume that when it is evaluated at @xmath87 , it will be the energy - momentum tensor for the regular matter . and when it is evaluated at @xmath88 , it will be the energy - momentum tensor for the primed matter . so we have @xmath101 it was explicitly pointed out by stephenson @xcite that when we use the full connections ( these connections are having both the symmetric and the anti - symmetric parts ) as dynamical variables , then the bianchi identities and vacuum eq . [ eq : sky ] together will imply a local covariant conservation law for the metric energy - momentum tensor @xmath102 this is valid for any generic metric and connections which satisfy vacuum eq . [ eq : sky ] . in particular when the metric is @xmath87 ( or @xmath88 ) with corresponding connection @xmath103 ( or @xmath104 ) , we will have two separate conservations laws @xmath105 eq . [ eq : eng_mom_tensors ] and eq . [ eq : sep_conser ] together will mean that the regular matter and the primed matter are separately conserved . therefore , a regular ( primed ) particle at @xmath7 will mean an energy - momentum tensor that is localized at @xmath7 and satisfies the first ( second ) conservation law . by integrating over a small spacetime region surrounding the test particle , a local covariant conservation law can be translated into an equation of motion for the test object which shares the same @xmath51 with the source but not affecting the metric @xcite . eq . [ eq : eng_mom_tensors ] and eq . [ eq : sep_conser ] will mean that the regular ( primed ) matter will move under the influence of the regular ( primed ) metric which is generated by regular ( primed ) matter . in other words the regular ( primed ) matter will interact only with the regular ( primed ) matter gravitationally . a little care should be paid in the situation of a source of a mixture of @xmath56 and @xmath89 : in the presence of a metric given in eq . [ eq : superposition ] , a regular test particle will have an energy - momentum tensor of the form of @xmath106 . because either @xmath87 or @xmath107 gives the same @xmath103 ( @xmath103 wo nt change under a constant scaling of the metric ) , @xmath106 will obey the same local covariant conservation law as @xmath108 . similar situation holds for a primed test particle . other than gravitational interactions , the regular matter and the primed matter are assumed to be identically the same in all other interactions . the geodesic equations are equivalent to the acceleration equations under the respective metric . and the respective acceleration produced by these metrics , when the speed of motion is small compared with the speed of light , can be calculated from @xmath109 and are respectively @xmath110 and @xmath111 . the @xmath112 and @xmath89 are introduced in order to make a parallel comparison between the newtonian gravitational force and the new gravitational force . we will call @xmath112 the primed gravitational constant , and @xmath89 the primed gravitational mass . this new gravitational force is attractive when @xmath84 is positive . let us now consider the interesting situation in which a regular matter of mass @xmath85 is sticking together with a primed matter of mass @xmath99 through non - gravitational interactions , and in which they are moving together in a circular motion of radius @xmath113 under the influence of a gravitation field produced by a source consisting of a regular mass @xmath56 and a primed @xmath89 . note that even though the regular matter and the primed matter may be very close together , they nevertheless occupy different spacetime points and hence their energy - momenta will not overlap , and hence they will interact with their own gravitational forces , and the centrifugal force will balance the combined gravitational forces @xmath114 and then we will arrive at a relationship holding between the rotation speed @xmath115 with the distance @xmath113 , @xmath116 there is an immediate application of eq . [ eq : v_squ ] to describe the kinematics of a stellar object which is made up of the regular matter @xmath85 and the primed matter @xmath99 , and is moving at a distance @xmath113 from the center of a spiral galaxy . the spiral galaxy is assumed to be composed of a disk solely of regular matter of mass @xmath56 and a halo composed of a uniformly distributed matter with a contamination of primed matter of mass density @xmath117 . because of @xmath118 where @xmath89 is the total primed mass inside @xmath113 , and because the arm structure of the regular matter presents its contribution in modified bessel functions form @xcite , the total contributions to @xmath119 by the spiral arms and the halo will sum up to @xmath120 \times ( \frac{m}{m+m ' } ) \\ & & + \frac { g^{\star } r^3}{r + g^{\star}r^3 } ( \frac{m'}{m+m ' } ) , \nonumber\end{aligned}\ ] ] where @xmath121 , @xmath122 are modified bessel functions of the first and second kinds . the @xmath123 is the speed coming from the newtonian force of the disk with @xmath124 @xcite . we will see immediately that this agrees exactly with the empirical formula given by salucci @xmath125 @xmath126 @xcite who analyzed a large number of spiral galaxies and drew up a formula to describe their rotation curves . we have also extracted the values of @xmath127 ( which is the product of the primed gravitational constant and the primed matter density in the halo ) and @xmath128 ( which is the ratio of the primed mass density to the regular mass density in the halo ) by fitting some well - known galaxies @xcite , and it is amazing to find that a universal value of @xmath129 and a universal ratio of @xmath130 fit very well with the observed results when @xmath113 ranges from 3 kpc to 30 kpc . note also that the value of @xmath131 and @xmath132 are more or less the same as those observed . the results are shown in fig . [ fig:1 ] and table [ table:1 ] . [ cols="^,^,^,^,^",options="header " , ] we also have to check for the influence of this diffuse halo medium on the motion of the planets in our solar system . with the values of the @xmath127 and @xmath128 given in the above , the observed rotation speeds of the planets fit well with the predictions of eq . [ eq : v_squ_new ] . figure [ fig:2 ] gives the planetary motions for various values of @xmath127 and @xmath128 . with different @xmath128 values . @xmath133 , @xmath134 and @xmath135 are for red , blue and green points respectively . [ fig:2],scaledwidth=47.0% ] the alert reader may find that , in the above discussions , we have already made the assumption that the stellar objects in the galaxies are composed solely of regular matter . this assumption can be understood in the following way . the primed matter always respond to the primed gravitational pull with a high rotation speed even when they are bound with some regular matter and hence are far harder for them to condense gravitationally into a star . though the stars contain no primed matter , the stars that are rotating at the outskirt of a spiral galaxy are , in fact , embedded in pockets of halo media and are rotating around the galaxy all together . the amount of regular matter in the halo is small when compared with that in the galactic bulge and spiral arms , and is thus neglected in the above discussions . next , let us turn to see what the primed matter does in explaining the large light deflections that are observed in intergalactic gravitational lensing . we shall take the galaxy cluster abell 1689 as our illustration . we shall regard abell 1689 as a cluster consisting of galaxies which are carrying their own individual halos with them . and this collection of galactic halos forms the halo of the cluster . since the galactic halos are always regarded as having a size of the order of 30 kpc , the size of the cluster will be very close to its halo size which is taken to be 300 kpc . the azimuthal angle swept by the light , when it travels from point @xmath136 to the point of closest approach @xmath137 , under the influence of gravity described by the metric @xmath138 is given by @xcite @xmath139^{-\frac{1}{2 } } \frac{dr}{r}.\ ] ] in the case of the tppn metric for a point source of primed mass @xmath89 , the angle swept is @xmath140^{\frac{1}{2}}},\ ] ] which can be readily integrated to give @xmath141 with @xmath142 = @xmath143 . we will immediately notice that the angle change will be @xmath144 when @xmath136 goes to infinity . that means no deflection for the light by a point source of the primed matter when it comes from infinity and then goes back to infinity . things will be different when we deal with a distributed source of the primed matter as we are going to show in the following . let @xmath136 be the radius of the halo of abell 1689 and @xmath137 be the closest approach from the cluster center . the light will see a point source of the primed matter of constant mass @xmath89 when it is moving beyond the halo . and it will see a point source of diminishing mass ( i.e. increasing @xmath142 ) when it enters the halo because it sees only the mass that lies inside @xmath113 . we claim that the deflection by a point source of mass @xmath89 is smaller than that by a uniformly distributed halo which has a total mass of @xmath89 , if the light penetrates into the halo during some time in its journey . the above claim is obvious by noting that @xmath145 so if the light penetrates into the cluster halo , the total azimuthal angle change will be greater than @xmath144 , and we will see light deflectng towards the center of the cluster . the actual angle swept , when light comes from the infinity , enters the halo at @xmath136 and reaches the point of closest approach at @xmath137 is given by @xmath146 the second term in the bracket in eq . [ eq : delta_phi_result_2 ] comes from the fact that we have to take out a @xmath144 in calculating the angle of deflection . the quantity inside the bracket of eq . [ eq : delta_phi_result_2 ] can be regarded as a change @xmath147 in @xmath148 due to a change @xmath149 in @xmath89 , and we get @xmath150 \delta m ' \nonumber \\ & & = \frac{3}{\sqrt{2 } } ( r - r_0)^ { \frac{3}{2 } } g^{\star -\frac{1}{2 } } r^{-\frac{5}{2}}. \nonumber \end{aligned}\ ] ] a plug in the data of @xmath136 = 300 kpc , @xmath151 = 30 kpc and @xmath127 = @xmath152 @xmath153 , will give us an angle of deflection , which is 2@xmath154 , a value of @xmath155 , which is many times of that expected from general relativity . note that the value of @xmath127 that we used in calculating the abell 1689 light deflection comes from the curve fittings of galactic rotations . again there seems to be a universal value for @xmath127 , as it should be . we should also note that the sun and the planets in the solar system are nt contaminated with the primed matter as we have explained in the above . the fundamental tests on general relativity will hence remain intact . there is another nice feature of this yang - mills gauge theory of gravity , when we use it to study the universe as a whole . this theory was shown to admit a cosmological solution of the form @xcite @xmath156 with primordial local torsion compoments , @xmath157 the @xmath137 and @xmath158 , with @xmath159 are integration constants arising from the integration of the equation of motion . we can interpret this solution as representing an expanding and accelerating universe when the influence of gravity dominates over the influence of matter and radiation . the role played by the primordial torsion is crucial here : the stretching on the universe by the metric is compensated by the twisting by the torsion . and from the metrical point of view , we look like living in a universe with a cosmological constant @xmath158 . and interesting enough , our torsion selects the spatially flat metric ( @xmath43 = 0 ) as the only accompanying metric @xcite . the spatially flat geometry of the universe is confirmed by wmap . for more information on the predictions of the yang - mills type gauge theory of gravity on the evolution of the primordial universe , the reader is referred to ref . we are fully aware that some of the terms and ideas used here have already appeared in the literatures . but we want to emphasize that they are appearing here in very different contexts . for example , we do not regard the world metric in our theory as a fundamental variable . instead , the world metric is just an arbitrary background of measuring clock and stick for our spacetime . the observed metric takes a particular form ( for example the schwarzschild metric ) simply because the affine gauge bosons require that particular metric in order to exist as a solution . as a result , we will not have spin-2 gravitons in our theory ; the 16 spin-1 vector bosons are the only propagating particles for gravitation . and the dynamics of gravity is fully governed by the @xmath14 yang - mills equation of motion . our theory is therefore a theory of vector gravity . it might be interesting to point out that there had been many works done in the direction called the gauge theories of gravity . yet most of them are not gauge theories in the sense of yang and mills . people are either reluctant to give up the metric ( or the vierbein ) as gauge potential because they might think that the metric is too important for gravity theory to be put in an auxiliary position . or they might be afraid to take up the gravitational lagrangian that is quadratic in the field strength tensor as yang and mills did , because of the fear that there could be spurious solutions that will upset the known gravitational observations @xcite . but an important factor that is preventing people to arrive earlier at a true gauge theory of gravitation based on @xmath14 , as we believe , is the wrong perception that the particle contents of the theory will not fit into experiments . it is one of our observations , as we have explained in the above , that our observed gravitational world is , in fact , a classical solution to the affine symmetric theory and is in a state of spontaneously broken symmetry . the remaining symmetry is the local lorentz symmetry . finally , a gravity theory with the lorentz group ( which is a subgroup of @xmath14 ) as the local gauge group can be obtained from our theory by simply restricting ourselves to the anti - symmetric components of the @xmath22 . only the generators @xmath16 and the first commutation relation of eq . [ eq : commu_rel ] will be used . and we shall have a yang - mills gauge theory of 6 gauge vector bosons . all the results given in this article will then remain valid , with the exceptions that the local lorentz symmetry will now be honored by both the lagrangian and the solutions to the equations of motion , and that the metric and the connections will now be compatible automatically . one price has to be paid , though , by gauging the lorentz group instead of the @xmath14 . the riemannian tensor will then contain terms which are of second derivatives in the metric , and the metric and the torsion ( which are functions of the vierbeins and the gauge potentials ) are independent variables . and gravity theories basing on lagrangians quadratic in the riemannian tensor may have to face the affections by many of the pathologies of higher derivative theories . from the solutions of the equation of motion , it seems that gauging the lorentz group is good enough to describe the present day physical phenomena of gravity . apart from staying away from the pathologies that we have just mentioned by gauging the lorentz group , we are venturing into @xmath14 because we believe that there might be something in physics , in addition to gravity , that can be explained by the full affine symmetric gauge theory @xcite . the yang - mills gauge theory of gravity has a richer structure than that of einstein s general theory of relativity . its richer number of solutions , both in the absence and in the presence of torsion allow us to describe more physical phenomena with it . the recently observed astronomical dark matter and dark energy phenomena seems to show that nature is enjoying the full use of this yang - mills gauge theory of gravity . in this appendix we will cast the vaccum stephenson equation and the vacuum stephenson - kilmister - yang equation in static spherical symmetric forms . the metric is taken to be the form of eq . [ eq : gen_metric ] , namely @xmath160 there are three independent , non - vanishing components for the left - hand - side of eq . [ eq : step ] . the @xmath161 component , @xmath162 the @xmath163 component , @xmath164 the @xmath165 component ( same as the @xmath166 component ) , @xmath167 there are two independent components for the left - hand - side of eq . [ eq : sky ] . they are @xcite @xmath168 and @xmath169 note that all these equations contain terms that are polynomials in @xmath170 and @xmath171 . subtracting eq . [ eq : rr_component ] from eq . [ eq : tt_component ] will give @xmath172 and will lead to @xmath173 the primes appearing as suffices in eq . [ eq : gen_metric_2 ] to eq . [ eq : subtract_eq_2 ] signify differentiations with respect to @xmath113 . eq . [ eq : subtract_eq_2 ] are the two solutions for the schwarzchild metric and the tppn metric , respectively . for halos with both regular matter and primed matter whose densities are decreasing as @xmath113 increases , the rhs of eq . [ eq : gen_metric_2 ] to eq . [ eq : sky_ind_2 ] will no longer be vanishing , and eq . [ eq : subtract_eq_2 ] will not hold . in that case , as explained in the following , an approximation to the solutions of the gravitational equations can be constructed as the superposition of @xmath87 and @xmath88 in the form of @xmath174 and if we use @xmath175 and @xmath176 to denote the @xmath161 and @xmath163 components of the metric of eq . [ eq : superposition_app ] then @xmath177 in the limit of @xmath178 . the bars and primes here are used to label the regular matter and the primed matter as we have mentioned in the article . note that the above equations of eq . [ eq : gen_metric_2 ] to [ eq : sky_ind_2 ] do not have @xmath83 and @xmath91 , which appear in eq . [ eq : superposition_app ] , as their coefficients because @xmath83 and @xmath91 are integration constants of the solutions . substituting eq . [ eq : b_and_d ] into eq . [ eq : tt_component ] , [ eq : rr_component ] , [ eq : thetatheta_component ] , [ eq : sky_ind_1 ] and [ eq : sky_ind_2 ] , and if we remember that @xmath179 and @xmath180 satisfy the original equations , then the left - hand - side of eq . [ eq : tt_component ] , [ eq : rr_component ] , [ eq : thetatheta_component ] , [ eq : sky_ind_1 ] and [ eq : sky_ind_2 ] will contain only terms that are proportional to @xmath96 or it higher orders . what that means is that for some given @xmath83 and @xmath84 , and some given @xmath179 and @xmath180 satisfying the original equations , @xmath175 and @xmath176 will satisfy the equations approximately as far as @xmath96 is very small . that means the superposition given in eq . [ eq : superposition_app ] is an approximate solution to the equations as far as @xmath96 is very small . the physical consequence is : at least in the limit of @xmath181 , a solution of the form of eq . [ eq : superposition_app ] which represents halos having mixtures of regular and primed matter exist in nature . similar arguments apply to the case when @xmath95 is very small . we would like to thank professor friedrich w. hehl , professor james m. nester , professor t. c. yuan , and professor daniel wilkins for various suggestions . in particular , we would like to thank professor nester for pointing out some errors in the previous version of our manuscript . 10 t. kaluza , math . phys . ( berlin ) * 1921 * 966 ( 1921 ) ; o. klein , z. phys . a * 37 * 895 ( 1926 ) . r. r. hsu and w. b. yeung , phys . b * 155 * 143 ( 1985 ) and references therein . r. utiyama , phys . rev . * 101 * 1597 ( 1956 ) . f. hehl , j. mccrea , e. mielke and y. neeman . reports * 258 * 1 ( 1995 ) and references therein . see h. coxeter , introduction to geometry . new york , john wiley and sons ( 1989 ) . the original affine transformations introduced by euler include the action of translations . here we are talking about local transformations without the translations , and shall borrow the term local affine transformations to mean our proposed transformations . f. klein , ann . math . * 43 * 63 ( 1893 ) . c. n. yang and r. l. mills , phys . rev . * 96 * 191 ( 1954 ) . j. w. moffat , j. of math . phys . * 29 * 1655 ( 1989 ) . infinite - component @xmath14 spinors are considered as matter fields in this paper and in @xcite . see also the discussions given in the followings . 16 looks like the relations between the spin connections and the levi - civita connections under the tetrad postulate . here we want to emphasize that we are just making a variable substitution for the yang - mills equation . j. yepez , arxiv : 1106.2037v1 ( 2011 ) . see also appendix ii of ref @xcite . a. palatini , rend.circ . palerno * 43 * 203 ( 1918 ) . h. weyl , ann . ( leipzig ) * 59 * , 101 ( 1919 ) . for a discussion on higher derivative theories , see k. s. stelle , phys . * d * * 16 * , 953 ( 1977 ) and references therein . c. n. yang , phys . lett . * 33 * , 445 ( 1974 ) . g. stephenson , nuovo cimento * ix * , 263 ( 1958 ) . f. w. hehl , erice lecture 1979 , plenum press , new york ( 1979 ) . y. yang and w. b. yeung , arxiv:1205.2690 ( 2012 ) . for an analysis on the occurrence of unphysical states in gravity theories with torsion , see h. j. yo and j. m. nester , int . j. mod . phys , * d * 747 ( 2002 ) , and references therein . g. stephenson , nuovo cimento * 9 * 263 ( 1958 ) . e. e. fairchild jr , phys . rev . * 14 * 384 ( 1976 ) . c. w. kilmister and d. j. newman , proc . * 57 * 851 ( 1961 ) . y. yang and w. b. yeung , arxiv:1210.0529 ( 2012 ) . m. q. chen , d. c. chern , r. r. hsu and w. b. yeung , proc roc ( taiwan ) * 9 * 10 ( 1985 ) . for another approaches on the relations between spontaneous breakdowns of symmetries and gravity , see y. neeman , and dj . sijacki , phys . lett * b * * 200 * 489 ( 1988 ) , and e. t. tomboulis , arxiv.1105.5848v1 ( 2010 ) see for example , y. neeman and dj . sijacki , ann . of phys . * 120 * 292 ( 1979 ) . a. h. thomspson , phys . * 34 * , 507 ( 1975 ) ; r. pavelle , phys . lett . * 34 * , 1114 ( 1975 ) . p. baekler , p. yasskin , general relativity and gravitation , * 16 * 1135 ( 1984 ) ; w - t ni , phys . rev . lett . * 35 * 319 ( 1975 ) ; erratum * 35 * 1748 ( 1975 ) ; e. e. fairchild , jr . , phys . rev . * d * , 384 ( 1976 ) . r. r. hsu and w. b. yeung , chinese j. phys.(taiwan ) , * 25 * 463 ( 1987 ) . a. zee and john e. kim , phys * d * 21 , 1938 ( 1980 ) . a. einstein , l. infeld and b. hoffmann , ann . math . * 41 * , 455 ( 1940 ) . a. toomre , astrophys . * j * * 138 * , 385 ( 1963 ) . p. salucci , a. lapi , c. tonini , g. gentile , i. yegorova and u. klein , mon . not . r. astron . 378 , * 41 * ( 2007 ) and references therein . y. sofue , @xmath182 . r. bottena , j. gerritsen , mon . not . r. astron soc . * 000 * 1 ( 1997 ) . y. m. lee , m. s. chun , the j. of the korean astron soc . * 22 * 31 ( 1989 ) . m. e. bacon , a. sharrar , eur . * 31 * 479 ( 2010 ) . s. mcgaugh , the astro . j. * 683 * 137 ( 2008 ) . s. weinberg , gravitation and cosmology , john wiley & sons . inc ( 1972 ) . h. chen , r. r. hsu and w. b. yeung , class . quantum grav . * 2 * , 927 ( 1985 ) . simple inspection on the equations list in this paper will reveal that a = @xmath183 , h = k = @xmath184 and f = g = 0 readily solves the equations . m. q. chen , d. c. chen , r. r. hsu and w. b. yeung , proceedings of the nat . council ( taiwan ) * 9 * , 10 ( 1985 ) . y. yang and w. b. yeung , arxiv:1303.3801 ( 2013 ) .

a large class of non - markovian quantum processes in open systems can be formulated through time - local master equations which are not in lindblad form . it is shown that such processes can be embedded in a markovian dynamics which involves a time dependent lindblad generator on an extended state space . if the state space of the open system is given by some hilbert space , the extended state space is the triple hilbert space which is obtained by combining the open system with a three state system . this embedding is used to derive an unraveling for non - markovian time evolution by means of a stochastic process in the extended state space . the process is defined through a stochastic schrdinger equation which generates genuine quantum trajectories for the state vector conditioned on a continuous monitoring of an environment . the construction leads to a continuous measurement interpretation for non - markovian dynamics within the framework of the theory of quantum measurement .

quant-ph0403117

an open quantum system is a certain distinguished quantum system which is coupled to another quantum system , its environment @xcite . a particularly simple way of describing an open system is obtained in the markovian approximation . in this approximation all memory effects due to system - environment correlations are neglected which usually leads to a markovian master equation , that is , to a linear first - order differential equation for the reduced density matrix @xmath2 of the open system with a time independent generator . generally , one demands that the generator is in lindblad form @xcite , which follows from the requirements of the conservation of probability and of the complete positivity of the dynamical map @xcite . a remarkable feature of markovian master equations in lindblad form is given by the fact that they allow a stochastic representation , also known as unraveling , by means of a stochastic schrdinger equation ( sse ) for the state vector of the open system @xcite . a sse generates the time evolution of the state vector which results from a continuous monitoring of the environment of the system @xcite . a specific realization @xmath3 of the sse is called a quantum trajectory : at each time @xmath4 the open system is known to be in a definite state @xmath5 under the condition that a specific readout of the monitoring of the system s environment is given . the reduced density matrix at time @xmath6 is therefore obtained if one averages the quantity @xmath7 over all possible quantum trajectories . this means that the relation @xmath8 $ ] holds , where the symbol @xmath9 denotes the ensemble average or expectation value . in the markovian case it is thus true that the environment acts as a quantum probe by which an indirect continuous observation of the system is carried out . the description by means of a markovian master equation in lindblad form is , however , only an approximation which uses the assumption of short correlation times . for strong couplings and low temperature environments memory effects can lead to pronounced non - markovian behavior . it is sometimes argued that the treatment of non - markovian processes by means of master equations necessarily requires solving integro - differential equations for the reduced density matrix . such equations arise , for example , in the application of the nakajima - zwanzig projection operator technique @xcite which leads to dynamic equations involving a retarded memory kernel and an integration over the past history of the system . however , the usage of another variant of the projection operator method allows in many cases the derivation of approximate or even exact non - markovian master equations for the reduced density matrix which are local in time . this method is known as the time - convolutionless ( tcl ) projection operator technique @xcite . the non - markovian character of the tcl master equation is reflected by the fact that its generator may depend explicitly on time and may not be in lindblad form . time - local equations which are of the form of the tcl master equation have also been derived by other means , e. g. , by path integral and influence functional techniques @xcite . a well known example is provided by the exact equation of motion for a damped harmonic oscillator coupled linearly to a bosonic reservoir @xcite . the fact that the tcl generator is generally not in lindblad form leads to several important mathematical and physical consequences . in particular , a stochastic unraveling of the tcl master equation of the form indicated above does not exist : any such process will automatically produce a master equation whose generator is in lindblad form . the question is therefore as to whether one can develop a general method for the construction of stochastic schrdinger equations for non - markovian dynamics which do have a physical interpretation in terms of continuous measurements . it is the purpose of this paper to show that this is indeed possible . our starting point is a time - local non - markovian master equation for the density matrix @xmath2 on some hilbert space @xmath0 with a time dependent and bounded generator . it will be demonstrated that the dynamics given by such a master equation can always be embedded in a markovian dynamics on an appropriate extended state space . the non - markovian dynamics thus appears as part of a markovian evolution in a larger state space . if one chooses the extended state space as the hilbert space of the total system , consisting of open system plus environment , this statement is of course trivial . however , it turns out that the embedding can be realized in a fairly simple , much smaller state space , namely in the tensor product space @xmath1 . in physical terms this is the state space of a composite quantum system which results if one combines the original open system on the state space @xmath0 with a further auxiliary three state system described by the state space @xmath10 . the open system could be , for example , a damped quantum particle interacting with a dissipative environment . the auxiliary system can then be realized through an additional internal degree of freedom of the particle which leads to a state space spanned by three basis states . it will be demonstrated that the dynamics in the extended state space follows a markovian master equation with a time dependent generator in lindblad form . the application of the standard unraveling of markovian master equations to the dynamics in the extended state space therefore yields a stochastic unraveling for the non - markovian dynamics . the resulting sse generates genuine quantum trajectories which do admit a physical interpretation in terms of a continuous observation carried out on an environment . the construction thus gives rise to a consistent measurement interpretation for non - markovian evolution in full agreement with the general setting of quantum measurement theory . the paper is structured as follows . section [ qmt - m ] contains a brief review of the continuous measurement theory for markovian dynamics . time dependent generators in lindblad form are introduced in sec . [ lindblad - gen ] , and sec . [ cmt - markov ] treats the corresponding continuous measurement unraveling . the quantum measurement theory for non - markovian evolution is developed in sec . [ qmt - nonmarkov ] . we introduce time - local non - markovian master equations in sec . the embedding of these equations in a markovian dynamics is constructed in sec . [ embedding ] , whereas the derivation of the continuous measurement unraveling is given in sec . [ sse - nonmarkov ] . the construction of the sse and its physical interpretation are illustrated by means of an example in sec . [ example ] . a series of interesting stochastic unravelings of non - markovian quantum dynamics is known in the literature . section [ conclu ] contains a discussion of our results and of the relations to alternative non - markovian sses , as well as some conclusions . we consider a density matrix @xmath11 on a state space @xmath12 which obeys a master equation of the form : @xmath13 + \sum_i j_i(t ) w(t ) j_i^{\dagger}(t ) \nonumber \\ & ~ & - \frac{1}{2 } \sum_i \big\ { j_i^{\dagger}(t)j_i(t),w(t ) \big\}.\end{aligned}\ ] ] the commutator with the hamiltonian @xmath14 represents the unitary part of the evolution and the lindblad operators @xmath15 describe the various decay channels of the system . in analogy to the terminology used for classical master equations , the expressions @xmath16 may be called gain terms , while the expressions @xmath17 , involving an anti - commutator , may be referred to as loss terms . both the hamiltonian @xmath14 and the operators @xmath15 are allowed to depend on time @xmath6 . the generator @xmath18 of the master equation may thus be explicitly time dependent and does not necessarily lead to a semigroup . we observe , however , that the superoperator @xmath19 is in lindblad form @xcite for each fixed @xmath20 . this means that @xmath21 is in the form of the generator of a quantum dynamical semigroup . the particular form of the generator derives from the requirements of complete positivity and of the conservation of the trace @xcite . under certain technical conditions which will be assumed to be satisfied here , one concludes that eq . ( [ lindblad - eq ] ) yields a 2-parameter family of completely positive and trace preserving maps @xmath22 @xcite . these maps can be defined with the help of the chronological time - ordering operator @xmath23 as @xmath24 , \qquad t \geq s \geq 0,\ ] ] and satisfy @xmath25 in terms of these maps the solution of the master equation ( [ lindblad - eq ] ) at time @xmath6 can be written as @xmath26 , where @xmath27 . thus , @xmath22 propagates the density matrix at time @xmath28 to the density matrix at time @xmath6 . each @xmath22 maps the space of density matrices into itself . this means that @xmath22 can be applied to any density matrix @xmath29 to yield another density matrix @xmath30 . the range of definition of the maps @xmath22 is thus the space of all density matrices and is independent of time . usually , one associates a markovian master equation with a time independent generator . we slightly generalize this notion and refer to eq . ( [ lindblad - eq ] ) as a markovian master equation with a time dependent lindblad generator @xmath18 . as mentioned already in the introduction the master equation ( [ lindblad - eq ] ) allows a stochastic unraveling through a random process in the state space @xmath12 . this means that one can construct a stochastic dynamics for the state vector @xmath31 in @xmath12 which reproduces the density matrix @xmath11 with the help of the expectation value @xmath32.\ ] ] to mathematically formulate this idea one writes a stochastic schrdinger equation ( sse ) for the state vector @xmath31 . an appropriate sse for which the expectation value ( [ expec ] ) leads to the master equation ( [ lindblad - eq ] ) is given by @xmath33 dn_i(t),\end{aligned}\ ] ] where we have introduced the nonlinear operator @xmath34 |\phi(t)\rangle.\ ] ] the term @xmath35 in eq . ( [ sse ] ) , which is proportional to the time increment @xmath36 , expresses the drift of the process . this drift contribution obviously corresponds to the nonlinear schrdinger - type equation @xmath37 whose linear part involves the non - hermitian hamiltonian @xmath38 the nonlinear part of the drift ensures that , although @xmath39 is non - hermitian , the norm is conserved under the deterministic time evolution given by eq . ( [ nse ] ) . the second term on the right - hand side of eq . ( [ sse ] ) represents a jump process leading to discontinuous changes of the wave function , known as quantum jumps . these jumps are described here with the help of the poisson increments @xmath40 which satisfy the relations : @xmath41 & = & ||j_i(t)|\phi(t)\rangle||^2dt . \label{def - dn2}\end{aligned}\ ] ] according to eq . ( [ def - dn1 ] ) the increments @xmath40 are random numbers which take the possible values @xmath42 or @xmath43 . moreover , if @xmath44 for a particular @xmath45 , we have @xmath46 for all @xmath47 . the state vector then performs the jump @xmath48 thus we see that @xmath49 is an integer - valued process which counts the number of jumps of type @xmath45 . we infer from eq . ( [ def - dn2 ] ) that @xmath50 occurs with probability @xmath51 . the jump described by ( [ jump ] ) thus takes place at a rate of @xmath52 . the case @xmath44 for all @xmath45 is realized with probability @xmath53 . in this case the state vector @xmath31 follows the deterministic drift described by eq . ( [ nse ] ) . summarizing , the dynamics described by eq . ( [ sse ] ) yields a piecewise deterministic process , i. e. , a random process whose realizations consist of deterministic evolution periods interrupted by discontinuous quantum jumps . since both the deterministic drift ( [ nse ] ) as well as the jumps ( [ jump ] ) do not change the norm , the whole process preserves the norm of the state vector . the formulation of the dynamics by means of a sse bears several numerical advantages over the integration of the corresponding density matrix equation ( [ lindblad - eq ] ) . this fact was the original motivation for the development of stochastic wave function methods in atomic physics and quantum optics ( for an example , see @xcite ) . what is important in our context is the fact that , additionally , the stochastic process given by the sse allows a physical interpretation in terms of a continuous measurement which is carried out on an environment of the system . to explain this point we consider a microscopic model in which the open system is weakly coupled to a number of independent reservoirs @xmath54 , one reservoir for each value of the index @xmath45 . each reservoir @xmath54 consists of bosonic modes @xmath55 which satisfy the commutation relations : @xmath56 = \delta_{ij}\delta_{\lambda\mu}.\ ] ] the lindblad operators @xmath15 appearing in the master equation ( [ lindblad - eq ] ) couple linearly to the reservoir operators @xmath57 where the @xmath58 are certain system frequencies , @xmath59 is the frequency of the mode @xmath55 of reservoir @xmath54 , and the @xmath60 are coupling constants . thus , the hamiltonian of our model is taken to be : @xmath61.\ ] ] we have included a factor @xmath62 , where @xmath63 is a typical relaxation rate of the system which will be introduced below . the combination @xmath64 is therefore dimensionless and the @xmath65 have the dimension of an inverse time , choosing units such that @xmath66 . the time evolution operator over the time interval @xmath67 will be denoted by @xmath68 . the correlation function of reservoir @xmath54 can be expressed through the spectral density @xmath69 , which is assumed to be the same for all reservoirs : @xmath70 here , @xmath71 denotes the vacuum state defined by @xmath72 for all @xmath45 and all @xmath73 . suppose that the state of the combined system ( open system plus reservoirs @xmath54 ) at some time @xmath6 is given by @xmath74 . at time @xmath75 this state has evolved into the entangled state @xmath76 . we consider @xmath77 to be a time increment which is small compared to the time scale of the systematic motion of the system , but large compared to the correlation time of the reservoirs . suppose further that at time @xmath78 a measurement of the quanta in the reservoir modes @xmath55 is carried out . according to the standard theory of quantum measurement @xcite the detection of a quantum in mode @xmath55 projects the state vector onto the state @xmath79 . the open system s state conditioned on this event thus becomes : @xmath80 where @xmath81 is the corresponding probability . if , on the other hand , no quantum is detected , one has to project the state vector onto the vacuum state which yields the open system s conditioned state @xmath82 where @xmath83 is the probability that no quantum is detected . in the born - markov approximation the above expression simplify considerably . we take a constant spectral density @xmath84 , corresponding to the case of broad band reservoirs with arbitrarily small correlation times . we further neglect the lamb shift contributions which lead to a renormalization of the system hamiltonian . the expression ( [ proj1 ] ) then becomes ( up to an irrelevant phase factor ) : @xmath85 this expression is seen to be independent of @xmath73 . therefore , the total probability of observing a quantum in reservoir @xmath54 is : @xmath86 the last two equation show that , conditioned on the detection of a quantum in reservoir @xmath54 , the system state carries out the jump described in ( [ jump ] ) , and that these jumps occur at a rate given by @xmath52 [ see eq . ( [ def - dn2 ] ) ] . for the case that no quantum is detected expression ( [ proj0 ] ) gives in the born - markov approximation : @xmath87 which leads to the drift contribution @xmath35 of eq . ( [ sse ] ) . the probability for this event is found to be @xmath88 considering that the detected quanta are annihilated on measurement ( quantum demolition measurement ) we see that for both alternatives described above the conditional state vector of the combined system after time @xmath77 is again a tensor product of an open system s state vector and the vacuum state of the environment . thus , we may repeat the measurement process after each time increment @xmath77 . in the limit of small @xmath77 we then get a continuous measurement of the environment and a resulting conditioned state vector of the open system that follows the sse ( [ sse ] ) . summarizing , the sse ( [ sse ] ) can be interpreted as resulting from a continuous measurement of the quanta in the environment . this measurement is an indirect measurement in which the jump ( [ jump ] ) of the state vector @xmath31 describes the measurement back action on the open system s state conditioned on the detection of a quantum in reservoir @xmath54 , while the nonlinear schrdinger equation ( [ nse ] ) yields the evolution of the state vector under the condition that no quantum is detected . the realizations of the process given by the sse , i.e. , the quantum trajectories thus allow a clear physical interpretation in accordance with the standard theory of quantum measurement . we investigate master equations for the density matrix @xmath2 of an open system which are of the following general form : @xmath89 \nonumber \\ & ~ & + \sum_{\alpha } \left [ c_{\alpha}(t ) \rho(t ) d_{\alpha}^{\dagger}(t ) + d_{\alpha}(t ) \rho(t ) c_{\alpha}^{\dagger}(t ) \right ] \nonumber \\ & ~ & -\frac{1}{2 } \sum_{\alpha } \left\ { d_{\alpha}^{\dagger}(t ) c_{\alpha}(t ) + c_{\alpha}^{\dagger}(t ) d_{\alpha}(t ) , \rho(t ) \right\}. \nonumber\end{aligned}\ ] ] the hamiltonian @xmath90 , the @xmath91 and the @xmath92 are given , possibly time dependent operators on the state space @xmath0 of the open system . the generator @xmath93 may thus again depend explicitly on time . the master equation ( [ gen - non - markov ] ) is however local in time since it does not contain a time integration over a memory kernel . the structure of @xmath93 was taken to ensure that the hermiticity and the trace of @xmath2 are conserved . if we choose @xmath94 the generator @xmath93 reduces to the form of a time dependent lindblad generator @xmath18 . the markovian master equation ( [ lindblad - eq ] ) is thus a special case of eq . ( [ gen - non - markov ] ) which will be referred to as non - markovian master equation . with an appropriate choice for the hamiltonian @xmath90 and the operators @xmath91 and @xmath92 , a large variety of physical phenomena can be described by master equations of the form ( [ gen - non - markov ] ) . for example , as mentioned in the introduction a master equation of this form arises when applying the tcl projection operator technique @xcite to the dynamics of open system . the basic idea underlying this technique is to remove the memory kernel from the equations of motion by the introduction of the backward propagator . under the condition of factorizing initial conditions one then finds a homogeneous master equation with a time - local generator @xmath93 . the latter can be determined explicitly through a systematic perturbation expansion in terms of ordered cumulants @xcite . specific examples are the tcl master equations describing spin relaxation @xcite , the spin - boson model @xcite , systems coupled to a spin bath @xcite , charged particles interacting with the electromagnetic field @xcite , and the atom laser @xcite . moreover , several exact time - local master equations of the form ( [ gen - non - markov ] ) are known in the literature which have been derived by other means . examples are provided by the master equations for non - markovian quantum brownian motion @xcite and for the nonperturbative decay of atomic systems , which will be discussed in sec . [ example ] . the existence of a homogeneous , time - local master equation requires , in general , that the initial state of the total system represents a tensor product state . for simplicity we restrict ourselves to this case since we intend to develop stochastic unravelings for pure states of the reduced open system . due to its explicit time dependence the generator @xmath93 of the master equation ( [ gen - non - markov ] ) does of course not lead to a semigroup . but even for a fixed @xmath95 the superoperator @xmath93 is , in general , not in lindblad form , by contrast to the property of the generator @xmath18 of the master equation ( [ lindblad - eq ] ) . to make this point more explicit we introduce operators @xmath96 and rewrite eq . ( [ gen - non - markov ] ) as : @xmath97 + { \mathcal{d}}_1(t)\rho(t ) + { \mathcal{d}}_2(t)\rho(t),\ ] ] where we have defined the superoperators @xmath98 , \\ { \mathcal{d}}_2(t)\rho & \equiv & -\sum_{\alpha } \left [ e_{\alpha}(t ) \rho e_{\alpha}^{\dagger}(t ) - \frac{1}{2}\left\{e_{\alpha}^{\dagger}(t)e_{\alpha}(t),\rho\right\ } \right].\end{aligned}\ ] ] one observes that @xmath99 is in lindblad form , while @xmath100 is not : the superoperator @xmath100 carries an overall minus sign and violates therefore the complete positivity of the generator . of course , we will assume in the following that the dynamics given by eq . ( [ gen - non - markov ] ) yields a dynamical map @xmath101 for all times considered , that is , we suppose that eq . ( [ gen - non - markov ] ) describes the evolution of true density matrices at time @xmath102 into true density matrices at time @xmath103 . however , this assumption does not imply that the propagation of an arbitrary positive matrix at time @xmath6 with the help of the master equation ( [ gen - non - markov ] ) necessarily leads to a positive matrix for future times . this is only guaranteed if we propagate a density matrix @xmath2 which results from the time evolution over the previous interval @xmath104 . a further important consequence of the form of the master equation ( [ gen - non - markov ] ) is that it does not allow a stochastic unraveling of the type developed in sec . [ cmt - markov ] . any unraveling of this kind would automatically produce a master equation with a time dependent lindblad generator . trying to construct an unraveling which leads to the contribution @xmath105 of the master equation , one would find a process with negative transitions rates , which is both unphysical and mathematically inconsistent . to overcome the difficulties in the development of a stochastic representation we are going to employ an interesting general feature of the non - markovian master equation ( [ gen - non - markov ] ) : even if the generator @xmath93 is not in lindblad form , it is always possible to construct an embedding of the non - markovian dynamics in a markovian evolution on a suitable extended state space . the precise formulation of this statement and its proof will be given in the following . the extended state space is obtained by combining the original open system on the state space @xmath0 with another auxiliary quantum system . the auxiliary system is a three state system whose state space @xmath10 is spanned by three basis states @xmath106 , @xmath107 and @xmath108 : @xmath109 the hilbert space @xmath12 of the combined system then becomes the triple hilbert space @xmath110 the extended state space is thus given by the tensor product of @xmath0 and @xmath10 , which in turn is isomorphic to the orthogonal sum of three copies @xmath111 , @xmath112 , @xmath113 of @xmath0 . this means that states @xmath114 in @xmath12 take the general form : @xmath115 where @xmath116 for @xmath117 . as a possible physical realization of the extended state space on may think of @xmath0 as the state space of a damped quantum particle with an additional internal degree of freedom which can be represented by a three level system we now regard eq . ( [ lindblad - eq ] ) as an equation of motion on the triple hilbert space , that is , @xmath11 is considered as a density matrix on @xmath12 governed by a master equation with time dependent lindblad generator @xmath18 . on the other hand , @xmath2 is a density matrix on @xmath0 satisfying the given non - markovian master equation ( [ gen - non - markov ] ) . we will assume in the following that the operators @xmath91 and @xmath92 are bounded . evolves into @xmath2 according to the given non - markovian master equation ( [ gen - non - markov ] ) . this evolution can be embedded in a markovian dynamics on the extended state space @xmath118 in which the density matrix @xmath119 evolves into @xmath11 following the master equation ( [ lindblad - eq]).[fig1 ] ] our aim is to show that by an appropriate choice of the hamiltonian @xmath14 and of the lindblad operators @xmath120 in eq . ( [ lindblad - eq ] ) one can always achieve that the density matrix @xmath2 on @xmath0 is connected to the density matrix @xmath11 on @xmath12 by means of the relation ( @xmath121 denotes the trace ) : @xmath122 here , we have defined @xmath123 which is an operator acting on @xmath0 . we can regard this operator @xmath124 as a matrix which is formed by the coherences ( off - diagonal elements ) of @xmath11 between states from the subspace @xmath111 and states from the subspace @xmath112 . this is the embedding theorem . it states that the non - markovian dynamics of @xmath2 can be expressed through the time evolution of a certain set of coherences @xmath124 of a density matrix @xmath11 on the extended space which follows a markovian dynamics ( see fig . [ fig1 ] ) . to proof the embedding theorem we first demonstrate that the relation ( [ def - rho ] ) can be achieved to hold at time @xmath102 . if @xmath125 is any initial density matrix on @xmath0 we define a corresponding density matrix on @xmath12 by @xmath126 \nonumber \\ & = & \rho(0 ) \otimes |\chi\rangle\langle\chi|,\end{aligned}\ ] ] where @xmath127\end{aligned}\ ] ] is a state vector of the auxiliary system . we thus obtain @xmath119 by combining the open system in the state @xmath125 with the three state system in the pure state @xmath128 . it is obvious that @xmath119 given by eq . ( [ def - w0 ] ) is a true density matrix on the triple hilbert space @xmath12 , i. e. , we have @xmath129 and @xmath130 . moreover , eq . ( [ def - w0 ] ) yields : @xmath131 which is eq . ( [ def - rho ] ) at time @xmath102 . to show that the relation ( [ def - rho ] ) is valid for all times @xmath4 we have to demonstrate that the right - hand side of this relation satisfies the non - markovian master equation ( [ gen - non - markov ] ) , provided the lindblad operators @xmath15 in eq . ( [ lindblad - eq ] ) are chosen appropriately . to simplify the presentation we first treat the case that the master equation ( [ gen - non - markov ] ) only involves a single operator @xmath132 and @xmath133 . thus we write , suppressing the time arguments , @xmath134 + c \rho d^{\dagger } + d \rho c^{\dagger } \nonumber \\ & ~ & -\frac{1}{2 } \left\ { d^{\dagger}c + c^{\dagger}d , \rho \right\}.\end{aligned}\ ] ] and the action of the operators @xmath135 defined in eqs . ( [ def - j1])-([def - j4]).[fig2 ] ] we define four time dependent lindblad operators for the master equation ( [ lindblad - eq ] ) : @xmath136 and the hamiltonian : @xmath137 \nonumber \\ & \equiv & h_s(t ) \otimes i_3,\end{aligned}\ ] ] where @xmath138 denotes the unit operator on the auxiliary space @xmath10 . in eqs . ( [ def - j3 ] ) and ( [ def - j4 ] ) we have introduced a time dependent operator @xmath139 on @xmath0 which will be defined below . according to the definitions ( [ def - j1 ] ) and ( [ def - j2 ] ) the operators @xmath140 and @xmath141 leave invariant the subspace @xmath142 which contains the states of the form @xmath143 . the operators @xmath144 and @xmath145 defined in eqs . ( [ def - j3 ] ) and ( [ def - j4 ] ) induce transitions @xmath146 and @xmath147 between the states of the auxiliary three state system . the extended state space and the action of the lindblad operators are illustrated in fig . [ fig2 ] . we have from the above definitions ( suppressing again the time arguments ) : @xmath148 and : @xmath149 with the lindblad operators defined in eqs . ( [ def - j1])-([def - j4 ] ) and with the hamiltonian given by eq . ( [ def - h ] ) the master equation ( [ lindblad - eq ] ) leads to : @xmath150 | 2 \rangle + \sum_{i=1}^4 \langle 1 | j_i w j_i^{\dagger } | 2 \rangle \nonumber \\ & & -\frac{1}{2 } \sum_{i=1}^4 \langle 1 | \big\ { j_i^{\dagger}j_i , w \big\ } | 2 \rangle.\end{aligned}\ ] ] on using eq . ( [ sumi - ji ] ) we find @xmath151 while eqs . ( [ def - j1])-([def - h ] ) yield @xmath152 | 2 \rangle & = & [ h_s , w_{12 } ] . \label{rel5}\end{aligned}\ ] ] employing the relations ( [ lossterm])-([rel5 ] ) in eq . ( [ eq - w-1 ] ) we arrive at : @xmath153 + c w_{12 } d^{\dagger } + d w_{12 } c^{\dagger } \nonumber \\ & & -\frac{1}{2 } \left\ { \omega^{\dagger}\omega + c^{\dagger}c + d^{\dagger}d , w_{12 } \right\}.\end{aligned}\ ] ] here we see the reason for our choice of the extended state space and of the lindblad operators @xmath135 . the operators @xmath140 and @xmath141 have been chosen in such a way that @xmath154 acts from the left and @xmath155 from the right on the coherences @xmath156 , or vice versa [ see eqs . ( [ rel1 ] ) and ( [ rel2 ] ) ] . on the other hand , since @xmath144 and @xmath145 induce transitions into the state @xmath108 of the auxiliary system , the corresponding gain terms @xmath157 and @xmath158 of the master equation ( [ lindblad - eq ] ) do not contribute towards the equation of motion ( [ eq - w-2 ] ) for the coherences @xmath156 [ see eq . ( [ rel34 ] ) ] . the subspace @xmath113 of the extended state space plays the role of a sink which will be used now to achieve that the loss terms of the master equation come out correctly . we observe that eq . ( [ eq - w-2 ] ) is already of a form which is similar to the desired master equation ( [ nmsa ] ) . these equations differ , however , with respect to the structure of the loss terms which are given by the terms containing the anti - commutator . to get an equation of motion of the desired form we now choose @xmath159 to be a solution of the equation @xmath160 which is equivalent to @xmath161 here , @xmath162 denotes the unit operator on @xmath0 and @xmath163 is a time dependent non - negative number . since @xmath164 is a positive operator a solution @xmath159 of eq . ( [ omega - def ] ) exists under the condition that the right - hand side of eq . ( [ omega - def ] ) is also a positive operator . thus , @xmath165 must be chosen in such a way that @xmath166 for all normalized state vectors @xmath167 in @xmath0 . we note that it is at this point that the assumption of bounded operators enters our construction . to make a definite choice we define @xmath165 to be the largest eigenvalue of the positive operator @xmath168 . this definition ensures that the inequality ( [ ineq ] ) is satisfied and that a solution @xmath159 of eq . ( [ omega - def ] ) exists . the solution of eq . ( [ omega - def ] ) is , in general , not unique . if @xmath159 is a solution , then also @xmath169 , where @xmath170 is an arbitrary unitary operator . changing @xmath159 into @xmath169 does , however , not influence the equation of motion ( [ eq - w-2 ] ) since only the combination @xmath164 enters this equation . in the language of quantum measurement theory the transformation @xmath171 is called a quantum operation @xcite . it describes the change of a density matrix @xmath172 under a generalized measurement whose outcome occurs with probability @xmath173 . we can thus say that the unitary operator @xmath170 , expressing our freedom in the choice of @xmath159 , affects the change of the system state @xmath172 , but not the probability of its occurrence . substituting eq . ( [ omega - eq ] ) into eq . ( [ eq - w-2 ] ) we get : @xmath174 + c w_{12 } d^{\dagger } + d w_{12 } c^{\dagger } \nonumber \\ & & -\frac{1}{2 } \left\ { d^{\dagger}c + c^{\dagger}d , w_{12 } \right\ } - a w_{12}.\end{aligned}\ ] ] we conclude from this equation that the trace of @xmath156 satisfies the equation @xmath175 using this fact as well as eq . ( [ eq - w-3 ] ) one immediately demonstrates that the expression on the right - hand side of eq . ( [ def - rho ] ) satisfies the desired master equation ( [ nmsa ] ) , which concludes the proof of the embedding theorem . the general case of an arbitrary number of operators @xmath91 and @xmath92 in eq . ( [ gen - non - markov ] ) can be treated in a similar way . to this end , one has to re - introduce the index @xmath176 and to carry out a summation over @xmath176 in the equations of motion . thus , for each value of @xmath176 we have a corresponding @xmath177 and an @xmath178 , as well as four @xmath179 , @xmath180 . finally we note that according to eq . ( [ eq - w-3 ] ) the operator @xmath181 satisfies the same differential equation as @xmath156 . since also @xmath182 , we conclude that @xmath183 for all times . it follows that we can write for any operator @xmath184 on @xmath0 : @xmath185 where @xmath186 is an operator on the auxiliary state space . this shows that the expectation value of all observables @xmath184 in the state @xmath2 can be determined through measurements on the state @xmath11 of the extended system . the embedding of the previous section enables us to construct a stochastic unraveling for the non - markovian dynamics given by the master equation ( [ gen - non - markov ] ) . since the master equation governing @xmath11 involves a time dependent lindblad generator we can use the sse developed in sec . [ cmt - markov ] for this purpose . the sse ( [ sse ] ) generates a stochastic process for the state vector @xmath31 in the triple hilbert space . employing the representation ( [ repr ] ) we write @xmath187 as shown in sec . [ cmt - markov ] the density matrix @xmath11 on the extended state space is reproduced through the expectation value @xmath188 $ ] , and the norm of the state vector is exactly conserved during the stochastic evolution : @xmath189 in accordance with eq . ( [ def - w0 ] ) the initial state of the process is taken to be of the form : @xmath190,\ ] ] where @xmath191 is a normalized random state vector in @xmath0 , @xmath192 , and @xmath193 is the fixed state vector of the auxiliary three state system defined in ( [ chi ] ) . we thus have @xmath194 and @xmath195 . hence , eq . ( [ def - rho ] ) at time @xmath102 gives @xmath196.\ ] ] the embedding theorem now reveals that @xmath2 is obtained from the stochastic evolution with the help of the relation [ see eq . ( [ def - rho ] ) ] : @xmath197 thus we have constructed a stochastic unraveling for the non - markovian dynamics . it is important to realize that our construction leads to an unraveling through a normalized stochastic state vector @xmath31 and that the process allows a definite physical interpretation in terms of a continuous measurement , as has been discussed in sec . [ cmt - markov ] . we remark that for the case @xmath198 the given master equation ( [ nmsa ] ) is already in time dependent lindblad form . our construction then yields @xmath199 [ see inequality ( [ ineq ] ) ] and @xmath200 , as well as @xmath201 . this means that the jump operators @xmath140 and @xmath141 are identical and that the decay channels @xmath144 and @xmath145 are closed . it follows that @xmath202 and that @xmath203 evolves in exactly the same way as @xmath204 , that is , we have @xmath205 . equation ( [ rho - stoch ] ) thus becomes @xmath206 . note that the factor @xmath207 is due to the normalization condition ( [ norm ] ) which yields @xmath208 . in the case @xmath198 our construction therefore reduces automatically to the standard unraveling of a master equation in lindblad form . as an example we discuss a model for the non - markovian decay of a two state system into the vacuum of a bosonic bath . the model serves to illustrate how to construct the embedding in a markovian dynamics , and how to interpret physically the quantum trajectories generated by the resulting sse . the interaction picture master equation of the model is given by @xmath209 \\ & ~ & + \gamma(t)\left ( \sigma_-\rho(t)\sigma_+ -\frac{1}{2}\left\ { \sigma_+\sigma_-,\rho(t ) \right\ } \right ) , \nonumber\end{aligned}\ ] ] where @xmath210 this is an exact master equation for the nonperturbative decay of a two state system with excited state @xmath211 and ground state @xmath212 , which interacts with a bosonic bath @xcite . @xmath213 and @xmath214 are the usual raising and lowering operators of the two state system . these operators couple linearly to the bath through an interaction hamiltonian of the form @xmath215 , where @xmath216 is a bath operator depending linearly on the annihilation operators of the bath modes . the real functions @xmath217 and @xmath218 are determined by the vacuum correlation function @xmath219 of the bath . an example will be discussed below [ see eq . ( [ corr - jc ] ) ] . the function @xmath217 describes a time dependent renormalization of the system hamiltonian induced by the coupling to the bath ( lamb shift ) . under the condition @xmath220 one can interpret the function @xmath218 as a time dependent decay rate of the excited state . but for certain spectral densities @xmath218 may become negative in certain time intervals such that the master equation ( [ jc - model ] ) is not in lindblad form and the generator is not completely positive . the master equation ( [ jc - model ] ) can however always be brought into the form of the non - markovian master equation ( [ nmsa ] ) by means of the definitions : @xmath221 to obtain the operator @xmath139 introduced in eq . ( [ omega - def ] ) we first have to determine the quantity @xmath163 which is defined to be the largest eigenvalue of @xmath168 . this operator is equal to @xmath42 for @xmath222 , and equal to @xmath223 for @xmath224 . thus we find @xmath225 and @xmath226 hence , eq . ( [ omega - def ] ) takes the form @xmath227 which leads to an obvious solution : @xmath228 the lindblad operators @xmath135 defined in eqs . ( [ def - j1])-([def - j4 ] ) are therefore given explicitly by : @xmath229 , \label{j1 } \\ j_2 & = & \sqrt{\frac{|\gamma|}{2 } } \sigma_- \otimes \big [ { \textrm{sign}}\gamma |1\rangle\langle 1| + |2\rangle\langle 2| \big ] , \label{j2 } \\ j_3 & = & \sqrt{a}\sigma_+ \otimes |3\rangle\langle 1| , \label{j3 } \\ j_4 & = & \sqrt{a}\sigma_+ \otimes |3\rangle\langle 2| . \label{j4}\end{aligned}\ ] ] considering times @xmath6 for which @xmath220 , we have @xmath230 [ see eqs . ( [ def - a ] ) ] and , hence , @xmath201 . this means that for @xmath220 the decay channels described by @xmath144 and @xmath145 are closed : the process only involves the jumps of the state vector given by the operators @xmath140 and @xmath141 . we infer from eqs . ( [ j1 ] ) and ( [ j2 ] ) that @xmath140 and @xmath141 induce downward transitions @xmath231 of the two state system ( action of @xmath232 ) . these transitions result from the projection of the system s state vector into the ground state @xmath212 conditioned on the detection of a quantum in reservoir @xmath233 or @xmath234 . for @xmath224 the operators @xmath140 and @xmath141 again induce downward transitions of the two state system and , at the same time , introduce a relative phase factor of @xmath235 between the states @xmath106 and @xmath107 of the auxiliary three state system . moreover , in the case @xmath224 the decay channels @xmath144 and @xmath145 are open : this enables the additional jumps of the state vector described by @xmath144 and @xmath145 . equations ( [ j3 ] ) and ( [ j4 ] ) show that @xmath144 and @xmath145 lead to upward transitions @xmath236 of the two state system ( action of @xmath237 ) with simultaneous transitions between the auxiliary states of the form @xmath146 or @xmath147 . a re - population of the excited state @xmath211 is thus possible through jumps into the auxiliary state @xmath108 , corresponding to the detection of a quantum in reservoir @xmath238 or @xmath239 . a detailed analysis of the process can be given in terms of the statistics of the quantum jumps . to this end , we note that the waiting time distribution for the sse ( [ sse ] ) is given by @xmath240 |\phi(t_0)\rangle \right|\right|^2,\ ] ] where @xmath241 is defined by eq . ( [ h - hat ] ) . @xmath242 is the probability that a jump takes place in the time interval @xmath243 , given that the previous jump occurred at time @xmath244 and yielded the state @xmath245 . note that this is a true cumulative probability distribution , i.e. @xmath242 increases monotonically with @xmath246 and satisfies @xmath247 . for our model we have [ see eqs . ( [ h - hat ] ) , ( [ def - h ] ) , ( [ sumi - ji ] ) and ( [ omega - eq ] ) ] : @xmath248 let us analyze the process starting from the initial state @xmath249 and investigate the occupation probability @xmath250 of the ground state . from the master equation it is clear that this quantity is given by the simple expression : @xmath251.\ ] ] if @xmath252 takes on positive and negative values , this is a non - monotonic function of time . our aim is to illustrate how the stochastic dynamics reproduces this behavior and to explain the physical picture provided by the unraveling . in the stochastic representation we have the formula [ see eq . ( [ rho - stoch ] ) ] : @xmath253 } { e[\langle\psi_2(t)|\psi_1(t)\rangle]}.\ ] ] we denote the moment of the first jump by @xmath246 , the moment of the second jump by @xmath254 . it follows from eq . ( [ wdf ] ) that the waiting time distribution for the first jump is given by @xmath255,\ ] ] and that the waiting time distribution for the second jump , given that the first jump took place at time @xmath246 , becomes @xmath256.\ ] ] we further denote the total number of jumps in the time interval @xmath104 by @xmath257 . since the process starts from the state ( [ init - jc ] ) the first jump is given by the application of @xmath140 or @xmath141 which project the state vector into the ground state . therefore , prior to the first jump we have latexmath:[$2\langle g we get latexmath:[$2\langle g during the time intervals in which @xmath224 a second jump described by @xmath144 or @xmath145 is possible by which the state vector leaves the manifold @xmath142 and lands in @xmath113 . once the state vector is in @xmath113 , no further jumps are possible . of course , we get again @xmath259 from these relations we find the expectation value @xmath260 } \nonumber \\ & = & \int_0^t dt_1 \dot{f}(t_1,0)[1-f(t , t_1 ) ] { \mathrm{sign}}\gamma(t_1 ) \\ & = & \exp\left[-\int_0^t dt_1 a(t_1)\right ] -\exp\left[-\int_0^t dt_1 |\gamma(t_1)| \right ] . \nonumber\end{aligned}\ ] ] here , the quantity @xmath261 is the probability that the first jump occurs in @xmath262 , while @xmath263 is the probability that no further jumps take place within @xmath264 . a similar analysis can be performed to obtain the expectation value of the quantity @xmath265 . one finds : @xmath266 thus we get : @xmath267 } \nonumber \\ & = & 1-f(t,0 ) + \int_0^t dt_1 \dot{f}(t_1,0)[1-f(t , t_1 ) ] { \mathrm{sign}}\gamma(t_1 ) \nonumber \\ & = & \exp\left[-\int_0^t dt_1 a(t_1)\right].\end{aligned}\ ] ] the term @xmath268 represents the no - jump probability , i. e. , the contribution from the event @xmath269 . the result ( [ y - t ] ) could have been obtained also directly from eq . ( [ trace ] ) . using finally eqs . ( [ x - t ] ) and ( [ y - t ] ) in eq . ( [ p - e ] ) we find , of course , the correct expression ( [ p - e - exact ] ) for the excited state probability . ) for the non - markovian decay of a two - state system . top : ground state probability @xmath270 obtained from a sample of @xmath271 quantum trajectories ( dots ) , and analytical solution ( continuous line ) . bottom : expectation value @xmath272 $ ] . parameters : @xmath273 , @xmath274 . [ fig3 ] ] we illustrate the above analysis by means of an example . figure [ fig3 ] shows the results of a monte carlo simulation of the sse ( [ sse ] ) for the case of a lorentzian spectral density which is detuned from the transition frequency of the two state system by an amount @xmath275 ( damped jaynes - cummings model ) . this leads to a bath correlation function of the form @xmath276 where @xmath277 is the markovian relaxation time and @xmath278 is the correlation time of the bath . the simulation was carried out in a nonperturbative regime : while the born - markov approximation requires that @xmath279 , the simulation uses @xmath280 . in this regime @xmath218 becomes negative for certain time intervals . these intervals can be seen in the figure as those intervals over which @xmath270 and @xmath272 $ ] decrease monotonically with time . this is a signature for the fact that transitions @xmath146 or @xmath147 become possible through the channels @xmath144 or @xmath145 . these channels are closed in the time intervals over which @xmath272 $ ] stays constant . the decrease of the ground state probability can be interpreted as due to virtual processes in which a quantum is emitted into the bath and re - absorbed at a later time . this is a clear non - markovian feature of the dynamics . in the stochastic unraveling this decrease results from the contributions of those quantum trajectories which involve at least one jump and for which the first jump at time @xmath246 occurred during a phase in which @xmath281 . the first jump then yields a negative contribution to the expectation values of @xmath282 and of @xmath283 as a result of the relative phase factor @xmath284 between the states @xmath106 and @xmath107 introduced by @xmath140 and @xmath141 [ see eqs . ( [ xsigma ] ) and ( [ ysigma1 ] ) ] . moreover , in a possible second jump the state vector leaves the manifold @xmath142 to end up in a state proportional to @xmath285 , which gives @xmath286 [ see eq . ( [ ysigma2 ] ) ] . the decrease of @xmath270 is therefore due to quantum trajectories for which a second jump is possible which leads to a re - excitation @xmath236 . thus we see that the virtual emission and re - absorption processes appear in the stochastic unraveling in the extended state space as certain real processes , namely as jumps with @xmath140 or @xmath141 involving a negative phase factor ( detection of a quantum in reservoir @xmath233 or @xmath234 ) , and as jumps into the auxiliary state @xmath108 ( detection of a quantum in reservoir @xmath238 or @xmath239 ) . this shows how the quantum memory effect of virtual emission and re - absorption processes is encoded and completely stored in the continuous measurement record . in this paper we have developed a general method for the derivation of stochastic unravelings for non - markovian quantum processes given by time - local master equations of the form ( [ gen - non - markov ] ) . the key point of the construction is the fact that such master equations always allow a markovian embedding in an extended state space with a rather simple structure , namely in the triple hilbert space @xmath1 . within this embedding the density matrix @xmath2 of the original open system is expressed through a certain set of coherences of the full density matrix @xmath11 on the extended state space . the transition to the extended state space can be viewed physically as the addition of a further degree of freedom which is realized by a three level system . this enables one to represent the given non - markovian dynamics by means of a suitable interaction with a markovian environment consisting of the reservoirs @xmath54 introduced in sec . [ cmt - markov ] . although the generator of the given non - markovian master equation needs not be in lindblad form , the corresponding dynamics in the extended state space is therefore governed by a time dependent lindblad generator of the form of eq . ( [ lindblad - eq ] ) . the lifting to an appropriate extended state space thus allows the derivation of stochastic schrdinger equations for non - markovian dynamics through a consistent application of the standard theory of quantum measurement . the sses obtained in this way generate genuine quantum trajectories with the physical interpretation of continuous measurements . the construction of sec . [ embedding ] provides a fairly general method for the markovian embedding of a given non - markovian dynamics : apart from the existence of the master equation and from the boundedness of the operators @xmath91 and @xmath92 , no assumption was made regarding the interaction hamiltonian , the spectral density , the reservoir state , its temperature , etc . the markovian embedding could therefore be useful in itself since it enables one to employ well established and developed concepts from the theory of completely positive maps and lindblad generators in the study of non - markovian master equations . formulating a non - markovian unraveling we made use of piecewise deterministic jump processes . in an electromagnetic environment this corresponds , for example , to direct photodetection . it should be clear , however , that our derivation allows any unraveling in the extended state space . alternatively one can use diffusion - type sses , which in a continuous measurement interpretation correspond to other detection schemes like homodyne or heterodyne photodetection @xcite . various stochastic unravelings for non - markovian dynamics have been suggested in the literature , involving both jump processes @xcite as well as sses with colored noise @xcite . the technique developed by imamoglu @xcite is related to the method of pseudo modes @xcite . it employs an approximate markovian embedding of a given non - markovian dynamics . this embedding is based on the assumption that the reservoir can be represented by means of an effective set of fictitious damped harmonic oscillator modes . the markovian embedding of the present paper is realized in an entirely different way by the introduction of the triple hilbert space , and avoids the expansion into pseudo modes . it should also be noted that in the present method the density matrix @xmath2 is _ not _ given by the partial trace of the density matrix @xmath11 in the extended state space . a further interesting method has been formulated by disi , gisin and strunz @xcite . these authors employ a nonlocal stochastic integro - differential equation for the state vector . as demonstrated by gambetta and wiseman @xcite it seems , however , that the nonlocal sse does not admit a continuous measurement interpretation within the framework of standard quantum measurement theory ( see also @xcite in this context ) . this means that measurements carried out at different times on the environment will influence the dynamics in a way which is incompatible with the stochastic process . the sse does therefore not generate genuine quantum trajectories in the sense it does for markovian dynamics . a number of unravelings of non - markovian dynamics has been proposed @xcite which are based on the idea of propagating a pair @xmath287 , @xmath288 of stochastic state vectors and of representing the reduced density matrix with the help of the expectation value @xmath289 $ ] . it is even possible to design an exact stochastic unraveling @xcite which neither requires the existence of a master equation nor a factorizing initial state . this method makes use of a pair of independently evolving product states in the state space of the total system . related stochastic wave function methods have also been formulated for the description of bosonic and fermionic many - body systems @xcite , and for the simulation of quantum gases @xcite by use of the positive p - representation @xcite . a measurement interpretation of these stochastic methods is however not available . a pair @xmath287 , @xmath288 of state vectors can be considered as an element of the double hilbert space @xmath290 which is the tensor product of @xmath0 and the state space of a two state system . in ref . @xcite a stochastic unraveling in the double hilbert space has been constructed . although this method has been demonstrated to provide a useful numerical tool , a continuous measurement interpretation seems again to be impossible . this is connected to the facts that not only the master equation in @xmath0 but also the master equation in the double hilbert space is generally not in lindblad form and that the process does not preserve the norm of the state vector . we mention finally some restrictions of the present theory . similar to the formulation of the lindblad theorem , we made use of the assumption of boundedness of the operators @xmath91 and @xmath92 in the non - markovian master equation ( [ gen - non - markov ] ) . this assumption excludes the immediate treatment of important cases , such as quantum brownian motion which involves the unbounded operators for position and momentum of the particle . however , what is really needed in the proof is that the inequality ( [ ineq ] ) is satisfied . provided the dynamics of the state vector is confined to an effective subspace of @xmath0 on which the right - hand side of inequality ( [ ineq ] ) is bounded , we can still define a finite @xmath291 and construct the embedding . a further restriction of the theory is that for certain models time - local master equation of the form ( [ gen - non - markov ] ) may not exist for very strong couplings . the latter can lead to singularities of the tcl generator and to a breakdown of the tcl expansion ( an example is discussed in @xcite ) . it is an important open problem whether a continuous measurement unraveling can be developed for such cases . 99 h. p. breuer and f. petruccione , _ the theory of open quantum systems _ ( oxford university press , oxford , 2002 ) . g. lindblad , commun . phys . * 48 * , 119 ( 1976 ) . v. gorini , a. kossakowski , and e. c. g. sudarshan , j. math * 17 * , 821 ( 1976 ) . v. gorini , a. frigerio , m. verri , a. kossakowski , and e. c. g. sudarshan , rep . * 13 * , 149 ( 1978 ) . h. carmichael , _ an open systems approach to quantum optics _ m18 of _ lecture notes in physics _ ( springer - verlag , berlin , 1993 ) . j. dalibard , y. castin , and k. mlmer , phys . 68 * , 580 ( 1992 ) . r. dum , p. zoller , and h. ritsch , phys . a * 45 * , 4879 ( 1992 ) . m. b. plenio and p. l. knight , rev . phys . * 70 * , 101 ( 1998 ) . n. gisin and i. c. percival , j. phys . a : math . gen . * 25 * , 5677 ( 1992 ) . h. m. wiseman and g. j. milburn , phys . a * 47 * , 642 ( 1993 ) . h. m. wiseman and g. j. milburn , phys . a * 47 * , 1652 ( 1993 ) . s. nakajima , progr . . phys . * 20 * , 948 ( 1958 ) . r. zwanzig , j. chem . phys . * 33 * , 1338 ( 1960 ) . s. chaturvedi and f. shibata , z. phys . b * 35 * , 297 ( 1979 ) . f. shibata and t. arimitsu , j. phys . . jap . * 49 * , 891 ( 1980 ) . a. royer , phys . rev . a * 6 * , 1741 ( 1972 ) . a. royer , phys . lett . a * 315 * , 335 ( 2003 ) . r. p. feynman and f. l. vernon , ann . ( n.y . ) * 24 * , 118 ( 1963 ) . a. o. caldeira and a. j. leggett , physica a * 121 * , 587 ( 1983 ) . f. haake and r. reibold , phys . a * 32 * , 2462 ( 1985 ) . b. l. hu , j. p. paz and y. zhang , phys . d * 45 * , 2843 ( 1992 ) . r. karrlein and h. grabert , phys . e * 55 * , 153 ( 1997 ) . e. b. davies and h. spohn , j. stat . * 19 * , 511 ( 1978 ) . r. alicki , j. phys . a : math * 12 * , l103 ( 1979 ) . y. castin and k. mlmer , phys . 74 * , 3772 ( 1995 ) . v. b. braginsky and f. ya . khalili , _ quantum measurement _ ( cambridge university press , cambridge , 1992 ) . chang and j. l. skinner , physica a * 193 * , 483 ( 1993 ) . l. d. blanga and m. a. despsito , physica a * 227 * , 248 ( 1996 ) . h. p. breuer , b. kappler , and f. petruccione , ann . ( n.y . ) * 291 * , 36 ( 2001 ) . h. p. breuer , eur . j. d ( 2004 ) , in press ( quant - ph/0309114 ) . h. p. breuer , d. burgarth , and f. petruccione , quant - ph/0401051 . h. p. breuer , d. faller , b. kappler , and f. petruccione , europhys . * 54 * , 14 ( 2001 ) . h. p. breuer and f. petruccione , phys . a * 63 * , 032102 ( 2001 ) . k. kraus , _ states , effects , and operations _ 190 of _ lecture notes in physics _ ( springer - verlag , berlin , 1983 ) . a. imamoglu , phys . a * 50 * , 3650 ( 1994 ) . m. w. jack and m. j. collett , phys . a * 61 * , 062106 ( 2000 ) . l. disi , n. gisin , and w. t. strunz , phys . a * 58 * , 1699 ( 1998 ) . w. t. strunz , l. disi , and n. gisin , phys . lett . * 82 * , 1801 ( 1999 ) . j. gambetta and h. m. wiseman , phys . rev . a * 66 * , 012108 ( 2002 ) . j. gambetta and h. m. wiseman , phys . a * 66 * , 052105 ( 2002 ) . b. m. garraway , phys . a * 55 * , 2290 ( 1997 ) . b. m. garraway , phys . a * 55 * , 4636 ( 1997 ) . j. gambetta and h. m. wiseman , phys . a * 68 * , 062104 ( 2003 ) . h. p. breuer , b. kappler , and f. petruccione , phys . a * 59 * , 1633 ( 1999 ) . j. t. stockburger and h. grabert , phys . 88 * , 170407 ( 2002 ) . h. p. breuer , phys . a * 69 * , 022115 ( 2004 ) . i. carusotto , y. castin , and j. dalibard , phys . a * 63 * , 023606 ( 2001 ) . o. juillet and ph . chomaz , phys . lett . * 88 * , 142503 ( 2002 ) . m. j. steel , m. k. olsen , l. i. plimak , p. d. drummond , s. m. tan , m. j. collett , d. f. walls , and r. graham , phys . a * 58 * , 4824 ( 1998 ) . p. d. drummond and j. f. corney , phys . a * 60 * , r2661 ( 1999 ) . d. drummond and c. w. gardiner , j. phys . a : math . * 13 * , 2353 ( 1980 ) . a. gilchrist , c. w. gardiner , and p. d. drummond , phys . rev . a * 55 * , 3014 ( 1997 ) .

we review the main results we have obtained in the area of high - energy elastic hadron scattering and presented in this series of _ workshops on hadron interactions_. after an introduction to some basic experimental and theoretical concepts , we survey the results reached by means of four approaches : analytic models , model - independent analyses , eikonal models and nonperturbative qcd . some of the ongoing researches and future perspectives are also outlined . * contents * ii.a physical quantities ii.b principles , theorems and high - energy bounds ii.c basic pictures ii.c.1 optical / geometrical picture ii.c.2 exchange picture iii.a derivative dispersion relations iii.b basic models iii.b.1 ensembles iii.b.2 analytic models iii.b.3 fits and results iii.c non - degenerate meson trajectories iii.c.1 extrema bounds for the pomeron intercept iii.c.2 idr , ddr and the subtraction constant iv.a differential cross sections iv.a.1 unconstrained fits and the eikonal iv.a.2 constrained fits and energy dependence iv.b total cross sections and slopes v.a geometrical model - inelastic channel v.b qcd - inspired models v.b.1 basic formalism v.b.2 dynamical gluon mass v.b.3 momentum scale vi.a stochastic vacuum model vi.b elementary amplitudes vi.b.1 correlators from lattice qcd vi.b.2 correlators from instanton approach vii.a experiments vii.b theory

hep-ph0411400

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` qcd nowadays has a split personality . it embodies hard and soft physics , both being hard subjects and the softer the harder . '' _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ yuri dokshitzer ( 2001 ) @xcite despite the great success of qcd as the field theory of hadronic interactions , there still remains some open questions and one of them is related to the hadron - hadron scattering at _ high - energies _ and _ small momentum transfer _ ( soft diffraction ) . the region of high energies is characterized by scattering of particles with center of mass energy @xmath0 gev @xmath1 @xmath2 ( the proton mass ) . from the experimental point of view , diffractive processes are associated with a slow increase of the total cross sections , the diffraction pattern in the differential cross section , and rapidity gaps in the plots of pseudo rapidity _ versus _ azimuthal angle . in the theoretical context , diffraction means that the initial and final states in the scattering process have the same quantum numbers and , therefore , the exchanged object " has the vacuum quantum numbers ( pomeron ) . the soft diffractive processes are generally classified as double diffraction dissociation , single diffraction dissociation and elastic scattering . introductory reviews on the area can be found in refs . @xcite . high - energy elastic hadron scattering is the simplest soft diffractive process and , at the same time a topical problem in high - energy physics . being associated with long distance phenomena perturbative qcd can not be applied . on the other hand , the standard non - perturbative approach starts with the ground state ( vacuum ) , proceeds with bound states ( mesons , barions ) and eventually reaches the scattering states . however , it is obvious that the vacuum is a non - trivial problem . moreover , even assuming some vacuum concept , to treat only one gluon field it is necessary to take into account more than 30 invariants , and all that becomes a typical problem of statistical physics , with specific technical approaches , such as monte carlo simulation ( lattice qcd ) . although bound states may be described , the point is that , presently , we do not know how to calculate elastic scattering amplitudes from a pure nonperturbative qcd formalism . at this stage , phenomenology certainly plays an important role in the search for connections between experimental data , model descriptions , and the possible development of new calculational schemes in the underlying theory ( qcd ) . here , however , we are faced with another kind of problem , namely , the wide variety of model descriptions , based on different ideas and approaches , not always giving enough support for the development of novel calculational schemes well founded on qcd . based on the above facts , our main strategy in the investigation of the elastic sector is to search for _ model independent information _ that may be extracted from the experimental data , through approaches that have well established bases on the general principles , theorems and bounds from axiomatic quantum field theory ( the _ analytic approach _ ) . simultaneously , we attempt to construct phenomenological models , in agreement with the above principles and connected , in some way , with the underlying dynamics of qcd . in this review , it is presented some results we have obtained in the area of elastic scattering in the last years , with focus on high - energy proton - proton ( @xmath3 ) and antiproton - proton ( @xmath4 ) elastic scattering . the manuscript is organized as follows . in sec . ii we recall some basic experimental and theoretical concepts , defining also our notation . in secs . iii , iv , v , and vi we present the main results we have obtained throughout the analytic approach , model independent analyses , eikonal models and nonperturbative qcd , respectively . in sec . vii we discuss some perspectives in the area , from both experimental and theoretical points of view . a summary and some final remarks are the contents of sec . in this section we recall the physical quantities that characterize the elastic scattering and shortly review some principles , high - energy theorems , and the main formulas associated with two basic pictures , usually referred as s - channel ( geometrical / optical picture ) and t - channel ( exchange picture ) @xcite . in elastic scattering , the connection between experimental data and theory is done by means of the _ invariant scattering amplitude _ , expressed in terms of two mandelstam variables , generally the center - of - mass ( c.m . ) energy squared @xmath5 and the four - momentum transfer squared @xmath6 : @xmath7 . it is expected that spin effects decrease as the energy increases ( for some recent results see @xcite ) , and neglecting spin , the physical quantities that characterize the elastic scattering process are the differential cross section , @xmath8 where @xmath9 is the c.m . momentum , the elastic integrated cross section , @xmath10 the total cross section ( optical theorem ) , @xmath11 the inelastic cross section @xmath12 the @xmath13 parameter , @xmath14 and the slope parameter , @xmath15_{t=0}.\end{aligned}\ ] ] the corresponding experimental data have been analyzed and compiled by the particle data group and can be found in ref . @xcite and quoted references . in what follows we shall be mainly interested in @xmath3 and @xmath4 data in the regions : @xmath16 gev @xmath17 tev and @xmath18 gev@xmath19 gev@xmath20 . in a particular analysis we shall also use the @xmath3 data at @xmath21 gev , in the region @xmath22 gev@xmath23 gev@xmath20 . some treatment of cosmic - ray information on @xmath3 total cross sections at @xmath24 6 - 40 tev is also presented . in figure 1 it is displayed the experimental information available on @xmath3 and @xmath4 total cross sections from accelerators and cosmic - ray experiments . from that plot , it is clear that the mathematical description of the increase of the total cross sections at the highest energies is an open problem . as we shall discuss , the study of the effects of the discrepant points at the highest energies is one of our goals . figure 2 shows the typical diffractive pattern that characterizes the differential cross section . we note that the data cover the region corresponding to 10 decades . in figure 3 it is displayed the slope parameter from @xmath3 and @xmath4 scattering as function of the energy and determined in the region of small momentum transfer . in what follows we shall refer to these three figures as indicative of the empirical behavior of the quantities involved . and @xmath4 from accelerator and cosmic - ray experiments ( for complete list of references and tables see @xcite).,width=340,height=340 ] elastic scattering at @xmath24 52.8 gev.,width=302,height=264 ] 0.20 gev@xmath25.,height=226 ] for our purposes , we recall some principles and theorems from axiomatic quantum field theories @xcite . the basic principles are : lorentz invariance , unitarity ( related with the conservation of probability ) , analyticity ( related to causality ) and crossing ( connecting particle - particle and particle - antiparticle interactions ) . analyticity and crossing allow the connections between real and imaginary parts of the scattering amplitude by means of dispersion relations . several rigorous theorems and bounds may be deduced from the basic principles and axiomatic quantum field theory . among them , the froissart - martin bound concerns the increase of the total cross section stating that @xmath26 the pomeranchuk theorem treats the difference between cross sections for particle - particle ( @xmath27 ) and particle - antiparticle scattering ( @xmath28 ) . the original form was deduced when it was believed that the cross section decreased to a constant value , and in this case @xmath29 as @xmath30 . after the discovery of the rising of the cross section , grunberg and truong obtained the generalized or revised form of the pomeranchuk theorem , stating that @xmath31 and this means that , if the froissart - martin bound is reached , then @xmath32 by expressing the cross sections in terms of crossing even ( @xmath33 ) and odd ( @xmath34 ) contributions , @xmath35 we have @xmath36 . therefore , @xmath37 if and only if @xmath38 . this possible odd contribution is named odderon and the case of even dominance at asymptotic energies is associated with the pomeron . nearly all the phenomenological models , able to describe the experimental data on elastic hadron scattering , are based on the optical / geometrical picture ( s - channel ) and/or the exchange picture ( t - channel ) . the corresponding formulas may be obtained from the partial waves representation of the scattering amplitude , @xmath39 p_l(\cos \theta ) , \nonumber\end{aligned}\ ] ] where @xmath40 is the phase shift . in what follows , we outline the main steps and formulas in both pictures . from the partial wave representation , one considers the high - energy limit and the semi - classical approximation , so that the discrete angular momentum @xmath41 may be replaced by the continuum impact parameter @xmath42 , @xmath43 in turn , the discrete phase shifts @xmath40 are replaced by the continuum eikonal function of @xmath42 and @xmath5 , @xmath44 and @xmath45 the scattering amplitude in this _ eikonal representation _ , with azimuthal symmetry assumed , reads @xmath46.\end{aligned}\ ] ] the quantity @xmath47 is named profile function . from unitarity this function is related to the probability that an inelastic event takes place at @xmath42 and @xmath5 , the inelastic overlap function : @xmath48 since in the eikonal representation @xmath49 for @xmath50 we have @xmath51 , which implies in an automatically unitarized representation . in this picture , from the partial wave representation , one considers the analytic continuation of the amplitude to complex angular momentum . in the asymptotic limit ( @xmath52 ) and with symmetry connecting the crossed channels one arrives at the watson - sommerfeld - gribov - regge representation for the scattering amplitude , expressed as a sum over the poles of the amplitude ( the regge poles ) , as outlined in what follows . as it is known at high energies the number of partial waves is large , and one way to circumvent that is to transform the sum of partial waves into a complex integral , and then use the residues theorem to obtain a new sum , but involving only the number of residues : @xmath53 detailed calculation allows one to obtain the following representation for the scattering amplitude , @xmath54 where @xmath55 is called the background integral . by considering the high - energy limit ( then @xmath56 ) and crossing ( exchange four - momenta @xmath57 ) we can replace the crossing channel variable ( @xmath58 ) @xmath59 also , @xmath60 x^l \quad { \rm for } \quad x \rightarrow \pm \infty , \nonumber\end{aligned}\ ] ] and grouping all the @xmath5-independent quantities in a function @xmath61 we have @xmath62 rearranging the terms we arrive at a descending asymptotic series in powers of @xmath5 , with leading contribution : @xmath63 where @xmath64 is the residue function , @xmath65 the signature factor and @xmath66 the trajectory function . this last function connects the spin and masses through the chew - frautschi plot , as exemplified in fig . 4 . in this picture the interaction of the colliding particles is basically interpreted in terms of exchanges of regge poles ( also regge cuts ) and the pomeron ( an _ ad hoc _ trajectory with intercept nearly above 1 ) . we note that , as constructed , the exchange picture is intended for asymptotic energies . the analytic approach for elastic hadron - hadron scattering is based on general principles and theorems from quantum field theory . it is characterized by analytical parametrizations for the imaginary part of the forward amplitude , together with the use of dispersion relation techniques . the central point is the simultaneous investigation of the total cross section ( imaginary part of the scattering amplitude , eq . ( 2 ) ) and the @xmath13 parameter ( connected with the real part of the amplitude , eq . ( 3 ) ) . for particle - particle and particle - antiparticle interactions , dispersion relations are consequences of the principles of analyticity and crossing . in this context , they correlate real and imaginary parts of crossing even ( @xmath33 ) and odd ( @xmath34 ) amplitudes , which in turn are expressed in terms of the scattering amplitudes for a given process and its crossed channel , for example , @xmath67 and @xmath68 : @xmath69 at high energies , the standard singly subtracted integral dispersion relations , with poles removed , are given by @xmath70 and @xmath71 where @xmath72 is the subtraction constant and , for @xmath3 and @xmath4 scattering , @xmath73 gev@xmath25 . in this section we review some results obtained through this approach . we start with the replacement of the above integral forms by derivative operators ( derivative dispersion relations ) and then we discuss the use of analytic models ( reggeons , pomeron , odderon ) for parametrizations involving the total cross section and the @xmath13 parameter , the determination of bounds for the soft pomeron intercept , and the practical role of the subtraction constant . in what follows we are mainly concerned with the @xmath3 and @xmath4 elastic scattering , since for particle and antiparticle interactions they correspond to the highest energy interval with available data and are the only set including the cosmic - ray information on total cross sections ( @xmath3 scattering ) . as commented before , the experimental data available on the total cross sections ( figure 1 ) are characterized by discrepant experimental information at the highest energies , and one of our aims is to investigate the effects of these discrepancies in the context of the analytic models . this concern permeates all the discussion in this section . the use of dispersion relations in the investigation of scattering amplitudes may be traced back to the end of fifties , when they were introduced in the form of _ integral _ dispersion relations ( idr ) . despite the important results that have been obtained since then , one limitation of the integral forms is their non - local character : in order to obtain the real part of the amplitude , the imaginary part must be known for all values of the energy . moreover , the class of functions that allows analytical integration is limited . in the last years , we have investigated the applicability of _ derivative _ dispersion relations ( ddr ) in place of integral forms @xcite . in reference @xcite we present a recent review on different results and statements related to this replacement , and a discussion connecting these different aspects with the corresponding assumptions and classes of functions considered in each case . in particular , we have shown that for the class of functions which are entire in the logarithm of the energy ( as is the case of analytic models at high energies ) it is possible to expand the integrand in the above formulas and by considering a high - energy approximation , represented by @xmath74 , to integrate term by term . in that case , as demonstrated in detail in @xcite , the derivative dispersion relations with one subtraction reads @xmath75\frac{{\mathop{\mathrm{im}}}f_+(s)}{s},\ ] ] @xmath76 \frac{{\mathop{\mathrm{im}}}f_-(s)}{s } , \label{eq:11}\ ] ] where the series expansion is implicit in the tangent operator . from this deduction one arrives to three formal results : ( 1 ) the subtraction constant is preserved when the idr are replaced by ddr and , therefore , in principle , can not be disregarded in fit procedures ; ( 2 ) except for the subtraction constant , the ddr with entire functions in the logarithm of the energy do not depend on any additional free parameter ; ( 3 ) the only approximation involved in the replacement concerns the lower limit in the idr ( 13 - 14 ) , namely @xmath77 , which represents a high - energy approximation . in the next two subsections we discuss some uses of the ddr with analytical models , and in the third subsection we return to the replacement of idr by ddr , investigating the important role of the subtraction constant from a practical point of view . in this subsection we make use of two basic and well known parametrizations for the total cross sections and investigate the effects of the discrepancies in the experimental information from cosmic - ray experiments . in the cosmic - ray region , @xmath78 tev , the discrepancies on the total cross section information are due to both experimental and theoretical uncertainties in the determination of @xmath79 from p - air cross sections . the situation has been recently reviewed in detail in @xcite , where a complete list of references , numerical tables and discussions are presented . from fig . 1 we see that , despite the large error bars in the cosmic - ray region , we can identify two distinct sets of estimations : one corresponding to the results by the fly s eye collaboration ( fly s eye ) together with those by the akeno collaboration ( akeno ) ; the other set associated with the results by gaisser , sukhatme , and yodh ( gsy ) together with with those by nikolaev ( nikolaev ) . taken separately these two sets suggest different scenarios for the increase of the total cross section , as previously discussed in @xcite . based on these considerations , it is important to investigate the behavior of the total cross section by taking into account the discrepancies that characterize the cosmic ray information . to this end , in @xcite we have considered _ two ensembles of data and experimental information _ , as follows : @xmath80 ensemble i : @xmath4 and @xmath3 accelerator data + akeno + fly s eye ; @xmath80 ensemble ii : @xmath4 and @xmath3 accelerator data + nikolaev + gsy . to some extent , ensemble i represents a kind of high - energy standard picture and ensemble ii a nonstandard one . with analytical parametrizations for @xmath3/@xmath4 total cross sections , the connections with the @xmath13 parameter , eq . ( 3 ) , are obtained by defining the associated crossing even and odd quantities , @xmath81 using the high - energy normalization for the optical theorem , @xmath82 and the ddr given by eqs . ( 15 ) and ( 16 ) . in @xcite we have considered two different parametrizations for the total cross sections , one introduced by donnachie and landshoff @xcite and other by kang and nicolescu @xcite . the main difference concerns the asymptotic limits , which allow the dominance of an even amplitude ( pomeron ) or the odd amplitude ( odderon ) , respectively . in this way , we may contrast these possibilities with the standard and non - standard pictures represented by ensembles i and ii . the _ donnachie - landshoff _ ( dl ) parametrization for the total cross sections is expressed by @xmath83 where the first contribution is associated with a single pomeron exchange ( universal ) and the second one with reggeon exchange . with the procedure explained above , we obtain the analytical connections with the @xmath13 parameter for @xmath3 and @xmath4 scattering : @xmath84s^{\epsilon } \nonumber \\ & + & \left[\frac{(y - z)}{2}\cot\left(\frac{\pi\eta}{2}\right ) - \frac{(y+z)}{2}\tan\left(\frac{\pi\eta}{2}\right)\right]s^{- \eta } , \nonumber\end{aligned}\ ] ] @xmath85s^{\epsilon } \nonumber \\ & + & \left[\frac{(z - y)}{2}\cot\left(\frac{\pi\eta}{2}\right ) - \frac{(y+z)}{2}\tan\left(\frac{\pi\eta}{2}\right)\right]s^{- \eta}. \nonumber\end{aligned}\ ] ] from the above formulas , since @xmath86 , this model predicts that , asymptotically ( @xmath52 ) , @xmath87 the parametrization for the total cross sections introduced by _ kang and nicolescu _ ( kn ) , under the hypothesis of the odderon , is given by @xmath88 @xmath89 and the connections with @xmath13 read @xmath90 @xmath91 differently from the previous case , this model predicts that the difference between the two cross sections is given by @xmath92 so that , if @xmath93 and/or @xmath94 , the total cross section difference may increase and @xmath79 may even become greater than @xmath95 , depending on the values and signs of @xmath96 and @xmath97 , which is formally in agreement with the theorems of sec . moreover , if @xmath96 and @xmath97 are sufficiently small , so that we may replace @xmath98 , then , asymptotically , @xmath99 this means that , depending on the fit results , there may be a change of sign in @xmath100 , with @xmath101 becoming greater than @xmath102 at some finite energy . therefore , the case of a crossing either in @xmath103 or @xmath13 is a sign of the odderon contribution in the imaginary or real part of the amplitude , respectively . we have performed 16 different fits through the program cern - minuit . in these fits we have used both ensembles i and ii and both the dl and kn models . for each of these four possibilities we have performed global and individual fits to @xmath103 and @xmath13 and , in each case , we either considered the subtraction constant @xmath72 as a free fit parameter , or assumed @xmath104 . all the results are presented and discussed in detail in ref . our main conclusions are the following : ( 1 ) despite the small influence from different cosmic - ray estimations , the results allow to extract an upper bound for the soft pomeron intercept : @xmath105 ; ( 2 ) although global fits present good statistical results , in general , this procedure constraints the rise of @xmath103 ; ( 3 ) the subtraction constant as a free parameter affects the fit results at both low and high energies ; ( 4 ) independently of the cosmic - ray information used and the subtraction constant , global fits with the odderon parametrization predict that , above @xmath106 gev , @xmath107 becomes greater than @xmath108 , and this result is in complete agreement with all the data presently available . that result is displayed in fig . 5 and we can infer @xmath109 at @xmath110 gev and @xmath111 at @xmath112 gev ( bnl rhic energies ) . and @xmath113 through the kn parametrization with @xmath114 and ensembles i ( dotted curves for @xmath3 and dashed for @xmath4 ) and ii ( solid curves for @xmath3 and dot - dashed for @xmath4 ) @xcite.,width=302,height=264 ] and @xmath113 through the kn parametrization with @xmath114 and ensembles i ( dotted curves for @xmath3 and dashed for @xmath4 ) and ii ( solid curves for @xmath3 and dot - dashed for @xmath4 ) @xcite.,width=302,height=264 ] the dl parametrization referred above assumes degeneracies between the secondary reggeons , imposing a common intercept for the @xmath115 ( @xmath116 ) and the @xmath117 ( @xmath118 ) trajectories ( see fig . more recently , analysis treating global fits to @xmath103 and @xmath13 have indicated that the best results are obtained with non - degenerate meson trajectories . in this case the forward scattering amplitude is decomposed into three reggeon exchanges , @xmath119 where the first term represents the exchange of a single soft pomeron , the other two the secondary reggeons and @xmath120 ( @xmath121 ) for @xmath3 ( @xmath4 ) amplitudes . using the notation @xmath122 , @xmath123 and @xmath124 for the intercepts of the pomeron and the @xmath115 and @xmath117 trajectories , respectively , the total cross sections , eq . ( 18 ) , for @xmath3 and @xmath4 interactions are written as @xmath125 and the connection with the @xmath13 parameter by means of ddr is similar to that displayed in the last subsection . making use of this parametrization , in this section we present the determination of extrema bounds for the pomeron intercept @xcite and a practical analysis on the replacement of idr by ddr together with a discussion on the role of the subtraction constant @xcite . in order to analyze the extrema effects in the soft pomeron intercept due to discrepancies in the experimental data , we performed a detailed analysis including the highest and the lowest values of the total cross section from both accelerators and cosmic - ray experiments . as it is well known , in the accelerator region , the conflict concerns the results for @xmath95 at @xmath126 tev reported by the cdf collaboration and those reported by the e710 and the e811 collaborations ( fig . 1 ) . in the cosmic - ray region , as we have discussed , the highest predictions for @xmath79 concern the result by gaisser , sukhatme , and yodh together with those by nikolaev . in order to treat the lowest estimations in the cosmic - ray region , we consider the results obtained by block , halzen , and stanev ( bhs ) , by means of a qcd - inspired model . as discussed in @xcite , the reason for this choice is that , although the extracted @xmath127 shows agreement with the akeno results , it is about @xmath128 mb below the fly s eye value at @xmath129 tev and therefore may be considered as a extreme lower estimate . all the numerical tables and references can be found in @xcite . in this case we have considered the following ensembles of experimental information . first we only consider accelerator data in two ensembles with the following notation : @xmath80 ensemble i : @xmath130 and @xmath131 data ( @xmath132 gev ) + cdf datum ( @xmath133 tev ) ; @xmath80 ensemble ii : @xmath130 and @xmath131 data ( @xmath132 gev ) + e710/e811 data ( @xmath133 tev ) . ensemble i represents the faster increase scenario for the rise of @xmath103 from accelerator data and ensemble ii the slowest one . these ensembles are then combined with the highest and lowest estimations for @xmath79 from cosmic - ray experiments , namely , the nikolaev and the gaisser , sukhatme , and yodh ( ngsy ) results and the block , halzen , and stanev ( bhs ) results , respectively . these new ensembles are denoted by @xmath80 i + ngsy @xmath80 ii + bhs as in the previous analysis , we have considered both individual fits to @xmath103 , and simultaneous fits to @xmath103 and @xmath13 , either in the case where the subtraction constant is considered as a free fit parameter or assuming @xmath104 . from this analysis , in the case of only accelerator data , we could infer the following upper and lower values for the pomeron intercept : @xmath134 ( global fits to ensemble i , with @xmath104 ) and @xmath135 ( individual fit to @xmath103 from ensemble ii ) , with bounds @xmath136 and @xmath137 , respectively . adding the cosmic - ray information , we inferred the following upper and lower values : @xmath138 ( individual fit to @xmath103 from ensemble i + ngsy ) and @xmath139 ( global fits to ensemble ii + bhs and @xmath72 as a free fit parameter or individual fit to @xmath103 from this ensemble ) , with bounds @xmath140 and @xmath141 , respectively . therefore we may infer the following _ extrema _ bounds for the soft pomeron intercept : @xmath142 figure 6 shows the total cross sections with parametrization ( 20 ) and the above extrema bounds , together with the experimental information available . extensions of these _ extrema bounds _ for the pomeron intercept to meson - p , gamma - p and gamma - gamma scattering have been discussed in @xcite . by means of global fits to total cross section data it is shown that these bounds are in agreement with the bulk of experimental data presently available , and extrapolations to higher energies indicate different behaviors for the rise of the total cross sections . we have also obtained new _ constrained bounds _ for the pomeron intercept from spectroscopy data ( chew - frautschi plots ) and have extended the analysis to baryon - p , meson - p , baryon - n , meson - n , gamma - p and gamma - gamma scattering @xcite . it is also presented tests on factorization and quark counting rules with both extrema and constrained bounds ( asymptotic energy region ) . in particular , at 14 tev ( cern lhc ) the extrema and constrained bounds allow to infer @xmath143 mb and @xmath144 mb , respectively . @xcite . as commented before , we have shown in ref . @xcite that for entire functions in the logarithm of the energy the only approximation involved in the replacement of idr by ddr concerns the lower limit @xmath145 in the idr : the high - energy condition is reached by assuming that @xmath77 in eqs . ( 13 - 14 ) . in that paper we have investigated the practical applicability of the ddr and idr in the context of the pomeron - reggeon parametrizations , with both degenerate and non - degenerate higher meson trajectories . by means of global fits to @xmath146 and @xmath113 data from @xmath3 and @xmath4 scattering , we have tested all the 16 important variants that could affect the fit results , namely the number of secondary reggeons , energy cutoff ( 5 and 10 gev ) , effects of the high - energy approximation connected with the subtraction constant and the analytic approach using both ddr and idr with fixed @xmath145 . our results led to the conclusion that the high - energy approximation and the subtraction constant affect the fit results at both low and high energies . this effect is a consequence of the fit procedure , associated with the strong correlation among the free parameters . a striking novel result concerns the practical role of the subtraction constant . we have shown that , with the pomeron - reggeon parametrizations , once the subtraction constant is used as a free fit parameter , the results obtained with the ddr and with the idr ( with finite lower limit , @xmath147 ) are the same up to 3 significant figures in the fit parameters and @xmath148 . this conclusion , as we have shown , is independent of the number of secondary reggeons ( dl or extended parametrization ) or the energy cutoff ( @xmath149 = 5 or 10 gev ) . in table i we display the fit results with the extended parametrization and cutoff at 10 gev . .simultaneous fits to @xmath103 and @xmath13 through the extended parametrization , @xmath150 10 gev ( 154 data points ) , with @xmath72 as a free parameter and using idr with lower limit @xmath151 and ddr @xcite . [ cols="^,^,^",options="header " , ] this kind of analysis is characterized by model independent parametrizations of the experimental data involved and the extraction of empirical or semi - empirical information that can contribute with the development of phenomenological models and the underlying theory . in this section we review some results we have obtained in the investigation of @xmath3 and @xmath4 differential cross section data ( unconstrained and constrained fits , as will be explained ) and the correlations between the experimental data on total cross section and the slope parameter ( eq . ( 4 ) ) . several authors have investigated elastic hadron scattering by means of parametrizations for the scattering amplitude and fits to the differential cross section data , eq . the extraction of the profile , eikonal and inelastic overlap functions in the @xmath42-space ( impact parameter ) and , in some special cases , the eikonal in the @xmath152-space , has led to important and novel results related with geometrical aspects ( radius , central opacity ) , differences between charge distributions and hadronic matter distributions , existence or not of eikonal zeros in the @xmath152-space and , more recently , connections with pomerons , reggeons and nonperturbative qcd aspects . in ref . @xcite we present a review and a critical discussion on the main results concerning this kind of analysis and also a wide list of references to outstanding works . the basic input in all these analyses is the parametrization of the scattering amplitude as a sum of exponentials in @xmath152 ( as empirically suggested by the diffractive pattern shown in fig . 2 ) and fits to the differential cross section data . this parametrization allows analytical expressions for the fourier transform of the amplitude , providing also analytical expressions for the quantities of interest in the @xmath42-space . in the next two subsections we review the results we have obtained by means of unconstrained fits ( fit parameters completely free , without extracted dependences on the energy ) @xcite , and discuss some research in course related to constrained fits ( including dependences on the energy which are based on empirical information ) @xcite . in the high energy region , @xmath15310 gev , differential cross section data are available at @xmath24 13.8 , 19.5 , 23.5 , 30.7 , 44.7 , 52.8 and 62.5 gev for @xmath3 scattering and at @xmath24 13.8 , 19.4 , 31 , 53 , 62 , 546 and 1800 gev for @xmath4 scattering . data from @xmath3 scattering also exists at @xmath24 27.5 gev and 5.5 @xmath154 14.2 gev@xmath25 ( but not on @xmath103 and @xmath13 ) , and as we shall show , that set plays a fundamental role in our analyse . see @xcite for a complete list of references . as discussed in @xcite two main problems are typical of model independent analysis of the differential cross sections : \(1 ) experimental data are available only over finite regions of the momentum transfer ( which in general are small , @xmath155 7 gev@xmath25 ) and the fourier transform demands integration from @xmath152 = 0 to infinity . this means that any fit is biased by extrapolations and although some extrapolated curves may look unphysical , they can not be excluded on mathematical grounds . \(2 ) the exponential parametrization allows analytical determination of the quantities in the @xmath42-space ( profile , inelastic , eikonal functions ) and also the statistical uncertainties , by means of error propagation from the fit parameters . however , in this case , the translation of the eikonal from b - space to the @xmath152-space can not be analytically performed and neither the error propagation ( through standard procedures ) . as a consequence , the unavoidable uncertainties from the fit extrapolations can not , in principle , be taken into account . in what follows we review a model independent approach able to minimize the above two problems . - in order to treat problem ( 1 ) we have used the following procedure @xcite . since it is known that for large @xmath156 the experimental data do not depend on the energy at @xmath16 gev @xmath157 gev and that there exist data at @xmath21 gev in the region @xmath22 gev@xmath23 gev@xmath20 , we have selected two ensembles of @xmath3 and @xmath4 differential cross section data : @xmath80 ensemble i : experimental data at each energy ; @xmath80 ensemble ii : ensemble i + data at @xmath21 gev . for the scattering amplitude we have introduced the following model independent analytical parametrization for both real and imaginary parts : @xmath158 @xmath159 with the experimental @xmath13 value at each energy the fits to the differential cross section data have been performed through the cern - minuit routine and the validity or not of ensemble ii is checked by means of the minuit output and standard statistical interpretation of the fit results ( dof , confidence levels ) . for @xmath3 scattering we have found that the data at @xmath24 13.8 gev are not compatible with ensemble ii . in the case of @xmath4 scattering none of the data sets are compatible with ensemble ii . therefore , in what follows , ensemble ii ( data at @xmath24 27.5 gev added ) corresponds only to @xmath3 scattering at 6 energies : 19.5 , 23.5 , 30.7 , 44.7 , 52.8 and 62.5 gev . from the error matrix ( variances and covariances ) , @xmath148 and confidence intervals , we infer the best values for the parameters and corresponding errors @xmath160 , @xmath161 . by means of standard error propagation , the uncertainties in the free parameters , @xmath162 , @xmath163 , ( @xmath164 = 1 , 2 , ... ) have been propagated to the scattering amplitude , and then to the differential cross section , providing @xmath165 by adding and subtracting the corresponding uncertainties we may estimate the confidence region associated with all the extrapolations , which can not be excluded on statistical grounds . a typical result with ensembles i and ii is illustrated in fig . 7 , for @xmath3 scattering at @xmath149= 23.5 gev . we see that , as expected , the effect of adding the experimental data at @xmath149= 27.5 gev ( when statistically justified ) is to reduce drastically the uncertainty region . that result will be fundamental in the extraction of the empirical information on the eikonal , as shown in what follows . differential cross section data , eq . ( 23 ) , at @xmath166 gev with ensembles i ( below ) and ii ( above ) @xcite.,width=302,height=264 ] - by means of the fourier transform , eqs.(7 - 8 ) , the parametrization ( 21 - 22 ) provides analytical expressions for the real and imaginary parts of the profile function , @xmath167 and @xmath168 , and also the associated uncertainties . from the fit results , together with error propagation , we have found that @xmath169^{2 } } \ll1 , \nonumber\end{aligned}\ ] ] and therefore , the imaginary part of the eikonal may be approximated by @xmath170 and the uncertainty @xmath171 determined directly from @xmath172 through propagation . the next step is to go to the momentum transfer space and concerns problem ( 2 ) : the fourier transform can not be performed analytically and therefore also the error propagation . for this reason we used a semi - analytical method as follows . expanding the above equation , we express the remainder of the series as @xmath173-\gamma_{r}(s , b)\end{aligned}\ ] ] and then fit the numerical points ( minuit ) by a sum of gaussians in the impact parameter space : @xmath174 a typical result is displayed in fig . 8 . by means of eq . ( 26 ) @xcite.,width=302,height=264 ] with this , the errors @xmath175 and @xmath176 may be propagated determining @xmath177 and then @xmath178 . in order to check the results and approximations , we performed also numerical integration through the nag routine . one of the main results extracted from this analysis is the statistical evidence of eikonal zeros in the momentum transfer space , first presented in @xcite . in order to investigate the position of the zeros and , mainly , to determine the uncertainties in its values , we consider the expected behavior of @xmath179 at large @xmath152 , namely @xmath180 . in fig . ( 9 ) we show a typical plot of the quantity @xmath181 as function of @xmath152 . the shaded areas correspond to the uncertainties obtained from error propagation . this example shows clearly the role and the effect of data at large values of the momentum transfer . in fact , within the uncertainties , ensemble i does not allow to infer a zero , but with ensemble ii , we find statistical evidence for the change of sign . ) for @xmath3 at @xmath182 gev with ensemble i ( above ) and ii ( below ) @xcite.,width=226,height=188 ] ) for @xmath3 at @xmath182 gev with ensemble i ( above ) and ii ( below ) @xcite.,width=226,height=188 ] from plots like that we can determine the positions of the zeros and the associated errors from the extrems of the uncertainty region ( in general not symmetrical ) . the position of the zero can also be obtained from the numerical method , but without uncertainties . in figure 10 it is shown the position of the zeros as function of the energy determined by means of both the semi - analytical ( with uncertainties ) and numerical ( without uncertainties ) methods . despite the systematic difference on the values with these methods , we may conclude that the position of the zero decreases as the energy increases . roughly , @xmath183 gev@xmath25 as @xmath184 gev . as reviewed in @xcite , there has been previous indication of eikonal zeros in the momentum transfer space , but without associated uncertainties . our first statistical evidence , published in 1997 , indicated the position of the zero at @xmath185 gev@xmath25 @xcite . in 2000 , experiments performed at the jefferson laboratory , on electron - proton scattering , have indicated an unexpected decrease of the ratio between the electric and magnetic proton form factors as the momentum transfer increases from 0.5 to 5.6 gev@xmath25 . moreover , extrapolations from empirical fits indicate a change of sign ( zero ) in this ratio , just at @xmath186 7.7 gev@xmath25 . since for @xmath3/@xmath4 scattering the eikonal is connected with the hadronic matter form factor ( see sec . ( 38 ) ) , the above results on the position of the zeros suggest novel and important insights on possible correlations between hadronic and electromagnetic interactions . we discuss that subject in @xcite , calling attention to the possibility of _ hadronic form factors depending on the energy_. we have also obtained the value of the imaginary part of the eikonal at zero momentum transfer , that is , the central opacity . the results are displayed in fig . 11 . as discussed in @xcite , one naive way to test these results is with the glauber model for the scattering involving hadrons @xmath187 and @xmath188 , and the optical theorem at the elementary level . in that case , the elementary cross section may be expressed by @xmath189 where @xmath190 and @xmath191 are the number of constituents in hadrons @xmath187 and @xmath188 , respectively . if we take @xmath192 we obtain @xmath193 mb at the isr region , a result in agreement with other estimations . and @xmath4 scattering @xcite.,width=302,height=264 ] in ref . @xcite we discuss the applicability of our results in the phenomenological context , outlining some connections with nonperturbative qcd and presenting a critical review on the main results concerning model - independent " analyses . despite the results obtained with the parametrization discussed in the last subsection , due to the fit procedure , we do not have the dependence on the energy of the free parameters @xmath194 , @xmath195 . presently , we are investigating that subject and we review here some preliminary results @xcite . the energy dependence has been introduced according to some empirical information : the increase of the total cross section and of the slope parameter with @xmath196 and @xmath197 , respectively ( see figs 1 and 3 ) . let us consider the standard exponential parametrization for the imaginary part of the amplitude , normalized as @xmath198.\end{aligned}\ ] ] at @xmath199 , from the optical theorem , eq . ( 18 ) , we expect a dependence of the parameters @xmath194 with @xmath196 , and the slope represented by the parameters @xmath195 with @xmath197 . these are the central choices in our approach . in order to treat @xmath3 and @xmath4 scatterings , in agreement with analyticity and crossing , we introduce crossing even and odd amplitudes and make use of the derivative dispersion relations , eqs . ( 15 ) and ( 16 ) , to connect real and imaginary parts of the amplitudes involved . specifically , for the imaginary part of the scattering amplitude we consider the parametrizations @xmath200,\end{aligned}\ ] ] @xmath201,\end{aligned}\ ] ] and , based on the above arguments , we introduce the following general dependences on the energy @xmath202 for @xmath3 scattering and @xmath203 for @xmath4 scattering , where @xmath204 . defining the crossing even ( + ) and odd ( - ) amplitudes , @xmath205 @xmath206 the corresponding real parts can be determined by means of the leading terms of the ddr , eqs . ( 15 - 16 ) , and so the corresponding real parts of the @xmath3 and @xmath4 amplitudes . with these analytic amplitudes we obtain the differential cross section : latexmath:[\ ] ] for @xmath323 or @xmath324 we have @xmath325 $ ] , where @xmath326 denotes a n - dimensional fourier transform . with the above formalism , once one has inputs for the correlation functions @xmath327 and @xmath328 , the elementary amplitude in the momentum transfer space , eq . ( 43 ) , may , in principle , be evaluated through eqs . ( 44 - 47 ) . it is important to stress that , as constructed , the formalism is intended for small momentum transfer ( @xmath329 gev@xmath25 ) and asymptotic energies @xmath52 . despite of these limitations , the investigation of soft high energy scattering at the energies presently available has led to satisfactory results @xcite . in this section we review the results we have obtained from inputs for the above correlators from lattice qcd @xcite . we also comment the research in course in the semi - classical context of instantons @xcite . numerical determinations of the above correlation functions , in limited interval of physical distances , exist from lattice qcd in both quenched approximation ( absence of fermions ) and full qcd ( dynamical fermions included ) @xcite . with the procedure described above ( see @xcite for all the calculational details ) , the elementary scattering amplitude in the momentum transfer space can be determined in numerical form . in order to obtain analytical expressions , suitable for investigating distinct contributions and also for phenomenological uses , we have parametrized these numerical points through a sum of exponentials in @xmath152 : @xmath330 the results are displayed in fig . ( 16 ) from both quenched approximation and full qcd , together with the corresponding exponential components . our main conclusions are the following @xcite : ( 1 ) the amplitudes decrease smoothly as the momentum transfer increases and they do not present zeros ; ( 2 ) the decreasing is faster when going from quenched approximation to full qcd ( with decreasing quark masses ) , and this effect is associated with the increase of the correlation lengths ; ( 3 ) the dynamical fermions generate two contributions in the region of small momentum transfer , which are of the same order at @xmath331 1 gev@xmath25 ( only one contribution is present in the case of quenched approximation ) . we understand that result ( 3 ) may suggest some kind of change in the dynamics at the elementary level , near @xmath331 1 gev@xmath25 and at asymptotic energies . if that is true , some signal could be expected at the hadronic level . one possibility is that this effect can be associated with the position of the dip ( or the beginning of the shoulder " ) in the hadronic ( elastic ) differential cross section data . the asymptotic condition embodied in our result indicates that @xmath331 1 gev@xmath25 seems to be in agreement with the limit of the shrinkage of the diffraction peak , empirically verified when the energy increases in the region 23 gev @xmath332 1.8 tev . by means of the stochastic vacuum formalism we also presently investigate the elementary amplitudes using correlators determined in the context of the _ constraint _ instanton approach , developed by a. dorokhov and collaborators @xcite . the basic picture is that of an instanton field dominating at small distances and decreasing exponentially at large distances in the physical vacuum . in ref . @xcite , we make use of suitable parametrizations for the correlators and investigate the effects of the contributions from the short and long range correlations in the determination of the full correlator . denoting those contributions as @xmath333 and @xmath334 , respectively , we introduce a dimensionless parameter @xmath335 in terms of the _ driven parameter _ @xmath336 and the _ size parameter _ @xmath337 @xcite . since @xmath336 is correlated with the relative contribution of each kind of correlator , we consider two extreme cases : 1 ) equal contributions ( weights 0.5 and 0.5 ) , corresponding to @xmath338 ; 2 ) almost pure instanton contribution ( weights 0.99 and 0.01 ) , corresponding to @xmath339 . in the lack of information on the long range component , and for our purpose , we consider parametrization in a gaussian form @xcite @xmath340 for the short range case , @xmath339 , we introduce the parametrization , @xmath341 the full correlator is then determined by @xmath342 for the long range case , @xmath338 , the parametrization takes the form @xmath343 and for the full correlator we have @xmath344 with the eqs . ( 49 ) and ( 50 ) we can calculate the elementary amplitude through the steps indicated in sec . the results are displayed in fig . ( 17 ) . a central result is that if the contribution from the long range correlator is small , that is , an almost purely instanton case , the corresponding elementary amplitude presents a minimum . in terms of the associated differential cross section , this implies a diffractive pattern in the momentum transfer space , a result already indicated in some phenomenological approaches @xcite . in the case of equal weights the amplitude decreases monotonically with the momentum transfer . we conclude that in the context of the instanton approach , the balance between the contributions of the short and long range correlators is a crucial point for the determination of the behavior of the elementary amplitudes . further investigation along this line can bring new important insights on the connection between instanton correlators and the physical quantities which characterize the high - energy hadronic scattering . in this section we outline some perspectives in the area of high - energy elastic hadron scattering from both experimental and theoretical point of views . certainly , some ideas may be biased by our own knowledge and our own personal view . from the experimental point of view the perspectives are very optimistic due to the new generation of experiments with both accelerators and cosmic - ray observatories . let us quickly summarize some projects in development . the upgrade of the fermilab tevatron machine , together with upgrades and new devices in the cdf ( collider detector facility ) and d0 detectors , are going to provide improved investigations on @xmath4 collisions at @xmath345 tev . although the main purpose of the experiment concerns hard diffraction , it will be possible to investigate elastic scattering in both high and low @xmath152 regions , the slope parameter , total cross section and single diffraction . of topical importance for soft physics , the new determination of the total cross section shall possibly bring a solution for the puzzle represented by the discrepant results around @xmath346 tev ( e710/e811 and cdf ) . at the relativistic heavy ion collider ( rhic ) in the brookhaven national laboratory , @xmath3 collisions are presently being investigated at energies never reached before : @xmath347 gev . the experiment `` total and differential cross sections and polarization effects in @xmath3 elastic scattering at rhic '' ( pp2pp ) plans to investigate both elastic scattering and diffraction dissociation ( single and double ) , in addition to spin effects . this will provide the first opportunity for direct comparison between @xmath3 and @xmath4 scattering at the highest collider energies . although in a bit longer term , at the cern large hadron collider ( lhc ) , the totem experiment ( total cross section , elastic scattering and diffraction dissociation at the lhc ) is specifically planned to study soft diffractive physics in @xmath3 collisions at @xmath348 tev . in particular , diffraction dissociation , total cross section and elastic scattering at large values of the momentum transfer will be investigated , up to @xmath349 gev@xmath25 . that will certainly allow discrimination and selection of various models and approaches , giving fundamental information at large momentum transfer ; for example , showing the existence or not of structures . moreover , this experiment will probably provide a final answer on the possible differences between @xmath3 and @xmath4 total cross sections and the correct power @xmath240 in the @xmath350 dependence of @xmath146 . the most energetic event detected in cosmic - ray experiments had @xmath351 ev , corresponding to an energy of @xmath352 joules ! goals of the auger project are the measurement of arrival direction and the energy and mass composition of cosmic rays above @xmath353 ev . for @xmath3 collisions this means @xmath354 above @xmath355 tev , nearly @xmath356 times the lhc energy . in addition to the astrophysical importance of the experiment , the measurement of the longitudinal development of showers will provide severe tests on hadronic interaction models . as a consequence , among others , the puzzles concerning the extraction of @xmath3 cross section from @xmath357 production cross section may receive better insights , allowing more precise determinations and at energies possibly never to be reached by accelerator machines . elastic hadron scattering ( and soft diffractive processes in general ) is a long distance phenomena and therefore we expect and look for a theoretical treatment via non - perturbative qcd . despite all the difficulties mentioned along this manuscript , we understand that two approaches deserve special attention . one of them concerns the approach by nachtmann and the stochastic vacuum model ( sec . vi.a ) @xcite . although under restrictive kinematic conditions ( momentum transfer of the order or below 1 gev@xmath25 , and asymptotic energies , @xmath358 ) the formalism has provided interesting results in the investigation of the physical quantities that characterize the elastic @xmath3 and @xmath4 scattering @xcite , in special the works by ferreira and pereira , connecting experimental observables and qcd parameters @xcite . attempts to implement the energy dependence in pure qcd grounds may be an important task for the near future . the other approach is associated with evidences for finite gluon propagator and running coupling in the infrared region and that is the case in some classes of solutions of the nonperturbative schwinger - dyson equations . in particular , in the solution proposed by cornwall @xcite the gluon acquires a dynamical mass leading to a freezing of the coupling constant in the infrared region . as referred before , natale and collaborators @xcite have discussed several phenomenological tests , reaching interesting results which have permitted the development of the formalism and the selection of adequate basic inputs . that opens a new way to investigate long distance phenomena with a finite calculational approach . we also understand that the connections between soft and semi - hard processes , typical of qcd - based models in the eikonal context , may bring new insights for the development of adequate calculational schemes in the nonperturbative treatment of high - energy elastic collisions . despite its simplicity , elastic hadron scattering constitutes a topical problem in high - energy physics . although the bulk of experimental data can be efficiently described in different phenomenological contexts , we are still facing the lack of a treatment and of a reasonable understanding of these processes based exclusively on qcd . our main strategy in investigating elastic scattering has been to look for descriptions based on the high - energy principles and theorems from axiomatic quantum field theory and , simultaneously , attempting to extract empirical " information from all the experimental data available . tests of discrepant data and their influence on the extracted information play a central role . in that way we hope to get feedbacks for theoretical development in nonperturbative and semi - hard qcd . we can summarize our main recent results as follows . in the context of the _ analytic approach _ , we have investigated the effects of discrepant experimental information on the total cross sections in both accelerator and cosmic - ray energy regions . by means of analytical fits , we have obtained extrema bounds for the soft pomeron intercept , namely @xmath359 and @xmath360 . we have also obtained novel constrained bounds for the intercept from spectroscopy data ( fitted regge trajectories from chew - fautschi plots ) and extended the analysis to several reactions . that information on the pomeron bounds may be important for phenomenological developments and projects for new experiments . we have also shown that the presence of the odderon in the real part of the elastic hadronic amplitude is not forbidden by the bulk of experimental data on @xmath103 and @xmath13 . in particular , the fit with the kang - nicolescu parametrization has indicated a crossing in @xmath113 , with @xmath361 becoming greater than @xmath362 above @xmath149 = 70 gev . that parametrization predicts @xmath363 ( rhic regions ) . detailed investigation on the applicability of ddr have shown that , once the subtraction constant is used as a free fit parameter , the ddr is equivalent to the idr with finite lower limit ( @xmath147 ) . that result was obtained for the class of entire functions in the logarithm of the energy ( typical of analytic models ) . in the context of _ model independent analyses _ , we have investigated the correlations between the experimental data on total cross section and the slope parameter . the parametrization introduced is based on the empirical behavior of these quantities and extrapolations to cosmic - ray energies may be useful in the determination of proton - proton total cross sections from proton - air cross sections . by means of a novel model independent fit procedure to the differential cross section data , we have found statistical evidence for eikonal zero in the momentum transfer space and that the position of the zeros decreases as the energy increases . the zero position shows agreement with the result recently obtained for the electromagnetic form factor and inferred from elastic electromagnetic @xmath364 scattering ( polarization transfer experiments ) . since our analysis concerns only hadronic interactions the results may bring new insights in the investigation of electromagnetic and hadronic form factors . we are also treating analytic fits with free parameters depending on the energy so as to develop a model independent predictive approach . in the context of _ eikonal models _ we have obtained connections between the elastic and inelastic channel by means of the geometrical picture , and correlating @xmath3 and @xmath4 scattering with contact interactions simulated by @xmath365 distributions and multiplicities . with the class of qcd - based or qcd - inspired models we have developed novel gluonic contributions by means of two approaches . one is based in solutions of the schwinger - dyson equations characterized by the dynamical gluon mass . the other one is intended to take into account the @xmath298- scale in updated gluonic distribution functions . certainly , the two approaches are not independent , and we presently investigate their simultaneous implementation . in the context of the _ stochastic vacuum model _ , we have obtained novel results for the elementary scattering amplitudes , making use of correlators determined either from lattice qcd or from constrained instantons . in both cases the elementary amplitudes present no zeros ( change of sign in the momentum transfer space ) . in the context of eikonal models , in which the eikonal is expressed in terms of the elementary amplitudes and form factors , ( 38 ) , this result corroborates the interpretation of the zero in the hadronic form factor , and therefore , its dependence on the energy . i am thankfull to y. hama and f.s . navarra for the invitation to present this review . for fruitful comments and discussions along these workshops i am particularly grateful to e. ferreira , y. hama and t. kodama . i am deeply thankful to those who have developed research programs under my supervision and are co - authors of all the works reviewed here : r.f . vila , p.c . beggio , s.d . campos , p.a.s . carvalho , e.g.s . luna , a.f . martini , j. montanha , and d.s . i am also thankful to a. di giacomo , p. valin , a. dorokhov , a.a . natale and r.c . rigitano for valuable discussions . this work was supported by fapesp ( contract n. 00/04422 - 7 ) . e. predazzi , in _ international workshop on hadron physics 98 _ , edited by e. ferreira , f.f.s . cruz and s.s . avancini ( world scientific , singapore , 1999 ) p. 80 - 175 ; v. barone and e. predazzi , _ high - energy particle diffraction _ ( springer - verlag , berlim , 2002 ) . a.f . martini , m.j . menon , j. montanha , zeros in the electromagnetic and hadronic form factors " , in _ ix hadron physics and viii relativistic aspects of nuclear physics - a joint meeting on qcd and qgp _ , edited by m.e . bracco , m. chiapparini , e. farreira , t. kodama , aip conf . 739 ( aip , new york , 2004 ) pp . 578 - 580 . carvalho , a.f . martini , m.j . menon , a.e . motter , in _ international workshop on hadron physics 98 , florianpolis , 1998 _ , edited by e. ferreira , f.f . de souza cruz and s.s . avancini ( world scientific , singapore , 1999 ) p. 326 . vila , s.d . campos , m.j . menon , and j. montanha , analytical fits to hadron - hadron differential cross section data at the diffraction peak " , in _ ix hadron physics and viii relativistic aspects of nuclear physics - a joint meeting on qcd and qgp _ , edited by m.e . bracco , m. chiapparini , e. farreira , t. kodama , aip conf . 739 ( aip , new york , 2004 ) pp . 529 - 531 . beggio , m.j . menon , and p. valin , phys . d * 61 * , 034015 ( 2000 ) ; in _ international conference on elastic and diffractive scattering - blois viii _ , edited by v.a . petrov and a.v . prokudin ( world scientific , singapore , 2000 ) pp . 371 - 376 . m. b. gay ducati , f. halzen and a. a. natale , phys . rev . * d48 * , 2324 ( 1993 ) ; f. halzen , g. krein and a. a. natale , phys . rev . * d47 * , 295 ( 1993 ) ; a. c. aguilar , a. mihara and a. a. natale , phys . rev . * d65 * , 054011 ( 2002 ) , int . a19 * , 249 ( 2004 ) ; a. c. aguilar , p. s. r. da silva and a. a. natale , phys . . lett . * 90 * , 152001 ( 2003 ) ; e.g.s . luna , a.f . martini , m.j . menon , a. mihara , a.a . preliminary results on the influence of a gluon mass in hadronic scattering " in _ ix hadron physics and viii relativistic aspects of nuclear physics - a joint meeting on qcd and qgp _ , edited by m.e . bracco , m. chiapparini , e. farreira , t. kodama , aip conf . 739 ( aip , new york , 2004 ) pp . 572 - 574 . o. nachtmann , `` high - energy collisions and nonperturbative qcd '' , hep - ph/96093 ; o. nachtmann , in _ international workshop on hadron physics 2000 _ , edited by f.s . navarra , m.r . robilotta , g. krein ( world scientific , 2001 ) p. 152 - 216 ; h.g . dosch , in _ international workshop on hadron physics 96 _ , edited by e. ferreira et al . ( world scientific , 1997 ) pp . 169 - 219 . dosch , phys . b * 190 * , 177 ( 1987 ) ; h.g . dosch , yu . a. simonov , phys . b * 205 * , 339 ( 1988 ) ; yu . a. simonov , nucl . b * 307 * , 512 ( 1988 ) ; * 204 * , 67 ( 1989 ) ; a. krmer , h.g . dosch , phys . b * 252 * , 669 ( 1990 ) ; * 272 * , 114 ( 1991 ) . e. ferreira and f. pereira , phys . d * 55 * , 130 ( 1997 ) ; phys . d , 179 ( 1997 ) ; phys . b * 399 * , 177 ( 1997 ) ; phys . d , 014008 ( 1998 ) ; phys . d , 077507 ( 2000 ) ; e. ferreira , braz . * 30 * , 282 ( 2000 ) . a. di giacomo and h. panagopoulos , phys . lett . b 285 ( 1992 ) 133 ; a. di giacomo , e. meggiolaro , and h. panagopoulos , nucl . phys . b 483 ( 1997 ) 371 ; m. delia , a. di giacomo , e. meggiolaro , phys . lett . b 408 ( 1997 ) 315 .

we show how accretion rate governs the physical properties of a sample of unobscured broad - line , narrow - line , and lineless active galactic nuclei ( agns ) . we avoid the systematic errors plaguing previous studies of agn accretion rate by using accurate accretion luminosities ( ) from well - sampled multiwavelength seds from the cosmic evolution survey ( cosmos ) , and accurate black hole masses derived from virial scaling relations ( for broad - line agns ) or host - agn relations ( for narrow - line and lineless agns ) . in general , broad emission lines are present only at the highest accretion rates ( ) , and these rapidly accreting agns are observed as broad - line agns or possibly as obscured narrow - line agns . narrow - line and lineless agns at lower specific accretion rates ( ) are unobscured and yet lack a broad line region . the disappearance of the broad emission lines is caused by an expanding radiatively inefficient accretion flow ( riaf ) at the inner radius of the accretion disk . the presence of the riaf also drives narrow - line and lineless agns to 10 times higher ratios of radio to optical / uv emission than broad - line agns , since the unbound nature of the riaf means it is easier to form a radio outflow . the ir torus signature also tends to become weaker or disappear from agns , although there may be additional mid - ir synchrotron emission associated with the riaf . together these results suggest that specific accretion rate is an important physical `` axis '' of agn unification , described by a simple model .

1103.0276

supermassive black holes ( smbhs ) are now known to be ubiquitous in the centers of all massive galaxies @xcite . smbhs grow in an `` active '' phase of accretion , during which they are observed as active galactic nuclei ( agn ) . agn growth is intimately tied to galaxy evolution , as evident in the well - studied correlations between smbh mass ( @xmath3 ) and properties of the host galaxy bulge ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? the agn phase is also hypothesized to regulate star formation in its host galaxy , with the galaxy feeding the black hole in turn ( e.g. * ? ? ? * ; * ? ? ? all massive galaxies are thought to experience episodic agn behavior in their lifetime @xcite . agns are generally classified by differences in their optical spectra . type 1 or broad - line agns have broad ( @xmath4 km s@xmath5 ) emission lines superimposed on blue unobscured continua in the uv / optical ( e.g. , * ? ? ? * ) , and are the most luminous persistent sources in the sky . type 2 or narrow - line agns lack broad emission lines and have weaker continua ( frequently dominated by their host galaxies ) , but have strong narrow emission lines , especially from forbidden transitions . narrow emission lines associated with nuclear activity can be distinguished from lines caused by star formation by studying the line ratios @xcite . the line ratio diagnostics work because the `` harder '' emission of an agn is more efficient at ionizing the surrounding gas and dust than star formation , and thus agns have stronger lines from high - energy forbidden transitions ( e.g. , @xmath65007 and @xmath66583 ) relative to the lower - energy hydrogen transitions ( e.g. , @xmath64861 and @xmath66563 ) . the subclass of `` low - ionization nuclear emission region '' agns ( liners , * ? ? ? * ) have narrow emission lines that are probably excited by some combination of ionization from both star formation and an agn @xcite . deep x - ray surveys have additionally revealed `` optically dull '' agns @xcite , which have bright x - ray emission but none of the broad or narrow emission line signatures of agn accretion . while many optically dull agns can be explained as type 2 agns diluted by prominent host galaxies @xcite , at least @xmath71/3 are undiluted but intrinsically optically weaker than other agns @xcite . the inferred x - ray column density @xmath8 can also be used to classify agns , with type 2 ( narrow - line ) agns typically more x - ray absorbed than type 1 ( broad - line ) agns . however x - ray and optical classifications differ for @xmath720% of objects @xcite . historically , type 2 and optically dull agns have been described as obscured versions of type 1 agn , with the broad emission line region ( blr ) hidden behind a partially opaque `` torus '' of gas and dust , while the narrow emission lines lie outside the torus ( e.g. , * ? ? ? the best evidence for this scenario is the observation that some type 2 agns have a `` hidden '' blr revealed by spectropolarimetry @xcite . however , recent observations have revealed several serious limitations of a simple unified model based solely on geometric obscuration . even in very deep spectropolarimetric observations , many type 2 agns show no hidden blr @xcite . observations suggest a lower @xmath9 limit in accretion rate for broad - line agns @xcite , although they remain incomplete at low accretion rates and low masses @xcite . the x - ray spectra are unabsorbed ( @xmath10 @xmath11 ) for 30 - 40% of type 2 agns @xcite , as well as most local liners ( * ? ? ? * and references therein ) and distant optically dull agns @xcite . several well - studied liners additionally lack the narrow fe k@xmath12 emission signature of a dusty torus @xcite . many type 2 agns and most optically dull agns have mid - ir colors like normal galaxies @xcite , in contrast to hot mid - ir colors of type 1 agns described by torus models @xcite . toroidal obscuration is additionally ruled out for some strongly varying type 2 @xcite and optically dull agns @xcite , since these objects have continua which vary on year timescales , well within the inferred light travel time dimension of any torus . several authors have proposed models which use different accretion rates as a cause of the differences between observed agns . @xcite suggest that the blr and `` torus '' are inner ( ionized ) and outer ( clumpy and dusty ) parts of the same disk - driven wind , and that this wind is no longer supported at low accretion rate ( see also * ? ? ? * ; * ? ? ? similarly , @xcite suggested that low accretion rates actually drive the disk wind within the last stable orbit of the smbh , meaning that the blr can not form . models for radiatively inefficient accretion ( e.g. , * ? ? ? * ) suggest that at @xmath13 , the accretion disk becomes truncated near the smbh , with a geometrically thick and optically thin disk at low radii , and a normal thin disk ( e.g. , * ? ? ? * ) at higher radii . such objects are predicted to lack strong emission lines ( both broad and narrow ) and have weak uv / optical emission , as observed in many optically weak low - luminosity agns @xcite and x - ray bright , optically dull agns @xcite . @xcite additionally show that x - ray hardness , generally attributed to x - ray absorption , may also result from the naturally x - ray hard spectrum expected from radiatively inefficient accretion . in this work we directly measure eddington ratios for a large , x - ray selected sample of broad - line , narrow - line , and lineless agns . the eddington ratio is a unitless measure of accretion power , defined as @xmath14 . ( with @xmath0 the intrinsic accretion luminosity ) . the sample is drawn from the cosmic evolution survey ( cosmos , * ? ? ? * ) x - ray agn sample @xcite , as described in section 2 . estimates of specific accretion rates are described in section 3 , with intrinsic accretion luminosity @xmath0 measured directly from fits to the multiwavelength continuum ( avoiding uncertain bolometric corrections ) and black hole masses from the broad line scaling relations ( for broad - line agns ) or the @xmath15 relations ( for narrow - line and lineless agns ) . in section 4 we show that broad emission lines are present at only high accretion rates ( @xmath16 ) , while narrow - line and lineless agns at lower accretion rates have cooler disks , stronger radio jets , and no torus ir signature . we present a `` cartoon '' model which summarizes our results in section 5 , with predictions for future observations in section 6 . we adopt a cosmology with @xmath17 , @xmath18 , @xmath19 throughout . measuring an accurate specific accretion rate requires accurate accretion luminosities and black hole mass estimates . in particular , sed measurements from optical / uv to x - ray are necessary to constrain intrinsic luminosities to within a factor of a few ( as we show in 3.1 ) . we select a sample of 348 agns from the cosmic evolution survey ( cosmos , * ? ? ? * ) field , which is based on the 1.7 deg@xmath20 hst / acs mosaic @xcite . these agns have multiwavelength data in the form of spitzer / irac , hst / acs , subaru / suprime - cam , galex , xmm - newton , and chandra observations , as described in table [ tbl : obsdata ] . spectroscopic identification and redshifts for these objects comes from archival sdss data , magellan / imacs and mmt / hectospec @xcite , and vlt / vimos observations @xcite . the sample is selected from the parent catalog of 1651 xmm - cosmos point sources with optical counterparts @xcite , limited by @xmath21 erg s@xmath5 @xmath11 . of these x - ray point sources , 649 objects with @xmath22 have high - confidence ( @xmath23 likelihood as correct ) identifications and redshifts from optical spectroscopy @xcite in cosmos . most of the x - ray point sources without spectroscopy were missed simply due to random slit placement constraints . the optical spectroscopy is @xmath790% complete to @xmath24 , although the completeness is redshift - dependent . for broad - line agns , the spectroscopic completeness is lower at @xmath25 , @xmath26 , and @xmath27 , especially at @xmath28 ( see figure 13 of * ? ? ? * ) . for narrow - line and lineless agns , spectroscopic completeness drops dramatically at @xmath29 , since at higher redshifts the 4000 break and the feature shift redward of the observed wavelength range . to ensure that x - ray objects with narrow - line and lineless spectra are bona - fide agns , we select only objects with @xmath30 erg s@xmath5 . this x - ray luminosity limit is generally used to separate agns from x - ray fainter starburst galaxies ( e.g. , * ? ? ? we also include seven broad - line agns without x - ray detection , six of which were selected by their spitzer / irac colors and one which is a serendipitous object from the bright zcosmos survey ( which selected targets based only on @xmath24 ) . while these 7 x - ray undetected agns do not come from a complete sample , we include them to gain a larger parameter space of agn spectral types and accretion rates ( in effect , when using their x - ray limits , they occupy the same @xmath31 parameter space as a few other x - ray detected agns ) . restricting narrow - line and lineless agns to be x - ray luminous and adding the 7 x - ray undetected broad - line agns makes a parent sample of 380 broad - line , 124 narrow - line , and 49 lineless agns ( 553 total ) with high - confidence redshifts and spectral identification . measuring accurate black hole masses additionally constrains the sample to certain redshift ranges . for type 1 agns , we require the presence of one of the , , or broad emission lines in the observed spectral range , effectively limiting broad - line agns with imacs or vimos spectra to @xmath32 , @xmath33 , and @xmath34 and objects with hectospec or sdss spectra to @xmath35 . for narrow - line and lineless agns , we estimate black hole mass from the @xmath36 relation , and so we require an accurate estimate of @xmath37 . for this we use the sample of objects in cosmos with morphological decompositions @xcite from the hst / acs images @xcite , which also effectively limits the narrow - line and lineless agns to @xmath38 ( beyond which the 4000 break shifts out of the acs-@xmath39 band and the host galaxy is much more difficult to detect ) . the accurate host measurements from @xcite additionally allow us to subtract the host component before computing the intrinsic bolometric luminosity . in general , the narrow - line and lineless agns are biased towards lower redshift and consequently higher mass , since agns grow over the cosmic time . the narrow - line and lineless agns have a mean redshift of 0.7 , while the broad - line agns have a mean redshift of 1.6 . the final sample of 348 agns includes 256 broad - line , 65 narrow - line , and 27 lineless agns . full multiwavelength data exist for @xmath40 of the agns in the sample in every wavelength region except the uv . x - ray data exist from both chandra and xmm - newton : we use the deeper chandra data when available , but the chandra observations cover only the central 0.8 deg@xmath20 of the cosmos field . for the 7 x - ray undetected broad - line agns , we use the 0.5 - 2 kev xmm flux limit ( @xmath41 erg s@xmath5 @xmath11 ) for their x - ray luminosity ( since these agns have @xmath42 , their bolometric luminosity is dominated by their optical / uv emission and completely neglecting their x - ray emission does not significantly change their bolometric luminosity estimate ) . we apply the zero - point offsets derived by @xcite to the ir - uv photometry . ccccrrrc x@xmath43 & chandra & 1.24 - 6.20 & 2000 - 10000 & @xmath44 & 79/92 & 228/256 & ( 1 ) + x@xmath43 & xmm & 1.24 - 6.20 & 2000 - 10000 & @xmath45 & 79/92 & 228/256 & ( 2 ) + x@xmath46 & chandra & 6.20 - 24.8 & 500 - 2000 & @xmath47 & 88/92 & 249/256 & ( 1 ) + x@xmath46 & xmm & 6.20 - 24.8 & 500 - 2000 & @xmath48 & 88/92 & 249/256 & ( 2 ) + fuv & galex & 1426 - 1667 & 7.44 - 8.63 & 25.7 & 27/92 & 131/256 & ( 3 ) + nuv & galex & 1912 - 2701 & 4.59 - 6.84 & 26.0 & 55/92 & 184/256 & ( 3 ) + u@xmath49 & cfht & 3642 - 4180 & 2.97 - 3.40 & 26.4 & 92/92 & 254/256 & ( 4 ) + b@xmath50 & subaru & 4036 - 4843 & 2.56 - 3.07 & 27.7 & 92/92 & 256/256 & ( 4 ) + g@xmath51 & subaru & 4347 - 5310 & 2.33 - 2.85 & 27.1 & 92/92 & 256/256 & ( 4 ) + v@xmath50 & subaru & 4982 - 5916 & 2.10 - 2.49 & 27.0 & 92/92 & 255/256 & ( 4 ) + r@xmath51 & subaru & 5557 - 6906 & 1.80 - 2.23 & 27.1 & 92/92 & 256/256 & ( 4 ) + i@xmath49 & cfht & 6140 - 9119 & 1.36 - 2.02 & 26.7 & 92/92 & 256/256 & ( 4 ) + f814w & hst / acs & 7010 - 8880 & 1.40 - 1.77 & 27.2 & 92/92 & 256/256 & ( 5 ) + z@xmath51 & subaru & 8544 - 9499 & 1.31 - 1.45 & 25.7 & 92/92 & 254/256 & ( 4 ) + j & ukirt & 11665 - 13223 & 0.94 - 1.06 & 23.8 & 92/92 & 256/256 & ( 4 ) + ks & cfht & 19900 - 23050 & 0.538 - 0.623 & 23.4 & 92/92 & 253/256 & ( 6 ) + irac1 & spitzer & 31557 - 38969 & 0.318 - 0.383 & 23.9 & 91/92 & 255/256 & ( 4 ) + irac2 & spitzer & 39550 - 49663 & 0.250 - 0.313 & 23.3 & 91/92 & 255/256 & ( 4 ) + irac3 & spitzer & 50015 - 63514 & 0.195 - 0.248 & 21.3 & 91/92 & 255/256 & ( 4 ) + irac4 & spitzer & 62832 - 91229 & 0.136 - 0.197 & 21.0 & 91/92 & 255/256 & ( 4 ) + 1.4 ghz & vla & @xmath52 & @xmath53 & 20@xmath54jy & 92/92 & 256/256 & ( 7 ) + x - ray absorption and optical / uv extinction could pose a challenge to measuring the intrinsic accretion power . the most heavily absorbed agns ( e.g. compton - thick agns with @xmath55 @xmath11 ) are entirely missed by our survey because they lack detectable x - ray emission ( e.g. , * ? ? ? but if an agn is moderately absorbed and still x - ray detected , we might expect its disk to appear cooler because the uv light is preferentially extincted , and its x - ray slope to appear harder because the soft x - rays are preferentially absorbed . some agns are also intrinsically reddened , decreasing their uv emission by a factor of 2 - 3 @xcite and causing us to underestimate their accretion disk emission . with absorbed soft x - rays and extincted disk emission , we could significantly underestimate @xmath56 . we use x - ray column density @xmath8 to characterize the obscuration properties of our agns . column density and optical extinction are roughly correlated , with @xmath57 cm@xmath20 @xcite . then at @xmath59 @xmath11 , optical magnitude should be extincted by @xmath60 ( @xmath61 mag ) . assuming a smc reddening law @xcite , as is most appropriate for agns , this optical extinction translates to a factor of @xmath71.2 extinction at 3000in the uv . @xcite showed that the @xmath62 relation varies by up to a factor of 30 because of unknown changes in the gas - to - dust ratio , grain size , and/or different physical locations of the optical and x - ray absorbing material . however for all agns in the @xcite sample with @xmath63 erg s@xmath5 , @xmath64 cm@xmath20 , meaning at @xmath65 @xmath11 even the maximum optical ( @xmath66-band ) extinction is a factor of 5 and the maximum uv ( 3000 ) extinction a factor of 30 . column density @xmath8 can be accurately measured for the 153 agns ( 93 broad - line , 38 narrow - line , 22 lineless agns ) in the sample which have @xmath67 xmm or chandra counts . ( with less than 40 counts , the spectral fitting does not always stably converge . ) we fit each x - ray spectrum as an intrinsically absorbed power - law with galactic absorption ( @xmath68 cm@xmath20 in the direction of the cosmos field ) , with the power - law slope and @xmath8 as free parameters . the best - fit @xmath8 value and its error are found using the @xcite statistic . we present @xmath8 and x - ray slope @xmath69 in figure [ fig : nhxslope ] . among the 153 agns with @xmath67 x - ray counts , there are 118 unobscured agns with @xmath59 @xmath11 ( 82 broad - line , 24 narrow - line , and 12 lineless agns ) . we restrict our main conclusions to this set of 118 unobscured agns for the remainder of this work . in this work we describe the specific accretion rate using the eddington ratio parameter , @xmath14 . here @xmath0 is the intrinsic luminosity , a measure of the total accretion luminosity which includes only light from the accretion disk and x - ray corona and excludes any reprocessed ir emission . while the reprocessed ir emission can represent a large fraction of the bolometric luminosity , especially for obscured agns , it may be anisotropic . most of our agns are unobscured ( see section 2.1 ) and we exclude the ir emission to avoid double - counting the agn emission . instead we use only the optical / uv and x - ray emission which comes directly from the disk and corona in the agn : in this work when using `` intrinsic '' luminosity we are always referring to the total of the disk ( optical / uv ) and corona ( x - ray ) emission , without the reprocessed ( ir ) emission . the eddington luminosity is derived from the black hole mass , with @xmath70 erg s@xmath5 . agn luminosity is powered by accretion rate , with @xmath71 . for a constant efficiency @xmath72 , the eddington ratio @xmath6 is equivalent to the specific accretion rate @xmath73 . for example , assuming @xmath74 the eddington accretion rate can be written @xmath75 m@xmath76 yr@xmath5 with @xmath77 . however there is good evidence that @xmath72 decreases at very low accretion rates @xmath78 ( e.g. , * ? ? ? indeed , in sections 4 and 5 we invoke a lower - efficiency ( radiatively inefficient ) accretion to explain the observational properties of @xmath2 agns . this means that the accretion power @xmath56 probably underestimates the accretion rate @xmath79 for our most weakly accreting agns with @xmath2 : e.g. , a measured accretion power of @xmath80 might correspond to @xmath81 . below we outline our methods for estimating black hole masses and bolometric luminosities from the data for the agns in our sample . table [ tbl : catalog ] presents the full catalog of @xmath0 , @xmath3 , and @xmath56 , and their associated errors , for our agns . we calculate the intrinsic luminosity from the full rest - frame near - ir to x - ray multiwavelength data . this avoids monochromatic bolometric corrections which are highly uncertain and probably depend on eddington ratio ( e.g. , * ? ? ? * ; * ? ? ? instead we measure intrinsic luminosity by integrating the best - fit accretion disk + x - ray power - law sed model . we compile the broad - band near - ir ( @xmath82 , @xmath83 ) , optical ( @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 , @xmath89 , @xmath90 ) , uv ( galex nuv & fuv ) , and x - ray ( 0.5 - 2 kev and 2 - 10 kev from chandra when available or xmm - newton ) data , for which the wavebands and limits are described in table [ tbl : obsdata ] . to avoid reprocessed mid - ir emission , which would double - count the intrinsic emission for an unobscured agn , we restrict the accretion disk fit to rest - frame @xmath91 ev ( @xmath92 ) . the radio emission is negligible in the total energy output of our agns . while narrow - band optical photometry also exists for our agns , its inclusion does nt appreciably change the best - fit multiwavelength sed from using only the broad - band data . the rest - frame near - ir and optical emission of narrow - line and lineless agns is dominated by the emission from the host galaxy . for these objects , accurate intrinsic luminosities require modeling and subtracting the host galaxy light . @xcite measured the host f814w luminosities from surface brightness fitting to the hst / acs data of our agns . we use this luminosity to scale a galaxy template from @xcite . lineless agns have early - type hosts , since their spectra lack the emission lines associated with a late - type star - forming galaxy , and so we use the `` ell5 '' early - type template from @xcite . the narrow - line galaxies in our sample typically have intermediate - type ( `` green valley '' ) hosts based on their morphologies @xcite and star formation rates @xcite , and so we use the `` s0 '' template of @xcite . we subtract the host contribution in each photometric band before performing our sed fit . the reddest ( `` ell2 '' ) and bluest ( `` sd '' ) normal galaxy templates of @xcite are additionally used as extreme hosts to estimate the possible error contribution from choosing the wrong host template ( described in 3.3 ) . it is possible that a few of the narrow - line and lineless agns might have very blue starbursting hosts , although such galaxies are uncommon at @xmath93 . an extremely blue , uv - emitting host would cause us to overestimate the accretion disk emission and consequently overestimate the accretion rate . since the narrow - line and lineless agns have lower accretion rates than broad - line agns ( as we discuss in section 4 ) , if their true accretion rates were even lower it would only strengthen our conclusions . it is also possible that very red , dusty hosts could cause us to underestimate the true accretion rates for narrow - line and lineless agns . however a dusty host should cause the agn to appear extincted , and our sample of agns generally has low measured absorption ( see section 2.1 ) . in addition , restricting our fitting to @xmath91 ev ( @xmath92 ) already means that a normal elliptical galaxy ( like our `` ell2 '' template ) contributes very little flux where we fit the accretion disk . while broad - line agns are likely to have some host contribution , we can not use surface brightness fitting to estimate their host luminosity because they are at high redshift and their point source overwhelms their extended emission @xcite . however at the peak of the accretion disk emission for a broad - line agn ( @xmath94 , or 4 ev ) the host galaxy contributes @xmath95% of the emission ( e.g. , * ? ? ? because we additionally restrict our accretion disk fitting to @xmath91 ev ( @xmath92 ) , we can assume that the error from not subtracting the host for broad - line agns is typically @xmath96dex . we shift the observed ( and host - subtracted , for narrow - line and lineless agns ) photometry to the rest - frame from the measured spectroscopic redshift and convert the magnitudes or fluxes to luminosities . we then fit an accretion disk model to the optical / uv emission within the range @xmath91 ev ( @xmath97 hz , or @xmath92 ) and a power - law representing the x - ray corona emission to the rest - frame x - ray data . we measure the total bolometric luminosity from the sum of the disk luminosity ( given by the analytic solution in equation 3 below ) and the power - law luminosity from @xmath98 kev ( where @xmath99 is the peak energy of the best - fit disk model ) . while the x - ray background requires a high - energy cutoff for agns in the few hundreds of kev @xcite , measurements of the cutoff energy exist for only @xmath715 agns and vary from 50 - 500 kev @xcite . we choose 250 kev as an intermediate value , although any cutoff from 50 - 500 kev does not greatly influence our results . our agns have typically flat x - ray spectra with @xmath100 , and so changing the x - ray cutoff energy by a factor of 0.2 - 2 effectively changes the integrated x - ray luminosity by the same factor of a few . because the x - ray and disk luminosities are roughly comparable ( see figure [ fig : sedfits ] ) , this results in less than a factor of two change in the total accretion luminosity : much less than the @xmath70.5 dex errors we compute for our estimated @xmath56 ( see 3.3 ) . we use the accretion disk model of @xcite , which improves upon a basic blackbody accretion disk by including a correction for relativistic effects . ( the @xcite model is the `` diskpn '' model of the xspec x - ray fitting software . ) this model is based on the pseudo - newtonian gravitational potential @xmath101 @xcite , where @xmath102 is the schwarzschild radius @xmath103 . from @xcite , the model takes the form : @xmath104 - 1}}\ ] ] where @xmath105 and we assume the innermost stable orbit @xmath106 . the temperature depends on radius as @xmath107^{1/4},\ ] ] with @xmath108 , and @xmath109 . the coefficient @xmath110 depends on inclination angle , coronal absorption , and the color to effective temperature ratio . rather than estimate these values , we assume that @xmath110 is a constant , computed by simply scaling the model to our data . @xmath111 is the sole free parameter . in our analyses below we refer to @xmath99 , the peak energy of the disk , rather than @xmath111 , and in general @xmath112 . we find the best - fit disk model in terms of @xmath111 by minimizing the @xmath113 function . while most of the best - fit disk models have significant emission at @xmath114 ev , we restrict the fit to @xmath91 ev to mitigate the effects of a contaminating torus and/or host galaxy light . note that the relation @xmath109 above means that the disk temperature is constrained not only by the photometry by also by the black hole mass . in practice this prevents our fits from resulting in unphysically hot accretion disks , since disks peaking at energies much higher than @xmath74 ev ( 3000 ) would require unphysically small black hole masses . this is especially important to note because about one - third of the sample lacks galex uv detections , and as a result the declining high - energy slope is not well constrained by the photometry for low redshift agns . the black hole mass error ( @xmath115 dex ) is used during the bootstrapped uncertainty measurements for the accretion disk temperature and luminosity . the total disk luminosity is calculated analytically ( see appendix a of * ? ? ? * ) : @xmath116 errors in both @xmath99 and @xmath117 are found by bootstrapping 1000 fits to the resampled data . to characterize the x - ray corona emission , we use the x - ray spectral fits described in 2.1 . each x - ray spectrum is fit as an intrinsically absorbed power - law with galactic absorption ( @xmath118 cm@xmath20 in the direction of the cosmos field ) . we use the photon index @xmath119 to represent the power - law slope , such that @xmath120 . figure [ fig : nhxslope ] shows that the typical @xmath121 , and we assume this slope for agns with too few x - ray counts for a good fit . we calculate the total x - ray luminosity by integrating the power - law model over @xmath98 kev ( where @xmath99 is the energy peak of the disk model ) , using the analytic solution : @xmath122\ ] ] the total bolometric luminosity is simply the sum of the integrated accretion disk and x - ray power - law components , @xmath123 . figure [ fig : sedfits ] shows a representative sample of broad - line , narrow - line , and lineless seds with model fits . note that emission lines and variability ( the various photometric data were taken over 3 years ) mean that our simple accretion disk model is not a perfect fit : some of the optical / uv data and differ from the model fit by up to 0.2 dex . however , such small errors in individual photometry points are negligible compared to the @xmath124 dex total errors we estimate for @xmath0 ( see figure [ fig : lbolmbh ] and section 3.3 ) . in general , the accretion disk plus x - ray power - law model provides an accurate , physically motivated fit to the data . for type 1 agns , we estimate black hole masses using the scaling relations of @xcite for the broad emission line and @xcite for the and broad emission lines . these relations estimate black hole mass from single - epoch spectra by employing the correlation between the radius of the broad emission line region and the continuum luminosity , @xmath125 , observed in local agn with reverberation mapping @xcite . in general , masses estimated from the scaling relations are accurate to @xmath126 dex @xcite and agree with local agn masses from dynamical estimators @xcite and the @xmath3-@xmath127 correlation @xcite . the scaling relations take the form of equation 5 , with @xmath128 in units of @xmath129 erg / s and @xmath130 in units of 1000 km / s ; @xmath131 , @xmath132 , and @xmath133 for ; @xmath134 , @xmath132 , and @xmath135 for ; @xmath136 , @xmath137 , and @xmath138 for . black hole masses for the type 1 agns with magellan / imacs or sdss spectra in cosmos are already published in previous work @xcite , and we repeat the same techniques for type 1 agns with vlt / vimos spectra . briefly , a power - law fit plus iron emission are fit to each agn . the continuum luminosity is estimated directly from the continuum fit , while the velocity widths are computed from gaussian fits to the continuum - subtracted emission lines . some objects also have black hole masses from @xcite ; for these objects , our masses are consistent with a random scatter of only @xmath70.4 dex : equivalent to the intrinsic scatter of the scaling relations ( see figure 3 of * ? ? ? @xcite showed that the scatter in @xmath3 from the scaling relations might decrease to 0.2 dex if radiation pressure is taken into account . replacing the scaling relations from equation 5 with those of @xcite would tighten the distribution of @xmath56 estimates for broad - line agns about @xmath140 . this has no impact on the @xmath3 estimates for narrow - line and lineless agns , and does not affect the difference in @xmath56 between the broad - line sample and the narrow - line and lineless agn sample . estimating black hole masses for agns without broad emission lines requires secondary estimators . we employ the relationship between @xmath3 and rest - frame @xmath110-band host bulge luminosity @xcite : @xmath141 with @xmath142 in units of erg s@xmath5 . the @xmath143 relation comes from the more fundamental @xmath15 relation , since rest - frame @xmath110 bulge luminosity is correlated with @xmath144 ( e.g. , * ? ? ? we add an additional @xmath145 term to the @xcite relation in order to account for the evolution in the @xmath146 ratio , @xmath147 @xcite . we measure rest - frame @xmath142 from the host galaxy template from the multiwavelength sed fit ( described above in 3.1 ) . the early - type template for the lineless agns is , by definition , bulge - dominated , and so @xmath148 . the s0 template used for the narrow - line agns , however , has a significant disk component , and so we take @xmath149 . the intrinsic error in the @xmath150 is 0.35 dex @xcite . we do not correct the @xmath3 estimates for any evolution in the @xmath151 relation because measuring of @xmath151 evolution has proved difficult due to significant biases in most tests @xcite . besides , while there is some evidence for evolution to @xmath152 @xcite , there is probably little or no evolution to @xmath153 @xcite and our narrow - line and lineless agns lie at @xmath93 . because we use different mass estimators for broad - line and narrow - line / lineless agns , it is important to demonstrate that the two methods agree . seven of our broad - line agns have detected host galaxies from the decompositions of @xcite , and for these agns we compare @xmath3 estimates from the broad - line scaling relations and from the host galaxy rest - frame @xmath142 in figure [ fig : checkmbh ] . the @xmath3 estimates from broad lines and @xmath142 agree within @xmath154 for all objects ( indeed , estimates for all objects but one agree within @xmath155 ) . in addition to the seven broad - line agns in our sample , both the broad - line and host galaxy @xmath3 estimators have been shown to produce consistent masses for nearby agns @xcite . it is particularly unlikely that either of the estimators is systematically off by a factor of 100 . therefore we are confident that the factor of 100 difference in @xmath56 for broad - line and narrow - line / lineless agns in section 4 ( see , for example , figure [ fig : acchistogram ] ) is a physical effect , robust beyond the choice of black hole mass estimator . we highlight the range and limitations of the agn sample in figure [ fig : lbolmbh ] , which shows bolometric luminosities and black hole masses for the broad - line , narrow - line , and lineless agns . objects in the upper left have the highest specific accretion rates , while those in the lower right are weakly accreting agns . while the total sample spans 3 orders of magnitude in both luminosity and black hole mass , our narrow - line and lineless agns are generally less luminous and more massive than broad - line agns . the lack of low - mass narrow - line and lineless agns is due to the selection limits of the survey : such objects are too faint to be detected in cosmos . it is suggestive that these higher mass narrow - line and lineless agns are at @xmath93 and are less luminous : this is consistent with downsizing , with more massive agns becoming less active at lower redshift @xcite . figure [ fig : lbolmbh ] shows that at a given mass or luminosity there are generally all types of agns present in our sample . for this reason we do not expect that the differences between broad - line and narrow - line / lineless agns are biased by selected samples from different masses or luminosities . in addition , despite the different redshifts of most broad - line and narrow - line / lineless agns , we do not expect their differences to be caused by redshift . there is evidence that agn obscuration properties depend on redshift @xcite , but these agns are unobscured . the agn central engine , meanwhile , does not change with redshift in terms of ionization parameters @xcite , spectral energy distributions @xcite , or metallicity @xcite . limiting the sample to @xmath93 , @xmath156 , or @xmath157 does not significantly change the differences between the broad - line and narrow - line / lineless agn samples seen in figures 5 , 6 , 7 , or 8 . we estimate errors for each of our specific accretion rates , propagating the errors from both the intrinsic luminosity estimate and the black hole mass estimate . our intrinsic luminosity is subject to three major uncertainties : * photometry errors , @xmath158 . we measure the error contribution of the photometry by bootstrapping , fitting our model sed to 1000 realizations of randomly drawn photometry values distributed according to the measurement errors . in general , @xmath159 dex . * errors in the host subtraction , @xmath160 . for broad - line agn , we do not subtract a host component and assume that any remaining galaxy light overestimates the intrinsic luminosity ( from the uv and x - ray ) by only @xmath96dex ( see 3.1 ) . for narrow - line and lineless agn we estimate @xmath160 from the difference in the resultant @xmath0 when using a very red ( `` ell2 '' ) and a very blue ( `` sd '' ) template from @xcite . since the accretion disk is fit only at @xmath161 ev where there is little host emission ( even from the `` sd '' galaxy ) , this error is usually insignificant ( @xmath162 dex ) . * incorrect @xmath0 resulting from extinction , @xmath163 . extinction will make the true @xmath117 greater than our estimate because optical / uv light will be missed , but will make the true @xmath164 lower than our estimate because the power - law slope will be too hard . because we restrict our main analyses to unobscured ( @xmath59 @xmath11 ) agns , we assume this error is @xmath96 dex ( see 2.1 ) . the black hole estimate is subject to two major uncertainties : * intrinsic errors in the @xmath3 relations , @xmath165 . for broad - line agn , the intrinsic error in the scaling relations is 0.4 dex @xcite , such that @xmath166 . for narrow - line and lineless agn , we use the @xmath167 relation , and its associated intrinsic scatter is @xmath168 dex @xcite , such that @xmath169 . these errors dominate the error in @xmath56 , except for highly absorbed agns with @xmath170 @xmath11 . * measurement error in the luminosity used in the scaling relation , @xmath171 . for broad - line agn , this is the measured continuum luminosity associated with the appropriate scaling relation , estimated by @xcite as @xmath172 dex . since @xmath173 , @xmath174 for broad - line agns . for other agns the @xmath171 comes from our measured @xmath175 . we estimate this error for the narrow - line and lineless agns from 1000 fits to the randomly subsampled data , and find that the error is generally insignificant compared to the intrinsic error ( @xmath176 dex ) . note the contribution from error in @xmath130 to @xmath3 in broad - line agns is also negligible , since for our agns @xmath177 @xcite . the total error in specific accretion rate , @xmath178 , is then given by : @xmath179 we measure the total error @xmath178 by bootstrapping , with 1000 fits to the resampled data . in each fit we allow all of the parameters above to vary according to their error . the intrinsic error in the @xmath3 relations ( @xmath165 ) dominates the error . the average errors are @xmath70.5 dex , compared to the @xmath74 dex range in @xmath56 for the agn in the sample . the distribution of @xmath56 for the 118 unobscured agns is shown in figure [ fig : acchistogram ] . it is immediately evident that unobscured narrow - line and lineless agns accrete much more weakly than broad - line agns , with specific accretion rates differing , on average , by @xmath72 orders of magnitude . this suggests that many narrow - line and lineless agns are not simply geometrically obscured versions of broad - line agns , but they instead have fundamentally different accretion physics which we examine in more detail below . the large @xmath70.5 dex errors in accretion rate artificially broaden the distributions , such that the intrinsic distributions are likely to be narrower than the histograms in figure [ fig : acchistogram ] appear ( although many @xmath180 narrow - line and lineless agns could be too faint for the cosmos x - ray and spectroscopy limits ) . the @xmath181 limit for broad - line agns could be partially explained by selection effects @xcite , since low accretion rates agns are typically less luminous . however at the highest masses ( @xmath182 ) , broad - line agns with @xmath183 must be very rare @xcite . meanwhile unobscured narrow - line and lineless agns are generally limited by @xmath184 . with low x - ray column densities and low accretion rates , these objects have similar properties to the `` naked '' type 2 agns of @xcite , which additionally lack reflected broad emission lines in spectropolarimetry ( see also * ? ? ? * ; * ? ? ? we expect that the x - ray unobscured low accretion rate agns would similarly lack reflected broad emission lines . our method can not accurately estimate @xmath56 for obscured agns , but following a unified model with geometric obscuration ( e.g. , * ? ? ? * ) , obscured narrow - line agns would likely have accretion rates comparable to our broad - line agns . we can compare the specific accretion rates and agn types with the physical parameters of our model fits : namely the ratio of disk to power - law emission , the peak energy of the accretion disk model , and the x - ray power - law slope . these quantities are particularly useful in unifying agn in terms of their accretion physics . figure [ fig : accretion ] shows the specific accretion rate with these parameters for each agn type . the values of @xmath31 can be roughly translated to values of @xmath185 , with @xmath186 $ ] @xcite . the left panel of figure [ fig : accretion ] shows tracks of @xmath187 , assuming @xmath188 kev and @xmath189 ( hotter disks and softer x - ray slopes increase @xmath185 ) . once again , narrow - line and lineless agns have lower specific accretion rates , and they also tend to have lower @xmath31 and @xmath99 . we can determine the significance of any differences in @xmath31 , @xmath99 , and @xmath119 between rapidly accreting broad - line agns and weakly accreting narrow - line and lineless agns by comparing their mean values and considering the scatter of each sample . given mean values @xmath190 and @xmath191 and associated scatters @xmath192 and @xmath193 for each set , the significance of their difference is given by @xmath194 ( where @xmath195 and @xmath196 are the numbers of agns in each sample . the broad - line agns have @xmath197 while the narrow - line and lineless agns have @xmath198 , so that their difference is marginally significant at @xmath199 . the difference in @xmath99 is more significant : the broad - line agns have @xmath200 and the narrow - line / lineless agns have @xmath201 , so that the difference is significant to @xmath202 . from this we can conclude that a transition from weakly accreting narrow - line and lineless agns to the rapidly accreting broad - line agns results in significantly hotter and marginally brighter emission from the accretion disk . there is no significant difference between x - ray slope @xmath119 for the different agn types : mean @xmath203 for rapidly accreting broad - line agns , and mean @xmath204 for weakly accreting narrow - line and lineless agns ( the difference is only @xmath205 significant ) . this is in contrast to the prediction of @xcite , who suggest that harder x - ray slopes are expected for radiatively inefficient accretion flows ( riaf ) expected at low accretion rates . the appearance of a riaf at inner radii might produce more x - ray emission , as we discuss in 4.1 below , but this emission probably has a similar power - law slope as the x - ray corona present in broad - line agns with high accretion rates . this is unsurprising , since both the riaf and the corona are thought to be ionized plasmas with x - ray emission from inverse compton scattering and/or bremsstrahlung . we can conclude that the onset of a riaf in unobscured narrow - line and lineless agns with accretion rates of @xmath206 do not cause harder x - ray power - law slopes . as accretion rate increases from lineless and narrow - line to broad - line agns , the disk temperature significantly increases and its brightness with respect to the x - rays marginally increases . an increase in temperature with accretion rate is expected for a thin accretion disk , which has @xmath207 @xcite . we additionally discuss below how the onset of a radiatively inefficient accretion flow could also cause apparent cooler disk emission . both an increase in temperature and in @xmath31 with accretion rates would contribute to the observed increase of @xmath185 ( the ratio of rest - frame uv to x - ray emission ) with accretion rate @xcite . in our previous work @xcite , we suggested that the increase of @xmath185 with accretion rate was due only to the disk luminosity decreasing with respect to the corona luminosity . while this is partly correct , the correlation is also caused by increasing disk temperatures at higher accretion rates . agns with @xmath183 are predicted to have radiatively inefficient accretion flows ( riafs ) near the central black hole @xcite . at such accretion rates , we can define a truncation radius @xmath208 where the collisional cooling time is comparable to the accretion time . beyond @xmath208 , accretion will remain in a standard geometrically thin and optically thick disk with a thermal blackbody spectrum ( e.g. , * ? ? ? * ) . however within @xmath208 , there are too few collisions to couple the ions and electrons and the gas becomes a two - temperature plasma . the electrons are cooled by bremsstrahlung , synchrotron , and compton up - scattering , while the ions remain at the virial temperature . this means the flow is geometrically thick and optically thin . the introduction of a truncation radius changes the @xmath209 assumption for the accretion disk model , since by definition @xmath210 . the peak energy of the best - fit accretion disk model is not very sensitive to the choice of @xmath211 , although larger inner radii change the shape of the model with additional red emission . at accretion rates @xmath212 , as in our sample , @xmath213 @xcite . using @xmath214 in the accretion disk model fitting in section 3.1 does nt change the best - fit values of @xmath99 , although it does result in slightly better fits . the marginal ( @xmath199 significant ) increase of @xmath31 with @xmath56 might also be caused by the onset of the riaf . as @xmath208 expands outwards , the disk emission decreases and the riaf emission increases . the riaf hot plasma emission is mostly x - ray bremsstrahlung and compton up - scattering ( like the corona ) , with an additional ir synchrotron component ( which we discuss in sections 4.3 ) . as accretion rate drops and @xmath208 increases , the rise of the riaf x - ray emission compared to the optical / uv disk emission is seen as a decrease of @xmath31 . indeed , local low - luminosity agns have even lower accretion rates and larger @xmath208 , with consequently lower @xmath31 ratios and cooler optical thin - disk emission @xcite . the transition to an inner riaf also causes the disappearance of broad emission lines at @xmath183 . @xcite was the first to elegantly show that the broad emission lines are only present above a critical accretion rate . however @xcite assumed that the innermost possible orbit was given by the @xcite thin - disk model , @xmath215 . here we follow their basic derivation , with the key difference that we use the riaf transition radius as the innermost orbit for the presence of a broad - line region . there is evidence that the broad emission line region is part of a disk wind ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? the positions of individual broad emission lines are stratified and set by the ionizing luminosity of the continuum ( e.g. * ? ? ? * ; * ? ? ? the base of the wind itself , however , is set by the radius at which the radiation pressure equals the gas pressure , defined by @xcite as : @xmath216 with @xmath217 in units of @xmath218 , @xmath219 in units of @xmath220 , @xmath12 is the viscosity parameter , and @xmath72 is the accretion efficiency . while riafs are expected to have strong outflows ( see section 4.2 ) , the riaf region is a high - temperature ionized plasma and so any associated disk wind would not emit broad emission lines in the uv / optical . thus the riaf truncation radius sets the innermost possible radius for the existence of a broad - line region . assuming that @xmath221 and rearranging equation 8 with @xmath222 , @xmath223 , and @xmath224 , this sets the minimum specific accretion rate for a broad line region as : @xmath225 with @xmath226 . we leave @xmath208 as a free parameter since it is poorly constrained , although the best - fit riaf models for @xmath227 agns suggest @xmath228 @xcite . as an agn drops below this minimum accretion rate , its broad lines disappear and only narrow lines ( or no lines ) are observed , as seen in the transition at @xmath229 transition in figures [ fig : acchistogram ] and [ fig : accretion ] . @xcite also predict that the disk wind associated with the blr will disappear below an accretion rate at which the outflowing velocity drops below the random velocity of the disk . @xcite measure a blr - disappearance accretion rate of @xmath230 from the low - luminosity local agns of @xcite , with @xmath231 and @xmath232 . in our sample ( as well as those of * ? ? ? * ; * ? ? ? * ) , the blr disappears at @xmath233 . for a typical bolometric luminosity of @xmath234 erg s@xmath5 ( also appropriate for the * * sample ) , and assuming the same @xmath235 , this instead corresponds to @xmath236 : a remarkable difference of 6 orders of magnitude . it is unlikely that the bolometric corrections of @xcite are incorrect by 6 orders of magnitude , and so we must conclude that the @xcite model does not describe the disappearance of the blr for high luminosity agns . instead a disk - wind model following @xcite best describes the blr disappearance as the radius of wind generation region moves within the inner riaf region . it must be noted that while disk wind models have had success in describing highly ionized emission and absorption lines in the uv @xcite , they have not been applied to optical emission lines . the broad emission line almost certainly forms in a higher density , lower ionization region than the and broad emission lines . in addition there is evidence that the dynamics of the broad emission line are wildly variable , with reverberation mapping indicating infalling , virialized , and outflowing emission regions in three agns @xcite . while we do find that broad tends to be present only for @xmath181 and so fits in the wind / riaf framework , we do not study and can not say if this line is described by the same physics . indeed , @xcite present several agns with broad emission and @xmath237 . this suggests that broad emission may have its origin outside the disk wind , although it is important to note that the accretion rates of @xcite rely on bolometric corrections to monochromatic luminosities and so may suffer from significant systematic uncertainties . the gas in a riaf is not gravitationally bound to the supermassive black hole because the ions are not losing energy through radiation . as a result , agns with riafs are predicted to have strong radio outflows @xcite . the coupling between a riaf and a strong radio outflow has been confirmed by observations of black hole binaries @xcite , and it is possible to translate these observations to agn scales ( e.g. * ? ? ? * ) . in figure [ fig : radio ] we show the agns of our sample with the ratio of radio luminosity to disk luminosity . note that since the radio emission is coincident with the x - ray point source we assume that it originates from the agn , but we can not strictly rule out other sources of radio emission ( e.g. , from star formation ) . the @xmath238 agns which are expected to have riafs tend to have higher ratios of radio to disk ( optical / uv ) luminosity . the mean @xmath239 for rapidly accreting broad - line agns is a factor of ten lower than the mean @xmath239 for narrow - line and lineless agns , and since the scatter in each sample is about @xmath70.5 dex this translates to a highly significant difference ( @xmath240 ) . the large scatter in the @xmath239 ratio at both high and low accretion rates is likely because the radio power is additionally dependent on properties like black hole spin and orientation . but the highly significant increase in @xmath239 for low accretion rate agns suggests that @xmath238 agns with riafs generally have relatively brighter radio emission . @xcite noticed a similar trend of increasing radio luminosity with decreasing accretion rate , using as a proxy for intrinsic luminosity ( e.g. , * ? ? ? * ; * ? ? ? many nearby radio galaxies are also measured to have low accretion rates and may even have their optical / uv emission dominated by synchrotron emission rather than a thermal disk @xcite . in general , the radiation and disk winds of agns are thought to cause feedback on galaxy scales by quenching star formation ( e.g. , * ? ? ? * ; * ? ? ? * ) , while radio jets are thought to cause larger - scale feedback which can heat the cores of galaxy clusters ( e.g. , * ? ? ? * ) and is observed as extended emission line regions @xcite . the fact that riafs tend to have stronger radio emission suggests that weakly accreting agns may remain important for large - scale radio - mode feedback despite their optical / uv and x - ray luminosities . this suggests that heating cluster cores may not require bright quasars , but can be accomplished by faint agns ( see also * ? ? ? @xcite similarly found that several nearby weakly accreting agns had most of their bondi accretion rates converted to radio outflows . a clumpy dust `` torus '' emits a unique power - law signature in the mid - ir from @xmath71 - 10@xmath54 m @xcite . this was first noticed observationally as a distinct agn locus in spitzer / irac color - color space @xcite , although @xcite show that power - law selection is the most effective way to select agn in the mid - ir . we compute the ir power - law slope in our agns from the host - subtracted observed irac photometry within the rest - frame wavelength range @xmath241 m , shown with accretion rate in figure [ fig : iraccrete ] . type 1 agns typically have @xmath242 ( @xmath243 in terms of the @xmath244 form used by * ? ? ? * ) , matching the predictions of clumpy dust models @xcite . about 10% of type 1 agns are `` hot - dust - poor '' and do not satisfy the @xmath245 selection criterion , about half of the narrow - line agns and only one of the lineless agns lack this torus signature . @xcite similarly found that many x - ray agns did not have a mid - ir power - law , although they did not track it with accretion rate . in our sample , the rapidly accreting broad - line agns have a mean @xmath246 , while the weakly accreting narrow - line and lineless agns have a mean @xmath247 , meaning that the two samples differ with high significance ( @xmath248 ) . a unified model based solely on geometrical obscuration suggests that narrow - line and lineless agns are obscured by the same torus present in broad - line agns ( e.g. , * ? ? ? instead the low accretion rate agns ( @xmath249 ) frequently lack the torus ir signature . in part , this may be because the torus power - law is simply being overwhelmed by the accretion disk sed at @xmath249 . at low accretion rates , the temperature of the disk decreases , and a disk with @xmath250 ev will peak at 1.2 @xmath54 m , emitting a power - law of @xmath251 at @xmath252 m . in a typical broad - line agn , the ir torus is roughly the same strength as the accretion disk ( * ? ? ? * see also figure [ fig : sedfits ] ) . since many @xmath249 agns in figure [ fig : iraccrete ] have @xmath253 , they must be dominated by the accretion disk emission and have , at best , very little emission from the torus . the weaker or missing torus in many @xmath249 agns can be described in a similar fashion to the vanishing disk - wind blr in section 4.1 . there is good evidence that the outer edge of the blr coincides with the inner edge of the clumpy dust @xcite . some authors additionally suggest that the blr and the clumpy dust `` torus '' are two components of the same wind driven off the accretion disk ( e.g. , * ? ? ? if the clumpy dust wind emerges from the disk at a similar radius to that calculated in section 4.1 , then we would expect the ir power - law signature to disappear at @xmath249 , just as the blr disappears . however many narrow - line agns with @xmath249 still have the negative ir power - law slopes , suggesting that there must be another source of mid - ir emission . either there is a distant source of clumpy dust beyond the expanding riaf , or there is mid - ir synchrotron emission in the riaf region at the base of the radio jet ( as observed by * ? ? ? figure [ fig : cartoonmodel ] presents a simple schematic outlining the changes in agns from high ( @xmath254 ) to low ( @xmath249 ) accretion rate . at the top is a broad - line agn with high accretion rate ( @xmath255 ) . at these high accretion rates the gas and dust falling into the black hole forms a thin accretion disk and a disk wind originates at @xmath256 . the broad emission lines are emitted in stratified regions along this wind based on the radiation pressure ( which ionizes and excites the wind material ) , with @xmath125 and high ionization lines ( e.g. , ) emitted from nearer radii than low ionization lines ( e.g. , ) @xcite . at higher radii , the disk wind forms clumpy dust @xcite . this dusty `` torus '' can obscure the agn along lines of sight near the disk , causing an observer to see an obscured narrow - line agn @xcite . the bottom of figure [ fig : cartoonmodel ] shows an agn with low accretion rate ( @xmath257 ) , characteristic of the unobscured narrow - line and lineless agns in our sample . the onset of a geometrically thick riaf changes the picture dramatically . because the disk wind radius is within the riaf , there are no broad emission lines . instead the dominant outflow is a radio jet , and agns with low accretion rates and riafs are typically more radio luminous than broad - line agns . the lack of a disk wind also means that there is not the typical clumpy dust `` torus '' seen in broad - line agns . however we can not rule out the presence of dust completely , as clumpy dust may come from another source besides the disk wind and some @xmath183 have the ir signature of hot dust . the multiwavelength data of cosmos provide many diagnostic capabilities , and we have argued that decreasing accretion rates lead to the onset of a riaf at @xmath258 and subsequently stronger radio jets , a weaker torus , and the disappearance of broad emission lines . the onset of a riaf also makes several predictions testable by future observations . in addition the simple model in section 5 can be more fully constrained by additional investigations . if the broad - line region is truly disappearing at @xmath258 then we would expect spectropolarimetry to reveal reflected broad emission lines in only high accretion rate ( @xmath259 ) narrow - line and lineless agns . spectropolarimetry of nearby agns shows a dichotomy based on accretion rate , although most authors place the change from hidden broad lines to `` true '' type 2 agn at @xmath260 @xcite . most likely , the difference results from the uncertain bolometric corrections used in these previous works , compared to the full modeled seds used here . mid - ir broad - band polarimetry could determine the cause of the negative ir power - law slopes in @xmath258 agns . if the clumpy dust `` torus '' is associated with the same wind that drives the broad line region , it should disappear in these objects . the mid - ir signature might instead be synchrotron radiation in the riaf at the base of the jet , which would appear polarized at the @xmath261 level ( e.g. , * ? ? ? if no polarization is detected , then we must conclude that clumpy dust exists at higher radii than the blr disk - wind , beyond the riaf region of @xmath258 agns . it is very difficult to measure accretion rates of partially or fully obscured agns , and such objects are generally missed by the x - ray and optical limits of this study . however we do make a few predictions for the accretion rates of various agns . if the torus is part of a disk - wind that vanishes at @xmath258 , then torus - obscured agns of the classical @xcite unified model will have only high accretion rates ( @xmath259 ) . obscuration by cooler dust associated with host galaxy star formation , as predicted by the observed redshift evolution in the narrow - line / broad - line agn ratio @xcite , could conceivably be present at any accretion rate ( although it may be limited by the ability of the dusty star formation to feed the black hole , * ? ? ? we might then expect that obscured agns with a strong mid - ir torus signature should have @xmath259 , while agns obscured by the cooler dust associated with host galaxy star formation might have a wider range of accretion rates . jrt acknowledges support from nsf / ddep grant # 0943995 , and with cdi acknowledges support from nsf grant # ast-0908044 . bck acknowledges support from nasa through hubble fellowship grant # hf-51243.01 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , for nasa , under contract nas 5 - 26555 . lcrcrrr 095728.34 + 022542.2 & bl & 1.54 & s & @xmath263 & @xmath264 & @xmath265 + 095740.78 + 020207.9 & bl & 1.48 & i & @xmath266 & @xmath267 & @xmath268 + 095743.33 + 024823.8 & bl & 1.36 & s & @xmath269 & @xmath270 & @xmath271 + 095749.02 + 015310.1 & nl & 0.32 & i & @xmath272 & @xmath273 & @xmath274 + 095750.20 + 022548.3 & bl & 1.24 & z & @xmath275 & @xmath276 & @xmath277 + 095752.17 + 015120.1 & bl & 4.16 & z & @xmath278 & @xmath279 & @xmath280 + 095752.17 + 015120.1 & bl & 4.17 & i & @xmath281 & @xmath282 & @xmath283 + 095753.49 + 024736.1 & bl & 3.61 & i & @xmath284 & @xmath285 & @xmath286 + 095754.11 + 025508.4 & bl & 1.57 & s & @xmath287 & @xmath288 & @xmath289 + 095754.70 + 023832.9 & bl & 1.60 & s & @xmath290 & @xmath291 & @xmath292 + 095755.08 + 024806.6 & bl & 1.11 & s & @xmath293 & @xmath294 & @xmath295 + 095755.34 + 022510.9 & bl & 2.74 & i & @xmath296 & @xmath297 & @xmath298 + 095755.48 + 022401.1 & bl & 3.10 & i & @xmath299 & @xmath300 & @xmath301 + 095759.50 + 020436.1 & bl & 2.03 & s & @xmath302 & @xmath303 & @xmath304 + 095759.91 + 021634.5 & bl & 1.54 & i & @xmath305 & @xmath306 & @xmath307 + 095801.61 + 020428.9 & bl & 1.23 & z & @xmath308 & @xmath309 & @xmath310 + 095802.10 + 021541.0 & od & 0.94 & i & @xmath311 & @xmath312 & @xmath313 + 095805.10 + 020445.8 & nl & 0.67 & i & @xmath314 & @xmath315 & @xmath316 + 095806.24 + 020113.8 & nl & 0.62 & i & @xmath317 & @xmath318 & @xmath319 + 095806.99 + 022248.5 & bl & 3.10 & i & @xmath320 & @xmath321 & @xmath322 + 095809.45 + 020532.4 & od & 0.61 & i & @xmath323 & @xmath324 & @xmath325 +

we present hst / wfpc2 linear ramp filter images of high surface brightness emission lines ( either [ oii ] , [ oiii ] , or h+[nii ] ) in 80 3cr radio sources . we overlay the emission line images on high resolution vla radio images ( eight of which are new reductions of archival data ) in order to examine the spatial relationship between the optical and radio emission . we confirm that the radio and optical emission line structures are consistent with weak alignment at low redshift ( ) except in the compact steep spectrum ( css ) radio galaxies where both the radio source and the emission line nebulae are on galactic scales and strong alignment is seen at all redshifts . there are weak trends for the aligned emission line nebulae to be more luminous , and for the emission line nebula size to increase with redshift and/or radio power . the combination of these results suggests that there is a limited but real capacity for the radio source to influence the properties of the emission line nebulae at these low redshifts ( ) . our results are consistent with previous suggestions that both mechanical and radiant energy are responsible for generating alignment between the radio source and emission line gas .

0710.3105

radio galaxies are an important class of extragalactic objects : they represent one of the most energetic astrophysical phenomena ; they may be used as probes of their environments ; and they are unique probes of the early universe @xcite . as the 3cr sample of radio galaxies is radio flux - density selected @xcite , it provides an unbiased optical sample to study the host galaxies of these radio - loud active galactic nuclei ( agn ) . the sample has been well studied at multiple wavelengths and previous ground - based studies have investigated the spatial coincidence of optical emission nebulae and the radio source @xcite . @xcite and @xcite first demonstrated the `` alignment effect '' at high redshift ( @xmath2 ) where the continuum optical and/or ir emission is aligned along the radio axis . an `` alignment effect '' for emission line gas was also shown by @xcite and @xcite . in this paper we focus on the alignment of the fine scale high surface brightness emission line gas imaged by the hubble space telescope ( hst ) with the radio emission . the primary mechanisms for the emission line gas alignment are thought to be shocks induced by the radio jet and photoionization from the central agn ( e.g. , * ? * ; * ? ? ? * ; * ? ? ? @xcite demonstrated the redshift dependence of the alignment effect - at @xmath3 there is no alignment , at @xmath4 some alignment is seen , and for @xmath5 essentially all the powerful radio galaxies show strong alignment . @xcite found alignment at low @xmath6 if one compares the emission in the quadrants containing the radio source to quadrants without radio emission . studies which attempt to break the redshift - radio power degeneracy suggest that the alignment depends on both redshift and radio power ( e.g. , * ? ? ? * ) . thus , the alignment depends on both the presence of extended gas in the environment as well as the ability of the radio source to influence its environment . in this paper we examine the relationship of the high surface brightness emission line gas and radio emission over a wide range of redshift ( @xmath7 ) , focusing on the transition redshift range @xmath1 where the alignment effect starts to turn on . using the wide field and planetary camera 2 ( wfpc2 ) on hst , we obtained images with resolutions of 0.@xmath805 0.@xmath81 , similar to or better than the synthesis imaging resolution obtained with the very large array ( vla ) at 5 ghz in the `` a '' configuration . the higher resolution allows us to probe denser gas in the center of these galaxies . in addition , the tunable linear ramp filter ( lrf ) permits narrow band imaging of a selected spectral feature for a wide range of redshifts , enabling studies of a specified emission line for a sample such as the 3cr covering a large range of redshift . @xcite conducted an initial study using this data , and published a subset of the data . here we expand on their analysis , presenting the complete set of detected images as well as statistical results . the paper is organized in the following fashion : we discuss the properties of the sample in @xmath9[sec : sample ] . @xmath9[sec : observations ] discusses the observations made and the data reduction methods used . in @xmath9[sec : data analysis ] we outline the analysis of the data . @xmath9[sec : properties ] contains a discussion of our sources and the results of our study . in @xmath9[sec : conclusions ] we conclude by discussing the implications of our results . appendix [ appendix ] contains notes for individual sources . our sample is drawn from a set of 100 3cr radio galaxies selected from the hst 3cr imaging survey . using earlier spectroscopic observations , objects with significant emission line flux were selected @xcite for this study to maximize detection . thus , our results may not be extendible to objects with very faint emission lines and our observations do not reach uniform surface brightness in the rest - frame of the object . observations were done in hst s `` snapshot '' mode , in which targets are randomly selected from the target list for observation . 80 sources were observed and constitute our sample , spanning a redshift of 0.017 @xmath10 1.406 ( see figure [ redshift_histogram ] ) . the sample contains 9 fris , 58 friis , 12 css sources , and 1 source of an unknown morphology . 22 of the radio galaxies have broad emission lines while 58 have narrow emission lines in their optical spectrum . arcsecond - scale resolution radio maps for the sample objects were obtained from a variety of sources ( referenced in figure captions ) . sources for which 5 or 8-ghz vla ` a ' array maps were not already available were reduced using archival data . the cleaning and self - calibration of the raw data was performed using the nrao s aips software package . we present new maps for the following sources : 3c124 , 3c135 , 3c284 , 3c303.1 , 3c341 , 3c368 , 3c379.1 , and 3c382 . table [ table : vla_obs_params ] lists the observation parameters for the new vla maps , including the date of observation , frequency and rms noise of the final map . snapshot images were taken using the lrfs on hst s wfpc2 instrument . there are 3 wf chips and 1 pc chip . the chip on which the lrf images a target varies with the selected wavelength . we have noted the chip upon which a source was imaged in table [ table : optical_obs ] . nearby objects ( @xmath11 ) were imaged in h@xmath0+[nii ] . objects at intermediate redshifts were imaged in [ oiii ] 5007 @xmath12 and objects at high redshift ( @xmath13 ) were imaged in [ oii ] 3727 @xmath12 . table [ table : optical_obs ] also lists which emission line was imaged for each source . all of the lrf observations consisted of a pair of 300 sec exposures taken in `` snapshot '' mode . the images were calibrated and combined using the crrej procedure to remove cosmic rays @xcite . because the design of the lrf does not allow direct observations of the flat field , the images were flat fielded using the f656n flat . note that due to the small size of the wfpc2 pixels , we are sensitive only to high surface brightness extended emission line gas . in order to obtain images of only the emission line region ( elr ) , the continuum contribution to the lrf images was subtracted . this was usually done using an image of the object in the broadband f702w filter . the continuum was scaled to the observed wavelength by assuming a 3000k blackbody spectrum for the continuum @xcite . in some cases the continuum was contaminated by emission lines . @xcite discussed the uncertainties in flux calibration due to contamination from emission lines . using standard line ratios , they found corrections from 1.002 to 1.36 ( median 1.09 ) . we have not applied corrections for emission line contamination to our data , but the values given by them indicate the uncertainty due to contamination . continuum images were not available for four sources , 3c98 , 3c273 , 3c289 , and 3c323.1 , and the subtraction could not be performed . these sources have been left out of the analysis , but their narrow - band lrf images are presented if the source was detected . we grouped the emission - line nebulae into 4 broad optical morphological categories : ` extended ' sources ( 21 sources ) , ` partially extended ' sources ( 33 sources ) , slightly or unresolved sources ( 12 sources ) , and non - detections ( 11 sources ) . four sources were eliminated due to the lack of a continuum image for subtraction . sources were classified as extended if the 3-@xmath14 contour was extended compared with the expected psf in the continuum subtracted image . ` partially extended ' sources are those which feature extensions smaller than 2@xmath8 but do not match the psf . they often resemble an unresolved source , with a small faint extension . we analyzed only the ` extended ' and ` partially extended ' sources , as we were able to determine reliable position angles and/or angular sizes for most of these objects . table [ table : optical_data ] lists the morphological categories for each object . to examine the spatial coincidence of the elr and the radio source , the radio and optical images for the ` extended ' sources were combined into overlays . in general , the images were registered using the radio core ( for the radio sources ) , and the centroid of the host galaxy ( obtained from the continuum images ) . one ` extended ' source ( 3c305 ) has a known dust lane , making centroiding on the continuum image unreliable . we attempted to determine a center through ellipse fitting to the outer portion of the galaxy , but this was unsuccessful due to the close companion galaxy . the registration in the image presented was done using centroiding on the continuum image . the overlays are shown in figures [ 3c46-montage ] - [ 3c381-overlay ] along with the continuum and emission - line images for each source . position angles ( pas ) for the ` extended ' sources were measured using the emission within the 3-@xmath14 contour , with the pa passing through the nucleus of the host galaxy in the direction of the longest extent of the emission line nebulae . pas for the partially extended sources were determined using iraf s ` ellipse ' task . radio pas were taken from the literature ( if available ) , or measured from hotspot - to - hotspot for friis or across the longest extent for fris on the radio maps . the measurement of the difference in the pas is independent of the spatial alignment of the radio and optical images as the position angles are independently determined from the images . a measure of alignment between the elr and the radio source was determined by taking the absolute value of the difference in position angles of the radio source and the elr . if @xmath15pa @xmath16 30 degrees , the elr and radio source were considered aligned . in some cases the elr displayed a position angle for the inner regions which differed from the large scale position angle . in most cases , the usage of an `` inner pa '' did not cause misaligned objects to be moved into the aligned category , so only the large - scale pa was used for the purposes of our analysis . emission line fluxes were measured using iraf s ` apphot ' set of tasks . luminosities and sizes were calculated using a cosmology of h@xmath17=71 km s@xmath18 , @xmath19@xmath20=0.27 , and @xmath19@xmath21=0.73 . values were obtained using an online cosmology calculator @xcite . in order to compare the luminosities from each source , the measured fluxes were converted to h@xmath0 fluxes using line ratios derived from observations of narrow - line radio galaxies @xcite . the ratios of [ oiii ] , [ oii ] , and h@xmath0+[nii ] to h@xmath0 were 3 , 1.23 , and 2.08 respectively . here we discuss the properties of the nebulae and their relationship to the radio source . individual source descriptions are provided in appendix [ appendix ] . in table [ table : radio_data ] we provide morphological information about the radio structure , including classifications , position angles , and angular sizes . we also present luminosities for each source . table [ table : optical_data ] contains the morphological information for the continuum - subtracted emission line images . this includes position angles , angular sizes , and fluxes . in addition , this table contains the v - band magnitudes for the host galaxies and the position angle of the galaxy . the detection statistics varied depending on the emission line observed . nine of ten sources ( 90% ) imaged in h@xmath0+[nii ] were detected . forty - three of forty - seven sources ( 91% ) imaged in [ oiii ] were detected , while fifteen of twenty - three ( 65% ) [ oii ] detections were made . the statistical significance of the observed relationships discussed in this section are given in table [ table : correlations ] . the kendall tau value is a measure of the correlation between the specified measurements . the spearman s rho value is a similar measure , describing how likely the two measurements are to increase or decrease concordantly . statistical values for non - correlations are given in table [ table : non - correlations ] . we find that 19 of 37 sources for which a pa could be measured showed alignment between the high surface brightness elr and the radio source to within 30 degrees ( figure [ delta_pa - redshift ] ) . we have 4 sources at @xmath5 which all show strong alignment . in the remaining 33 sources at @xmath1 , 15 show alignment . the median values of the position angle difference in bins of redshift do not decrease significantly as a function of @xmath6 out to 0.6 . however , the upper envelope of the position angle differences does seem to decrease with redshift . we clearly see a weaker alignment effect at low redshift than is seen at the high redshifts @xmath5 . our high resolution hst data show that the lack of strong alignment at @xmath1 is not due to a lack of spatial resolution . we investigated the possibility of the misaligned objects being predominantly broad line radio galaxies ( blrgs ) . according to the unified scheme @xcite , blrgs are not expected to show alignment as the jet would be pointing almost along the line of sight . however , blrgs make up only 11 of the sources with measured position angle differences , and several of the blrgs show alignment ( 3c234 , 3c268.3 , 3c277.1 ) . thus , orientation of the radio jet close to the line of sight is not responsible for the misalignment at @xmath1 . a potential concern is that much of the low redshift observations ( @xmath22 ) are obtained using the [ oiii ] line which is higher ionization than the [ oii ] and h@xmath0+[nii ] . if the high ionization gas has very different morphology and position angle than the lower excitation gas , this might reduce the strength of the alignment at low redshift . however , observations of radio galaxies in which both [ oii ] and [ oiii ] are obtained show that the estimated large scale position angles are consistent ( e.g. * ? ? ? * ) . another possible bias is the extent of the elrs for different emission lines . studies of single objects in multiple emission lines indicate typical ratios in the extended emission of @xmath231 - 6 for [ o iii]/[o ii ] and @xmath231 - 4 for [ o iii]/h@xmath0 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? thus , as the transition between lines is made , the elr should be smaller for [ o ii ] than for [ o iii ] . as will be discussed in @xmath9 [ ssec : elr - size_l ] , the opposite is found . over the redshift range of our sample ( 0.017 @xmath10 1.406 ) , the expected ( 1+z)@xmath24 dimming of surface brightness is a factor of @xmath2330 - though the details will depend on the actual surface brightness profiles . thus , both the transition from [ oiii ] to [ oii ] and the ( 1+z)@xmath24 effect should result in apparently smaller elr at higher redshift . despite this , we see emission line size increasing with redshift ( see @xmath9 [ ssec : elr - size_l ] ) , suggesting that this result is real . in order to investigate the possible effect of an instrumental sensitivity selection effect on the detection of alignment , we redshifted the faintest level in the emission line images to see what would be detected if located at @xmath25 ( just prior to the onset of strong alignment ) . none of the misaligned sources showed any alignment when viewed using these modified levels . this indicates that the alignment effect is not due to a selection effect of the sensitivity of the instrument , but rather to some other inherent attribute of the source and its environment . the emission line images from hst were compared with previous studies using ground based imaging ( figure [ fig : hst - ground ] ) . morphologies and position angles were generally consistent between the two sets consistent with the hypothesis that the nebulae position angles do not exhibit large changes in position angle as a function of size scale . however , because of the lower surface brightness sensitivity of the hst / wfpc2 camera , the hst images show only the brighter inner regions of the nebulae . thus , the hst measured sizes and emission line fluxes tend to be smaller than those measured on ground based images . figure [ all_delta_pas ] shows a comparison of the distribution of position angle differences from ground - based data @xcite with our hst data . there is a suggestion that the hst data shows an alignment effect which is weaker than seen in the ground - based data at the same redshifts . the kolmogorov - smirnov ( ks ) test indicates that the distribution of position angle differences in the hst data is consistent with a uniform distribution of angles , while the ground based data are not ( see table [ table : ks - results ] ) . if larger samples confirm this result , this may be due to the hst data mainly being sensitive to the higher surface brightness , more compact emission , suggesting that the alignment effect is mainly produced in fainter , more extended emission line gas . . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] + . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] + . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] . the middle row is for @xmath26 and the bottom row is for @xmath4.,title="fig : " ] + we computed the ` size ratio ' comparing the relative projected linear sizes of the radio source and the high surface brightness emission line region . the linear sizes for fr ii and css sources were measured from hotspot to hotspot , and the linear sizes for fr i sources were measured across the overall extent . linear sizes for the emission line regions were measured roughly along the long axis ( if such an axis existed ) . comparing the relative sizes of the radio source and elr with the relative alignment indicates that sources with similar radio and elr sizes are more frequently aligned than sources in which the elr is much smaller than the radio source ( figure [ size_ratio - delta_pa ] ) . this is true at all redshifts ( figure [ size_ratio - delta_pa ] ) . we also note that all six of the small sources ( d @xmath27 15 kpc , i.e. , css ) are well aligned ( figure [ delta_pa - radio_size ] ) . all six of these have size ratios between @xmath230.3 and @xmath23 6 . these results suggest that the alignment at low redshift results when the source is smaller / younger and of a similar size as the nebula . this is consistent with suggestions that the alignment is due to mechanical energy input from the radio source ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . at low redshifts , suitable gas clouds are likely to exist only on the scale of the host galaxy . . the sources with size ratios near unity are found at both high and low redshifts.,title="fig:",width=302 ] . the sources with size ratios near unity are found at both high and low redshifts.,title="fig:",width=302 ] in addition , 13 of the 29 large sources ( size greater than 50 kpc ) also show alignment to within 30@xmath28 . these sources are much bigger than their emission line nebulae ( factors of 10 - 100 ) . some of this alignment may be produced by chance if the position angles differences in at least some low @xmath6 sources are randomly distributed . however , some of the alignment may be real since we suspect the effect is starting to turn on at these redshifts . diagnostic line ratios and kinematics indicate that shocks are important when the radio source size is comparable to the aligned emission line nebula size and that photoionization dominates when the radio source is much larger than the nebula ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? thus , in these large sources , the alignment may be produced by photoionization from the central agn along the ionization cone " . recombination times for warm gas ( t @xmath2310@xmath29 k ) are roughly 10@xmath30 years @xcite while the lifetimes of the radio sources are roughly 10@xmath31 10@xmath32 years , indicating the ionizing mechanism must be on - going throughout the lifetime of the source . the size of the high surface brightness elr increases with increasing radio power ( figure [ emission_size - radio_luminosity ] ) and/or with increasing redshift ( figure [ emission_size - redshift ] ) . both correlations are somewhat weaker for just the @xmath1 objects . at higher redshifts , dense gas clouds are found on increasingly larger scales culminating in the large ly-@xmath0 halos surrounding powerful radio galaxies @xcite . of course , the radio luminosity will also increase with redshift in flux density limited samples , meaning the correlations may be related . studies which attempt to break the redshift - radio power degeneracy suggest that the alignment depends on both redshift and radio power ( e.g. , * ? ? ? * ) . the more powerful radio sources can provide both more mechanical energy and more ionizing photons thus both the presence of gas on larger scales as well as the larger energy input from the radio source could result in the emission line size increasing with redshift and radio power . comparing the luminosity of the elr with the relative alignment ( figure [ emission_luminosity - delta_pa ] ) indicates aligned sources have more luminous elrs . the correlation becomes weaker for the redshift range @xmath1 . we find that the correlation is further weakened if we also exclude @xmath33 ( figure [ emission_luminosity - delta_pa ] ) . the existence of the ( weak ) correlation at @xmath1 suggests that the radio source provides additional energy to the emission line gas along the direction of the radio axis and thus , there is a ` real ' though weak alignment effect at these low redshifts . . right : same plot as that on the left , but for sources with @xmath4 only . the probability of this being a non - correlation is 0.0429 ( kendall tau ) . , title="fig:",width=302 ] . right : same plot as that on the left , but for sources with @xmath4 only . the probability of this being a non - correlation is 0.0429 ( kendall tau ) . , title="fig:",width=302 ] we present wfpc2 lrf images of selected emission lines in a sample of 3cr radio galaxies with bright emission lines . we detect clumpy , high surface brightness emission line gas with a wide range of morphologies . we examine the properties of the emission line gas and its relationship to the radio source , and in particular the nature of the alignment effect at @xmath1 where the effect is beginning to become important . we find there is a weak alignment effect in our low @xmath6 sample ( @xmath1 ) . the effect is clearly weaker than seen at high redshifts and is also somewhat weaker than seen at similar redshifts in ground - based data . this suggests that the alignment effect is dominated by the more extended lower surface brightness emission line gas which falls below our detection threshold in wfpc2 . the alignment at low redshift is strong when the radio source and the nebula are both of order galactic scales ( e.g. , css sources ) . this implies there is mechanical energy input from the radio source as it propagates through the ambient gas in the host galaxy ( e.g. , * ? ? ? * ; * ? ? ? there is a weak trend for the nebulae in the aligned sources ( at all radio / optical size ratios ) to have higher line luminosity . this suggests that the radio source preferentially provides energy to gas which is along the radio source axis . this is likely to be predominantly mechanical energy when the radio source and nebulae have similar sizes , and predominantly photoionization when the radio source is much larger than the nebula ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? there is a weak trend for the size of the high surface brightness emission line nebulae to increase with increasing radio power and/or redshift over the range @xmath1 . this may be due to both the increasing presence of gas on large scales and the ability of the more powerful radio sources to provide additional mechanical energy and/or ionizing photons as a function of redshift . thus , there is evidence for a modest excess of aligned sources at low redshift ( @xmath1 ) . the alignment is strong when the radio source and emission line nebula are both on galactic scales . there are also weak trends for the aligned emission line nebulae to be more luminous , and for the emission line size to increase with redshift and radio power . the combination of these results suggests that there is a limited but real capacity for the radio source to influence the properties of the emission line nebulae at these low redshifts . support for program 5957 was provided by nasa through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas 5 - 26555 . 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 research has also made use of nasa s astrophysics data system bibliographic services . we are grateful to the referee for comments which helped to clarify the presentation of the results . _ facilities : _ , , c. e. and garrington , s. t. , 1995 , , 112 , 235 , j. r. , 1984 , , 210 , 611 r. r. j. , 1984 , , 278 , 499 d. j. , capetti a. , fanti r. , morganti r. , robinson a. , & spencer r. , 2000 , , 120 , 2284 s. a. , heckman t. m. , bridle a. , van breugel w. j. m. , & miley g. k. , 1988 , , 68 , 643 s. a. , heckman t. , 1989a , , 336 , 681 s. a. , heckman t. , 1989b , , 336 , 702 a. s. , 1962 , , 68 , 163 p. n. , rttgering h. j. a. , & longair m. s. , 2000 , , 311 , 23 j. a. , 1996 , wfpc2 instrument handbook . stsci , baltimore j. a. , martel a. r. , mcmaster m. , sparks w. b. , baum s. a. , macchetto f. , & mccarthy p. j. , 2002 , new astronomy review , 46 , 181 k. c. , miley g. k. , & van breugel w. , 1987 , , 329 , 604 d. , fanti c. , fanti r. , schilizzi r. t. , & spencer , r. e. , 1995 , , 295 , 27 s. , baum s. a. , sparks w. b. , biretta j. , golombek d. , macchetto f. , mccarthy p. , & miley g. k. , 1996 , , 107 , 621 w. h. , odea c. p. , baum s. a. , sparks w. b. , biretta j. , de koff s. , golombek d. , lehnert m. d. , macchetto f. , mccarthy p. , & miley g. k. , 1997 , , 110 , 191 w. h. , odea c. p. , baum s. a. , & barthel p. d. , 1999 , , 526 , 27 , j. s. & peacock , j. a. , 1993 , , 263 , 936 b. l. , & riley j. m. , 1974 , , 167 , 31p k. j. , best p. n. , rawlings s. , longair m. s. , cotter g. , rttgering h. j. a. , & eales s. , 2002a , , 337 , 1381 k. j. , best p. n. , rttgering h. j. a. , rawlings s. , cotter g. , & longair m. s. , 2002b , , 337 , 1407 k. j. , best p. n. , longair m. s. , & rttgering h. j. a. , 2005 , , 359 , 1393 n. , sparks w. b. , miley g. k. , & macchetto f. , 1995 , , 296 , 339 , a. t. , 1978 , , 223 , 56 a. , odea c. p. , gelderman r. , de vries w. h. , axon d. j. , barthel p. d. , baum s. a. , capetti a. , fanti r. , koekemoer a. m. , morganti r. , & tadhunter c. n. , 2005 , , 436 , 493 j. p. , pooley g. g. & riley , j. m. , 1986 , , 222 , 753 , m. l. & homan , d. c. , 2005 , , 130 , 1389 e. , garrington s. t. , spencer r. e. , akujor c. e. , muxlow t. w. b. , sanghera h. s. , & fanti c. , 1998 , , 299 , 467 , c. d. , 1968 , 145 , 31 a. r. , baum s. a. , sparks w. b. , wyckoff e. , biretta j. a. , golombek d. , macchetto f. d. , mccarthy p. j. , de koff s. , & miley g. k. , 1998 , bulletin of the american astronomical society , 30 , 890 a. r. , baum s. a. , sparks w. b. , wyckoff e. , biretta j. a. , golombek d. , macchetto f. d. , de koff s. , mccarthy p. j. , & miley g. k. , 1999 , , 122 , 81 p. j. , van breugel w. , spinrad h. , & djorgovski s. , 1987 , , 321 , l29 p. j. , 1993 , , 31 , 639 p. j. , spinrad h. , & van breugel w. , 1995 , , 99 , 27 p. j. , & van breugel w. , 1989 , in the epoch of galaxy formation , proceedings of the nato advanced research workshop , dordrecht : kluwer , ed . carlos s. frenk . volume 264 , p.57 p. j. , miley g. k. , de koff s. , baum s. a. , sparks w. b. , golombek d. , biretta j. , & macchetto f. , 1997 , , 112 , 415 e. , rocca - volmerange b. , 2002 , , 383 , 46 l. m. , hardcastle m. j. , & riley , j. m. , 2006 , , 372 , 113 s. g. , roberts l. , & hutchings j. b. , 1995 , , 99 , 349 k. , valtonen m. j. , kotilainen j. , & jaakkola t. , 1993 , , 413 , 453 c. p. , de vries w. h. , koekemoer a. m. , baum s. a. , morganti r. , fanti r. , capetti a. , tadhunter c. n. , barthel p. d. , axon d. j. , & gelderman r. , 2002 , , 123 , 2333 d. e. , 1989 , astrophysics of gaseous nebulae and active galactic nuclei . research supported by the university of california , john simon guggenheim memorial foundation , university of minnesota , et al . mill valley , ca , university science books , 1989 , 422 p. , r. a. , 1982 , , 87 , 859 , r. , gower , a. c. , hutchings , j. b. , talon , s. , duncan , d. , & ross , g. , 1993 , , 86 , 365 s. , saunders r. , 1991 , , 349 , 138 a. , shone d. l. , akujor c. e. , browne i. w. a. , murphy d. w. , pedelty j. , rudnick l. , & walsh , d , 1995 , , 110 , 213 m. , van breugel w. , rttgering h. , de vries w. , stanford s. a. , dey a. , lacy m. , bland - hawthorn j. , dopita m. , & miley , g. , 2003 , , 592 , 755 , m. a. , stockton , a. , lilly , s. j. , hammer , f. , & le fevre , o. , 1992 , , 385 , 61 t. g. , tadhunter c. n. , axon d. j. , & robinson a. , 2000 , , 317 , 922 , 2003 , , 340 , 705 g. , sinha r. p. , & hilldrup , k. , 1984 , , 208 , 813 , d. , 2005 , , 130 , 2513 c. m. , & padovani p. , 1995 , , 107 , 803 r. , rttgering , h. j. a. , miley , g. k. , & hunstead r. w. , 1997 , , 317 , 358 m .- p . , & vron p. , 2003 , , 412 , 399 , e. l. , 2006 , , 118 , 1711w erg @xmath34 @xmath35 . both images have the same levels . radio map from @xcite.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . radio map from @xcite.,title="fig : " ] erg @xmath34 @xmath35 the radio map is from @xcite . ] erg @xmath34 @xmath35 ( note : in radio maps 3c84 is highly core dominated . a worthwhile 5 or 8 ghz radio map could not be located , so only the emission line image is presented . ) ] erg @xmath34 @xmath35 . ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . the radio map courtesy of david floyd ( private communication ) . ] erg @xmath34 @xmath35 . both images have the same levels . the radio map courtesy of david floyd ( private communication).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map courtesy of david floyd ( private communication).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . radio map from @xcite.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . radio map from @xcite.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . radio map from j. p. laing ( image obtained from the 3crr atlas ( http://www.jb.man.ac.uk/atlas).).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . radio map from j. p. laing ( image obtained from the 3crr atlas ( http://www.jb.man.ac.uk/atlas).).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map is from @xcite.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map is from @xcite.,title="fig : " ] erg @xmath34 @xmath35 . the radio map is from @xcite . ] erg @xmath34 @xmath35 . ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels.,title="fig : " ] erg @xmath34 @xmath35 . the radio map was provided by j. p. laing ( image obtained from the 3crr atlas ( http://www.jb.man.ac.uk/atlas ) . ) . ] erg @xmath34 @xmath35 . ] erg @xmath34 @xmath35 . the radio map was provided by j. p. laing ( image obtained from the 3crr atlas ( http://www.jb.man.ac.uk/atlas ) ) . ] erg @xmath34 @xmath35 . both images have the same levels . the radio map was provided courtesy of david floyd ( private communication).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map was provided courtesy of david floyd ( private communication).,title="fig : " ] erg @xmath34 @xmath35 . ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . , title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map was provided by david floyd ( private communication).,title="fig : " ] erg @xmath34 @xmath35 . both images have the same levels . the radio map was provided by david floyd ( private communication).,title="fig : " ] lrrrr 3c124 & 1987 - 08 - 17 & 4.885 ghz . & 0.1 ghz & 0.63 + 3c135 & 1990 - 05 - 25 & 8.465 ghz . & 0.025 ghz & 0.50 + 3c284 & 1985 - 01 - 11 & 4.885 ghz . & 0.1 ghz & 5.76 + 3c303.1 & 1991 - 08 - 11 & 8.415 ghz . & 0.1 ghz & 0.78 + 3c341 & 1996 - 12 - 14 & 8.435 ghz . & 0.1 ghz & 6.87 + 3c368 & 1995 - 07 - 31 & 4.535 ghz . & 0.1 ghz & 0.32 + 3c379.1 & 1987 - 10 - 10 & 4.873 ghz . & 0.05 ghz & 0.78 + 3c382 & 1983 - 09 - 20 & 4.885 ghz . & 0.1 ghz & 3.55 + [ table : vla_obs_params ] crrcccc 3c6.1&0.840&&wf2&[oii]&yes & 300 + 3c17&0.219&-22.137&wf2&[oiii]&yes & 2x140 + 3c31&0.016&&wf2&h@xmath0,[nii]&yes & 2x140 + 3c36&1.301&&wf4&[oii]&yes & 300 + 3c42&0.395&&wf3&[oiii]&yes & 300 + 3c46&0.437&&wf4&[oiii]&yes & 300 + 3c49&0.621&&wf2&[oii]&marginal detection & 300 + 3c55&0.734&&wf3&[oii]&no & 300 + 3c63&0.175&&pc1 ( f588n)&[oiii]&yes & 2x140 + 3c78&0.028&&pc1 ( f673n)&h@xmath0,[nii]&yes & 2x140 + 3c83.1&0.025&&pc1 ( f673n)&h@xmath0,[nii]&yes & 2x140 + 3c84&0.017&&wf2&h@xmath0,[nii]&yes & 2x140 + 3c93&0.358&-23.247&wf2&[oiii]&yes & 2x140 + 3c93.1&0.244&-21.423&wf3&[oiii]&yes & 300 + 3c98&0.030&-20.155&pc1&[oiii]&yes & + 3c103&0.330&&wf2&[oiii]&yes & 300 + 3c109&0.305&-22.971&wf3&[oiii]&yes & 2x140 + 3c111&0.048&-18.564&wf3&[oiii]&no & 2x140 + 3c124&1.083&&wf4&[oii]&yes & 300 + 3c135&0.127&-21.848&wf2&[oiii]&yes & 2x140 + 3c136.1&0.064&&wf4&[oiii]&yes & 2x140 + 3c138&0.759&-24.524&pc1 ( f656n)&[oii]&no & 2x140 + 3c169.1&0.633&&wf2&[oii]&yes & 300 + 3c171&0.238&-21.282&wf2&[oiii]&yes & 300 + 3c172&0.519&&pc1&[oiii]&yes & 300 + 3c175&0.768&-26.795&pc1 ( f658n)&[oii]&yes & 2x140 + 3c184.1&0.118&-21.676&wf2&[oiii]&yes & 2x140 + 3c192&0.059&-19.980&wf3&[oiii]&no & 2x140 + 3c196.1&0.198&&wf4&[oiii]&yes & 2x140 + 3c207&0.684&-24.934&pc1 ( f631n)&[oii]&yes & 2x140 + 3c213.1&0.194&&wf4&[oiii]&yes & 300 + 3c219&0.174&-22.158&pc1 ( f588n)&[oiii]&yes & 2x140 + 3c220.1&0.620&&wf2&[oii]&yes & 300 + 3c223&0.136&-21.950&wf2&[oiii]&yes & 2x140 + 3c227&0.086&-21.610&wf4&[oiii]&yes & 2x140 + 3c234&0.184&-22.463&wf4&[oiii]&yes & 2x140 + 3c244.1&0.428&-23.187&wf4&[oiii]&yes & 300 + 3c249.1&0.311&-25.310&pc1 ( f656n)&[oiii]&yes & 2x140 + 3c263&0.646&-26.612&wf2&[oii]&yes & 2x140 + 3c264&0.020&-20.697&wf2&h@xmath0,[nii]&yes & 2x140 + 3c268.1&0.973&&wf4&[oii]&no & 300 + 3c268.2&0.362&&wf2&[oiii]&yes & 300 + 3c268.3&0.371&-21.479&wf2&[oiii]&yes & 300 + 3c268.4&1.400&-26.602&pc1&[oii]&no & 2x140 + 3c273&0.158&-26.516&wf3&[oiii]&yes & + 3c277.1&0.320&-23.172&pc1&[oiii]&yes & 2x140 + 3c277.3&0.085&&wf4&h@xmath0,[nii]&yes & 2x140 + 3c280&0.997&-22.101&wf3&[oii]&no & 300 + 3c284&0.239&&wf2&[oiii]&yes & 2x140 + 3c289&0.967&&wf4&[oii]&no & + 3c297&1.406&-24.533&pc1&[oii]&yes & 300 + 3c299&0.367&&wf2&[oiii]&yes & 300 + 3c303.1&0.267&-22.147&wf3&[oiii]&yes & 300 + 3c305&0.041&-22.521&wf3&[oiii]&yes & 2x140 + 3c305.1&1.132&&wf3&[oii]&yes & 2x140 + 3c321&0.096&-22.194&wf3&[oiii]&yes & 2x140 + 3c323.1&0.264&-23.929&wf3&[oiii]&yes & + 3c330&0.550&-21.504&wf3&[oii]&yes & 300 + 3c332&0.151&-23.258&wf3&[oiii]&yes & 2x140 + 3c338&0.029&&pc1 ( f673n)&h@xmath0,[nii]&yes & 2x140 + 3c341&0.448&&wf4&[oiii]&yes & 300 + 3c343.1&0.750&-22.632&wf3&[oii]&yes & 300 + 3c349&0.205&&wf2&[oiii]&yes & 300 + 3c352&0.805&&pc1 ( f673n)&[oii]&yes & 300 + 3c356&1.079&-22.815&wf4&[oii]&no & 300 + 3c368&1.130&&wf3&[oii]&yes & 300 + 3c379.1&0.256&&pc1 ( f631n)&[oiii]&yes & 2x140 + 3c380&0.691&-26.302&wf3&[oii]&yes & 2x140 + 3c381&0.160&-22.157&wf3&[oiii]&yes & 2x140 + 3c382&0.058&-21.651&wf3&[oiii]&yes & 2x140 + 3c390.3&0.056&-21.582&wf3&[oiii]&yes & 2x140 + 3c402&0.025&&pc1 ( f673n)&h@xmath0,[nii]&no & 2x140 + 3c433&0.101&-20.512&wf3&[oiii]&yes & 2x140 + 3c436&0.215&&wf2&[oiii]&yes & 2x140 + 3c445&0.056&-21.192&wf3&[oiii]&no & 2x140 + 3c449&0.017&&wf2&h@xmath0,[nii]&yes & 2x140 + 3c452&0.081&-21.243&wf4&[oiii]&yes & 2x140 + 3c454.3&0.860&-27.601&pc1&[oii]&yes & 2x140 + 3c458&0.290&&wf3&[oiii]&yes & 300 + 3c465&0.029&-22.190&pc1 ( f673n)&h@xmath0,[nii]&yes & 2x140 + [ table : optical_obs ] crrrrcrrr 3c6.1&&&&&&partially res . & & + 3c17&&41&0.35&41.2&&partially res.&42 & 2 + 3c31&&&1.01&38.4&&partially res.&144 & 4 + 3c42&&&0.06&41.0&&partially res.&153 & 2 + 3c46&2.13&26&0.29&41.8&17&extended&177 & 2 + 3c83.1&&&&&&partially res.&166 & 4 + 3c84&15.67&see notes&59.14&40.3&180&extended&100 & 4 + 3c93.1&&65&0.35&41.3&&partially res.&132 & 2 + 3c93&&95&0.19&41.4&&partially res . & & + 3c103&0.87&49&0.01&39.9&61&partially res.&34 & 2 + 3c109&2.34&25&1.72&42.2&39&extended&145 & 1 + 3c124&1.08&4&0.13&41.8&15&extended & & + 3c135&3.71&54&1.05&41.1&35&extended&141 & 2 + 3c136.1&0.53&175&0.00013&36.6&35&partially res.&117 & 4 + 3c169.1&&&0.33&41.7&&partially res . & & + 3c171&3.44&85&1.38&41.8&23&extended&165 & 1 + 3c172&0.73&16&0.07&41.4&77&partially res . & & + 3c184.1&2.64&24&0.01&38.7&&partially res.&40 & 2 + 3c196.1&0.67&53&0.06&40.3&15&partially res.&57 & 1 + 3c213.1&&&0.08&40.4&&partially res.&161 & 2 + 3c220.1&&&0.06&40.9&&partially res . & & + 3c223&2.1&128&2.29&41.5&43&extended&50 & 1 + 3c234&2.37&79&9.85&42.4&32&extended&80 & 2 + 3c244.1&3.47&133&0.66&42.1&25&partially res.&76 & 2 + 3c249.1&2.08&80&2.27&42.3&112&extended & & + 3c264&&&0.49&38.3&&partially res.&152 & 1 + 3c268.2&1.8&15&0.44&41.8&56&extended&164 & 2 + 3c268.3&3.2&146&0.42&41.8&26&extended&144 & 2 + 3c277.1&1.85&124&1.4&42.1&52&partially res.&126 & 3 + 3c277.3&&&0.21&39.2&0&partially res.&170 & 4 + 3c284&2.9&74&0.56&41.4&30&extended&151 & 2 + 3c299&4.24&62&1.19&42.2&70&extended&47 & 2 + 3c303.1&1.56&140&0.95&41.8&38&extended&169 & 2 + 3c305&6.37&46&13.06&41.2&35&extended&74 & 1 + 3c305.1&&33&0.05&41.4&0&partially res.&24 & 3 + 3c321&6.21&108&5.8&41.6&42&extended&18 & 1 + 3c330&1&103&0.06&40.8&47&partially res . & & + 3c341&1.03&172&0.27&41.8&55&extended&17 & 2 + 3c349&&36&0.16&40.7&61&partially res.&14 & 2 + 3c352&&8&0.05&41.1&0&partially res . & & + 3c368&6.74&24&0.59&42.5&24&extended & & + 3c379.1&0.72&43&0.06&40.6&15&extended & & + 3c380&&&0.56&42.0&0&partially res.&131 & 3 + 3c381&5.69&150&2.35&41.7&31&extended&156 & 2 + 3c382&2.14&112&13.16&41.5&49&partially res.&85 & 4 + 3c390.3&0&62&11.34&41.4&0&partially res.&97 & 1 + 3c433&5.87&135&1.49&41.1&55&partially res.&147 & 1 + 3c436&&&0.26&41.0&0&partially res.&3 & 2 + 3c449&1.7&174&0.37&38.0&44&partially res.&1 & 4 + 3c452&1.58&125&0.71&40.5&30&partially res.&101 & 4 + 3c454.3&&&0.86&42.2&&partially res . & & + 3c458&&75&0.07&40.7&&partially res . & & + [ table : optical_data ] ccrcrcr 3c6.1 & fr ii & 25.8 & 26 & 13 & 35.7 + 3c17 & fr ii & 30.0 & 147 & 5 & 34.4 + 3c31 & fr i & 1833.0 & 159 & 11 & 32.0 + 3c36 & fr ii & 9.0 & 20 & 13 & 35.9 + 3c42 & fr ii & 28.0 & 132 & 5 & 34.8 + 3c46 & fr ii & 163.0 & 68 & 5 & 34.8 + 3c49 & css & 0.98 & 84 & 6 & 35.2 + 3c55 & fr ii & 69.0 & 94 & 13 & 35.7 + 3c63 & fr ii & 22.0 & 34 & 5 & 34.2 + 3c78 & fr i & 210.0 & 51 & 11 & 32.5 + 3c83.1 & fr i & 680.0 & 96 & 11 & 32.6 + 3c84 & fr i & 492.0 & 162 & 11 & 32.6 + 3c93 & fr ii & 34.7 & 44 & 17,99 & 34.8 + 3c93.1 & css & 0.3 & 165 & 4,5 & 34.2 + 3c103 & fr ii & 88 & 159 & 5,8 & 35.0 + 3c109 & fr ii & 96.0 & 143 & 5 & 34.8 + 3c111 & fr ii & 220.0 & 62 & 11 & 33.5 + 3c124 & css & 1.3 & 7 & 13 & 35.8 + 3c135 & fr ii & 130.0 & 70 & 5 & 33.9 + 3c136.1 & fr ii & 522.0 & 108 & 11 & 33.1 + 3c138 & css & 0.6 & 70 & 1 & 35.8 + 3c169.1 & fr ii & 38.0 & 137 & 8 & 35.1 + 3c171 & fr ii ( dtsb ) & 30.0 & 99 & 2 & 34.5 + 3c172 & fr ii & 103.0 & 37 & 13 & 35.2 + 3c175 & fr ii & 48.0 & 55 & 10,99 & 35.7 + 3c184.1 & fr ii & 167.0 & 157 & 3 & 33.7 + 3c192 & fr ii & 192.0 & 124 & 3 & 33.2 + 3c196.1 & css & 4.0 & 43 & 5 & 34.3 + 3c207 & fr ii & 11.4 & 90 & 14 & 35.4 + 3c213.1 & fr ii & 43.0 & 162 & 5 & 33.8 + 3c219 & fr ii & 184.0 & 40 & 5 & 34.5 + 3c220.1 & fr ii & 29.7 & 79 & 13 & 35.4 + 3c223 & fr ii & 300.0 & 164 & 5 & 33.8 + 3c227 & fr ii & 246.0 & 86 & 11 & 33.7 + 3c234 & fr ii & 110.0 & 64 & 5 & 34.5 + 3c244.1 & fr ii & 52.0 & 168 & 5 & 35.1 + 3c249.1 & fr ii & 12.7 & 54 & 99 & 34.5 + 3c263 & fr ii & 43.1 & 120 & 19 & 35.4 + 3c264 & fr i & 91.0 & 40 & 11 & 32.4 + 3c268.1 & fr ii & 46.0 & 83 & 13 & 36.0 + 3c268.2 & fr ii & 96.0 & 21 & 5 & 34.6 + 3c268.3 & css & 1.0 & 161 & 5 & 34.7 + 3c268.4 & fr ii & 10.2 & 42 & 2,18 & 36.1 + 3c277.1 & css & 1.5 & 131 & 6,99 & 34.4 + 3c277.3 & fr ii & 29.0 & 158 & 11 & 33.2 + 3c280 & fr ii & 12.9 & 82 & 1,10 & 36.1 + 3c284 & fr ii & 176.0 & 101 & 5 & 34.3 + 3c297 & & 4.0 & 167 & 13 & 36.1 + 3c299 & fr ii & 12.0 & 60 & 5 & 34.7 + 3c303.1 & css & 1.9 & 130 & 6 & 34.2 + 3c305 & css & 14.0 & 44 & 10 & 32.8 + 3c305.1 & fr ii & 7.9 & 11 & 13 & 35.5 + 3c321 & fr ii & 309.0 & 131 & 3 & 33.5 + 3c330 & fr ii & 62.0 & 62 & 13 & 35.5 + 3c332 & fr ii & 81.0 & 20 & 5 & 33.8 + 3c338 & fr i & 115.0 & 85 & 11 & 32.9 + 3c341 & fr ii & 71.0 & 50 & 5 & 34.9 + 3c343.1 & css & 1.0 & 97 & 13 & 35.5 + 3c349 & fr ii & 82.0 & 142 & 5 & 34.2 + 3c352 & fr ii & 10.2 & 164 & 13 & 35.5 + 3c356 & fr ii & 72.0 & 162 & 13 & 35.9 + 3c368 & fr ii & 8.5 & 17 & 7,13 & 36.0 + 3c379.1 & fr ii & 76.0 & 161 & 5 & 34.2 + 3c380 & css & 1.0 & 145 & 2,9 & 36.1 + 3c381 & fr ii & 69.0 & 4 & 5 & 34.1 + 3c382 & fr ii & 179.0 & 50 & 11 & 33.2 + 3c390.3 & fr ii & 231.0 & 145 & 3 & 33.5 + 3c402 & fr i & 528.0 & 163 & 11 & 32.1 + 3c433 & fr ii & 58.0 & 172 & 5 & 34.2 + 3c436 & fr ii & 105.0 & 172 & 5,10 & 34.4 + 3c445 & fr ii & 576.0 & 171 & 11 & 33.3 + 3c449 & fr i & 1742.0 & 10 & 11 & 31.9 + 3c452 & fr ii & 277.0 & 79 & 11 & 33.9 + 3c454.3 & css & 10.0 & 128 & 2,16 & 35.7 + 3c458 & fr ii & 161 & 75 & 12,15 & 34.6 + 3c465 & fr i & 375.0 & 125 & 11 & 32.9 + [ table : radio_data ] lccccccc emission line luminosity vs. delta pa&-0.6577&0.040&-0.474&0.0044 + emission line luminosity vs. delta pa ( @xmath1)&-0.5417&0.0258&-0.404&0.0222 + emission line luminosity vs delta pa ( @xmath4 ) & -0.6338 & 0.0219 & & + emission line luminosity vs. host mag.&-0.8734&0.0003&-0.611&0.0005 + emission line luminosity vs. radio luminosity&1.095&0.0000&0.740&0.0000 + emission line nebulae size vs. radio luminosity&0.7177&0.0039&0.487&0.0067 + emission line nebulae size vs. radio luminosity ( @xmath1)&0.6483&0.0119&0.442&0.0172 + emission line nebulae size vs. z&0.6976&0.0050&0.499&0.0055 + emission line nebulae size vs. z ( @xmath1)&0.6253&0.0152&0.458&0.0136 + delta pa vs. radio size&0.6050 & 0.0105&0.455&0.0080 + delta pa vs. size ratio&1.0099&0.0001&0.684&0.0003 + [ table : correlations ] lccccc delta pa vs. z&-0.3874&0.0913&-0.238&0.1540 + delta pa vs. emission line nebulae asymmetry&-0.3126&0.2243&-0.222&0.2327 + emission line nebulae luminosity vs radio size&-0.3624&0.0423&-0.280&0.0330 + emission line nebulae size vs radio size&-0.3484&0.1686&-0.283&0.1216 + [ table : non - correlations ] lccc this paper ( hst ) , @xmath36 & 7 & 0.169 & 98.8 + this paper ( hst ) , @xmath37 & 13 & 0.157 & 90.5 + this paper ( hst ) , @xmath38 & 26 & 0.152 & 58.8 + @xcite , @xmath36 & 15 & 0.171 & 77.0 + @xcite , @xmath37 & 12 & 0.44 & 1.90 + @xcite , @xmath38 & 30 & 0.343 & 0.168 + [ table : ks - results ] 3c46 ( figs . [ 3c46-montage ] & [ 3c46-overlay ] ) is a large fr ii radio source . the emission line nebula and the radio source are not well aligned . the optical continuum image seems to show a double structure . however , the line emission is predominantly in the southern portion . 3c84 ( figs . [ 3c84-montage ] & [ 3c84-overlay ] ) has a diffraction cross in the emission line image . a sufficiently deep 5 or 8 ghz radio map was not available and could not be found in archival data , so an overlay is not shown . the line emission is symmetrically distributed within the host galaxy so there is no meaningful position angle . 3c124 ( figs [ 3c124-montage ] & [ 3c124-overlay ] ) has an emission line region which is very closely aligned with the radio source . the exact center of the host galaxy is not well determined for our images . 3c135 ( figs [ 3c135-montage ] & [ 3c135-overlay ] ) features an emission line region which is closely aligned with the overall radio source . the radio source is very large compared to the emission line region . the emission line region is elongated along the radio jet axis and features a disconnected region of emission towards the northwest hotspot . 3c171 ( figs [ 3c171-montage ] & [ 3c171-overlay ] ) the emission line region traces the jets , linking the core to the hotspots . there is emission around the east hotspot . the elr structure has been discussed in detail in @xcite . 3c234 ( figs . [ 3c234-montage ] & [ 3c234-overlay ] ) is aligned with the radio source and the core shows slight extension along the major axis of the nebula . there are also filaments extending almost perpendicular to the radio axis . the broad lines in 3c234 are highly polarized so it is not surprising the properties of this source are similar to that of nlrgs @xcite . 3c249.1 ( figs [ 3c249.1-montage ] & [ 3c249.1-overlay ] ) has an elr which does not have a preferential alignment with the radio source . it features two regions of emission which are disconnected from the nuclear emission at the radio core . 3c268.2 ( figs . [ 3c268.2-montage ] & [ 3c268.2-overlay ] ) shows an emission line nebula which is closely aligned with the radio source . the elr is asymmetrical around the center of the host galaxy ( measured from the broadband continuum image ) . there is little to no line emission detected from the nucleus in our images . the core was not detected in the radio maps , however the location was inferred using the wcs coordinates given on ned . as the core was not detected in the radio map , we are unable to examine spatial coincidence of the elr and the radio core , but are still able to measure the alignment of the elr and radio source . 3c268.3 ( figs . [ 3c268.3-montage ] & [ 3c268.3-overlay ] ) is an aligned css source @xcite . the emission extends beyond the radio source . registration of the radio core and the emission line image was problematic due to the lower resolution of the optical image . 3c268.3 shows broad lines , indicating this source may not be in the plane of the sky . 3c284 ( figs . [ 3c284-montage ] & [ 3c284-overlay ] ) has and elr which is composed of three separate components in our images . the pa along the overall structure is not aligned with the source . however , taking an `` inner pa '' of the central structure gives the alignment with the radio source , so we consider this elr aligned with the radio source . 3c303.1 ( figs . [ 3c303.1-montage ] & [ 3c303.1-overlay ] ) has an emission line nebula which is of similar size in our wfpc images as the radio source . they are closely aligned . the emission line region is symmetrical about the nucleus and is elongated along the radio axis . 3c305 ( figs . [ 3c305-montage ] & [ 3c305-overlay ] ) the emission line region is aligned with the jets and hotspots . however , the emission extends beyond the extent of the radio source , limiting the probability of shock ionization as a primary mechanism for transferring energy to ambient gas @xcite . 3c341 ( figs . [ 3c341-montage ] & [ 3c341-overlay ] ) does not show alignment between the emission line nebula and the radio source . the size of the elr is comparable to a resolution element in the radio map , making examination of spatial coincidence problematic .

we study the one - loop sensitivity of the ( ) vertex to the new massive gauge bosons predicted by the minimal model , which have unusual couplings to the standard model ( sm ) gauge bosons . a gauge - fixing procedure covariant under the group was introduced for these new gauge bosons ( dubbed bileptons ) in order to generate gauge - invariant green functions . the similarities between this procedure and the nonconventional quantization scheme of the background field method are discussed . it is found that , for relatively light bileptons , with a mass ranging from to , the radiative corrections to the form factors associated with the vertex can be of the same order of magnitude than the sm one . in the case of heavier bileptons , their contribution is smaller by about one and two orders of magnitude than their sm counterpart .

hep-ph0508166

the @xmath0 ( @xmath6 ) one - loop structure has been considerably studied in the literature not just because it may constitute a mechanism through which physics beyond the fermi scale may show up , but also due to some theoretical issues concerning its dependence on the gauge - fixing scheme . it turns out that the conventional gauge - fixing procedures give rise to ill - behaved off - shell green functions that may display inadequate properties such as a nontrivial dependence on the gauge - fixing parameter , an increase larger than the one observed in physical amplitudes at high energies , and the appearance of unphysical thresholds . the on - shell green functions can represent physical amplitudes as they are independent on the gauge - fixing procedure , such as occurs with the static electromagnetic properties of the @xmath7 boson @xcite , but gauge independence is lost if at least one external particle becomes virtual . although off - shell green functions are generally gauge dependent , the @xmath8-matrix elements to which they contribute must be gauge independent . this is the case of the off - shell @xmath0 vertex , which is just a piece of some physical process such as the @xmath9 and @xmath10 reactions . nonetheless , it would be interesting if one was able to study the sensitivity to radiative corrections of the @xmath0 coupling , and other sm couplings as well , without invoking some particular @xmath8-matrix element . the concepts of gauge invariance and gauge independence are two essential ingredients of gauge systems , though the former is not necessarily present at the quantum level . while gauge invariance plays a central role when defining the classical action of the system , once the latter is quantized one must invariably invoke an appropriate gauge - fixing procedure to define a nondegenerate action , which means that gauge invariance is to be broken explicitly . the resultant action is not gauge invariant , though it is invariant under brst symmetry @xcite . as a consequence , the green functions derived from this action can not satisfy simple ( qed - like ) ward identities , but they do satisfy more elaborate slavnov - taylor identities that are dictated by brst symmetry . also , green functions contain much unphysical information that is removed provided a physical observable is considered . contrary to green functions , which are highly dependent on the gauge - fixing procedure , physical amplitudes have no such dependence , thereby being gauge independent . there are thus some subtle mechanisms that conspire to produce nontrivial cancelations between the green functions defining a physical observable . it is clear that a nonconventional quantization scheme must be applied in order to generate gauge - invariant green functions , which in turn can be obtained from a gauge - invariant quantum action @xmath11 . in this respect , the background - field - method ( bfm ) @xcite is meant to construct manifest gauge - invariant quantum actions from which well - behaved green functions satisfying simple ward identities can be derived . this method , implemented at the level of generating functionals , relies on the decomposition of the gauge fields into two parts : the quantum field @xmath12 and the background ( classical ) field @xmath13 , i.e. @xmath14 . in the generating functional only the quantum fields are integrated out , whereas the background fields are treated as sources . this means that only the quantum fields can circulate inside the loops . this method allows one to introduce a gauge - fixing procedure for the quantum fields without spoiling the gauge invariance of the quantum action with respect to the classical fields . although it is necessary to define a gauge - fixing procedure for both the quantum and the classical fields in order to define @xmath8-matrix elements , only a gauge - fixing scheme for the quantum gauge fields is required to define general off - shell green functions . the quantum action is invariant under ordinary gauge transformations of the classical fields , while the quantum fields transform as the adjoint representation of the group in consideration . in other words , the so constructed action @xmath15 $ ] is degenerate with respect to the background fields but nondegenerate with respect to the quantum fields . the green functions derived from the quantum action @xmath16 $ ] are gauge invariant in the sense that they satisfy simple ward identities , but it is worth stressing that they are still dependent on the gauge parameter @xmath17 that characterizes the gauge - fixing scheme used for the quantum fields , and so there is no gauge independence . the bfm has proved useful in many applications @xcite , simplifying both technically and conceptually the calculation of radiative corrections . as already mentioned , green functions arising from a conventional quantum action ( a brst - invariant but gauge - noninvariant one ) contain a lot of unphysical information that is removed once they are inserted into some physical observable . some of this unwanted information can be removed at the level of the generating functional through the bfm formalism , which allows one to construct a gauge - invariant quantum action from which gauge - invariant green functions can be obtained . although the resultant green functions satisfy simple ward identities , they are not gauge independent . thus far there is still no known mechanism yielding both gauge - invariant and gauge - independent green functions directly from the generating functional , although there is already a diagrammatic method meant for this purpose , the so - called pinch technique ( pt ) @xcite . this method relies on constructing well - behaved green functions by combining some individual contributions from self - energies , vertex , and box diagrams , which usually appear in physical processes . in general , the feynman rules used in this diagrammatic approach are derived from a conventional effective action , though those derived from the bfm have also been used for a deeper study of the method self - consistence @xcite . although the pt was first introduced for the study of pure yang - mills theories at the one - loop level @xcite , it has already been applied to theories with spontaneous symmetry breaking ( ssb ) @xcite , including the study of self - energies @xcite and trilinear vertices @xcite involving the electroweak gauge bosons . a complete calculation of the one - loop contribution to the @xmath0 vertex from the electroweak bosons was presented in refs . @xcite : it was intended to show that the vertex functions satisfy a simple ward identity , which establishes a relationship between this vertex and the @xmath7 self - energy . more specifically , ref . @xcite discusses the gauge independence of the form factors associated with the @xmath18 vertex for off - shell photon and on - shell @xmath7 bosons . afterwards , an important connection between the pt and the bfm was established @xcite at the one - loop level by showing that the green functions calculated via the bfm feynman rules coincide with those obtained through the pt for the specific value @xmath19 . more recently , the pt was extended to the two - loop level in the context of both the yang - mills @xcite and the electroweak sectors @xcite , and the one - loop connection to the bfm was established too . a step toward a nondiagrammatic formulation of the pt via the powerful batalin - vilkovisky quantization method @xcite was presented in ref . this framework was used to generalize the pt at any order of perturbation theory @xcite , and it was meant to show that the link between the pt green functions and those obtained via the bfm along with the feynman - thooft gauge remains at all orders of perturbation theory @xcite . the reason for such a link remains a puzzle , though it is worth noting that the feynman - thooft gauge yields no unphysical thresholds . establishing such a connection at any order of perturbation theory is very important for practical purposes because one can simply use the bfm feynman - thooft gauge ( bfmfg ) to calculate gauge - independent off - shell amplitudes , which happens to be much less cumbersome than the use of the pt . although in conventional quantization schemes the quantum action of the theory is not gauge invariant , it is still possible to introduce gauge invariance with respect to a subgroup of such a theory . this scheme is particulary useful when the quantum fluctuations of the gauge fields associated with this subgroup are deemed negligible . for instance , it would be interesting to assess the virtual effects of the heavy physics lying beyond the fermi scale on the sm green functions in a @xmath3-covariant manner , in which case it is only necessary to introduce a quantization scheme for the heavy fields since the sm fields would only appear as external legs . this is indeed the philosophy behind the effective lagrangian approach widely used in the context of the electroweak theory , where it is assumed that the new physics effects must respect the @xmath3 symmetry . in a specific theory beyond the sm , a @xmath3-invariant effective lagrangian can be constructed by introducing a @xmath3-covariant gauge - fixing procedure for the heavy gauge bosons in order to integrate them out in the generating functional . in analogy with the bfm , the gauge - fixing procedure for the heavy gauge fields must involve the @xmath3-covariant derivative given in the representation in which the heavy fields transform under this group . this is the reason why such gauges , which were first introduced by fujikawa in the context of the sm @xcite , are called nonlinear or covariant gauges . in this case , the @xmath7 propagator is defined in a covariant way under the electromagnetic @xmath20 group , so the vertex functions associated with the @xmath18 interaction and the @xmath7 self - energy satisfy a simple ward identity . the most general renormalizable structure of this gauge - fixing procedure has been discussed from the brst - symmetry standpoint in @xcite , and a discussion about the difficulties on implementing the faddeev - popov method ( fpm ) has been presented too @xcite . this gauge - fixing procedure has proved a valuable tool in radiative corrections as it simplifies considerably the loop calculations @xcite . the method has also been used to quantize yang - mills theories without ssb @xcite . we will show below that , within some specific models , it is possible to use a nonlinear gauge to parametrize in an @xmath3-invariant way the impact of new physics on the sm green functions . in particular , we will show that it is possible to use this class of gauges to estimate the one - loop effects of new heavy gauge bosons on the @xmath0 vertex . we are interested in the sensitivity of the @xmath0 vertex to the new heavy gauge bosons predicted by the so - called minimal @xmath21 model @xcite , which is based on the @xmath22 gauge group . apart from predicting signals of new physics at the tev scale , this model introduces unique features that have been the focus of great interest recently @xcite such as a possible approach to the solution of the family replication problem . in this model , the lepton spectrum is the same as the sm one , but it is accommodated in @xmath23 antitriplets ; the quark sector is also arranged in the fundamental representation of this group , which requires the introduction of three new quarks . in order to endow all the particles with mass , a higgs sector composed by three triplets and one sextet of @xmath23 is required , though only one of the triplets is needed to break down @xmath24 into @xmath3 at the new physics scale @xmath25 . in the first stage of ssb , there emerge singly and doubly charged gauge bosons in a doublet of the @xmath26 group , as well as a new neutral boson @xmath27 . the new charged gauge bosons were dubbed bileptons because they carry two units of lepton number . the three exotic quarks and a cp even higgs boson do not couple to the @xmath7 gauge boson since they emerge as singlets of @xmath26 , and get their mass at the @xmath28 scale . thus , at this scale , the @xmath18 and @xmath29 vertices can only receive contributions from the bileptons . the fact that the @xmath26 group is totally embedded in @xmath23 gives rise to unusual couplings between the bileptons and the sm gauge fields , which arise via the electroweak covariant derivative since the bileptons transform as the fundamental representation of @xmath26 . it turns out that these couplings do not involve any mixing angle and are similar both in strength and lorentz structure to those couplings existing between the sm gauge bosons themselves , as opposed to the gauge bosons appearing in other sm extensions . our main goal is to estimate , in a @xmath3-invariant way , the sensitivity of the @xmath0 vertex to the bileptons . to this end , we introduce a @xmath3-covariant gauge - fixing procedure for the bileptons , which leads to an invariant quantum action . we will show below that the resulting @xmath18 and @xmath29 green functions are gauge invariant and satisfy simple ward identities . another feature worthwhile to emphasize is that the fpm fails when it is attempted to be used in conjunction with this class of gauges : the resultant theory is not renormalizable @xcite . instead of using this method , we will present a discussion based on brst symmetry @xcite , which is a powerful formalism adequate not only to quantize yang - mills theories with broader gauge - fixing procedures , as the nonlinear ones , but also to quantize more general gauge systems . as we will see below , our quantization scheme incorporates the main ingredient of the bfm , namely , the gauge invariance of the quantum action , which turns loop calculations into a somewhat simple task . the rest of the paper has been organized as follows . in sec . [ model ] a brief description of the minimal @xmath21 model is presented . particular emphasis is given to the yang - mills sector . [ gauge ] a @xmath3-covariant gauge - fixing procedure for the bileptons is presented along with a discussion on the advantages of using the brst formalism instead of the fpm . [ cal ] is devoted to present the one - loop amplitudes for the @xmath18 and @xmath29 vertices , whereas in secs . [ dis ] and [ con ] we discuss our results and present the conclusions . the @xmath30 model has been discussed to some extent in the literature @xcite . we will only focus on those features that are relevant for the present discussion . in particular , we will concentrate on the first stage of ssb , when the @xmath21 group is broken down into the sm group . the complete higgs sector is comprised by three triplets and one sextet of @xmath23 , but only the following triplet is necessary to break @xmath2 into @xmath3 : @xmath31 where @xmath32 is a doublet of @xmath3 with hypercharge 3 . in the fundamental representation of @xmath2 , the covariant derivative can be written as @xmath33 where @xmath34 @xmath35 are the gell - mann matrices and @xmath36 . the generators are normalized as @xmath37 . when @xmath38 develops a vacuum expectation value , @xmath39 , the exotic quarks , one physical neutral scalar , and the new gauge bosons acquire masses , whereas the remaining scalar multiplets give mass to the sm particles . the first stage of ssb is accomplished by @xmath40 according to the following scheme : six generators are broken ( @xmath41 for @xmath42 ) and the remaining ones leave invariant the vacuum ( @xmath43 for @xmath44 ) . notice that @xmath45 , so the hypercharge can be identified with the following linear combination of broken generators : @xmath46 . at this stage of ssb there appear one single charged bilepton and one doubly charged one defined by @xmath47 which have the following mass @xmath48 according to the quantum number assignment , these fields accommodate in one doublet of @xmath3 with hypercharge 3 : @xmath49 as far as the scalar triplet is concerned , the two components of the @xmath3 doublet with hypercharge 3 , @xmath32 , coincide with the pseudo - goldstone bosons associated with the bilepton doublet : @xmath50 finally , the third component of the triplet contains the physical higgs boson and the pseudo - goldstone boson associated with @xmath27 : @xmath51 the bilepton masses receive new contributions at the fermi scale , when @xmath3 is broken into @xmath20 , which yields an upper bound on the splitting between the square bilepton masses : @xmath52 therefore , @xmath53 and @xmath54 can not not be very different : one of them can not become arbitrarily large while the other one remains fixed . in fact , the bilepton masses become nearly degenerate when they are much larger than @xmath55 . in addition , the theoretical constraint @xmath56 obtained from matching the gauge couplings constants at the @xmath2 breaking scale yields an upper bound on the bilepton masses of the order of 1 tev @xcite . therefore , our estimate for the green functions associated with the @xmath0 vertex at the @xmath28 scale would not become spoiled by the new contributions arising at the @xmath57 scale . as far as the remaining gauge bosons are concerned , the gauge fields @xmath58 and @xmath59 mix to produce a massive field @xmath60 , and a massless gauge boson @xmath61@xcite . the latter is associated with the @xmath62 group . at the @xmath28 scale , the @xmath27 field does not couple with the @xmath7 boson , though it can couple to a @xmath7 boson pair at the fermi scale via @xmath63 mixing @xcite . therefore , although five massive gauge bosons emerge at the @xmath28 scale , along with three exotic quarks and one higgs boson , only the bileptons couple to the @xmath7 gauge boson . these interactions are dictated by the @xmath3 symmetry , and emerge entirely from the yang - mills sector associated with the @xmath64 group . the corresponding lagrangian is composed of the following three @xmath3-invariant pieces @xcite : @xmath65 where @xmath66 is the electroweak yang - mills lagrangian : @xmath67 where we have made the association @xmath68 , for @xmath44 . @xmath69 encompasses the interactions between the sm gauge bosons and the new ones , it can be written as @xmath70,\end{aligned}\ ] ] where we have introduced the definitions @xmath71 and @xmath72 . in addition , @xmath73 stands for the covariant derivative associated with the electroweak group . the first two terms of this lagrangian induce a diversity of couplings between the sm gauge bosons and the bileptons . finally , the term @xmath74 induces the interactions between the @xmath27 boson and the bileptons : @xmath75 note that each term in the last two lagrangians is separately invariant under the electroweak group . as evident from above , the one - loop level contributions to the @xmath18 and @xmath29 vertices arise only from the first two terms appearing in @xmath69 . a gauge - fixing procedure covariant under the @xmath3 group will be introduced below for the bilepton sector . it has already been mentioned that the fpm fails to quantize yang - mills theories possessing more general supplementary conditions than the linear ones . this stems from the fact that the fpm leads to an action which is bilinear in the ghost and antighost fields since they arise essentially from the integral representation of a determinant . this is not however the most general situation that can arise since an action including quartic ghost interactions at the tree level is still consistent with brst symmetry and the power counting criterion of renormalization theory . it turns out that the fpm does succeed when applied to linear gauges because quartic ghost interactions can not arise from loop effects due to antighost translation invariance , , with @xmath76 arbitrary constant parameters . ] which stems from the fact that the antighost fields appear only through their derivatives . however , this symmetry is lost in the case of nonlinear gauges since the gauge - fixing functions depend on bilinear terms of gauge fields . these terms are responsible for the presence of ultraviolet - divergent quartic - ghost interactions at one - loop level . this means that renormalizability becomes ruined when the fpm is attempted to be used in the context of nonlinear gauges . it is thus convenient to discard the fpm and building up instead the most general action consistent with brst symmetry and renormalization theory . brst symmetry arises naturally from the field - antifield formalism @xcite , which has proved a powerful tool in quantizing gauge systems . in the case of yang - mills systems , the gauge - fixed brst action has a simple structure @xcite : @xmath77 where @xmath78 is the gauge - invariant classical action , @xmath79 is the brst operator , and @xmath80 is the fermion action . @xmath81 would be nondegenerate if a gauge - fixing procedure for all the gauge fields of the model was introduced . since we are only interested in the virtual effects of the bileptons , a gauge - fixing procedure for these fields is only necessary . furthermore , we use a gauge - fixing procedure covariant under the @xmath3 group because we are interested in preserving such a symmetry . the resultant @xmath81 action is nondegenerate with respect to the bilepton fields , boson can be introduced without affecting the @xmath82-invariance of @xmath81 . ] but degenerate with respect to the electroweak fields . as a consequence , a @xmath82-invariant quantum action can be constructed out of which gauge - invariant green functions , @xmath83 and @xmath84 , satisfying simple ward identities , can be derived . more specifically , we introduce a fermion action defined as follows : @xmath85 , \ \ \ \bar{a}=4,5,6,7 ; \ \ \ b , c=1,\cdots 8,\ ] ] where @xmath86 , @xmath87 , and @xmath88 are the gauge - fixing functions , the antighost fields , and the auxiliary scalar fields associated with the @xmath89 gauge fields , respectively . in addition , @xmath90 are the ghost fields associated with the @xmath12 fields , @xmath91 are the @xmath23 structure constants , and @xmath92 is the gauge parameter . note that the @xmath93 term can not arise from the fpm , though its presence is necessary to obtain renormalizability when the gauge - fixing functions are nonlinear . using the usual brst transformations , we obtain for the @xmath80 variation @xmath94.\ ] ] on the other hand , since the auxiliary fields @xmath88 appear quadratically , they can be integrated out in the generating functional . since the coefficients of the quadratic terms do not depend on the fields , their integration is equivalent to applying the equations of motion to the gauge - fixed brst action . once these steps are done , we obtain an action defined by the following @xmath3-invariant lagrangian @xmath95 where @xmath96 is the gauge invariant lagrangian of the @xmath21 model , whereas @xmath97 and @xmath98 arise from the action @xmath99 . the former is the gauge - fixing term , which can be written as @xmath100 and @xmath98 represents the ghost sector : @xmath101 while the first term in this lagrangian does arise when the fpm is used , the remaining ones are new and must be preserved if a nonlinear function @xmath86 is introduced . if a linear gauge is used , these terms can be removed after invoking antighost - translation invariance . we are now ready to introduce the most general @xmath82-covariant @xmath86 functions , and we will take advantage of the fact that every coupling involving at least one pseudo - goldstone boson and every coupling with gauge freedom can be modified leaving unaltered the @xmath8 matrix . the most general renormalizable gauge - fixing functions consistent with this symmetry can be written as @xmath102 notice that these gauge - fixing functions are nonlinear in both the vector and the scalar sectors . to be fully aware of the covariant structure of these gauge - fixing functions , it is convenient to express them in terms of the mass eigenstates fields . using the definitions @xmath103 we can write @xmath104 where @xmath105 is the @xmath3-covariant derivative given in the doublet representation . from this expression , it is evident that @xmath106 transforms as @xmath107 or @xmath32 , i.e. as a doublet of @xmath3 with hypercharge 3 . as a consequence , the gauge - fixing term @xmath97 is manifestly invariant under this group . as will become evident below , it is convenient to decompose the lagrangian @xmath97 into three gauge - invariant terms : @xmath108 where @xmath109,\\ { \cal l}_{gf3}&=&\frac{ig\sqrt{3}\sqrt{1 - 4s^2_w}}{2c_w\xi}z'_\mu [ ( d_\nu y^\nu)^\dag y^\mu - y^{\mu \dag}(d_\nu y^\nu)]\nonumber \\ & & -\frac{g^2\sqrt{3}\sqrt{1 - 4s^2_w}}{2\sqrt{2}c_w}z'_\mu(\phi^{0*}y^{\mu \dag}\phi_y+\phi^0 \phi^\dag_y y^\mu)\nonumber \\ & & -\frac{3g^2(1 - 4s^2_w)}{4c^2_w\xi}z'_\mu z'_\nu y^{\mu \dag}y^\nu.\end{aligned}\ ] ] we would like to discuss the dynamics of @xmath110 and @xmath111 . the first term appearing in @xmath110 , which is invariant under the electroweak group , not only allows to define the bilepton propagators but also modifies nontrivially the couplings between the bileptons and the electroweak gauge bosons appearing in the @xmath112 lagrangian . when the two lagrangians are combined , they lead to trilinear and quartic vertices : @xmath113\big \},\end{aligned}\ ] ] @xmath114+\nonumber \\ & & q^v_{y^{++}}\big [ v^\mu ( y^{--}_{\mu \nu}y^{++\nu}-y^{++}_{\mu \nu}y^{--\nu})-v_{\mu \nu}y^{--\mu}y^{++\nu}+\frac{1}{\xi}v^{\mu}(y^{++}_\mu \partial_\nu y^{--\nu}-y^{--}_\mu \partial_\nu y^{++\nu})\big],\end{aligned}\ ] ] @xmath115 @xmath116\nonumber \\ & & + ( q^v_{y^{++}}-2q^v_{y^+ } ) \big[w^{+\nu}(y^{--}_\mu y^+_\nu - y^{--}_\nu y^+_\mu)+w^{-\nu}(y^{++}_\mu y^-_\nu - y^{++}_\nu y^-_\mu)\big]\nonumber \\ & & + \frac{1}{\xi } \big[q^v_{y^+}(y^{-- } _ \nu w^{+\nu}y^+_\mu + y^{++}_\nu w^{-\nu } y^-_\mu)+q^v_{y^{++}}(y^{--}_\mu w^+_\nu y^{+\nu}+y^{++}_\mu w^-_\nu y^{-\nu})\big]\big \},\end{aligned}\ ] ] where @xmath117 , @xmath118 , and @xmath119 for @xmath120 , whereas @xmath121 , @xmath122 , and @xmath123 , for @xmath124 . also , @xmath125 in any case . the respective vertex functions are shown in fig . it is worth emphasizing that , as required by @xmath3 symmetry , the vertex functions associated with the trilinear couplings @xmath126 and @xmath127 share the same lorentz structure : @xmath128 it is not hard to show that it satisfies the following simple ward identity @xmath129 where @xmath130 and @xmath131 are two point vertex functions given by @xmath132 as far as the quartic vertices are concerned , they are characterized by the following vertex functions : @xmath133 @xmath134 feynman rules for the trilinear and quartic vertices involving the bileptons and sm gauge fields in the @xmath3-covariant @xmath135-gauge.,width=240 ] on the other hand , the scalar part of @xmath110 allows one to define unphysical masses for the pseudo - goldstone bosons associated with the bileptons , and it also modifies some unphysical couplings of the higgs potential . it is worthwhile to discuss the dynamics of the @xmath136 lagrangian . this term has a strong impact on the higgs kinetic - energy sector associated with the @xmath38 triplet as it helps to remove some unphysical vertices . note that the latter can be decomposed into three @xmath3-invariant pieces : @xmath137 where @xmath138,\ ] ] and @xmath139.\ ] ] the @xmath140 lagrangian is not relevant for the present discussion as it is only composed by the interactions involving the @xmath27 boson , so we refrain from presenting it here . the lagrangian @xmath141 gives rise to the interactions between the pseudo - goldstone bosons and the electroweak gauge bosons . these interactions , which are dictated by the electroweak group , contribute to the @xmath18 and @xmath29 couplings at the one - loop level , so the corresponding feynman rules are shown in fig . [ fig2 ] . on the other hand , the @xmath142 term is responsible for the appearance of the bilinear terms @xmath143 and @xmath144 , as well as the unphysical trilinear and quartic couplings @xmath145 , @xmath146 , @xmath147 , and @xmath148 . when the @xmath136 lagrangian is taken into account , all these couplings vanish . in fact , after adding up these two terms , we obtain @xmath149,\ ] ] where some surface terms were ignored . needless to say that the absence of these unphysical vertices renders great simplicity for some loop calculations . as far as the ghost sector is concerned , the following definitions for the ghost fields @xmath150 and similar expressions for the antighost fields , allow us to express the corresponding lagrangian as follows : @xmath151\nonumber \\ & & + \frac{ig}{\sqrt{2}}y^\dag_\mu m_cd^\mu \bar{c}_y+\frac{ig}{2}y^\dag_\mu { \cal m}_c\bar{c}_y-\frac{\xi g}{2}\big[\phi^{0*}\phi^0c^\dag_y\bar{c}_y+\phi^0\phi^\dag_ym_c\bar{c}_y-(c^\dag_y\phi_y)(\phi^\dag_y\bar{c}_y)\big]\nonumber \\ & & + \frac{i\sqrt{2}}{\xi}\big[(\bar{m}_cc_y+m_c\bar{c}_y)^\dag ( d_\mu y^\mu)-(d_\mu y^\mu)(\bar{m}_cc_y+m_c\bar{c}_y)\big]\nonumber \\ & & -g\big[\phi^\dag_y(\bar{m}_cc_y+m_c\bar{c}_y)\phi^0+\phi^{0*}(\bar{m}_cc_y+m_c\bar{c}_y)^\dag \phi_y\big]+{\rm h.c.}\nonumber \\ & & -\frac{1}{2}f^{\bar{a}bc}f^{cde}\bar{c}^{\bar{a}}\bar{c}^bc^dc^e,\end{aligned}\ ] ] where @xmath152 they have the same quantum numbers that @xmath107 and @xmath32 . in addition , @xmath153 @xmath154 where @xmath155 and @xmath156 stands for the covariant derivative given in the adjoint representation of @xmath23 . the @xmath157 matrix is obtained from @xmath158 after replacing the ghost fields by antighost fields . under the electroweak group , @xmath158 transforms as @xmath159 , with @xmath160 . a similar transformation holds for @xmath157 and @xmath161 . as a consequence , @xmath98 is invariant under the @xmath3 group . both the pseudo - goldstone boson and the ghost sectors contribute to the vertex @xmath0 via trilinear and quartic couplings . as shown in fig . [ fig2 ] , the feynman rules arising from each sector are identical because each sector is @xmath3 invariant by its own . as a consequence , the trilinear vertices @xmath162 and @xmath163 satisfy simple ward identities : @xmath164 where @xmath165 , @xmath8 stands for a commutative ( pseudo - goldstone boson ) or anticommutative ( ghost ) charged scalar , and @xmath166 stands for the two - point vertex functions @xmath167 . feynman rules for the trilinear and quartic vertices involving sm gauge fields and scalar unphysical particles ( pseudo - goldstone bosons and ghosts ) in the @xmath82-covariant @xmath135-gauge . in this gauge , the @xmath7 and @xmath168 couplings to pseudo - goldstone bosons and ghosts coincide.,width=288 ] the prospect of the nlc and clic @xcite , have triggered the interest in the @xmath169 reaction , and motivated by this we will focus on the @xmath0 vertex with the two @xmath7s on - shell and @xmath168 off - shell . even in this case there is no reason to expect a gauge - invariant amplitude arising from a conventional quantization scheme . we will show below that this is indeed the case . retaining just the transverse degrees of freedom of @xmath168 , the vertex function for the @xmath0 coupling can be written as @xcite : @xmath170 + 2\delta \kappa_v ( q_\beta g_{\alpha \mu}-q_\alpha g_{\beta \mu } ) + \frac{4\delta q_v}{m^2_w}\big(p_\mu q_\alpha q_\beta-\frac{1}{2}q^2p_\mu g_{\alpha \beta}\big)\big\},\end{aligned}\ ] ] where @xmath171 we have dropped the cp - odd terms since they do not arise at the one - loop level in the minimal @xmath21 model . our notation and conventions are depicted in fig . [ fig3 ] . in the sm , the tree level values are @xmath172 , @xmath173 , and @xmath174 . in the case of the on - shell @xmath18 vertex , @xmath175 and @xmath176 are related to the @xmath7 magnetic dipole moment @xmath177 and the electric quadrupole moment @xmath178 : @xmath179 the trilinear @xmath0 vertex . the large circle denotes loop contributions and the arrows denote the flow of momenta.,width=192 ] as already noted , we are interested in the impact of the bileptons on the @xmath180 and @xmath181 form factors , which characterize the radiative corrections to the @xmath0 vertex . our results will only be an estimate since we are only considering the new physics effects at the @xmath28 scale , but our approach has the advantage of invariance under the @xmath3 group , which greatly simplifies the calculations . at the @xmath28 scale , only the bileptons couple to the @xmath7 boson since they arise as a doublet of the electroweak group . we already presented the feynman rules given in a @xmath135-gauge scheme and their @xmath82-covariant nature was displayed via simple ward identities . the one - loop amplitude of the @xmath0 vertex will also be gauge invariant , though gauge dependent , which also occurs when the bfm is applied . in other words , gauge - invariant quantum actions render gauge - invariant but not gauge - independent green functions . however , motivated by the link between the bfmfg and the pt , we will present our results in the feynman - thooft gauge . the generic feynman diagrams contributing to the @xmath0 vertex are shown in fig . it turns out that the bileptons contribute through all these diagrams , but the only nonvanishing contribution of scalar particles arises from the triangle graphs . in addition , owing to the separate @xmath3 invariance of the ghost and scalar sectors , the ghost - antighost contribution is exactly minus twice the one coming from the pseudo - goldstone bosons . taking into account all these facts , the total amplitude can be written as @xmath182 with @xmath183 the loop amplitude , which is the same for both @xmath18 and @xmath29 vertices . it is worth mentioning that the associated green functions differ only by the factor @xmath184 , just as occurs at the tree - level . this means that @xmath3 invariance is preserved at the one - loop level . thus the loop amplitude @xmath183 for on - shell @xmath7 bosons must satisfy the simple ward identity @xmath185 which can be verified once the loop integrals are solved explicitly . the explicit calculation shows that the @xmath186 and @xmath181 form factors are given by @xmath187q^2c_0 \nonumber \\ & + & 4x_y ( 4x_w-1)^2\big[b_0(3)-b_0(1)\big]+\left[26x^2_w+32x_y x_w(1 - 4x_w)+x_w\right]\big[b_0(1)-b_0(2)\big]\bigg\},\end{aligned}\ ] ] @xmath188q^2c_0\nonumber \\ & + & 4x_w\left(x_w(6x^2_w-5x_w+8x_y-1)-2x_y\right)\big[b_0(1)-b_0(2)\big]+4x_wx_y(4x_w-1)^2\big[b_0(2)-b_0(3)\big ] \nonumber \\ & -&x_w(4x_w-1)\left(1 + 2x_w(6x_w-1)\right)\bigg\},\end{aligned}\ ] ] where we have introduced the definitions @xmath189 , @xmath190 , @xmath191 , and @xmath192 . @xmath193 and @xmath194 stand for the following passarino - veltman scalar functions : @xmath195 , @xmath196 , @xmath197 , and @xmath198 . feynman diagrams for the @xmath0 vertex in the @xmath3-covariant gauge . the pseudo - goldstone bosons and ghosts contribute through an identical set of diagrams , but only the triangle ones give a nonvanishing contribution to the form factors @xmath180 and @xmath181.,width=288 ] the @xmath9 reaction will play an essential role in future researches at @xmath199 colliders : it will provide relevant information for our knowledge of the sm such as a more precise determination of the @xmath7 mass and its width decays , and it will also open up the possibility for detecting new physics effects via the distinctive @xmath200-channel contribution from the @xmath0 vertex . in the sm , the radiative corrections to the @xmath9 process have been widely studied both for on - shell @xcite and off - shell @xmath7 gauge bosons @xcite . new physics effects have also been studied in a model - independent manner using the effective lagrangian approach @xcite . beyond the sm , the reaction @xmath9 has been analyzed in technicolor theories @xcite and supersymmetric models @xcite . as for the radiative corrections to the @xmath0 vertex , they have received considerably attention . in the sm , the one - loop amplitudes were calculated using the conventional quantization scheme along with the feynman - thoof gauge @xcite . as emphasized in that work , the resultant amplitudes are not gauge invariant , which is evident from the presence of infrared divergences and the bad high - energy behavior of the @xmath180 form factor @xcite . in contrast , it was found that @xmath201 is well - behaved . shortly afterwards , these vertices were revisited by papavassiliou and philippides @xcite in a gauge - invariant way via the pt . the form factor @xmath180 obtained by these authors disagrees from that presented in @xcite , though there is agreement for @xmath181 . it was found that the radiative corrections to @xmath180 are of the order of @xmath202 , whereas @xmath181 is about one order of magnitude below . for instance , @xmath175 goes from @xmath203 for @xmath204 gev to @xmath205 for @xmath206 gev @xcite , whereas @xmath207 ranges from @xmath205 to @xmath208 in the same energy range @xcite . as far as experimental measurements are concerned , the constraint @xmath209 was obtained from cern @xcite and fermilab data @xcite . it is expected that this constraint is substantially improved at the cern large hadron collider ( lhc ) . even more , it has been argued that a deviation at the @xmath203 level might be measured at nlcs . as a consequence , only @xmath180 would be at the reach of nlcs , though the appearance of new physics effects may improve this situation . this is not the case however for one of the more popular sm extensions , namely , supersymmetry , which yields similar or smaller contributions than the sm ones @xcite . we turn now to our results . the @xmath180 and @xmath181 form factors depend on @xmath210 , @xmath211 , and @xmath55 . as for @xmath210 , it can take both positive ( time - like ) and negative ( space - like ) values . however , motivated by the prospect of nlcs , the form factors will be evaluated for @xmath212 gev . it is also evident that for relative low energies , where the sm contribution is dominant , those contributions of very heavy bileptons will be highly suppressed . however , when @xmath213 and @xmath211 are of the same order , it is expected that the new physics contributions become more relevant . one interesting feature of the minimal @xmath21 model is the constraint @xmath214 obtained from theoretical arguments , which in turn translates into @xmath215 tev in the minimal 331 model @xcite , though it can be relaxed by introducing a more complex higgs sector . we will consider this upper constraint for the bilepton mass . on the other hand , the most stringent lower bound @xmath216 gev @xcite arises from muonium - antimuonium conversion . it has been argued however that this bound can be evaded in a more general context since it relies on very restrictive assumptions @xcite . other strong limit , @xmath217 gev , arises from fermion pair production and lepton flavor violating decays @xcite , and the bound @xmath218 gev , valid only for the singly charged bilepton mass , was derived from limits on the muon decay width @xcite . we would like to stress that all these bounds depend on several assumptions , and in principle there still remains the possibility of lighter bileptons . although it is quite unlikely the existence of a relatively light bilepton , as way of illustration , we will concentrate on the range @xmath219 . also , although the contribution of a bilepton is expected to become more significant when its mass is of the same order of magnitude than that of @xmath213 , we will present results in the range 100 gev @xmath220 gev . the @xmath180 and @xmath201 form factors are shown in figs . [ deltak ] and [ deltaq ] for @xmath221 , @xmath222 , @xmath223 , and @xmath224 . we have included the values that arise above the threshold @xmath225 only by completeness , as in such a situation it would be more appropriate to study the direct production of bilepton pairs rather than their virtual effects . from those figures we can see that the @xmath180 and @xmath181 signs are reversed , a situation also observed in the sm model . for larger values of @xmath211 the form factors increase with the energy , although both of them approach asymptotically to zero for very large @xmath226 after reaching an extremum . this situation is similar to what is observed in the sm after the pt is implemented @xcite . from fig . [ deltak ] , we can see that @xmath180 ranges between @xmath205 and @xmath227 for a relatively light bilepton with a mass in the range @xmath228 . these values are of the same order of magnitude than the sm contribution . as far as @xmath181 is concerned , it ranges from @xmath205 to @xmath227 in the same @xmath211 range , which means that this form factor has essentially the same behavior than the sm contribution . also , we can see that a more heavy bilepton , with a mass in the range @xmath229 , yields both @xmath180 and @xmath181 at the @xmath230 level , and they increase smoothly when the energy increases . this means that for very heavy bileptons , @xmath180 and @xmath181 are one order of magnitude smaller than the respective sm radiative correction . notice also that the inequality @xmath231 always holds , which is opposite to what is observed in the sm . this apparent contradiction stems from the fact that , in our case , the new physics effects are of decoupling nature : the form factors vanish in the limit of a very large mass @xmath211 . indeed , @xmath181 is always of decoupling nature since it arises from a nonrenormalizable dimension - six operator . in contrast , @xmath186 can be sensitive to nondecoupling effects since it is associated with a renormalizable lorentz structure of dimension four . its nondecoupling nature is well known from the sm @xcite and some of its extensions @xcite . the @xmath180 form factor , in units of @xmath230 , as a function of the center - of - mass energy @xmath232 , in units of gev , for several values of @xmath211 . the solid line is for the real part of @xmath180 and the dashed line for its imaginary part.,width=384 ] the same as in fig . [ deltak ] but for the @xmath181 form factor.,width=384 ] finally , we would like to comment our results within the context of the effective lagrangian framework . in this scheme , the electroweak lagrangian is extended with nonrenormalizable operators of dimension higher than four respecting the @xmath3 symmetry . in particular , anomalous contributions to the @xmath180 and @xmath181 form factors are induced by the following @xmath3-invariant dimension - six operators @xmath233 where @xmath234 is the sm higgs doublet and @xmath235 , with @xmath236 the pauli matrices . the @xmath237 constants , which parametrize the details of the underlying physics , could be determined once the fundamental theory is known . in addition , @xmath238 is the new physics scale . it turns out that the @xmath239 and @xmath240 operators induce contributions to @xmath180 and @xmath181 , respectively . also , it has been shown that these operators can only be induced by the fundamental theory at one - loop or higher orders @xcite . assuming that these operators are induced at the one - loop level in the full theory , the associated @xmath237 constant must contain a factor of @xmath241 together with an additional coefficient @xmath242 or @xmath243 for each gauge field . taking the bilepton mass as the new physics scale , it is natural to assume @xmath244 where @xmath245 and @xmath246 stand for the dimensionless loop functions , whose structure depends on the details of the underlying physics . since the new physics effects are of decoupling nature , the loop functions @xmath245 and @xmath246 are expected to be of order @xmath247 at most . under this assumption , a straightforward evaluation shows that @xmath180 goes from @xmath248 to @xmath249 for @xmath229 , whereas @xmath181 ranges from @xmath250 to @xmath251 . this simple qualitative discussion shows that our results are in agrement with those expected in a decoupling scenario of physics beyond the fermi scale . there is a plenty of good reasons to expect the appearance of new physics beyond the sm , but it still remains a mystery how and where this class of effects would show up . any new particles would arise by direct production if there is enough energy available , or through their virtual effects on some observable . the last scenario seems to be the most promising if the new particles have masses much larger than the fermi scale . in this case , high precision measurements are needed in order to detect any deviation from the sm predictions . we have examined this possibility via the radiative corrections to the @xmath18 and @xmath29 vertices , which would play a special role at nlc experiments . the one - loop contribution to these couplings from the new gauge bosons predicted by the minimal @xmath21 model was studied in a @xmath82-invariant way by introducing a nonlinear quantization method . this scheme , even though conventional in the sense that it is based on brst symmetry , enables one to assess the new physics effects predicted by the 331 model on the sm green functions through a quantum action that is invariant under the electroweak group . special emphasis was put on discussing the similarities between our quantization method and the bfm . the main ingredient shared by both methods is that they allow one to construct gauge - invariant quantum actions . it is worth emphasizing however that while the bfm can be used at all energies , as gauge - invariance is preserved with respect to the gauge group of the full theory , the one presented here is only appropriate to study heavy physics effects on low - energy ( sm ) green functions . in the latter case the complete quantum action is only invariant under a subgroup of the theory . the @xmath82 invariance of the loop amplitudes associated with the @xmath18 and @xmath29 vertices was showed and special emphasis was put on the inherent simplicity of the calculation . our results show that for a relative light bilepton , with mass in the range @xmath252 , both @xmath180 and @xmath181 take values within the range of the sm contribution . on the other hand , for a more heavy bilepton , with mass in the range @xmath229 , the form factors remain essentially uniform and are of the order of @xmath230 . it means that they are , respectively , two and one order of magnitude smaller than their sm counterpart . it was also shown that our results are in agreement with the expectations arising from a decoupling scenario of new physics , as argued in the light of the effective lagrangian approach . our results suggest that it would be necessary a high experimental sensitivity in order to detect the virtual effects of new massive gauge bosons because it is hard that any heavy excitations arising from a renormalizable full theory could be much larger than the sm radiative corrections . g. couture and j. n. ng , z. phys . c * 35 * , 65 ( 1987 ) ; g. couture , j. n. ng , j. l. hewett , and t. g. rizzo , phys . d * 36 * , 859 ( 1987 ) ; g. tavares - velasco and j. j. toscano , phys . d * 69 * , 017701 ( 2004 ) ; c. l. bilchak , r. gastmans , and a. van proeyen , nucl * b273 * , 46 ( 1986 ) ; g. couture , j. n. ng , j. l. hewett , and t. g. rizzo , phys . d * 38 * , 860 ( 1988 ) ; a. b. lahanas and v. c. spanos , phys . b * 334 * , 378 ( 1994 ) ; t. m. aliev , _ ibid . _ * 155 * , 364 ( 1985 ) ; a. arhrib , j. l. kneur , and g. moultaka , _ ibid . _ * 376 * , 127 ( 1996 ) ; n. k. sharma , p. saxena , s. singh , a. k. nagawat , r. s. sahu , phys . d * 56 * , 4152 ( 1997 ) . b. s. dewitt , phys . rev . * 162 * , 1195 ( 1967 ) ; _ dynamical theory of groups and fields _ , gordon and breach , new york ( 1965 ) ; c. j. isham , r. penrose , and d. w. sciama , _ quantum gravity 2 _ , oxford university press , oxford ( 1981 ) ; g. thooft , acta universitatis wratislavensis * 368 * , 345 ( 1976 ) ; m. t. grisaru , p. van nieuwenhuizen , and c. c. wu , phys . rev . d * 12 * , 3203 ( 1975 ) ; h. kluberg - stern and j. b. zuber , phys . d * 12 * , 482 3159 ( 1975 ) ; d. g. boulware , phys . d * 23 * , 389 ( 1981 ) ; c. f. hart , phys . d * 28 * , 1993 ( 1983 ) ; l. f. abbott , nucl . b185 * , 189 ( 1981 ) ; acta phys . pol . b * 13 * , 33 ( 1982 ) ; l. f. abbott , m. t. grisaru , and r. k. schaefer , nucl . b229 * , 372 ( 1983 ) . see for instance : s. j. gates jr . , m. t. grisaru , m. rocek , and w. siegel , _ superspace , frontiers in physics 58 _ , the benjamin - cummings publishing company , advanced book program , reading , massachusetts ( 1983 ) and references therein ; s. ichinose and m. omote , nucl b203 * , 221 ( 1982 ) ; d. m. capper and a. maclean , nucl . phys . * b203 * , 413 ( 1982 ) ; p. s. howe , g. papadopoulos and k. s. stelle , nucl . phys . * b296 * , 26 ( 1988 ) ; k. a. meissner , acta phys . b * 17 * , 409 ( 1986 ) ; j. antikainen , m. chaichian , n. r. pantoja , and j. j. salazar , phys . b * 242 * , 412 ( 1990 ) ; k .- kondo , int . j. mod . phys . a * 16 * , 1303 ( 2001 ) ; c. becchi and r. collina , nucl . * b562 * , 412 ( 1999 ) ; qun wang _ e * 10 * , 483 ( 2001 ) ; j. p. bornsen and a. e. m. van de ven , nucl . b657 * , 257 ( 2003 ) ; d. j. toms , phys . rev . d * 27 * , 1803 ( 1983 ) ; m. omote and s. ichinose , phys . d * 27 * , 2341 ( 1983 ) ; s. mukhi , nucl . phys . * b264 * , 640 ( 1986 ) ; a. rebhan , z. phys . c * 30 * , 309 ( 1986 ) ; p. a. grassi , nucl . phys . * b462 * , 524 ( 1996 ) ; xiao - yuan li and yi liao , phys . b * 356 * , 68 ( 1995 ) ; p. a. grassi , nucl . phys . * b560 * , 499 ( 1999 ) ; a. denner , g. weiglein , and s. dittmaier , phys . b * 333 * , 420 ( 1994 ) ; nucl . . suppl . * b37 * , 87 ( 1994 ) ; nucl . phys . * b440 * , 95 ( 1995 ) ; z. bern and a. g. morgan , phys . rev . d * 49 * , 6155 ( 1994 ) ; l. g. cabral - rosetti , j. bernabeu , j. vidal , and a. zepeda , eur . j. c * 12 * , 633 ( 2000 ) . the most general gauge - fixing functions of @xmath135 type , nonlinear in both the vector and scalar sectors , were introduced in u. cotti , j. l. daz - cruz , and j. j. toscano , phys . b * 404 * , 308 ( 1997 ) and their implications on the ghost sector discussed with detail in j. g. mndez and j. j. toscano , rev . mex . fs . * 50 * , 346 ( 2004 ) . m. bace and n. d. hari dass , ann . * 94 * , 349 ( 1975 ) ; m. b. gavela , g. girardi , c. malleville , and p. sorba , nucl . phys . * b193 * , 257 ( 1981 ) ; n. m. monyonko , j. h. reid , and a. sen , phys . b * 136 * , 265 ( 1984 ) ; n. m. monyonko and j. h. reid , phys . d * 32 * , 962 ( 1985 ) ; g. jikia and a. tkabladze , phys . b * 323 * , 453 ( 1994 ) ; _ ibid . _ * 332 * , 441 ( 1994 ) ; g. jikia , phys . lett . b * 298 * , 224 ( 1993 ) ; nucl . b405 * , 24 ( 1993 ) ; _ ibid . _ * 412 * , 57 ( 1994 ) ; d. a. dicus and c kao , phys . d * 49 * , 1265 ( 1994 ) ; a. abbasabadi , d. bowser - chao , d. a. dicus , and w. w. repko , phys . d * 52 * , 3919 ( 1995 ) ; j. m. hern ' andez , m. a. p ' erez , g. tavares - velasco , and j. j. toscano , phys . d * 60 * , 013004 ( 1999 ) ; u. cotti , j. l. daz - cruz and j. j. toscano , phys . d * 62 * , 035009 ( 2000 ) ; j. l. daz - cruz , j. hernndez - snchez , and j. j. toscano , phys . b * 512 * , 339 ( 2001 ) ; j. hernndez - snchez , m. a. prez , g. tavares - velasco , and j. j. toscano , phys . d * 69 * , 095008 ( 2004 ) . h. min , t. lee , and p. y. pac , phys . d * 32 * , 440 ( 1985 ) ; h. hata and i. niigata , nucl phys . * b 389 * , 133 ( 1993 ) ; k .- i . kondo , phys . d * 58 * , 105019 ( 1998 ) ; t. shinohara , t. imai , and k .- kondo , int . a * 18 * , 5733 ( 2003 ) . for some of the most recent works on the phenomenology and some theoretical topics of @xmath21 models see : alex g. dias , c. a. de s. pires , v. pleitez , and p. s. rodrigues da silva , arxiv : hep - ph/0503192 ; alex g. dias , a. doff , c. a. de s. pires , and p. s. rodrigues da silva , arxiv : hep - ph/0503014 ; g. tavares - velasco and j. j. toscano , phys . d * 70 * , 053006 ( 2004 ) ; alex g. dias , phys . rev . d * 71 * , 015009 ( 2005 ) ; j. alexis rodrguez and m. sher , phys . d * 70 * , 117702 ( 2004 ) ; alex g. dias , r. martnez , and v. pleitez , eur . j. c * 39 * , 101 ( 2005 ) ; j. c. montero , v. pleitez , and m. c. rodriguez , phys . d * 70 * , 075004 ( 2004 ) . k. j. f. gaemers and g. j. gounaris , z. phys . c * 1 * , 259 ( 1979 ) ; k. hagiwara , r. d. peccei , d. zeppenfeld , and k. hikasa , nucl . b282 * , 253 ( 1987 ) ; u. baur and d. zeppenfeld , _ ibid . _ * b308 * , 127 ( 1988 ) ; * b325 * , 253 ( 1989 ) . b304 * , 463 ( 1988 ) ; c .- rim ahn , m. e. peskin , b. w. lynn , and s. b. selipsky , _ ibid . _ * b309 * , 221 ( 1988 ) ; j. fleischer , f. jegerlehner , and m. zralek , z. phys . c * 42 * , 409 ( 1989 ) ; j. fleischer , k. kolodziej , and f. jegerlehner , phys . d * 47 * , 830 ( 1993 ) ; w. beenakker _ b410 * , 245 ( 1993 ) ; w. beenakker , a. denner , s. dittmaier , and r. mertig , phys . lett . b * 317 * , 622 ( 1993 ) . a. denner and t. sack , z. phys . c * 45 * , 439 ( 1990 ) ; d. y. bardin , w. beenakker , and a. denner , phys . b * 317 * , 213 ( 1993 ) ; r. g. stuart , _ ibid . _ * 262 * , 113 ( 1991 ) ; a. aeppli , g. j. van oldenborgh , and d. wyler , nucl . * b428 * , 126 ( 1994 ) ; w. beenakker and a. denner , int . j. mod . phys . a * 9 * , 4837 ( 1994 ) ; k. melnikov and o. yakovlev , nucl * b471 * , 90 ( 1996 ) ; w. beenakker , a. p. chapovsky , and f. a. berends , phys . b * 411 * , 203 ( 1997 ) ; a. denner , s. dittmaier , and m. roth , nucl . phys . * b519 * , 39 ( 1998 ) ; phys . lett . b * 429 * , 145 ( 1998 ) ; w. beenakker , f. a. berends , and a. p. chapovsky , nucl . phys . * b548 * , 3 ( 1999 ) ; a. denner , s. dittmaier , m. roth , and d. wackeroth , _ ibid . _ * b 560 * , 33 ( 1999 ) ; phys . b * 475 * , 127 ( 2000 ) . s. alam , phys . d * 50 * , 124 ( 1994 ) ; * 50 * , 148 ( 1994 ) ; * 50 * , 174 ( 1994 ) ; s. alam , k. hagiwara , s. kanemura , r. szalapski , and y. umeda , phys . rev . d * 62 * , 095011 ( 2000 ) ; t. hahn , nucl . b609 * , 344 ( 2001 ) ; k. hagiwara , s. kanemura , m. klasen , and y. umeda , phys . d * 68 * , 113011 ( 2003 ) . j. alitti _ et al . _ , b * 277 * , 194 ( 1992 ) ; k. ackerstaff _ et al . _ , b * 397 * , 147 ( 1997 ) ; k. ackerstaff _ et al . _ , j. c * 2 * , 597 ( 1998 ) ; p. abreu _ et al . b * 397 * , 158 ( 1997 ) ; m. acciarri _ et al . _ , b * 398 * , 223 ( 1997 ) ; * 403 * , 168 ( 1997 ) ; * 413 * , 176 ( 1997 ) ; r. barate _ _ , _ ibid . _ * 422 * , 369 ( 1998 ) . et al . _ , lett . * 74 * , 1936 ( 1995 ) ; * 75 * , 1017 ( 1995 ) ; * 78 * , 4536 ( 1997 ) ; s. abachi _ _ , _ ibid . _ * 75 * , 1034 ( 1995 ) ; * 78 * , 3634 ( 1997 ) ; * 75 * , 1023 ( 1995 ) ; * 77 * , 3303 ( 1996 ) ; b. abbott _ et al . _ , _ ibid . _ * 79 * , 1441 ( 1997 ) ; s. abachi _ et al . _ , d * 56 * , 6742 ( 1997 ) .

since the pioneering work of canham and helfrich , variational formulations involving curvature - dependent functionals , like the classical willmore functional , have proven useful for shape analysis of biomembranes . we address minimizers of the canham - helfrich functional defined over closed surfaces enclosing a fixed volume and having fixed surface area . by restricting attention to axisymmetric surfaces , we prove the existence of global minimizers . * keywords : * willmore functional , helfrich functional , biomembranes , direct method in the calculus of variations . * ams subject classification * : 49q10 , 49j45 ( 58e99 , 53c80 ) .

1202.1979

for compact surfaces @xmath0 embedded in @xmath1 , the canham - helfrich functional is defined by @xmath2 where the integration is with respect to the ordinary 2-dimensional area measure , @xmath3 is the sum of the principal curvatures of @xmath0 , i.e. , twice the mean curvature , @xmath4 is the gaussian curvature , @xmath5 are constant bending rigidities and @xmath6 is a given spontaneous curvature . we prove the existence of a global minimizer for @xmath7 , in the class of finite systems of axisymmetric surfaces , under the constraints that the total area and the total enclosed volume of the surfaces are fixed . biological membranes and their shapes have attracted attention from researchers across many areas of mathematics . for example , membranes connect classical problems of differential geometry involving willmore s functional to studies of shape configurations of biological cells in physics and biology ( see , e.g. , @xcite , @xcite ) . more recently , researchers in both mathematical analysis and scientific computing have directed efforts to understanding _ multiphase _ membranes , where phase transitions and pattern formation can be observed ( see , e.g.@xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) . the modeling of multiphase membranes has numerous applications associated with artificial membranes in pharmacology and bioengineering ( e.g. , @xcite ) . in his seminal work @xcite , canham proposed the functional , in the case @xmath8 , in order to model the elastic bending energy of biological membranes formed by a double layer of phospholipids . when immersed in water , these molecules , which are composed by a hydrophilic head and a hydrophobic tail , spontaneously aggregate in order to shield the tails from water , forming a closed bilayer with the heads pointing outwards . since the thickness of a layer is generally three to four orders of magnitude smaller than the size of the observed cells or vesicles , the bilayer is usually approximated as a two - dimensional surface @xmath0 embedded in @xmath1 . the functional @xmath7 is the most general example of energy which is quadratic in the principal curvatures . the parameter @xmath6 , added by helfrich @xcite , accounts for an asymmetry in the composition of the layers and gives rise to a spontaneous curvature of the membrane in absence of other constraints . the bending rigidities @xmath9 and @xmath10 are also material - dependent parameters . under the simplifying assumption that the membrane is homogeneous , we choose @xmath11 and @xmath10 constant . in reality , phases with different levels of aggregation and different rigidities are observed @xcite . there are two natural constraints associated with the membrane configuration . since lipid membranes are inextensible , the total area of the membrane should be fixed . on the other hand , the membrane is permeable to water but not to dissolved ions . the resulting osmotic pressure leads then to a constraint on the volume enclosed by the membrane , which can therefore be regarded as constant @xcite . from the point of view of our analysis , the constraint on the area plays a crucial role in obtaining a priori bounds and compactness . the constraint on the volume , instead , does not add any property or difficulty , but it is the combination of these two constraints that makes highly nontrivial the problem of determining the minimizer . another important feature of membranes is that they can undergo topological changes , for example , a spherical vesicle can shrink at the equator and eventually split into two vesicles ( fission ) or a small dome can rise from a point of the surface and grow into a new entity which separates from the original one ( budding ) , see e.g. ( * ? ? ? * section 3.9 ) and @xcite . in order to be able to describe these kind of phenomena , we do not impose restrictions on the number of components of the minimizers . helfrich s functional can be regarded as a generalization of the classical willmore functional , defined by @xmath12 in the seminal paper @xcite , leon simon proved that for each @xmath13 there exists a compact embedded real analytic torus in @xmath14 which minimizes @xmath15 among compact embedded surfaces of genus 1 . following the direct method of the calculus of variations , he first shows that sequences of minimizers are compact in the sense of measures , and then proves that the limit measure is actually an analytic surface . simon s proof of regularity relies on the invariance of willmore functional under conformal transformations and on the fact that a minimizing surface @xmath0 must satisfy @xmath16 . owing to the presence of the spontaneous curvature @xmath17 and to the combined area and volume constraints , helfrich s functional _ is not conformally invariant _ , and for general values of area and volume , we only know that @xmath18 . therefore , though measure - compactness can be easily transferred to our case , simon s method for regularity can not be employed , and we have to find a different approach . regarding minimization with area and volume constraints in the case @xmath8 , existence of genus 0 minimizers with fixed isoperimetric ratio was recently proved in @xcite , while @xcite gives a complete classification of smooth critical points with low energy . existence of minimizers for functionals with weak second fundamental form in @xmath19 was addressed also in @xcite , using the theory of varifolds ( see also the end of section [ ssec : discussion ] ) . however , in contrast to the mean curvature vector , the scalar mean curvature @xmath3 does not have a variational characterization and there is no definition of scalar mean curvature for an arbitrary integral varifold . this obstacle can be removed using the generalized gauss graphs introduced in @xcite and developed in @xcite . compactness and lower - semicontinuity properties allow to obtain a minimizer , but it is not trivial to understand whether the limit , which in general is only a rectifiable current , is actually a classical surface . this is certainly true in the case of one - dimensional curves in @xmath20 , see for example , @xcite , @xcite and @xcite , but the question remains open for surfaces in @xmath1 . a different approach , based on a new formulation for the euler - lagrange equation of willmore functional , was introduced in @xcite . one of the results therein is a new proof of the existence of minimizers . in the attempt to apply this new method to helfrich functional , the same difficulties as above appear , in particular , the lack of an equivalent of li - yau minimality condition ( * ? ? ? * theorem 6 ) for helfrich functional necessitates another approach in order to guarantee that minimizers are embedded . regarding the existence of minimizers for the constrained willmore functional , in @xcite it is proven existence and regularity of axisymmetric solutions of the euler - lagrange equations with symmetric boundary conditions . in @xcite , adopting the techniques introduced in @xcite , the author proves the existence of minimizers with prescribed isoperimetric ratio . we are not aware of any extension of these methods to the constrained helfrich functional . in the present work , we give an answer that is only partial , since we restrict to axisymmetric surfaces . we note that our result can not be obtained from the above - mentioned results for @xmath21-regular curves , since if a curve @xmath22 generates a surface with bounded helfrich energy @xmath7 , it is not true in general that @xmath22 is @xmath21-regular ( see section [ ssec : discussion ] below ) . the class of axisymmetric surfaces is probably the most interesting from the point of view of applications . in fact seifert ( ( * ? ? ? * section 3.1.4 ) ) notes that it turns out that in large regions of the interesting parameter space the shape of lowest energy is indeed axisymmetric for vesicles of spherical topology " . we actually conjecture that _ for any given area and volume satisfying the isoperimetric inequality , and for any constant spontaneous curvature , the problem of minimizing in the class of embedded , constrained surfaces has a solution , and it is axisymmetric_. after this work was completed , we became aware of the preprint @xcite , which studies the @xmath23-limit of a diffuse - interface approximation of helfrich s functional for two - phase axisymmetric surfaces , and where many of the same technical difficulties that we encounter are addressed . we are independently treating the sharp - interface case of two - phase axisymmetric surfaces in @xcite . an _ axisymmetric surface _ , in the context of the present paper , is a surface @xmath0 obtained by rotating a curve @xmath22 , contained in the @xmath24 plane in @xmath1 , around the @xmath25-axis . since , following the direct method of the calculus of variations , we want to find the minimum of @xmath7 as the limit of a sequence of minimizers , we need the class of possible minimizers to be closed with respect to a reasonable topology . simple curves alone are thus not sufficient , as a curve as in figure [ fig1]-left can be obtained as a uniform limit of smooth simple curves . the curves falling in the two classes introduced below are regular enough to allow for a definition of a generalized helfrich energy , surface area , and enclosed volume , and at the same time are closed under the convergence induced by @xmath7 . [ [ notation . ] ] notation . + + + + + + + + + let @xmath26\to { \mathbb{r}}^2 $ ] , @xmath27 , be a plane curve of class @xmath28 . denote @xmath29 . let @xmath30)=\{{\gamma}(t ) : t\in [ a , b]\}$ ] be the trace of @xmath22 and let @xmath31 be its length . we mostly parametrize @xmath22 on the interval @xmath32 $ ] with constant speed @xmath33 , in some cases , where specified , we use the arc - length parameterization @xmath34 on the interval @xmath35 $ ] . [ def : g0 ] a curve @xmath36\to { \mathbb{r}}^2 $ ] belongs to the class of _ curves generating a genus-0 surface with bounded weak curvature _ if and only if @xmath37 [ def : g1 ] a curve @xmath36\to { \mathbb{r}}^2 $ ] belongs to the class of _ curves generating a genus-1 surface with bounded weak curvature _ if and only if @xmath38,\\ & { \gamma}(0 ) = { \gamma}(1),\quad \dot { \gamma}(0 ) = \dot { \gamma}(1),\quad { \gamma}_1(t ) > 0\quad \forall\,t\in [ 0,1 ] . \label{eq : condpos1 } \ ] ] @xmath39 at 130 320 @xmath25 at -5 340 @xmath40 at 235 0 @xmath39 at 300 320 @xmath25 at 0 340 @xmath40 at 350 0 we note that the condition @xmath41 can not be imposed in definition [ def : g0 ] , since a curve with @xmath42 could be obtained as a continuous limit of curves in either ( see , e.g. , figure [ fig1]-left ) or . let @xmath22 be a curve as in or . by rotating @xmath22 around the @xmath25-axis we obtain the surface @xmath0 parametrized by : @xmath43,\qquad ( t,\theta)\in [ 0,1]\times [ 0,2\pi].\ ] ] if a surface @xmath0 admits the parametrization , we say that @xmath0 _ is generated by _ @xmath22 . @xmath0 at 385 250 @xmath25 at 185 330 @xmath22 at 345 120 a standard computation ( see section [ ssec : surrev ] below ) , shows that _ if _ a curve @xmath22 generates a smooth surface @xmath0 , the 2-dimensional surface area , the enclosed volume , and the principal curvatures of the generated surface are given by @xmath44 since @xmath45 and @xmath46 , the helfrich energy can then be written as @xmath47 if the surface generated by @xmath22 in or is not smooth , as is the case of figure [ fig1 ] , we _ define _ the _ generalized _ 2-dimensional surface area , enclosed volume , principal curvatures and helfrich energy of the generated surface by the quantities in - . since the integral of the gaussian term of the energy is constant for a surface of fixed genus ( see section [ ssec : bonnet ] ) , it is often disregarded in the analysis of minimizers . nonetheless , since we are not imposing a fixed genus , nor a fixed number of components , we can not drop this term . furthermore , we note that the gaussian term is expected to play an important role in the case of multiphase membranes @xcite . the main result of this paper is the following . [ th : main ] let @xmath48 be given such that @xmath49 assume that @xmath50 , @xmath51 such that @xmath52 . let @xmath53 denote the set of finite families @xmath54 , for some @xmath55 ( not fixed ) , of axisymmetric surfaces generated by disjoint curves in @xmath56 , as in definition [ def : g0 ] and definition [ def : g1 ] , and satisfying the generalized area and volume constraints @xmath57 let @xmath7 be the helfrich energy functional defined in and let @xmath58 then the problem @xmath59 has a solution . condition ensures that the constraints satisfy the isoperimetric inequality , so that the set @xmath60 is not empty . when is an equality , the only element in @xmath60 is the sphere of area @xmath61 , if it is a strict inequality , @xmath60 contains an infinite number of elements . the range of the parameters @xmath9 and @xmath10 specified in the assumptions of theorem [ th : main ] is the mathematical range for which @xmath7 is positive definite on the principal curvatures , or in other words , for which @xmath62 controls the full squared norm of the second fundamental form of @xmath0 , which is absolutely crucial for any kind of analysis . on the other hand , the physical range in which these parameters are typically found is contained in the one we assume , see e.g. @xcite and @xcite ( note that the latter cites the former , but inverting numerator and denominator , by mistake ) . note that the functional @xmath63 does not depend on the reciprocal position of the components @xmath64 . therefore , by translation along the vertical axis , we can transform a system with self intersections into one with the same energy and without crossings , thus avoiding unphysical situations . [ rem : index ] _ on the index @xmath65 of a curve . _ even if a curve has a smooth parametrization , it can generate a surface of revolution with singularities , which can not represent any physical lipid bilayer ( figure [ fig1 ] ) . a way to restrict to physical surfaces is to prescribe the index of the system of generating curves . if @xmath22 is a closed curve , @xmath66 , let @xmath65 be the index of @xmath22 with respect to @xmath67 ( * ? ? ? * chapter ii , section 1.8 ) . if @xmath22 is not closed , @xmath68 and @xmath69 , we can extend it symmetrically with respect to the @xmath25-axis in order to define its index . for a system of surfaces @xmath54 , generated by @xmath70 , and @xmath71 define @xmath72 note that if @xmath73 is a smooth connected bounded open set and @xmath22 is counterclockwise parametrization of @xmath74 , then @xmath75 and @xmath76 . more generally , points with index 1 represent the internal volume of a vesicle also for surfaces generated by curves which are not the parametrization of a boundary ( as in figure [ fig1]-left ) . in order to eliminate situations like figure [ fig1]-right , we can look for minimizers in the class of systems of surfaces @xmath77 such that , additionally , @xmath78 the advantage in using the index as a condition , is its continuity with respect to uniform convergence of the curves , and thus its compatibility with the convergence induced by the bound @xmath79 ( see section [ ssec : plan ] below and definition [ def : systconv ] ) , so that the proof of theorem [ th : main ] can be directly used to prove existence of minimizers in @xmath60 satisfying . on the other hand , by removing the condition on the index , the minimizer of helfrich functional could actually be found in non - embedded surfaces , like delauney surfaces or the wente torus . at 247 190 at 233 171 at 224 151 at 213 131 @xmath22 at 310 135 @xmath80 at 165 5 @xmath81 at 243 5 @xmath25 at 3 212 @xmath40 at 354 5 @xmath22 at 270 40 @xmath25 at 160 390 _ on the lack of @xmath21-regularity . _ in corollary [ cor : reg ] we prove that any curve that generates a surface with bounded helfrich energy is @xmath21-regular on any stretch at positive distance from the @xmath25-axis . since the area element vanishes on the @xmath25-axis , the second derivative of a curve in ( g0 ) need not be square - integrable near the intersection with the @xmath25-axis , and therefore we can not expect to control the @xmath19-norm of @xmath82 . loosely speaking , the reason why the regularity of a surface @xmath0 does not imply the same regularity for the generating curve is simply the fact that a function ( e.g. , @xmath83 ) can be integrable in a neighborhood of the origin in @xmath20 , but not in @xmath84 . for example , let @xmath85 and consider the curve defined by @xmath86.\ ] ] clearly , @xmath87;{\mathbb{r}}^2)$ ] , @xmath34 , @xmath88 , and a quick computation shows that @xmath89 , @xmath90 , as @xmath91 . therefore , @xmath92 are square - integrable with respect to the area measure , on @xmath93 . _ on the generalized area and varifolds . _ returning to the example in figure [ fig1]-left , we note that there could be two ways to describe the area of the middle annulus of the revolution surface generated by this curve . if we see it as a single layer of membrane , it should simply measure @xmath94 . on the other hand , if we obtained this curve as a limit of a sequence of simple curves , where the vertical distance between the layers of the membrane collapsed to zero in the stretch between @xmath80 and @xmath81 , it should be seen as a double layer , and measure @xmath95 . since we are imposing a constraint on the total area of the membrane , and the objects we wish to describe are _ closed _ vesicles , the second interpretation , in which the horizontal stretch represents two overlapping layers , should be preferred . this is a drawback of modeling a three - dimensional object ( i.e. a membrane with positive thickness ) as a two - dimensional one : the only way we have to represent adjacent layers is to allow them to overlap , even though overlapping is not a physical possibility for the original three - dimensional membranes . using parametrized curves to model membranes goes naturally in this second direction . the generalized area defined in , in fact , following the parametrization , counts the multiplicity of every self - intersection . in this respect , we obtain the same result as if we described the curves with varifolds . without going into details ( which can be found in @xcite ) , we may think of the weight - measure of an integral varifold @xmath96 associated to a 2-rectifiable set @xmath97 as @xmath98 , where @xmath99 is the two - dimensional hausdorff measure , and @xmath100 is a measurable @xmath101-valued function , which represents the _ density _ of @xmath96 . for example , if the surface @xmath0 generated by @xmath22 as in figure [ fig1]-left is obtained as limit of smooth surfaces @xmath102 , we would have @xmath103 , @xmath104 , in the sense of measures , @xmath105 for @xmath106 and @xmath107 , and @xmath108 for @xmath109 , counting the horizontal stretch twice , exactly as definition does with curves . in section [ ssec : bonnet ] we recall the main relations between mean curvature , gaussian curvature and the euler characteristic , and we show that helfrich s energy controls the @xmath19-norm of the principal curvatures , with respect to the area measure . in section [ ssec : surrev ] we derive the main geometrical quantities for surfaces of revolution . in the following subsections we study which properties of a general generating curve can be obtained from the @xmath19 bound on the principal curvatures , we estimate the length of the curves ( section [ ssec : lengthbd ] ) , we control the regularity of the tangents on the @xmath25-axis ( section [ sec : card ] ) , and the total variation of @xmath110 ( section [ ssec : oscbd ] ) . the proof of theorem [ th : main ] is given in section [ sec : existence ] , following the direct method of calculus of variations . the standard procedure consists in showing that sub - level sets of the functional @xmath63 are compact , so that from a minimizing sequence @xmath111 it is possible to extract a converging subsequence @xmath112 , and then proving that @xmath63 is lower - semicontinuous , so that @xmath113 is a global minimum for @xmath63 . the crucial ingredients , which are not prescribed by the general direct method and have to be chosen for each specific problem , are the set of minimizers , which we discuss at the beginning of section [ sec : existence ] , and the topology with respect to which compactness and lower - semicontinuity must be verified . regarding the topology , it is natural to expect ( or request ) at least strong @xmath114 convergence for the generating curves , in order to preserve the surface area in the limit . moreover , it is straightforward to show that the second fundamental form of @xmath102 is uniformly bounded in @xmath19 , but _ only with respect to the surface area measure_. hence the necessity to study simultaneously convergence of the curvatures ( expressed via @xmath115 ) and of the area measure @xmath116 . a suitable tool for this purpose is provided by the _ measure - function pairs _ introduced in @xcite . the main definitions and theorems regarding measure - function pairs are recalled in section [ ssec : mt ] . the main body of the proof consists then in the lower - semicontinuity result ( proposition [ prop : lscs ] ) and in the compactness result ( proposition [ prop : compsurf ] ) . continuity of the constraints follows from the choice of the topology , and it is described in section [ ssec : constr ] . all these steps are summarized in section [ ssec : proof ] , where the complete proof of theorem [ th : main ] is given . let @xmath117 be the euler - poincar characteristic of a compact surface @xmath0 , see , e.g. , ( * ? ? ? * proposition 3 , section 4 - 5 ) . recall that every compact connected surface is homeomorphic to a sphere with a certain number @xmath118 of handles , and the number @xmath119 is called the _ genus _ of @xmath0 . the gauss - bonnet theorem states that if @xmath0 has no boundary , then @xmath120 since we are dealing with surfaces of revolution , we are interested only in two cases : * curves in ( g0 ) ( figure [ fig2]-left ) , which generate surfaces homeomorphic to a sphere , hence @xmath121 , @xmath122 , and @xmath123 , and * curves in ( g1 ) ( figure [ fig2]-right ) , which generate surfaces homeomorphic to a torus , hence @xmath124 , @xmath125 , and @xmath126 . the next lemma contains the fundamental coercivity estimate for helfrich s functional . it can be considered a standard observation ( see e.g. @xcite ) , but we report the proof for completeness . [ lemma : totalbound ] let @xmath0 be generated by @xmath127@xmath56 , and let @xmath50 , @xmath51 such that @xmath52 . then there exists @xmath128 such that @xmath129 let @xmath130 , note that @xmath131 and the coefficient in front of the last term is positive if and only if @xmath132 . for all @xmath133 it holds @xmath134 and thus @xmath135 choosing @xmath50 , @xmath136 and @xmath133 such that @xmath137 , we get @xmath138 where @xmath139 and @xmath140 . integrating on @xmath0 we obtain the thesis . we note that since we restrict to surfaces of revolution , the genus of which can only be 0 or 1 , we could extend the range of parameters to @xmath141 . indeed , if @xmath142 , by gauss - bonnet theorem @xmath143 for a family of surfaces @xmath144 , we can then find a constant @xmath128 such that @xmath145 since in physical applications the parameters are as in the assumptions of lemma [ lemma : totalbound ] ( see , e.g. , @xcite , @xcite ) , we will not use this estimate on the cardinality of the system , and rely only on . from this section onwards , we restrict to surfaces of revolution . we start by deriving the geometrical quantities involved in the study of helfrich s functional . these computations can be found also , e.g. , in ( * ? ? ? * section 3 - 3 , example 4 ) ( with opposite orientation ) . with the parametrization @xmath146,\qquad ( t,\theta)\in [ 0,1]\times [ 0,2\pi],\ ] ] we compute the tangent vectors @xmath147 , \\ r_\theta:=\frac{\partial}{\partial \theta}r(t,\theta)&= \big[-{\gamma}_1(t)\sin \theta,\ { \gamma}_1(t)\cos \theta,\ 0\big].\end{aligned}\ ] ] note : @xmath148 , i.e. , the tangents are always orthogonal . the first fundamental form is given by @xmath149= \left [ \begin{array}{cc } r_t \cdot r_t & r_t \cdot r_\theta\\ r_\theta \cdot r_t & r_\theta \cdot r_\theta \end{array } \right ] = \left [ \begin{array}{cc } |\dot { \gamma}(t)|^2 & 0 \\ 0 & { \gamma}_1(t)^2 \end{array } \right],\ ] ] @xmath150 note that the first fundamental form does not depend on the longitude parameter @xmath100 . the normal vector can be oriented inwards or outwards , depending on the direction of @xmath22 . @xmath151\\ & = \frac{1}{|\dot { \gamma}(t)|}\big [ -\dot { \gamma}_2(t ) \cos \theta,-\dot { \gamma}_2(t ) \sin \theta,\ \dot { \gamma}_1(t)\big].\end{aligned}\ ] ] for the computation of the second fundamental form we make use of a constant - speed parametrization @xmath152\\ n_\theta:= \frac{\partial}{\partial \theta}n(t,\theta ) & = \frac{1}{|\dot { \gamma}(t)| } \big [ \dot { \gamma}_2(t ) \sin \theta,-\dot { \gamma}_2(t ) \cos \theta,\ 0\big]\end{aligned}\ ] ] @xmath153=-\left [ \begin{array}{cc } n_t \cdot r_t & n_t \cdot r_\theta\\ n_\theta \cdot r_t & n_\theta \cdot r_\theta \end{array } \right ] = \frac{1}{|\dot { \gamma}| } \left [ \begin{array}{cc } \ddot { \gamma}_2 \dot { \gamma}_1 -\ddot { \gamma}_1 \dot { \gamma}_2 & 0 \\ 0 & { \gamma}_1 \dot { \gamma}_2 \end{array } \right].\ ] ] we can then express the gaussian curvature @xmath154 ( twice ) the mean curvature @xmath155 and the principal curvatures @xmath156 note that @xmath157 is just the curvature of @xmath22 , with the sign depending on the orientation . let @xmath158}$ ] be the one - dimensional lebesgue measure restricted to the closed interval @xmath32 $ ] , define the radon measure @xmath159}.\ ] ] the area of the generated surface is given by @xmath160 and the enclosed volume by @xmath161 owing to , and recalling that we are using a constant - speed parametrization , @xmath162 since @xmath7 and the total area are invariant under reparametrizations of @xmath22 , we can write the last inequality in the case of arc - length parametrization , obtaining the crucial estimate @xmath163 in this subsection we derive a uniform bound on the length of the generating curves . [ lemma : lengthbound ] let @xmath22 be curve generating a revolution surface @xmath0 as in or . then @xmath164 in order to obtain the left inequality , we compute @xmath165}|{\gamma}_1(t)|\leq 2\pi\,\ell({\gamma}){\textup{diam}(\sigma)}.\ ] ] regarding the right inequality , it is not restrictive to assume that @xmath22 is parametrized by arc - length , on the interval @xmath166 $ ] , where @xmath167 , so that @xmath168 we compute @xmath169 by either ( g0 ) or ( g1 ) above , @xmath170 , and @xmath171 for all @xmath172 , so the first term of the right - hand side of becomes @xmath173_{t=0}^{t = l } \\ & = \int_0^l { \gamma}_1(t)\ddot { \gamma}_1(t)\,dt \\ & = \int_0^l \sqrt{{\gamma}_1(t)}\left(\sqrt{{\gamma}_1(t)}\ddot { \gamma}_1(t)\right)\,dt \\ & \leq \left ( \int_0^l { \gamma}_1(t)\,dt \right)^{1/2}\left ( \int_0^l { \gamma}_1(t)\ddot { \gamma}_1 ^ 2(t)\,dt \right)^{1/2}\\ & \stackrel{\eqref{eq : defarea}}{= } \left ( \frac{|{\sigma}|}{2\pi } \right)^{1/2}\left(\frac{1}{2\pi } \int_0^l |\ddot { \gamma}(t)|^2\ , 2\pi{\gamma}_1(t)\,dt \right)^{1/2}\\ & \stackrel{\eqref{eq : k1arcl}}{= } \frac{\sqrt { |{\sigma}|}}{2\pi}\left ( \int_{\sigma}k_1 ^ 2\ , da \right)^{1/2}.\end{aligned}\ ] ] on the other hand , the last term of gives @xmath174}|\dot { \gamma}_2(t)| \int_0^l |\dot { \gamma}_2(t)|\ , dt\\ & \leq |\dot { \gamma}| \int_0^l \sqrt{{\gamma}_1(t)}\frac{|\dot { \gamma}_2(t)|}{\sqrt{{\gamma}_1(t)}}\ , dt\\ & \stackrel{\eqref{eq : dotgarcl}}{\leq } \left ( \int_0^l { \gamma}_1(t)\,dt \right)^{1/2}\left ( \int_0^l \frac{\dot { \gamma}_2 ^ 2(t)}{{\gamma}_1(t)}\,dt \right)^{1/2}\\ & = \frac{1}{2\pi}\left ( |{\sigma}| \int_{\sigma}k_2 ^ 2\ , da \right)^{1/2}. \end{aligned}\ ] ] noting that @xmath175 we get the following bound for a system of curves . [ cor : realbd ] let @xmath176 , @xmath177 be a finite family of curves in or generating the revolution surface @xmath64 . then @xmath178 [ def : gengen ] we say that @xmath36\to{\mathbb{r}}^3 $ ] is a _ generalized generator _ if @xmath22 is lipschitz - continuous , @xmath179 and @xmath180 for almost every @xmath181 , @xmath182 and @xmath183 in particular , implies that @xmath184 . [ sec : card ] [ lemma : w2loc ] let @xmath22 be as in definition [ def : gengen ] . for every subinterval @xmath185\subset [ 0,1]\cap \{{\gamma}_1>0\}$ ] @xmath186;{\mathbb{r}}^2).\ ] ] it holds @xmath187 } \{{\gamma}_1(s)\ } \int_a^b|\ddot { \gamma}|^2\ , dt.\ ] ] thus , @xmath188 and @xmath189 . by standard sobolev inclusions , @xmath190 , and there is a unique function which extends @xmath191 to @xmath192;{\mathbb{r}}^2)$ ] . we denote this extension by @xmath191 . in particular , if @xmath193(g1 ) , then @xmath194 and @xmath191 has a unique extension to @xmath195;{\mathbb{r}}^2)$ ] . [ lemma : tg ] let @xmath22 be as in definition [ def : gengen ] . let @xmath196 $ ] be such that @xmath197 , @xmath198 for all @xmath199 . then , the limits of @xmath200 as @xmath201 and @xmath202 exist , and @xmath203 moreover , either @xmath204 or @xmath205 it is not restrictive to assume @xmath206 as before . for any @xmath207 @xmath208 since @xmath209 and @xmath210 belong to @xmath211 by , we can define an absolutely continuous function @xmath212 satisfying @xmath213 by and , for every @xmath133 there exists a @xmath214 such that @xmath215 thus , the limit of @xmath216 as @xmath217 exists . we can now prove that this limit is 0 . recall that @xmath218 and @xmath219 by hypothesis , then for all @xmath133 we have @xmath220 } { \gamma}_1(s)= \max_{s\in [ a , a+{\varepsilon } ] } \int_a^s \dot { \gamma}_1({\sigma})\ , d{\sigma}\leq { \varepsilon}|\dot { \gamma}|={\varepsilon},\ ] ] and @xmath221 since the integrand is nonnegative , we conclude that @xmath222 the proof of the corresponding statement for the limit as @xmath202 is identical . the statement on the limit of @xmath110 follows by the assumption @xmath223 and by the continuity of @xmath191 in @xmath224 obtained in lemma [ lemma : w2loc ] . [ cor : reg ] under the assumptions of lemma [ lemma : tg ] @xmath225;{\mathbb{r}}^2).\ ] ] since @xmath226 in @xmath224 , by lemma [ lemma : w2loc ] it holds @xmath227 . by lemma [ lemma : tg ] , @xmath191 has a continuous extension to @xmath228)$ ] . [ lemma : oscbd ] let @xmath22 be as in definition [ def : gengen ] . let @xmath229 be such that @xmath198 for all @xmath230 , then @xmath231 and @xmath232 by lemma [ lemma : w2loc ] , or corollary [ cor : reg ] , we can assume that @xmath233;{\mathbb{r}}^2)$ ] and @xmath234 . recall that @xmath235 and @xmath236 . on @xmath237 we compute @xmath238 defining @xmath239 we have @xmath240 let @xmath241 , @xmath242 . the unique minimizer of the above problem is the solution of the euler - lagrange equation @xmath243 by integration , we compute @xmath244 define @xmath245 , imposing the boundary conditions we get @xmath246 note that if @xmath247 in @xmath224 , then @xmath110 is constant in @xmath224 , and is trivially satisfied . if @xmath248 in a point @xmath249 , by continuity it is positive in an open interval containing t , and therefore @xmath250 . then , @xmath251 in is well - defined and , in particular @xmath252 on @xmath253 . the minimum value is then given by @xmath254 since @xmath255 by inserting the minimum value in and multiplying by @xmath256 we obtain @xmath257 noting that @xmath258 and taking the square root , we obtain . in order to prove , we follow the same computations , inverting the role of @xmath110 and @xmath259 . let @xmath260 we have @xmath261 let @xmath262 , @xmath263 . by the same computations as in and following lines , the minimum value is then given by @xmath264if @xmath265 , then is trivially true . if @xmath266 , by inserting this minimum value in and multiplying by @xmath256 we obtain @xmath267 using the simple inequality @xmath268 , we finally obtain @xmath269 taking the square root , we obtain . in the case @xmath270 , @xmath218 , @xmath271 in @xmath224 , @xmath272 we will make use of lemma [ lemma : oscbd ] , instead of the simpler estimate obtained in this remark , because it allows for a rigorous treatment of the set @xmath253 and of the jump points for @xmath110 . for @xmath273 we denote the integer part of @xmath40 by @xmath274 . [ lemma : finiteg10 ] let @xmath22 be as in definition [ def : gengen ] . then , there can not be more than @xmath275 intervals @xmath276 such that @xmath277 as a consequence , @xmath278 let @xmath279 $ ] be an interval such that @xmath280 by corollary [ cor : reg ] we have that @xmath227 , @xmath233;{\mathbb{r}}^2)$ ] . since @xmath281 , by lemma [ lemma : oscbd ] @xmath282 therefore , there can not be more than @xmath275 intervals satisfying . in particular , by lemma [ lemma : tg ] , there can not be more than @xmath283 points where @xmath284 . the results obtained in this section imply that a generalized generator @xmath22 with bounded helfrich energy is either in ( g1 ) , if @xmath226 and @xmath22 is closed , or it can be decomposed into a finite number of curves in ( g0 ) . more precisely , we have the following result . [ cor : split ] if @xmath22 as in definition [ def : gengen ] generates @xmath0 and satisfies @xmath68 , then there exist @xmath285 and curves @xmath286 generating @xmath64 and satisfying - for @xmath287 , such that @xmath288 moreover , the generated surfaces @xmath64 admit a @xmath28-regular parametrization . by lemma [ lemma : finiteg10 ] , @xmath289 therefore , there exist @xmath290 such that @xmath291 . for @xmath287 , let @xmath292 and define the curves @xmath293\to { \mathbb{r}}^2,\qquad \eta_i({{\tau}}):={\gamma}\left(\ell_i{{\tau}}+ t_{i-1}\right).\]]it is immediate to check that @xmath286 satisfies , and , while is ensured by corollary [ cor : reg ] . the latter and lemma [ lemma : tg ] also imply that @xmath64 has a @xmath28-regular parametrization . [ def : systems ] we say that a system of surfaces @xmath113 belongs to the class @xmath294 of systems of generalized surfaces if note that the number of components may depend on the choice of parametrization . for example , a system of two spheres @xmath298 touching in one point on the @xmath25-axis can be parametrized by one generator @xmath22 of length @xmath299 or by two generators @xmath300 of length @xmath301 . in section [ ssec : surrev ] we proved that generalized generators @xmath22 are piecewise-@xmath28 and are not differentiable only in a finite number of points on the @xmath25-axis . therefore , the number of components we are interested in is the ( minimum ) number of @xmath28 components , that is @xmath302 we recall that area , volume , and helfrich energy of a system @xmath54 are simply defined as the sum of the correspondent _ generalized _ quantities over all the components of the system , i.e. , @xmath303 we turn now to the suitable notion of convergence for such systems . we recall that a sequence of radon measures @xmath304 is said to converge weakly-@xmath305 to @xmath306 if @xmath307 for every @xmath308 . we define the space of @xmath67-summable functions with respect to a positive radon measure @xmath309 as @xmath310 [ def : weakags ] following ( * ? ? ? * definition 5.4.3 ) , given a sequence of measures @xmath311 converging weakly-@xmath305 to @xmath309 , we say that a sequence of ( vector ) functions @xmath312 converges weakly to a function @xmath313 , and we write @xmath314 in @xmath315 , provided @xmath316 for every @xmath317 . for @xmath318 , we say that a sequence of ( vector ) functions @xmath319 converges _ weakly _ to a function @xmath320 , and we write @xmath314 in @xmath321 , provided @xmath322 for @xmath318 , we say that a sequence of ( vector ) functions @xmath319 converges _ strongly _ to a function @xmath320 , and we write @xmath323 in @xmath321 , if holds and @xmath324 [ lemma : moser ] let @xmath325 such that @xmath326 . suppose that @xmath304 and @xmath309 are radon measures on @xmath84 such that @xmath327 and that @xmath328 , @xmath329 , @xmath330 , @xmath331 be such that @xmath332 then @xmath333 [ th : ags ] let @xmath318 , let @xmath304 and @xmath309 be radon measures on @xmath84 such that @xmath327 , and let @xmath319 be a sequence converging _ weakly _ to a function @xmath320 in the sense of definition [ def : weakags ] . then @xmath334 for every convex and lower - semicontinuous function @xmath335.$ ] [ def : systconv ] we say that a sequence of systems of surfaces @xmath336 as in definition [ def : systems ] converges to a finite system @xmath337 if @xmath113 admits a parametrization @xmath338 and there exist @xmath339 and @xmath340 such that @xmath341 for all @xmath342 and such that * for each @xmath343 the generating curves @xmath344 converge in the following sense @xmath345;{\mathbb{r}}^2 ) , \label{eq : convg}\\ \dot { \gamma}^n_i & \to \dot { \gamma}_i\quad \text{strongly in } l^2{\left((0,1);{\mathbb{r}}^2\right ) } , \label{eq : convgd}\\ \ddot { \gamma}^n_i & { \rightharpoonup}\ddot { \gamma}_i \quad \text{weakly in $ l^2(\mu_{{\gamma}^n};{\mathbb{r}}^2)$ , in the sense of \eqref{eq : weakagsp } } ; \label{eq : convgdd } \end{aligned}\ ] ] * for each @xmath346+@xmath347 @xmath348 with this definition of convergence , the passage to the limit of the area and volume constraints is straightforward . let @xmath350 be a sequence of systems converging to a system @xmath351 in the sense of definition [ def : systconv ] , then @xmath352 indeed , by and , using the fact that the for the last @xmath353 components @xmath354 strongly in @xmath355 , @xmath356 and @xmath357 let @xmath362 . first of all , we notice that , according to definition [ def : systconv ] , since @xmath363 , for all @xmath364 big enough @xmath365 therefore , it is enough to show that @xmath366 for @xmath367 . this is reasonable , since by definition [ def : systconv]-(ii ) the @xmath368-th component , for @xmath369 , vanishes as measure . let then @xmath370 be fixed , we write @xmath371 for the generating curve and @xmath372 for the generated revolution surface . define the family of measures @xmath373 } \in rm({\mathbb{r}}).\ ] ] by and @xmath374}\qquad \text{in } rm({\mathbb{r}}).\ ] ] recall that @xmath375 ( @xmath376 and @xmath377 are defined analogously ) and @xmath378 let @xmath379 be a subsequence realizing the liminf of @xmath7 , i.e. , such that @xmath380 for sake of notation , we drop the index @xmath381 from the subsequence in the rest of the proof . all we need to prove is that @xmath382 in the sense of definition [ def : weakags ] . indeed , if and hold , by theorem [ th : ags ] and @xmath383 and by lemma [ lemma : moser ] @xmath384 therefore , for every component @xmath385 @xmath386 and to conclude the proof , it is sufficient to notice that @xmath387 [ lemma : k1ags ] let @xmath388\to{\mathbb{r}}^2 $ ] be a sequence of generating curves for admissible surfaces @xmath102 , and assume that @xmath389 as in - . then @xmath390 in the sense of definition [ eq : weakags ] , where @xmath391 and @xmath392 are defined @xmath309-a.e . note that , in case ( i ) , we can assume that @xmath393 denote by @xmath394 the @xmath395 rotation of the function @xmath396 , i.e. , @xmath397 . then , for all @xmath398 @xmath399 by and @xmath400 therefore , @xmath401 converges strongly in @xmath402 to @xmath403 , according to definition [ def : weakags ] . by and lemma [ lemma : moser ] @xmath404 regarding @xmath392 , let @xmath398 , then @xmath405 so that , by , @xmath406 by and we also have that @xmath407 this concludes the proof of lemma [ lemma : k1ags ] which completes the proof lower - semicontinuity statement . let @xmath409 be a family of finite systems generated by curves in ( g0 ) or ( g1 ) . throughout this step , we assume that there exist constants @xmath410 such that for all @xmath411 @xmath412 moreover , owing to lemma [ lemma : lengthbound ] and , the diameter of every component of a system satisfying is bounded by a constant which depends only on the data of the problem . consequently , since we are interested only in the shape of the components , and not in their relative position in the space , it is not restrictive to assume that there exists @xmath413 such that for all @xmath411 @xmath414 for short , we say that a finite system @xmath113 generated by curves in ( g0 ) and ( g1 ) satisfying and is an _ admissible system_. note that in we are not fixing the total area , we are requiring only an upper bound , thus allowing for more flexibility in the proofs . note also that owing to the isoperimetric inequality no separate bound on the volume is necessary , and owing to the restrictive geometry of revolution surfaces the only possible genus is 0 or 1 . if @xmath22 is in the class ( g0 ) , by lemma [ lemma : tg ] we have @xmath417 if @xmath22 is in ( g1 ) , then @xmath418)$ ] by corollary [ cor : reg ] , and therefore there must be two points @xmath419 $ ] such that @xmath420 ( e.g , choose @xmath421 and @xmath422 : since @xmath22 is closed and differentiable , the tangents in these points are horizontal and have opposite orientation ) . by , and lemma [ lemma : oscbd ] , for every @xmath177 @xmath423}|\dot { \gamma}_1(s)-\dot { \gamma}_1(t)| \leq \ell({\gamma}_i)\int_0 ^ 1 { k}_{1,i}^2 + { k}_{2,i}^2\ , d\mu_{{\gamma}_i},\ ] ] @xmath424 by there is a constant @xmath425 , depending only on the data , such that summing over @xmath368 @xmath426 we conclude that @xmath427 . * there exists a subsequence @xmath429 and a generalized generator @xmath22 as in definition [ def : gengen ] such that @xmath429 converges to @xmath22 in the sense of - , + or * there exists a point @xmath430 such that @xmath431 strongly in @xmath432 . proof of case ( ii ) and convergence in case ( i ) . by , lemma [ lemma : lengthbound ] and , there exists a constant @xmath128 such that @xmath435}|{\gamma}^n(t)| + |\dot { \gamma}^n(t)|\leq \max_{t\in [ 0,1]}|{\gamma}^n(t)| + |\ell({\gamma}^n)|\leq c\ ] ] for every @xmath411 . therefore , by arzel - ascoli theorem and weak-@xmath305 compactness in @xmath436 there exists a subsequence , which we do not relabel , and a there exists a continuous limit curve @xmath36\to { \mathbb{r}}^2 $ ] such that @xmath437;{\mathbb{r}}^2),\label{eq : prop1:0}\\ \dot { \gamma}^n & { \stackrel{*}{\rightharpoonup}}\dot { \gamma}\quad \text { weakly-$*$ in } l^\infty((0,1);{\mathbb{r}}^2)\label{eq : prop1:1}.\end{aligned}\ ] ] moreover , up to extracting a further subsequence , since @xmath438 is just a bounded sequence of real numbers , we can assume that @xmath439 . the point now is to prove that @xmath440 . if @xmath441 , by lower - semicontinuity of the @xmath19-norm with respect to the weak topology @xmath442 i.e. , @xmath443 strongly in @xmath444 and @xmath445 strongly converges in @xmath432 to a constant @xmath22 . if , by absurd @xmath226 , then there exists @xmath133 such that @xmath446 for all @xmath364 big enough and thus ( as in ( * ? ? ? * lemma 3.1 ) ) @xmath447 and therefore , @xmath448 , which contradicts . this covers case @xmath449 of proposition [ prop : compsurf ] . we also note that @xmath450 indeed , the only if " part follows directly from lemma [ lemma : lengthbound ] and , while the if " part is a consequence of , and . in the following steps , in order to prove case @xmath451 , we assume @xmath452 convergence in case ( i ) . we have to show that @xmath453 recall that the total variation of a function @xmath454\to { \mathbb{r}}$ ] is defined as @xmath455}(f):=\sup\left\ { \sum_{j=0}^{n-1}|f(t_{i+1})-f(t_i)| : \ { 0=t_0 , \ldots , t_n=1\ } \text { is a partition of } [ 0,1 ] \right\}.\ ] ] by lemma [ lemma : oscbd ] , for any interval @xmath229@xmath456 therefore , by and , there exists a constant @xmath128 such that @xmath455}(\dot { \gamma}_1^n)\leq \frac{|\dot { \gamma}^n|}{4\pi}\int_0 ^ 1 ( { k}_1^n)^2 + ( { k}_2^n)^2\ , d\mu_{{\gamma}^n}\leq c.\ ] ] by standard compactness in bounded variation spaces ( see , e.g. , @xcite ) , there exists a subsequence @xmath457 and a limit function @xmath458 such that @xmath459 since @xmath110 is also bounded in @xmath460 @xmath461 and by we can identify the weak limit @xmath110 with the strong limit @xmath462 . we have thus proved @xmath463 since , by hypothesis , @xmath464 for all @xmath465 , @xmath466 for all @xmath467 $ ] . let @xmath468:{\gamma}_1(t)=0\}.\ ] ] note that @xmath469 is a compact subset of @xmath32 $ ] , so , in particular , it is lebesgue - measurable . assume , by contradiction , that @xmath470 owing to the bound on the energy , there is a constant @xmath128 such that @xmath471 by @xmath472 is bounded from above and we can find a new constant @xmath473 such that @xmath474 which implies that @xmath475 , strongly in @xmath476 . since @xmath477 , by @xmath478 strongly in @xmath476 . since strong convergence implies pointwise convergence ( up to extracting a subsequence ) , we obtain that @xmath479 for a.e . @xmath480 . on the other hand , since every @xmath480 is a point of minimum for @xmath481 , it must be @xmath482 for a.e . we have thus obtained a contradiction and we conclude that @xmath483 . for @xmath484 define @xmath485:{\gamma}_1(t)>\frac 2h \right\}.\ ] ] since @xmath486 and @xmath32\backslash e \subseteq \cup_{h \in { \mathbb{n}}}a_h$ ] , we have that @xmath487\backslash a_h|=0.\ ] ] our aim is now to show that @xmath488 in fact , since @xmath489\backslash a_h|,\ ] ] then and imply that @xmath490 and thus @xmath491 in order to prove , note that since @xmath492 converges uniformly to @xmath493 , for every @xmath494 there is @xmath339 such that @xmath495 let now @xmath496 be fixed , @xmath497 given as above and let @xmath498 . by , for all @xmath499 @xmath500 as we did for @xmath501 , we can then control @xmath502 and conclude . strong convergence of @xmath501 and @xmath503 , in particular , implies @xmath504 convergence in case ( i ) . the proof of convergence follows from the strong-@xmath19 convergence of @xmath505 . integrating by parts on @xmath506 , for all @xmath507 @xmath508 by and we can pass to the limit as @xmath509 and integrate back by parts @xmath510 in the same way , we get that @xmath511 limits and prove convergence . let the area and volume constraints @xmath512 be given , such that the isoperimetric inequality is satisfied . let the set @xmath53 and the functional @xmath63 be given as in the statement of theorem [ th : main ] . let @xmath513 , where @xmath514 , be a sequence of systems of revolution surfaces such that @xmath515 note that by @xmath516 since @xmath111 satisfies @xmath79 , for a suitable @xmath517 , by lemma [ lemma : fincard ] there is a constant @xmath128 such that @xmath518 for all @xmath411 . we can thus extract a subsequence ( not relabeled ) and find @xmath295 , such that @xmath519 for all @xmath364 . by corollary [ cor : realbd ] the total length of the curves generating @xmath111 is uniformly bounded and therefore it is not restrictive to assume that there exists @xmath413 such that every @xmath111 is contained in the ball of radius @xmath81 centered at the origin . thus @xmath111 is _ admissible _ in the sense of section [ ssec : compact ] . we can therefore apply the compactness result , i.e. proposition [ prop : compsurf ] . up to extracting @xmath520 subsequences , we find generalized generators @xmath521 and real numbers @xmath522 , for @xmath523 , such that , as @xmath524 @xmath525 in the sense of convergence - and @xmath526 strongly in @xmath432 . denoting by @xmath113 the system of surfaces generated by @xmath527 , by the lower - semicontinuity proposition [ prop : lscs ] @xmath528 so that , by , @xmath529 . to conclude , we have to show that @xmath530 , or , more precisely , that @xmath113 can be ( re)parametrized by curves in ( g0 ) or ( g1 ) , which satisfy the area and volume constraints . by section [ ssec : constr ] , the constraints are continuous with respect to strong convergence in @xmath432 , so that @xmath531 and @xmath532 let @xmath22 be a generator of @xmath113 , we distinguish two cases . if @xmath198 for all @xmath467 $ ] , then @xmath22 is the limit of a sequence @xmath533 ( g1 ) , which implies that @xmath22 is closed ; by lemma [ lemma : w2loc ] @xmath22 satisfies and we infer that @xmath193 ( g1 ) . in the second case , @xmath534 . since it s not restrictive to assume that @xmath68 , we can apply corollary [ cor : split ] and conclude that @xmath22 can be reparametrized as a union of curves in ( g0 ) , which generate a family of surfaces with the same area , enclosed volume , and helfrich energy as the one generated by @xmath22 . the proof of theorem [ th : main ] is thus complete . as a final note , recalling remark [ rem : index ] on the index of a system of curves , if we restrict the minimization to the class of systems of disjoint curves such that @xmath535 for almost every @xmath536 , by continuity of the index under uniform convergence , we obtain that the index takes values 0 or 1 also for the limit system . in particular , this implies that the minimizer is without self - crossings . b. venkatesan , j. polans , j. comer , s. sridhar , d. wendell , a. aksimentiev , and r. bashir . lipid bilayer coated al@xmath538o@xmath539 nanopore sensors : towards a hybrid biological solid - state nanopore . , 13(4):671682 , 2011 .

construction of high rate space time block codes ( stbcs ) with low decoding complexity has been studied widely using techniques such as sphere decoding and non maximum - likelihood ( ml ) decoders such as the qr decomposition decoder with m paths ( qrdm decoder ) . recently ren et al . , presented a new class of stbcs known as the block orthogonal stbcs ( bostbcs ) , which could be exploited by the qrdm decoders to achieve significant decoding complexity reduction without performance loss . the block orthogonal property of the codes constructed was however only shown via simulations . in this paper , we give analytical proofs for the block orthogonal structure of various existing codes in literature including the codes constructed in the paper by ren et al . we show that codes formed as the sum of clifford unitary weight designs ( cuwds ) or coordinate interleaved orthogonal designs ( ciods ) exhibit block orthogonal structure . we also provide new construction of block orthogonal codes from cyclic division algebras ( cdas ) and crossed - product algebras ( cpas ) . in addition , we show how the block orthogonal property of the stbcs can be exploited to reduce the decoding complexity of a sphere decoder using a depth first search approach . simulation results of the decoding complexity show a 30% reduction in the number of floating point operations ( flops ) of bostbcs as compared to stbcs without the block orthogonal structure .

1210.3449

consider a minimal - delay space - time coded rayleigh quasi - static flat fading mimo channel with full channel state information at the receiver ( csir ) . the input output relation for such a system is given by @xmath0 where @xmath1 is the channel matrix and @xmath2 is the additive noise . both @xmath3 and @xmath4 have entries that are i.i.d . complex - gaussian with zero mean and variance 1 and @xmath5 respectively . the transmitted codeword is @xmath6 and @xmath7 is the received matrix . the ml decoding metric to minimize over all possible values of the codeword @xmath8 is @xmath9 [ ld_stbc_def ] @xcite : a linear stbc @xmath10 over a real ( 1-dimensional ) signal set @xmath11 , is a finite set of @xmath12 matrices , where any codeword matrix belonging to the code @xmath10 is obtained from , @xmath13 by letting the real variables @xmath14 take values from a real signal set @xmath15 where @xmath16 are fixed @xmath12 complex matrices defining the code , known as the weight matrices . the rate of this code is @xmath17 complex symbols per channel use . we are interested in linear stbcs , since they admit sphere decoding ( sd ) @xcite and other qr decomposition based decoding techniques such as the qrdm decoder @xcite which are fast ways of decoding for the variables . designing stbcs with low decoding complexity has been studied widely in the literature . orthogonal designs with single symbol decodability were proposed in @xcite , @xcite , @xcite . for stbcs with more than two transmit antennas , these came at a cost of reduced transmission rates . to increase the rate at the cost of higher decoding complexity , multi - group decodable stbcs were introduced in @xcite , @xcite , @xcite . another set of low decoding complexity codes known as the fast decodable codes were studied in @xcite . fast decodable codes have reduced sd complexity owing to the fact that a few of the variables can be decoded as single symbols or in groups if we condition them with respect to the other variables . fast decodable codes for asymmetric systems using division algebras have been reported @xcite . the properties of fast decodable codes and multi - group decodable codes were combined and a new class of codes called fast group decodable codes were studied in @xcite . a new code property called the _ block - orthogonal _ property was studied in @xcite which can be exploited by the qr - decomposition based decoders to achieve significant decoding complexity reduction without performance loss . this property was exploited in @xcite to reduce to the average ml decoding complexity of the golden code @xcite and also in @xcite to reduce the worst - case complexity of the golden code with a small performance loss . while the other low decoding complexity stbcs use the zero entries in the upper left portion of the upper triangular matrix after the qr decomposition , these decoders utilize the zeroes in the lower right portion to reduce the complexity further . the contributions of this paper are as follows : * we generalize the set of sufficient conditions for an stbc to be block orthogonal provided in @xcite for sub - block sizes greater than 1 . * we provide analytical proofs that the codes obtained from the sum of clifford unitary weight designs ( cuwds ) @xcite exhibit the block orthogonal property when we choose the right ordering and the right number of matrices . * we provide new methods of construction of bostbcs using coordinate interleaved orthogonal designs ( ciods ) @xcite , cyclic division algebras ( cdas ) @xcite and crossed product algebras ( cpas ) @xcite along with the analytical proofs of their block orthogonality . * we show that the ordering of variables of the stbc used for the qr decomposition dictates the block orthogonal structure and its parameters . * we show how the block orthogonal property of the stbcs can be exploited to reduce the decoding complexity of a sphere decoder which uses a depth first search approach . * we provide bounds on the maximum possible reduction in the euclidean metrics ( em ) calculation during sphere decoding of bostbcs . * simulation results show that we can reduce the decoding complexity of existing stbcs by upto 30% by utilizing the block orthogonal property . the remaining part of the paper is organized as follows : in section [ sec2 ] the system model and some known classes of low decoding complexity codes are reviewed . in section [ sec3 ] , we derive a set of sufficient conditions for an stbc to be block orthogonal and also the effect of ordering of matrices on it . in section [ sec4 ] , we present proofs of block orthogonal structure of various existing codes and also discuss some new methods of constructions of the same . in section [ sec5 ] , we discuss a method to reduce the number of em calculations while decoding a bostbc using a depth first search based sphere decoder and also derive bounds for the same . simulation results for the decoding complexity of various bostbcs are presented in section [ sec6 ] . concluding remarks constitute section [ sec7 ] . _ notations : _ throughout the paper , bold lower - case letters are used to denote vectors and bold upper - case letters to denote matrices . for a complex variable @xmath18 , denote the real and imaginary part of @xmath18 by @xmath19 and @xmath20 respectively . the sets of all integers , all real and complex numbers are denoted by @xmath21 and @xmath22 , respectively . the operation of stacking the columns of @xmath23 one below the other is denoted by @xmath24 . the kronecker product is denoted by @xmath25 , @xmath26 and @xmath27 denote the @xmath28 identity matrix and the null matrix , respectively . for a complex variable @xmath18 , the @xmath29 operator acting on @xmath18 is defined as follows @xmath30.\ ] ] the @xmath29 operator can similarly be applied to any matrix @xmath31 by replacing each entry @xmath32 by @xmath33 , @xmath34 , @xmath35 , resulting in a matrix denoted by @xmath36 . given a complex vector @xmath37^{t}$ ] , @xmath38 is defined as @xmath39^{t}$ ] . for any linear stbc with variables @xmath40 given by ( [ ld_stbc ] ) , the generator matrix @xmath41 @xcite is defined by @xmath42 where @xmath43^{t}$ ] . in terms of the weight matrices , the generator matrix can be written as @xmath44.\ ] ] hence , for any stbc , can be written as @xmath45 where @xmath46 is given by @xmath47 and @xmath43,$ ] with each @xmath48 drawn from a 1-dimensional ( pam ) constellation . using the above equivalent system model , the ml decoding metric can be written as @xmath49 using @xmath50 decomposition of @xmath51 , we get @xmath52 where @xmath53 is an orthonormal matrix and @xmath54 is an upper triangular matrix . using this , the ml decoding metric now changes to @xmath55 if we have @xmath56 , $ ] where @xmath57 are column vectors , then the @xmath58 and @xmath59 matrices have the following form obtained by the gram - schmidt orthogonalization : @xmath60 , \ ] ] where @xmath61 are column vectors , and @xmath62,\ ] ] where @xmath63 and for @xmath64 @xmath65 a brief overview of the known low decoding complexity codes is given in this section . the codes that will be described are multi - group decodable codes , fast decodable codes and fast group decodable codes . in case of a multi - group decodable stbc , the variables can be partitioned into groups such that the ml decoding metric is decoupled into submetrics such that only the members of the same group need to be decoded jointly . it can be formally defined as @xcite , @xcite , @xcite : [ multi_group_decodability ] an stbc is said to be @xmath66-group decodable if there exists a partition of @xmath67 into @xmath66 non - empty subsets @xmath68 such that the following condition is satisfied : @xmath69 whenever @xmath70 and @xmath71 and @xmath72 . if we group all the variables of the same group together in , then the @xmath59 matrix for the sd @xcite , @xcite in case of multi - group decodable codes will be of the following form : @xmath73,\ ] ] where @xmath74 is a square upper triangular matrix . now , consider the standard sd of an stbc . suppose the @xmath59 matrix as defined in turns out to be such that when we fix values for a set of symbols , the rest of the symbols become group decodable , then the code is said to be fast decodable . formally , it is defined as follows : [ fast_decodability ] an stbc is said to be fast sd if there exists a partition of @xmath75 where @xmath76 into @xmath66 non - empty subsets @xmath68 such that the following condition is satisfied for all @xmath77 @xmath78 whenever @xmath79 and @xmath80 and @xmath81 where @xmath82 and @xmath83 are obtained from the @xmath50 decomposition of the equivalent channel matrix @xmath56 = \textbf{q}\textbf{r}$ ] with @xmath57 as column vectors and @xmath84 $ ] with @xmath61 as column vectors as defined in . hence , by conditioning @xmath85 variables , the code becomes @xmath66-group decodable . as a special case , when no conditioning is needed , i.e. , @xmath86 , then the code is @xmath66-group decodable . the @xmath59 matrix for fast decodable codes will have the following form : @xmath87,\ ] ] where @xmath88 is an @xmath89 block diagonal , upper triangular matrix , @xmath90 is a square upper triangular matrix and @xmath91 is a rectangular matrix . fast group decodable codes were introduced in @xcite . these codes combine the properties of multi - group decodable codes and the fast decodable codes . these codes allow each of the groups in the multi - group decodable codes to be fast decoded . the @xmath59 matrix for a fast group decodable code will have the following form : @xmath92,\ ] ] where each @xmath93 will have the following form : @xmath94,\ ] ] where @xmath95 is an @xmath96 block diagonal , upper triangular matrix , @xmath97 is a square upper triangular matrix and @xmath98 is a rectangular matrix . block orthogonal codes introduced in @xcite are a sub - class of fast decodable / fast group decodable codes . they impose an additional structure on the variables conditioned in these codes . an stbc is said to be block orthogonal if the @xmath99 matrix of the code has the following structure : @xmath100,\ ] ] where each @xmath101 is a block diagonal , upper triangular matrix with @xmath102 blocks @xmath103 , each of size @xmath104 and @xmath105 are non - zero matrices . the low decoding complexity codes described in section [ sec2 ] utilize the zero entries in the upper triangular matrix @xmath99 , in the breadth first or depth first search decoders such as the sphere decoder or the qrdm decoder to achieve decoding complexity reduction . the fast sphere decoding complexity @xcite of an stbc is governed by the zeros in the upper left block of the @xmath99 matrix and does not exploit the zeros in the lower right blocks . the zeros in the lower right block can be used to reduce the average decoding complexity of the code where the average decoding complexity refers to the average number of floating operations performed by the decoder . the zeros in the lower right block are also utilized in some non ml decoders such as the qrdm decoder @xcite or the modified sphere decoder @xcite to reduce the decoding complexity of the code . the structure of block orthogonal matrix was defined in . in general , the size of block diagonal matrices , @xmath106 s , and the upper triangular blocks in these matrices can be arbitrary . similar to @xcite , we consider only the case that @xmath106s have the same size , @xmath107 , and the upper triangular blocks in @xmath106s each have the same size @xmath108 . hence , a block orthogonal code can be represented by the parameters @xmath109 : * @xmath110 : the number of matrices @xmath106 in @xmath99 ; * @xmath102 : the number of blocks in the block diagonal matrix @xmath106 - denoted by @xmath111 , @xmath112 ; * @xmath113 : the number of diagonal entries in the matrices @xmath111 . a set of sufficient conditions for an stbc to be a bostbc with the parameters @xmath114 are described below : first a condition for the stbc to be block orthogonal with parameters @xmath115 is given . the case for @xmath116 will be given subsequently . [ bostc_lemma1 ] @xcite consider an stbc of size @xmath117 with weight matrices @xmath118 , @xmath119 . let @xmath120 , ~~ \mathcal{b}_{i } = \left[\begin{array}{cc } \textbf{b}_{i}^{r } & -\textbf{b}_{i}^{i}\\ \textbf{b}_{i}^{i } & \textbf{b}_{i}^{r}\\ \end{array}\right]\ ] ] and @xmath121 _ { 2 t \times 2n_{t}}$ ] , @xmath122 _ { 2 t \times 2n_{t}}$ ] , @xmath123 , @xmath124 and @xmath125 . this stbc has block orthogonal structure @xmath115 if the following conditions are satisfied : * @xmath126 is of dimension @xmath127 . * @xmath128 and @xmath129 for @xmath123 . * @xmath130 and @xmath131 for @xmath132 and @xmath72 . * @xmath133 for @xmath132 and @xmath72 where @xmath134 and each element ( tuple ) of @xmath135 includes four uniquely permuted scalars drawn from @xmath136 . the set of conditions for an stbc to have a block orthogonal structure with parameters @xmath137 is now given . [ bostc_lemma2 ] @xcite let the @xmath99 matrix of an stbc with weight matrices @xmath138 , @xmath139 be @xmath140,\ ] ] where @xmath141 is a @xmath89 block - orthogonal matrix with parameters @xmath142 , @xmath143 is an @xmath144 matrix and @xmath145 is a @xmath107 upper triangular matrix . the stbc will be a block orthogonal stbc with parameters @xmath146 if the following conditions are satisfied : * the matrices @xmath139 are hurwitz - radon orthogonal . * the matrix @xmath143 is para - unitary , i.e. , @xmath147 . the authors in @xcite only discuss the conditions for the block orthogonal codes with parameters @xmath146 . these conditions can be easily derived for bostbcs with parameters @xmath109 as well . we first derive the conditions for @xmath148 . [ bostc_lemma3 ] consider an stbc of size @xmath149 with weight matrices @xmath150 , @xmath151 . let the @xmath59 matrix for this stbc be of the form @xmath140,\ ] ] where @xmath141 and @xmath152 are @xmath153 upper triangular matrices , @xmath143 is an @xmath153 matrix . the stbc will have a block orthogonal structure with parameters @xmath154 if the following conditions are satisfied : * the matrices @xmath155 are @xmath102-group decodable with @xmath113 variables in each group , i.e. , @xmath155 can be partitioned into @xmath102 sets @xmath156 , each of cardinality @xmath113 such that @xmath157 for all @xmath158 , @xmath159 , @xmath160 . * the matrices @xmath161 are @xmath102-group decodable with @xmath113 variables in each group , i.e. , @xmath161 can be partitioned into @xmath102 sets @xmath156 , each of cardinality @xmath113 such that @xmath162 for all @xmath163 , @xmath164 , @xmath160 . * the set of matrices @xmath165 are such that the @xmath59 matrix obtained has full rank . * the matrix @xmath166 is a block diagonal matrix with @xmath102 blocks of size @xmath108 . proof is given in appendix [ proof_bostc_lemma3 ] . [ bostc_lemma4 ] let the @xmath99 matrix of an stbc with weight matrices @xmath138 , @xmath161 be @xmath140,\ ] ] where @xmath141 is a @xmath89 block - orthogonal matrix with parameters @xmath167 , @xmath143 is an @xmath168 matrix and @xmath145 is a @xmath153 upper triangular matrix . the stbc will be a block orthogonal stbc with parameters @xmath169 if the following conditions are satisfied : * the matrices @xmath161 are @xmath102-group decodable with @xmath113 variables in each group , i.e. , @xmath161 can be partitioned into @xmath102 sets @xmath156 , each of cardinality @xmath113 such that @xmath162 for all @xmath163 , @xmath164 , @xmath160 . * the set of matrices @xmath170 are such that the @xmath59 matrix obtained has full rank . * the matrix @xmath166 is a block diagonal matrix with @xmath102 blocks of size @xmath108 . proof is given in appendix [ proof_bostc_lemma4 ] . we now show that the block orthogonality property depends on the ordering of the weight matrices or equivalently the ordering of the variables . if we do not choose the right ordering , we will be unable to get the desired structure . [ golden_code ] let us consider the golden code @xcite given by : @xmath171,\ ] ] where @xmath172 , @xmath173 , @xmath174 , @xmath175 and @xmath176 for @xmath177 . if we order the variables ( and hence the weight matrices ) as @xmath178 $ ] , then the @xmath59 matrix for sd has the following structure @xmath179,\ ] ] where @xmath180 denotes non zero entries . this ordering of variables has presented a @xmath181 block orthogonal structure to the @xmath99 matrix . now , if we change the ordering to @xmath182 $ ] , then the @xmath59 matrix for sd has the following structure @xmath183,\ ] ] where @xmath180 denotes non zero entries . this ordering of variables has presented a @xmath184 block orthogonal structure to the @xmath99 matrix . we can also have an ordering which can leave the @xmath99 matrix bereft of any block orthogonal structure such as @xmath185 $ ] . the structure of the @xmath99 matrix in this case will be @xmath186,\ ] ] also note that we have many entries @xmath187 even when the @xmath188-th and the @xmath189-th weight matrices are hr orthogonal such as for cases @xmath190 and @xmath191 etc . code constructions for block orthogonal stbcs with various parameters were presented in @xcite . it was shown via simulations that these constructions were indeed block orthogonal with the aforementioned parameters . we provide analytical proofs for the block orthogonal structure of some of these constructions which include also other well known codes such as the bhv code @xcite , the silver code @xcite and the srinath - rajan code @xcite . we first study some basics of cuwds and ciods . @xcite linear stbcs can be broadly classified as unitary weight designs ( uwds ) and non unitary weight designs ( nuwds ) . a uwd is one for which all the weight matrices are unitary and nuwds are defined as those which are not uwds . clifford unitary weight designs ( cuwds ) are a proper subclass of uwds whose weight matrices satisfy certain sufficient conditions for @xmath66-group ml decodability . to state those sufficient conditions , let us list down the weight matrices of a cuwd in the form of an array as shown in table [ cuwd_table ] . .structure of cuwds [ cols="^,^,^,^ " , ] [ cuwd_table ] all the weight matrices in one column belong to one group . the weight matrices of cuwds satisfy the following sufficient conditions for @xmath66-group ml decodability . * @xmath192 . * all the matrices in the first row except @xmath193 should square to @xmath194 and should pair - wise anti - commute among themselves . * the unitary matrix in the @xmath188-th row and the @xmath189-th column is equal to @xmath195 . the cuwd matrix representation for these matrices for a system with @xmath196 transmit antennas are given below @xcite . let @xmath197 , ~~ \sigma_{2 } = \left[\begin{array}{cc } 0 & j\\ j & 0\\ \end{array}\right ] , ~~ \sigma_{3 } = \left[\begin{array}{cc } 1 & 0\\ 0 & -1\\ \end{array}\right].\ ] ] the representations of the clifford generators are given by : @xmath198 @xmath199 @xmath200 @xmath201 where @xmath202 . the weight matrices of the cuwd for a rate-1 , four group decodable stbc can be derived as follows . let @xmath203 for @xmath204 . let @xmath205 . the weight matrices are now given by @xmath206 for @xmath207 , @xmath208 and where @xmath209 is the binary representation of @xmath210 . coordinate interleaved orthogonal designs ( ciods ) were introduced in @xcite . [ ciod_def ] a ciod for a system with @xmath196 transmit antennas in variables @xmath48 , @xmath211 , @xmath212 even , is a @xmath213 matrix @xmath214 , such that @xmath215,\ ] ] where @xmath216 and @xmath217 are complex orthogonal designs of size @xmath218 and @xmath219 . we now show that stbcs obtained as a sum of rate-1 , four group decodable cuwds exhibit the block orthogonal structure with parameters @xmath220 . [ bostc_cuwd_const ] _ construction i : _ let @xmath221 be a rate-1 , four group decodable stbc obtained from cuwd @xcite with weight matrices @xmath222 . let @xmath223 be an @xmath224 matrix such that the set of weight matrices @xmath225 yield a full rank @xmath59 matrix . then the stbc given by @xmath226 will exhibit a block orthogonal structure with parameters @xmath227 . proof is given in appendix [ app_r_mat_struct_const_1 ] . [ bostc_cuwd_ex ] let us consider the bhv code given by : @xmath228 where @xmath229 and @xmath229 take the alamouti structure , and @xmath230,$ ] @xmath231 $ ] and @xmath232^{t } = \textbf{u}\left [ s_{3 } , s_{4}\right]^{t},$ ] where @xmath233 is a unitary matrix chosen to maximize the minimum determinant . in this case , as per the above construction , @xmath234 . hence , the bhv code is a bostbc with parameters @xmath235 . in this section , we show the block orthogonality property of two constructions from either cyclic division algebras or crossed product algebras over the field @xmath236 . [ bostc_const_2 ] _ construction ii : _ let @xmath23 be an stbc with weight matrices @xmath237 and @xmath238 for the variables @xmath239 $ ] and @xmath240 $ ] respectively such that @xmath241 for @xmath242 . let the weight matrices be chosen such that the @xmath59 matrix has full rank . then the code @xmath23 exhibits the block orthogonal property with parameters @xmath243 if we take the ordering of weight matrices as @xmath244 . proof is given in appendix [ app_r_mat_struct_const_2 ] . [ bostc_cda_ex ] consider any stbc obtained from the cyclic division algebra ( cda ) @xcite over the base field @xmath236 . the structure of such an stbc will be @xmath245,\ ] ] where @xmath246 . the weight matrices of this stbc satisfy the properties of the construction above . hence , this is a bostbc with parameters @xmath247 . the next construction is a special case of the previous construction . [ bostc_const_3 ] _ construction iii : _ let @xmath229 be a two group decodable stbc with weight matrices @xmath237 and @xmath238 for the variables @xmath248 $ ] and @xmath249 $ ] respectively such that @xmath241 for @xmath250 . let @xmath223 be a matrix such that the set of weight matrices @xmath251 yield a full rank @xmath59 matrix . then the stbc given by @xmath252 will exhibit a block orthogonal structure with parameters @xmath253 . proof is given in appendix [ app_r_mat_struct_const_3 ] . [ bostc_cda_2g_ex ] consider the golden code as given in example [ golden_code ] . if we consider , @xmath254,\ ] ] and @xmath223 as @xmath255,\ ] ] we can see that the golden code is a bostbc with parameters @xmath256 . in this section we show that the bostbcs that can be obtained from ciods @xcite . [ bostc_ciod_const ] _ construction iv : _ let @xmath257 be a rate-1 ciod with weight matrices @xmath258 . let @xmath223 be a matrix such that the set of weight matrices @xmath259 yield a full rank @xmath59 matrix . then the stbc given by @xmath260 will exhibit a block orthogonal structure with parameters @xmath261 . proof is given in appendix [ app_r_mat_struct_const_4 ] . [ bostc_ciod_ex ] consider the @xmath262 code constructed by srinath et al . in @xcite given by @xmath263,\ ] ] if we consider , @xmath264,\ ] ] and @xmath223 as @xmath265,\ ] ] we see that the code is a bostbc with parameters @xmath256 . in this section we describe how we can achieve decoding complexity reduction for bostbcs . also we show how the block orthogonal structure helps in the reduction of the euclidean metric ( em ) calculations and the sorting operations for a sphere decoder using a depth first search algorithm . we also briefly present the implications of the block orthogonal structure for qrdm decoders as discussed in @xcite . the sphere decoder under consideration in this section will be the depth first search algorithm based decoder with schnorr - euchner enumeration and pruning as discussed in @xcite . we first consider the case of @xmath266 block orthogonal code . consider a bostbc with parameters @xmath267 . the structure of the @xmath99 matrix for this code is as mentioned in with two blocks @xmath268 and @xmath269 . this code is fast sphere decodable , i.e. , for a given set of values of variables in sub - blocks @xmath270 , @xmath271 , we can decode the variables in @xmath272 and @xmath273 , @xmath274 , independently . the ml decoding complexity of this code will be @xmath275 . due to the structure of the block orthogonal code , we can see that the variables in the blocks @xmath270 and @xmath276 , @xmath274 , are also independent in the sense that the em calculations and the schnorr - euchner enumeration based sorting operations for the variables in @xmath270 are independent of the values taken by the variables in @xmath276 . we illustrate this point with an example . [ bostbc_em_ind_ex ] consider a hypothetical bostbc having the parameters @xmath277 with variables @xmath278 . the @xmath59 matrix for this bostbc will be of the form @xmath279\ ] ] the first two levels of the search tree for the sphere decoder are shown in in figure [ fig : bostbc_em_ind ] with the variables assumed to be taking values from a 2-pam constellation - a. as it can be seen from the figure , irrespective of the value taken by @xmath280 , the edge weights ( euclidean metrics ) for the variable @xmath281 remain the same . from example [ bostbc_em_ind_ex ] we can see that instead of calculating the em repeatedly , we can store these values in a look up table when they are calculated for the first time and retrieve them whenever needed . this technique of avoiding repeated calculations by storing the previously calculated values is known as _ memoization _ @xcite . this approach reduces the number of floating point operations ( flops ) significantly . consider a bostbc with parameters @xmath282 . the structure of the @xmath99 matrix for this code is as mentioned in . consider the block @xmath283 , @xmath284 of the @xmath59 matrix . for a given set of values for the variables in the blocks @xmath285 , @xmath286 , we can see that the variables in the blocks @xmath287 and @xmath288 , @xmath274 , are independent as seen in the case of @xmath148 . hence , we can use memoization here as well in order to reduce the number of em calculations and sorting operations . we calculate the maximum possible reduction in the number of em values calculated and the memory requirements for the look up tables in this section . first we consider the case of @xmath148 . considering a @xmath289 bostbc , we first calculate the memory requirements for storing the em values . let each of the variables of the stbc take values from a constellation of size @xmath290 . the number of em values that need to be stored for a single sub - block @xmath270 , @xmath291 , is @xmath292 these values will need to be stored for @xmath293 such sub - blocks . the total memory requirement for the block @xmath269 is , @xmath294 we now find the maximum number of reductions possible for the em calculations for this bostbc . this will occur when all the nodes are visited in the depth first search . for the block @xmath269 , the number of em calculations for a code without the block orthogonal structure would be @xmath295 for a bostbc , if we use the look up table , we would be performing the em calculations only once per each of the sub - block . for @xmath102 sub - blocks , the number of em calculations will be @xmath296 we therefore perform only a small percentage of em calculations if the code exhibits a block orthogonal structure . we call the ratio of the the number of em calculated for a bostbc to the number of em calculated if the stbc did not possess a block orthogonal structure as euclidean metric reduction ratio ( emrr ) given by @xmath297 which is a decreasing function of @xmath102 , @xmath290 and @xmath298 . considering a @xmath300 bostbc , we first calculate the memory requirements for storing the em values . the memory requirement per sub - block @xmath287 , @xmath291 , of any block @xmath301 , @xmath302 , under consideration is the same as that of the case of the sub - block @xmath270 in the @xmath148 case . this is so because , for a given set of values for the variables in the blocks @xmath285 , @xmath303 , the memory requirement for the sub - block @xmath287 can be calculated in the similar way as it was calculated for @xmath270 for the @xmath148 case . hence , the memory requirements for a block @xmath301 for a given set of values for the variables in the blocks @xmath285 is the same as that of @xmath269 in the @xmath148 case . @xmath304 we can reuse the same memory for another set of given values of the variables of @xmath285 , as the previous em values will not be retrieved again as the depth first search algorithm does not revisit any of the previously visited nodes ( i.e. , any previously given set of values for the variables in the tree ) . hence , we can write , @xmath305 for @xmath306 . since there are @xmath307 such blocks , the total memory requirement for storing the em values will be @xmath308 we now find the maximum number of reductions possible for the em calculations for this bostbc . this will occur when all the nodes are visited in the depth first search . for blocks other than @xmath268 , the number of em calculations for a code without the block orthogonal structure would be @xmath309 for a bostbc , if we consider the block @xmath301 and for a given set of values for the variables in @xmath285 , @xmath310 , if we use the look up table , we would be performing the em calculations only once per each of the sub - block . for @xmath102 sub - blocks , the number of em calculations will be @xmath311 these calculations need to be repeated for all the @xmath312 values of the variables in @xmath285 . @xmath313 the em calculations for all the blocks is given by @xmath314 the emrr in this case will be @xmath315 we can see that the ratio of the reduction of operations is independent of @xmath316 and dependent only on @xmath102 and @xmath298 . in this section we review the simplified qrdm decoding method which exploits the block orthogonal structure of a code as presented in @xcite . the traditional qrdm decoder is a breadth first search decoder in which @xmath317 surviving paths with the smallest euclidean metrics are picked at each stage and the rest of the paths are discarded . if @xmath318 for a block orthogonal code with parameters @xmath282 , then the qrdm decoder gives ml performance . the simplified qrdm decoder utilizes the block orthogonal structure of the code to find virtual paths between nodes , which reduces the number of surviving paths to effectively @xmath319 , to reduce the number of euclidean metric calculations . for details of how this is achieved , refer to @xcite . the maximum reduction in decoding complexity bound for a qrdm decoder is given by @xmath320 in all the simulation scenarios in this section , we consider quasi - static rayleigh flat fading channels and the channel state information ( csi ) is known at the receiver perfectly . any stbc which does not have a block orthogonal property is assumed to be a fast decodable stbc which is conditionally @xmath102 group decodable with @xmath298 symbols per group , but not possessing the block diagonal structure for the blocks @xmath321 . we first plot the emrr for bostbcs with different parameters against the snr . figures [ fig : stbc_241_em ] and [ fig : stbc_222_em ] show the plot of @xmath322 vs snr for a @xmath235 bostbc ( examples - silver code , bhv code ) with the symbols being drawn from 4-qam , 16-qam and 64-qam . we can clearly see that the reduction in the emrr with the increasing size of signal constellation as explained in section [ cplxity_redn_mem_req ] . it can also be seen that a larger value of @xmath102 gives a lower emrr if we keep the product @xmath323 constant . figure [ fig : stbc_242_em ] shows the plot of @xmath322 vs snr for a @xmath324 bostbc ( examples - @xmath325 code from pavan et al @xcite ) with the symbols being drawn from 4-qam and 16-qam . notice that the @xmath324 bostbc offers a lower emrr as compared to the @xmath326 bostbc due to the higher value of @xmath298 , as explained in section [ cplxity_redn_mem_req ] . we now compare the total number of flops performed by the sphere decoder for a bostbc against that of an stbc without a block orthogonal structure for various snrs . figures [ fig : stbc_241_flops ] , [ fig : stbc_222_flops ] , [ fig : stbc_242_flops ] show the plot of number of flops vs snr for a @xmath235 bostbc , a @xmath327 bostbc and a @xmath324 bostbc respectively with the symbols being drawn from 4-qam , 16-qam and 64-qam for the first two figures and from 4-qam and 16-qam for the last one . we can see that the bostbcs offer around 30% reduction in the number of flops for the @xmath235 and @xmath324 bostbcs and around 15% for the @xmath327 bostbc at low snrs . for a bostbc with parameters @xmath326,width=624,height=259 ] for a bostbc with parameters @xmath328,width=624,height=259 ] for a bostbc with parameters @xmath329,width=624,height=259 ] , width=624,height=259 ] , width=624,height=259 ] , width=624,height=259 ] the primary difference between the depth first and the breadth first ( qrdm ) approach is the variation of the emrr with respect to snr . as seen in the figures from section [ ml_simulations ] , the effect of the block orthogonal property reduces as the snr increases in the depth first sphere decoder . this is owing to the schnorr - euchner enumeration and pruning of branches . as the snr increases , the decoder needs to visit fewer number of nodes in order to find the ml solution and hence the emrr also tends to 1 . however , in the case of a breadth first search algorithm , all the nodes need to be visited in order to arrive to a solution . hence the emrr is independent of the snr in the breadth first search case . to reduce the number of nodes visited , only @xmath317 paths are selected in the qrdm algorithm to reduce complexity . the value of @xmath330 chosen needs to be varied with snr in order to get near ml performance . in this paper we have studied the block orthogonal property of stbcs . we have shown that this property depends upon the ordering of weight matrices . we have also provided proofs of various existing codes exhibiting the block orthogonal property . a method of exploiting the block orthogonal structure of the stbcs to reduce the sphere decoding complexity was also given with bounds on the maximum possible reduction . 1 b. hassibi and b. hochwald , high - rate codes that are linear in space and time , ieee trans . inf . theory , vol . 48 , no . 7 , pp . 1804 - 1824 , july 2002 . e. viterbo and j. boutros , a universal lattice code decoder for fading channels , ieee trans . theory , vol . 5 , pp . 1639 - 1642 , july 1999 . t. p. ren , y. l. guan , c. yuen and e. y. zhang , block - orthogonal space time code structure and its impact on qrdm decoding complexity reduction , ieee journal of selected topics in signal processing , vol . 5 , issue 8 , pp . 1438 - 1450 , nov . v. tarokh , h. jafarkhani and a. r. calderbank , space - time block codes from orthogonal designs , ieee trans . theory , vol . 5 , pp . 1456 - 1467 , july 1999 . liang , orthogonal designs with maximal rates , ieee trans . theory , vol.49 , no . 2468 - 2503 , oct . o. tirkkonen and a. hottinen , square - matrix embeddable space - time block codes for complex signal constellations , ieee trans . theory , vol . 2 , pp . 384 - 395 , feb . d. n. dao , c. yuen , c. tellambura , y. l. guan and t. t. tjhung , four - group decodable space - time block codes , ieee trans . signal processing , vol . 424 - 430 , jan . 2008 . s. karmakar and b. s. rajan , multigroup decodable stbcs from clifford algebras , ieee trans . theory , vol . 223 - 231 , jan . 2009 . s. karmakar and b. s. rajan , high - rate , multi - symbol - decodable stbcs from clifford algebras , ieee transactions on inf . theory , vol . . 2682 - 2695 , jun . e. biglieri , y. hong and e. viterbo , on fast - decodable space - time block codes , ieee trans . theory , vol . 2 , pp . 524 - 530 , feb . r. vehkalahti , c. hollanti and f. oggier , fast - decodable asymmetric space - time codes from division algebras , available online at arxiv , arxiv:1010.5644v1 [ cs.it ] . t. p. ren , y. l. guan , c. yuen and r. j. shen , fast - group - decodable space - time block code , proceedings ieee information theory workshop , ( itw 2010 ) , cairo , egypt , jan . 6 - 8 , 2010 , available online at http://www1.i2r.a-star.edu.sg/cyuen/publications.html . m. o. sinnokrot and j. barry , fast maximum - likelihood decoding of the golden code , ieee transactions on wireless commun . , vol . 9 , no . 26 - 31 , jan . 2010 . j. c. belfiore , g. rekaya and e. viterbo , the golden code : a 2x2 full - rate space time code with non - vanishing determinants , ieee trans . theory , vol . 1432 - 1436 , apr . 2005 . s. kahraman and m. e. celebi , dimensionality reduction for the golden code with worst - case complexity of @xmath331 , available online at http://istanbultek.academia.edu/sinankahraman . g. s. rajan and b. s. rajan , multi - group ml decodable collocated and distributed space time block codes , ieee trans . theory , vol . 56 , no . 7 , pp . 3221 - 3247 , july 2010 . z. ali khan md . , and b. s. rajan , single symbol maximum likelihood decodable linear stbcs , ieee trans . theory , vol . 5 , pp . 2062 - 2091 , may 2006 . b. a. sethuraman , b. s. rajan and v. shashidhar , full - diversity , high - rate space - time block codes from division algebras , ieee trans . theory , vol . 2596 - 2616 , oct 2003 . v. shashidhar , b. s. rajan and b. a. sethuraman , information - lossless space - time block codes from crossed - product algebras , ieee trans . theory , vol . 9 , pp . 39133935 , sep 2006 . o. damen , a. chkeif , and j.c . belfiore , lattice code decoder for space - time codes , ieee communication letters , vol . 161 - 163 , may 2000 . g. r. jithamithra and b. s. rajan , minimizing the complexity of fast sphere decoding of stbcs , available online at arxiv , arxiv:1004.2844v2 [ cs.it ] , 22 may 2011 . c. hollanti , j. lahtonen , k. ranto , r. vehkalahti and e. viterbo , on the algebraic structure of the silver code : a 2 2 perfect space - time code with non - vanishing determinant , in proc . of ieee inf . theory workshop , porto , portugal , may 2008 . k. p. srinath and b. s. rajan , low ml - decoding complexity , large coding gain , full - rate , full - diversity stbcs for 2x2 and 4x2 mimo systems , ieee journal of selected topics in signal processing : special issue on managing complexity in multiuser mimo systems , vol . 916 - 927 , dec . 2009 . t. h. cormen , c. e. leiserson , r. l. rivest , c. stein , introduction to algorithms , third edition , mit press , sep 2009 . following the system model in section [ sec2 ] , we have the equivalent channel matrix @xmath332 as @xmath333 ~=~ \left [ \textbf{h}_{1 } ~ ... ~ \textbf{h}_{l } ~ \textbf{h}_{l+1 } ~ ... ~ \textbf{h}_{2l}\right ] $ ] . we know from theorem 2 of @xcite that , if any two weight matrices @xmath334 and @xmath335 are hurwitz - radon orthogonal , then the @xmath188-th and the @xmath189-th columns of the @xmath336 matrix are orthogonal . due to the conditions on the weight matrices , we have that @xmath337 and @xmath338 are block diagonal with @xmath102 blocks , each of size @xmath104 . under @xmath339 decomposition , @xmath340 with @xmath341 $ ] with @xmath342 and @xmath343 $ ] as mentioned . it can be seen from lemma 2 of @xcite that the matrix @xmath268 is block diagonal with @xmath102 blocks , each of size @xmath104 . we can now write , @xmath344 @xmath345 simplifying , @xmath346 now , if @xmath347 is block diagonal with @xmath102 blocks of size @xmath104 each @xmath348 is block diagonal with @xmath102 blocks of size @xmath104 each . since @xmath269 is upper triangular and full rank , this means that @xmath269 is block diagonal with @xmath102 blocks of size @xmath104 each . following the system model in section [ sec2 ] , we have the equivalent channel matrix @xmath349 as @xmath333 ~=~ \left [ \textbf{h}_{1 } ~ ... ~ \textbf{h}_{l } ~ \textbf{h}_{l+1 } ~ ... ~ \textbf{h}_{l+l}\right ] $ ] . we know from theorem 2 of @xcite that , if any two weight matrices @xmath350 and @xmath351 are hurwitz - radon orthogonal , then the @xmath188-th and the @xmath189-th columns of the @xmath336 matrix are orthogonal . due to the conditions on the weight matrices , we have that @xmath338 is block diagonal with @xmath102 blocks , each of size @xmath104 . under @xmath339 decomposition , @xmath340 with @xmath341 $ ] with @xmath352 and @xmath353 and @xmath343 $ ] as mentioned . we can now write , @xmath344 @xmath345 simplifying , @xmath346 now , if @xmath347 is block diagonal with @xmath102 blocks of size @xmath104 each @xmath348 is block diagonal with @xmath102 blocks of size @xmath104 each . since @xmath269 is upper triangular and full rank , this means that @xmath269 is block diagonal with @xmath102 blocks of size @xmath104 each . according to construction i , the structure of the stbc is @xmath354 where @xmath229 is a rate-1 four group decodable stbcs obtained from cuwds as described in section [ cuwds ] . let the @xmath99 matrix for this code have the following structure : @xmath140,\ ] ] where @xmath268 , @xmath355 and @xmath269 are @xmath356 matrices . from @xcite , it can be easily seen that @xmath357 has a block diagonal structure with four blocks , and each block of the size @xmath358 . @xmath359,\ ] ] where @xmath360 , @xmath361 is a @xmath358 given by . @xmath362,\ ] ] ' '' '' [ r1_struct_prop_const_1 ] the non - zero blocks of the matrix @xmath268 are equal i.e. , @xmath363 , for @xmath364 . it is sufficient for us to prove that @xmath365 and @xmath366 for @xmath364 , @xmath367 and @xmath368 . the proof is by induction . we first consider the case of @xmath369 . we also recall @xcite that @xmath370 now , for we have , @xmath371 since @xmath372 and @xmath373 for @xmath374 . for we have , @xmath375 since @xmath376 . now we prove equations and for arbitrary @xmath189 . we prove this by induction . let the equations hold true for all @xmath377 . we now have for equation , @xmath378 which follows from the induction hypothesis and the fact that @xmath379 for @xmath380 . for equation , @xmath381 \\ % \end{equation * } % \begin{equation * } & = \frac{1}{2\parallel \textbf{r}_{j } \parallel } \left [ tr\left ( \check{\textbf{h } } \check{\textbf{a}}_{j } \check { \textbf{a}}_{k}^{t } \check { \textbf{h}}^{t}\right ) - \sum_{l=1}^{j-1 } \left\langle \textbf{q}_{l } , \textbf{h}_{j}\right\rangle \left\langle \textbf{q}_{l } , \textbf{h}_{k}\right\rangle\right ] \\ % \end{equation * } % \begin{equation * } & = \frac{tr\left ( \check{\textbf{h } } \check{\textbf{a}}_{j } \check{\textbf{a}}_{4\left ( i-1\right ) \lambda + 1 } \check { \textbf{a}}_{4\left ( i-1\right ) \lambda + 1}^{t } \check { \textbf{a}}_{k}^{t } \check { \textbf{h}}^{t}\right)}{2\parallel \textbf{r}_{4\left ( i-1\right ) \lambda + j } \parallel } \\ & \quad - \frac{1}{2\parallel \textbf{r}_{4\left ( i-1\right ) \lambda + j } \parallel}\sum_{l=1}^{j-1 } \left\langle \textbf{q}_{4\left ( i-1\right ) \lambda + l } , \textbf{h}_{4\left ( i-1\right ) \lambda + j}\right\rangle.\\ & \qquad \left\langle \textbf{q}_{4\left ( i-1\right ) \lambda + l } , \textbf{h}_{4\left ( i-1\right ) \lambda + k}\right\rangle % \end{equation * } % \begin{equation*}\end{aligned}\ ] ] @xmath382 the matrix @xmath355 is key for the block orthogonality property of the stbc in question . it is required to be para - unitary for achieving this property . the structure of the matrix @xmath355 for construction i is described in the following proposition . [ e_struct_prop_const_1 ] the matrix @xmath355 is of the form @xmath383,\ ] ] where @xmath384 , @xmath177 are @xmath358 matrices and @xmath385 is a @xmath358 permutation matrix given by @xmath386.\ ] ] let us represent the matrix @xmath355 using @xmath358 blocks as : @xmath387,\ ] ] we first prove that @xmath388 for @xmath364 . the proof is by induction on the rows of the matrix @xmath389 . the first row entries of the matrix @xmath389 are given by @xmath390 and for the matrix @xmath391 are given by @xmath392 due to the construction of the stbc , we have @xmath393 , for @xmath394 . using this , we get @xmath395 now , let us assume that row @xmath396 of @xmath391 is equal to the row @xmath396 of @xmath389 for all @xmath397 . the @xmath189-th row of @xmath389 is given by @xmath398 and the @xmath189-th row of @xmath391 is given by @xmath399 @xmath400 we now prove that @xmath401 . the proofs for the matrices @xmath402 and @xmath403 are very similar . first step is to prove that @xmath404 . the proof is by induction on the rows of the matrix @xmath405 . the first row entries of the matrix @xmath405 are given by @xmath406 and for the matrix @xmath407 are given by @xmath408 due to the construction of the stbc , we have @xmath393 , for @xmath394 . using this , we get @xmath409 now , let us assume that row @xmath396 of @xmath405 is equal to the row @xmath396 of @xmath407 for all @xmath397 . the @xmath189-th row of @xmath405 is given by @xmath410 and the @xmath189-th row of @xmath407 is given by @xmath411 @xmath412 we now prove that @xmath413 . the proof is by induction on the rows of the matrix @xmath407 . the first row entries of the matrix @xmath407 are given by @xmath408 due to the construction of the stbc , we have @xmath393 , for @xmath394 . using this , we get @xmath414 we need to show that this is equal to @xmath415 . @xmath416 substituting the values of the weight matrices from for @xmath417 , @xmath418 and @xmath419 , and simplifying , we see that it is sufficient to show that @xmath420 or equivalently , @xmath421 since @xmath422 and @xmath102 are one s complement of each other in the binary representation , we have , @xmath423 therefore we have , @xmath424 the equality for @xmath402 and @xmath403 can be shown similarly . [ r2_struct_prop_const_1 ] the matrix @xmath269 is block diagonal with @xmath425 blocks , each of size @xmath358 . for the matrix @xmath269 to be block diagonal with @xmath425 blocks , each of size @xmath358 , we need to satisfy the following conditions * the matrices @xmath426 form a four group decodable stbc with @xmath427 variables per group * the matrix @xmath355 is such that @xmath347 is block diagonal with @xmath425 blocks , each of size @xmath358 . since the matrices @xmath428 form a four group decodable stbc with @xmath427 variables per group , it is easily seen that the matrices @xmath426 also form a four group decodable stbc with @xmath427 variables per group as @xmath429 \textbf{m}^{h } = \textbf{0}$ ] for @xmath188 and @xmath189 in different groups . we now introduce some notation before we address the structure of the matrix @xmath430 . let @xmath396 be an integer such that @xmath431 . we denote by @xmath432 , the binary representation of @xmath433 using @xmath434 bits . let @xmath435 denote the bitwise xor operation between any two binary numbers . now , we turn to the structure of the matrix @xmath355 . from proposition [ e_struct_prop_const_1 ] , we know the structure of the matrix @xmath355 . computing @xmath347 , we see that for it to be block diagonal with @xmath425 blocks , each of size @xmath358 , it is sufficient to show that the matrices @xmath436 are symmetric with identical entries on the diagonal for @xmath437 . the entries of @xmath436 are given by @xmath438 expanding and simplifying , we get @xmath439 where @xmath440 @xmath441 and @xmath442 is given by @xmath443 for @xmath444 and @xmath445 . we now see that for every @xmath396 , there exists a unique @xmath446 such that @xmath447 as @xmath448 \check { \textbf{a}}_{\lambda + l}^{t}\check { \textbf{m}}^{t } \check { \textbf{h}}^{t}\right)\\ % \end{equation * } % \begin{equation * } & = \left\langle \textbf{h}_{m^ { ' } } , \textbf{h}_{4 \lambda + l } \right\rangle,\end{aligned}\ ] ] where @xmath449 . similarly , for every @xmath450 , there exists a unique @xmath451 such that @xmath452 where @xmath453 . we can now write , @xmath454 if @xmath455 . let @xmath456 . @xmath457 is given by , @xmath458 . therefore , we can see that @xmath459 is symmetric . using the above arguments , it is also easly seen that the diagonal elements of the matrix @xmath459 are identical . hence , we have shown that the matrix @xmath269 is block diagonal with @xmath425 blocks , each of size @xmath358 . the stbc @xmath461 can be written as @xmath462 where @xmath463 . tweaking the system model in section [ sec2 ] , we can get a generator matrix for this stbc as @xmath464.\ ] ] hence , can be written as @xmath465 where @xmath466 is given by @xmath467 and @xmath43,$ ] with each @xmath48 drawn from a 2-dimensional constellation . it can be easily seen that @xmath468 . let the qr decomposition of the complex matrix @xmath469 yield matrices @xmath470 and @xmath471 . using the relation : if @xmath472 , then @xmath473 , we can see that @xmath474 . the qr decomposition of a complex matrix yields a unitary @xmath58 matrix and an upper triangular matrix @xmath59 with real diagonal entries . hence , the diagonal entries of the matrix @xmath471 are real . since @xmath474 , we ll have @xmath475 for @xmath476 . hence , the stbc @xmath461 exhibits a block orthogonal property with parameters @xmath477 . let the @xmath99 matrix for this code have the following structure : @xmath140,\ ] ] where @xmath268 , @xmath355 and @xmath269 are @xmath478 matrices . from @xcite , it can be easily seen that @xmath268 has a block diagonal structure with two blocks , and each block of the size @xmath479 . @xmath480,\ ] ] where @xmath481 and @xmath482 are @xmath479 upper triangular matrices . [ r1_struct_prop_const_3 ] the non - zero blocks of the matrix @xmath268 are equal i.e. , @xmath483 . proof is similar to the proof of proposition [ r1_struct_prop_const_1 ] . the structure of the matrix @xmath355 is described in the following proposition . [ e_struct_prop_const_3 ] the matrix @xmath355 is of the form @xmath484,\ ] ] where @xmath384 , @xmath177 are @xmath479 matrices . proof is similar to the proof of proposition [ e_struct_prop_const_1 ] . [ r2_struct_prop_const_3 ] the matrix @xmath269 is block diagonal with @xmath485 blocks , each of size @xmath479 . proof is similar to the proof of proposition [ r2_struct_prop_const_1 ] . as only rate-1 ciods are considered in this construction , this can only be done for either @xmath486 ciods or @xmath487 ciods . the structure of the @xmath59 matrix obtained from the @xmath262 ciod is the same as the structure of @xmath59 matrix obtained from the construction iii . the proof of the structure is also the same as given in appendix [ app_r_mat_struct_const_4 ] . we now consider the structure of the @xmath59 matrix obtained from using a @xmath487 ciod . let the @xmath99 matrix for this code have the following structure : @xmath140,\ ] ] where @xmath268 , @xmath355 and @xmath269 are @xmath488 matrices . from @xcite , it can be easily seen that @xmath268 has a block diagonal structure with @xmath425 blocks , and each block of the size @xmath486 . @xmath359,\ ] ] where @xmath360 are @xmath486 upper triangular matrices for @xmath489 .

in this paper , we show that any unknotting tunnel for a two bridge knot is isotopic to either one of known ones . this together with morimoto sakuma s result gives the complete classification of unknotting tunnels for two bridge knots up to isotopies and homeomorphisms .

math9911254

let @xmath0 be a knot in the 3sphere @xmath1 . the _ exterior _ of @xmath0 is the closure of the complement of a regular neighborhood of @xmath0 , and is denoted by @xmath2 . a _ tunnel _ for @xmath0 is an embedded arc @xmath3 in @xmath1 such that @xmath4 . then we denote @xmath5 by @xmath6 , where we regard @xmath3 as obtained from @xmath6 by a radial extension . let @xmath7 , @xmath8 be tunnels for @xmath0 . we say that @xmath7 and @xmath8 are _ homeomorphic _ if there is a self homeomorphism @xmath9 of @xmath2 such that @xmath10 . we say that @xmath11 and @xmath12 are _ isotopic _ if @xmath13 is ambient isotopic to @xmath14 in @xmath2 . we say that a tunnel @xmath3 for @xmath0 is _ unknotting _ if @xmath15int @xmath16 is a genus two handlebody . we note that the unknotting tunnels for @xmath0 is essentially the genus 2 heegaard splittings of @xmath2 ; if @xmath3 is an unknotting tunnel , then we can obtain a genus 2 heegaard splitting @xmath17 , where @xmath18 is a regular neighborhood of @xmath19 in @xmath2 , and @xmath20 , and every genus 2 heegaard splitting of @xmath2 is obtained in this manner . moreover , such heegaard splittings are isotopic ( homeomorphic resp . ) if and only if the corresponding unknotting tunnels are isotopic ( homeomorphic resp . ) . we say that a knot @xmath0 is a 2bridge knot if @xmath0 admits a ( genus zero ) 2bridge position , that is , there exists a genus zero heegaard splitting @xmath21 of @xmath1 such that @xmath22 is a system of 2string trivial arcs in @xmath23 @xmath24 . it is known that each 2bridge knot admits six unknotting tunnels as depicted in figure 1.1 or figure 3.1 ( see @xcite , or @xcite ) . figure 1.1 then the purpose of this paper is to prove : [ main ] every unknotting tunnel for a non - trivial 2bridge knots is isotopic to one of the above six unknotting tunnels . we note that the isotopy , and homeomorphism classes of the above tunnels are completely classified by morimoto sakuma @xcite and y.uchida @xcite , and that it is known that the unknotting tunnels for a trivial knot are mutually isotopic ( see , for example @xcite ) . hence these results together with the above theorem give the complete classification of isotopy , and homeomorphism classes of unknotting tunnels for two bridge knots . throughout this paper , we work in the differentiable category . for a submanifold @xmath25 of a manifold @xmath0 , @xmath26 denotes a regular neighborhood of @xmath25 in @xmath0 . let @xmath27 be a manifold embedded in a manifold @xmath28 with dim@xmath29dim@xmath28 . then @xmath30 denotes the frontier of @xmath27 in @xmath28 . for the definitions of standard terms in 3dimensional topology , we refer to @xcite . let @xmath28 be a compact 3manifold , @xmath31 a union of mutually disjoint arcs or simple closed curves properly embedded in @xmath28 , @xmath32 a surface embedded in @xmath28 , which is in general position with respect to @xmath31 , and @xmath33 a simple closed curve with @xmath34 . a surface @xmath35 in @xmath28 is a _ @xmath31disk _ , if @xmath35 is a disk intersecting @xmath31 in at most one transverse point . we say that @xmath36 is @xmath31_inessential _ if @xmath36 bounds a @xmath31disk in @xmath32 . we say that @xmath36 is @xmath31_essential _ if it is not @xmath31inessential . we say that a disk @xmath35 is a @xmath31_compressing disk _ for @xmath32 if ; @xmath35 is a @xmath31disk , and @xmath37 , and @xmath38 is a @xmath31essential simple closed curve in @xmath32 . the surface @xmath32 is @xmath31_compressible _ if it admits a @xmath31compressing disk , and it is @xmath31_incompressible _ if it is not @xmath31compressible . let @xmath39 , @xmath40 be surfaces embedded in @xmath28 such that @xmath41 , or @xmath42 . we say that @xmath39 and @xmath40 are @xmath31_parallel _ , if there is a submanifold @xmath27 in @xmath28 such that @xmath43 is homeomorphic to @xmath44 as a pair , where @xmath45 is a nion of points in @xmath46 , and @xmath39 ( @xmath40 resp . ) is the closure of the component of @xmath47 containing @xmath48 ( @xmath49 resp . ) if @xmath41 , or @xmath39 ( @xmath40 resp . ) is the surface corresponding to @xmath48 ( @xmath49 resp . ) if @xmath42 . the submanifold @xmath27 is called a _ @xmath31parallelism _ between @xmath39 and @xmath40 . we say that @xmath32 is @xmath31_boundary parallel _ if there is a subsurface @xmath50 in @xmath51 such that @xmath32 and @xmath50 are @xmath31parallel . we say that @xmath32 is _ @xmath31essential _ if @xmath32 is @xmath31incompressible , and not @xmath31boundary parallel . let @xmath52 be an arc properly embedded in @xmath32 with @xmath53 . we say that @xmath52 is _ @xmath31inessential _ if there is a subarc @xmath54 of @xmath55 such that @xmath56 , and @xmath57 bounds a disk @xmath35 in @xmath32 such that @xmath58 , and @xmath52 is _ @xmath31essential _ if it is not @xmath31inessential . we say that @xmath32 is _ @xmath31boundary compressible _ if there is a disk @xmath59 in @xmath28 such that @xmath60 is an @xmath31essential arc in @xmath32 , and @xmath61 . let @xmath39 , @xmath40 be mutually disjoint surfaces in @xmath28 which are in general position with respect to @xmath31 . we say that @xmath39 and @xmath40 are _ @xmath31isotopic _ if there is an ambient isotopy @xmath62 @xmath63 of @xmath28 such that ; @xmath64 ; @xmath65 , and ; @xmath66 for each @xmath67 . * genus @xmath68 @xmath69bridge position * let @xmath70 be a system of mutually disjoint arcs properly embedded in @xmath28 . we say that @xmath71 is a _ system of @xmath69string trivial arcs _ if there exists a system of mutually disjoint disks @xmath72 in @xmath28 such that , for each @xmath73 @xmath74 , we have ( 1 ) @xmath75 , and ( 2 ) @xmath76 is an arc , say @xmath77 , such that @xmath78 . let @xmath79 be a system of 2string trivial arcs in a 3ball @xmath80 . the pair @xmath81 is often refered as _ 2string trivial tangle _ , or a _ rational tangle_. let @xmath0 be a link in a closed 3manifold @xmath28 . let @xmath82 be a genus @xmath68 heegaard splitting . then the next definition is borrowed from @xcite . we say that @xmath0 is in a _ ( genus @xmath68 ) @xmath69bridge position _ ( with respect to the heegaard splitting @xmath83 ) if @xmath84 ( @xmath85 resp . ) is a system of @xmath69string trivial arcs in @xmath86 ( @xmath80 resp . ) . in this paper , we abbreviate a genus 0 @xmath69bridge position to an @xmath69bridge position . a knot @xmath0 is called an _ @xmath69bridge knot _ if it admits an @xmath69bridge position . it is known that the @xmath87bridge positions of a 2bridge knot @xmath0 are unique up to @xmath0isotopy ( see @xcite,@xcite , or section 7 of @xcite ) . [ 2-bridge k - reducible ] we say that a genus @xmath68 bridge position of @xmath0 with respect to @xmath83 is _ weakly @xmath0reducible _ if there exist @xmath0compressing disks @xmath88 , @xmath89 for @xmath90 in @xmath86 , @xmath80 respectively such that @xmath91 . the genus @xmath68 bridge position of @xmath0 with respect to @xmath83 is _ strongly @xmath0irreducible _ if it is not weakly @xmath0reducible . * remark*it is known that the 2bridge positions of a 2bridge knot are strongly @xmath0irreducible ( see proposition 7.5 of @xcite ) . for a 2bridge knot @xmath0 we can obtain four genus one 1bridge positions of @xmath0 as follows . let @xmath83 be the heegaard splitting which gives the 2bridge position , and @xmath92 , @xmath93 , @xmath94 , @xmath95 the closures of the components of @xmath96 , where @xmath97 ( @xmath98 resp . ) is contained in @xmath86 ( @xmath80 resp . ) . let @xmath99 , @xmath100 , @xmath101 , and @xmath102 . then each @xmath103 is a solid torus and it is easy to see that @xmath77 is a trivial arc in @xmath103 ( @xmath104 ) . hence , @xmath105 gives genus one 1bridge position of @xmath0 . moreover , by using @xmath92 , @xmath93 , @xmath95 for @xmath94 , we can obtain other three genus one 1bridge positions of @xmath0 . let @xmath0 be a knot with a genus one 1bridge position with respect to @xmath105 . let @xmath106 , @xmath107 be tunnels for @xmath0 embedded in @xmath108 , @xmath109 respectively as in figure 2.1 . it is easy to see that @xmath106 , @xmath107 are unknotting tunnels , and we call them the _ unknotting tunnels associated to the genus one 1bridge position_. in section 8 of @xcite , it is shown that every genus one 1bridge position for a non - trivial 2bridge knot is obtained as above . hence , by definition ( see also figure 3.1 ) , it is easy to see : [ 1to2 ] let @xmath106 , @xmath107 be unknotting tunnels associated to a genus one 1bridge position of a 2bridge knot @xmath0 . then one of @xmath106 , @xmath107 is isotopic to @xmath110 or @xmath111 , and the other is isotopic to either @xmath112 , @xmath113 , @xmath114 or @xmath115 figure 2.1 let @xmath3 be an unknotting tunnel for @xmath0 . let @xmath116 , @xmath117 . note that @xmath118 is a genus two heegaard splitting of @xmath1 . we say that the heegaard splitting @xmath119 is _ weakly @xmath0reducible _ if there exist @xmath0compressing disks @xmath120 , @xmath121 properly embedded in @xmath122 , @xmath123 respectively such that @xmath124 . the splitting is _ strongly @xmath0irreducible _ if it is not weakly @xmath0reducible . [ weakly reducible ] if @xmath125 is weakly @xmath0reducible , then either @xmath0 is a trivial knot or @xmath0 admits a genus one 1bridge position , where @xmath3 is isotopic to one of the unknotting tunnels associated to the 1bridge position . let @xmath120 ( @xmath126 ) , @xmath121 ( @xmath127 ) be a pair of @xmath0compressing disks which gives weak @xmath0irreducibility . * claim 1 * we may suppose that @xmath120 ( @xmath121 resp . ) is non - separating in @xmath122 ( @xmath123 resp . ) . * proof of claim 1*suppose that @xmath121 is separating in @xmath123 . then @xmath121 cuts @xmath123 into two solid tori , say @xmath108 , @xmath109 . by exchanging the suffix , if necessary , we may suppose that @xmath128 . then take a meridian disk @xmath129 in @xmath109 such that @xmath130 . we may regard @xmath129 as a ( non - separating essential ) disk in @xmath123 , and we have @xmath131 . by regarding @xmath129 as @xmath121 , we see that we may suppose that @xmath121 is non - separating in @xmath123 . suppose that @xmath120 is separating in @xmath122 . since @xmath0 does not intersect @xmath120 in one point , we have @xmath132 . the disk @xmath120 cuts @xmath122 into two solid tori @xmath133 , @xmath134 , where @xmath0 is a core circle of @xmath133 . if @xmath135 , then the above argument works to show that there exists a non - separating meridian disk for @xmath122 giving weak @xmath0reducibility together with @xmath121 . if @xmath136 , then we take a meridian disk @xmath137 for @xmath133 such that @xmath138 , and @xmath137 intersects @xmath0 transversely in one point . we may regard @xmath137 a ( non - separating essential ) @xmath0disk in @xmath122 , and we have @xmath139 . by regarding @xmath137 as @xmath120 , we see that we may suppose that @xmath120 , @xmath121 are non - separating in @xmath122 , @xmath123 respectively . now we have the following two cases . * case 1*@xmath132 . let @xmath140 be the solid torus obtained from @xmath122 by cutting along @xmath120 . since @xmath141 is non - separating in @xmath142 and @xmath1 does not contain non - separating 2sphere , we see that @xmath141 is an essential simple closed curve in @xmath143 . since @xmath1 does not contain non - separating 2sphere or punctured lens spaces , @xmath141 is a longitude of @xmath140 , and , hence , there is an annulus @xmath86 in @xmath140 such that @xmath144 . then @xmath145 gives a disk bounding @xmath0 , and this shows that @xmath0 is a trivial knot . * case 2*@xmath146 . let @xmath147 , @xmath148 , @xmath149 , and @xmath150 . note that @xmath93 is a core with respect to a natural 1handle structure on @xmath27 . it is easy to see that @xmath92 is a trivial arc in @xmath108 . let @xmath151 . we regard @xmath93 as an arc properly embedded in @xmath109 . * claim 2*@xmath109 is a solid torus and @xmath93 is a trivial arc in @xmath109 . * proof of claim 2*let @xmath152 be the solid torus obtained from @xmath123 by cutting along @xmath121 and @xmath153 . by the arguments in case 1 , we see that @xmath154 is a longitude of @xmath152 . hence @xmath155 is a 3ball and @xmath93 is a trivial arc in @xmath155 . since @xmath123 is obtained from @xmath155 by identifying two disks in @xmath156 corresponding to the copies of @xmath121 , we see that @xmath109 is a solid torus , and @xmath93 is a trivial arc in @xmath109 . hence we see that @xmath105 gives a genus one 1bridge position of @xmath0 . by the construction of @xmath108 , we see that @xmath3 is isotopic to an unknotting tunnel associated to @xmath105 . in @xcite , rubinstein - scharlemann introduced a powerful machinery called _ graphic _ for studying positions of two heegaard surfaces of a 3manifold . successively , dr . osamu saeki and the author introduced an orbifold version of their setting , and showed that the results similar to rubinstein - scharlemann s hold in this setting @xcite . in this section , we quickly review the arguments and apply it to compare decomposing 2spheres giving 2bridge positions , and genus 2 heegaard splittings obtained from an unknotting tunnel for a 2bridge knot . let @xmath0 be a 2bridge knot , that is , there exists a genus zero heegaard splitting @xmath21 of @xmath1 such that @xmath22 is a 2strings trivial arcs in @xmath23 @xmath24 . then the unknotting tunnels @xmath110 , @xmath111 are contained in @xmath157 , @xmath158 respectively as in figure 3.1 . figure 3.1 there is a diffeomorphism @xmath159 such that @xmath160 is the decomposing 2sphere @xmath90 , and that @xmath161 for some @xmath162 , @xmath163 , @xmath164 , @xmath165 . let @xmath3 be an unknotting tunnel for @xmath0 . let @xmath166 , @xmath167 , @xmath117 , and @xmath168 a spine of @xmath123 such that each vertex has valency 3 . note that @xmath118 is a genus two heegaard splitting of @xmath1 . then there is a diffeomorphism @xmath169 . let @xmath170 , and @xmath171 . then for a fixed small constant @xmath172 , we may suppose that @xmath173 looks as one of the following , where @xmath174 or @xmath175 , and @xmath176 . 1 . @xmath173 consists of two transverse simple closed curves @xmath177 , @xmath178 which are @xmath0essential in @xmath179 , and inessential in @xmath180 . @xmath173 consists of a simple closed curve @xmath36 and a figure @xmath181 @xmath182 such that ; @xmath36 is @xmath0essential in @xmath179 , and inessential in @xmath180 , and @xmath182 is arising from a saddle tangency . @xmath173 consists of three transverse simple closed curves @xmath177 , @xmath178 , and @xmath183 such that ; @xmath177 and @xmath178 bound pairwise disjoint @xmath0disks in @xmath179 each of which contains a puncture from @xmath0 , @xmath177 and @xmath178 are parallel in @xmath180 , and ; @xmath183 is @xmath0essential in @xmath179 and inessential in @xmath180 , 4 . @xmath173 consists of two transverse simple closed curves @xmath177 , @xmath178 , and a figure @xmath181 , @xmath182 such that ; @xmath177 and @xmath178 bound pairwise disjoint @xmath0disks in @xmath179 each of which contains a puncture from @xmath0 , @xmath177 and @xmath178 are parallel in @xmath180 , and ; @xmath182 is arising from a saddle tangency . @xmath173 consists of four transverse simple closed curves @xmath177 , @xmath178 , @xmath184 , and @xmath185 such that @xmath177 , @xmath178 , @xmath184 , @xmath185 bound mutually disjoint @xmath0disks in @xmath179 each containing a puncture from @xmath0 , and @xmath177 and @xmath178 ( @xmath184 and @xmath185 resp . ) are pairwise parallel in @xmath180 . moreover , for a fixed @xmath186 , if we move @xmath187 from @xmath188 to @xmath189 , then the intersection @xmath190 ( @xmath191 resp . ) is changed as @xmath192 . figure 3.2 then , by the arguments in section 4 of @xcite , we see that by an arbitrarily small deformation of @xmath193 , and @xmath194 which does not alter @xmath195 \cup [ 1-\varepsilon , 1)}$ ] , and @xmath196}$ ] , we may suppose that the maps are pairwise generic , that is : there is a stratification of @xmath197 which consists of four parts below . regions : : region is a component of the subset of @xmath197 consisting of values @xmath198 such that @xmath179 and @xmath180 intersect transversely , and this is an open set . edges : : edge is a component of the subset consisting of values @xmath198 such that @xmath179 and @xmath180 intersect transversely except for one non - degenerate tangent point . the tangent point is either a center or a saddle . edge is a 1dimensional subset of @xmath199 . crossing vertices : : crossing vertex is a component of the subset consisting of points @xmath200 such that @xmath179 and @xmath180 intersect transversely except for two non - degenerate tangent points . crossing vertex is an isolated point in @xmath199 . in a neighborhood of a crossing vertex , four edges are coming in , where one can regard the crossing vertex as the intersection of two edges . birth - death vertices : : birth - death vertex is a component of the subset consisting of points @xmath198 such that @xmath179 and @xmath180 intersect transversely except for a single degenerate tangent point . in particular , there is a parametrization @xmath201 of @xmath202 such that @xmath203 , and @xmath204 . birth - death vertex is an isolated point in @xmath199 , and in a neighborhood of a birth - death vertex , two edges are coming in , with one from center tangency , the other from saddle tangency . let @xmath205 be the union of edges and vertices above . by the above , @xmath205 is a 1complex in @xmath199 . then we note that as in section 3 of @xcite , @xmath205 naturally extends to @xmath206 . here we note that , by the configurations ( 1 ) @xmath207 ( 5 ) above , @xmath205 looks as in figure 3.3 near the bottom corners of @xmath202 . we note that the arguments in section 6 of @xcite which uses labels on the regions hold without changing proofs in this setting . hence the argument in the proof of proposition 5.9 of @xcite which uses a simplicial map to a certain complex ( called @xmath0 in @xcite ) works in our setting , and this shows ( note that @xmath208 is always strongly @xmath0irreducible ( remark of definition [ 2-bridge k - reducible ] ) ) . figure 3.3 [ prop5 - 6 ] suppose that @xmath118 is strongly @xmath0irreducible , and @xmath0 is not a trivial knot in @xmath1 . then there is an unlabelled region in @xmath209 . and we also have ( see corollary 6.22 of @xcite ) : [ corollary : essential intersection ] suppose that @xmath118 is strongly @xmath0irreducible and @xmath0 is not a trivial knot in @xmath1 . then , by applying @xmath0isotopy , we may suppose that @xmath90 and @xmath210 intersect in non - empty collection of simple closed curves which are @xmath0essential in @xmath90 , and essential in @xmath210 . in this section , we give a proof of theorem [ main ] . for the statements and the proofs of lemmas b-1 , c-1 , c-2 , c-3 , d-2 , d-3 , d-4 which are used in this section , see appendix of this paper . let @xmath0 be a non - trivial 2bridge knot and @xmath110 , @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath3 , @xmath208 , @xmath118 be as in the previous section . we note that the closures of @xmath212 consist of two disks with each intersecting @xmath0 in two points , and annuli . since the disks are contained in @xmath122 , @xmath211 consists of even number of components . the proof of theorem [ theorem2 ] is carried out by the induction on the number of the components . as the first step of the induction , we show : * proof*since @xmath120 is separating in @xmath122 , the component of @xmath214 corresponding to @xmath154 is separating in @xmath142 . hence , by lemma c-2 , we see that @xmath86 is compressible or boundary parallel in @xmath123 . suppose that @xmath86 is compressible in @xmath123 . since @xmath1 does not contain non - separating 2sphere , we see that @xmath121 is also separating in @xmath122 , and , hence , @xmath120 and @xmath121 are pairwise parallel in @xmath122 . let @xmath215 be the annulus in @xmath210 such that @xmath216 . by exchanging suffix , if necessary , we may suppose that @xmath215 is properly embedded in @xmath157 . since each component of @xmath217 is an unknotted arc , we see that @xmath215 is an unknotted annulus in @xmath157 , and this implies that @xmath86 and @xmath215 are parallel in @xmath157 , and , hence , in @xmath123 ie , @xmath86 is boundary parallel . * proof*assume , without loss of generality , that there is a @xmath0compressing disk @xmath222 for @xmath120 . note that since @xmath223 consists of two points , @xmath224 and @xmath154 are parallel in @xmath225 . let @xmath226 be the annulus in @xmath120 bounded by @xmath227 . let @xmath137 be the disk in @xmath120 bounded by @xmath224 . then we have the following two cases . we consider the 2sphere @xmath229 in @xmath122 . let @xmath230 be the 3ball in @xmath122 bounded by @xmath229 . since @xmath0 does not contain a local knot in @xmath122 , we see that @xmath231 is an unknotted arc properly embedded in @xmath230 . hence there is an ambient isotopy of @xmath1 which moves @xmath231 to an arc in @xmath120 joining @xmath232 , and which does not move @xmath233 . on the other hand , @xmath234 is a component of the strings of the trivial tangle @xmath235 . this shows that @xmath0 is a trivial knot , a contradiction . in this case , we first consider the disk @xmath236 . by a slight deformation of @xmath236 , we obtain a @xmath0compressing disk @xmath237 for @xmath121 such that @xmath238 is contained in @xmath157 . then , by the argument as in case ( a ) , we see that @xmath0 is a trivial knot , a contradiction . let @xmath241 , @xmath242 be the closures of the components of @xmath243 such that @xmath244 @xmath24 . then @xmath245 is a torus with one hole properly embedded in @xmath158 . by lemma d-2 , we may suppose , by exchanging suffix if necessary , that there is a @xmath0compressing disk @xmath222 for @xmath241 such that @xmath246 , and @xmath247 consists of a point . we consider the genus one surface @xmath242 properly embedded in @xmath158 . by lemma b-1 , we see that @xmath242 is @xmath0compressible in @xmath158 . let @xmath237 be the @xmath0compressing disk for @xmath242 . now we have the following subsubcases . by the @xmath0incompressibility of @xmath121 ( claim 2 ) , we see that @xmath248 ie , @xmath249 consists of a point . then @xmath250 cuts @xmath251 into a 2string trivial tangle which is @xmath0isotopic to @xmath252 . hence @xmath3 is isotopic to @xmath110 . * proof*suppose that @xmath254 . then , by compressing @xmath242 along @xmath237 , we obtain a disk @xmath255 properly embedded in @xmath158 such that @xmath256 , and @xmath255 separates the components of @xmath257 . let @xmath258 , @xmath259 be the closures of the components of @xmath260 such that @xmath261 , @xmath262 . then we can isotope @xmath263 rel @xmath264 in @xmath265 to an arc in @xmath266 without moving @xmath217 . since @xmath120 and @xmath121 are @xmath0parallel in @xmath122 , this shows that @xmath0 is a trivial knot , a contradiction . let @xmath267 be the closure of the component of @xmath268 such that @xmath269 . note that @xmath267 is a solid torus in @xmath158 with @xmath270 . by regarding @xmath267 as a very thin solid torus , we may suppose that @xmath271 consists of a disk @xmath272 intersecting @xmath0 in one point . then @xmath273 is an annulus @xmath274 @xmath275 . * proof*assume that @xmath274 is compressible in @xmath123 . then , by compressing @xmath274 , we obtain a disk @xmath276 in @xmath123 such that @xmath277 . since @xmath272 intersects @xmath0 in one point , @xmath272 is a non - separating disk in @xmath122 . hence , we see that @xmath278 is a non - separating 2sphere in @xmath1 , a contradiction . * proof * assume that @xmath86 is not boundary parallel . since @xmath1 does not contain non - separating 2sphere , we see that @xmath86 is incompressible in @xmath123 . hence , by lemma c-1 , we see that there is an essential disk @xmath35 for @xmath123 such that @xmath281 , and that @xmath35 cuts @xmath123 into two solid tori @xmath108 , @xmath109 , where @xmath282 . moreover , since @xmath1 does not contain a punctured lens space , we see that each component of @xmath214 represents a generator of the fundamental group of the solid torus @xmath108 . however this contradicts lemma c-3 . by lemma d-3 , there is a @xmath0boundary compressing disk @xmath59 for @xmath120 or @xmath121 , say @xmath120 , such that @xmath284 . let @xmath241 be the closure of the component of @xmath285 which is a torus with two holes . let @xmath286 . then @xmath59 is a compressing disk for @xmath108 . let @xmath255 be the disk obtained by compressing @xmath108 along @xmath59 , and @xmath129 a disk obtained by pushing @xmath287 slightly into @xmath288 . we may regard @xmath129 is properly embedded in @xmath158 . suppose that @xmath129 is @xmath0compressible in @xmath158 . then we can show that @xmath0 is a trivial knot by using the argument as in the proof of claim 1 of case 1.2.2 . hence @xmath129 is @xmath0incompressible in @xmath158 . hence , by lemma b-1 ( 3 ) , either @xmath129 and @xmath121 are @xmath0parallel or @xmath289 bounds a 2string trivial tangle in @xmath122 , which is not a @xmath0parallelism between @xmath121 and @xmath129 . * proof*by lemma b-1 ( 2 ) , we see that @xmath129 and @xmath290 bounds a @xmath0parallelism in @xmath158 . hence , by isotopy , we can move @xmath90 to the position such that @xmath291 , and @xmath292 . then , by applying the argument of case 2.1.1 with regarding @xmath121 , @xmath129 as @xmath120 , @xmath121 respectively , we see that @xmath3 is isotopic to @xmath111 . let @xmath293 be a compressing disk for @xmath120 or @xmath121 , say @xmath120 , in @xmath122 . then @xmath294 and @xmath154 are parallel in @xmath225 , and let @xmath295 be the annulus in @xmath120 bounded by @xmath296 . let @xmath297 be a disk properly embedded in @xmath122 which is obtained by moving @xmath298 slightly so that @xmath299 . * proof*assume that @xmath301 . then we may regard @xmath302 is a @xmath0compressing disk for @xmath121 in @xmath122 . then , by using the arguments in case ( a ) of the proof of claim 2 of case 1 , we can show that @xmath0 is a trivial knot , a contradiction . * proof of assertion*let @xmath133 be a sufficiently small regular neighborhood of @xmath303 , and @xmath304 . note that @xmath134 is a handlebody , because @xmath110 is an unknotting tunnel for @xmath0 . let @xmath237 be a non - separating essential disk properly embedded @xmath134 . * proof*assume that @xmath310 . let @xmath311 be the solid torus obtained by cutting @xmath133 along @xmath305 . note that @xmath311 is a regular neighborhood of @xmath0 . since @xmath237 is non - separating in @xmath134 , and @xmath1 does not contain a non - separating 2sphere , @xmath312 is an essential simple closed curve in @xmath313 , and @xmath312 is not contractible in @xmath311 . this shows that @xmath0 bounds a disk which is an extension of @xmath237 . hence @xmath0 is a trivial knot , a contradiction . hence @xmath307 consists of a number of arcs joining points in @xmath314 to points in @xmath220 . here , by using cut and paste arguments , we remove the components of @xmath315 which are inessential in @xmath220 . * proof*assume that @xmath316 is contained in @xmath122 . since @xmath317 , this implies that @xmath110 is contained in a regular neighborhood of @xmath0 , contradicting the fact that @xmath110 is an unknotting tunnel . * proof*assume that @xmath319 . then we can show that there is a non - separating disk @xmath320 properly embedded in @xmath123 such that @xmath321 by using the argument as in the proof of claim 1 of the proof of proposition [ weakly reducible ] . then by using the argument as in the proof of claim 2 above , we can show that @xmath0 is a trivial knot , a contradiction . * proof*assume that @xmath322 consists of a point . let @xmath316 be the disk in @xmath237 bounded by @xmath36 . then @xmath323 and @xmath324 intersects in one point , and this shows that @xmath325 is a trivial arc in @xmath2 , a contradiction . let @xmath326 . we call the boundary component of @xmath327 corresponding to @xmath312 the _ outer boundary_. other boundary components of @xmath328 (: the components of @xmath329 ) are called _ inner boundary components_. let @xmath330 be the solid torus obtained by cutting @xmath122 along @xmath297 . let @xmath36 be an inner boundary component which is outermost with respect to the intersection @xmath331 , that is : let @xmath316 be the disk in @xmath237 bounded by @xmath36 . let @xmath334 be the components of @xmath331 , which are located on @xmath328 in this order , where @xmath335 @xmath336 cobounds a square @xmath337 in @xmath328 . let @xmath338 . let @xmath339 be the image of @xmath220 in @xmath330 . note that @xmath339 is a torus with two holes let @xmath340 . then by the minimality condition , we see that each @xmath341 is an essential arc properly embedded in @xmath339 . * proof*assume that @xmath345 are mutually parallel in @xmath339 . then we can take a simple closed curve @xmath183 in @xmath220 such that @xmath183 intersects @xmath323 transversely in one point , and @xmath346 is ambient isotopic to @xmath341 in @xmath339 . let @xmath311 be a regular neighborhood of @xmath347 in @xmath122 such that @xmath348 . note that @xmath311 is a solid torus , and @xmath324 wraps around @xmath313 longitudally @xmath69 times . this show that the 3sphere contains a lens space with fundamental group a cyclic group of order @xmath69 , a contradiction . by subclaim 1 , we see that we can take simple closed curves @xmath349 , @xmath350 in @xmath220 such that @xmath351 , @xmath352 @xmath353 intersects @xmath323 transversely in one point , @xmath354 is ambient isotopic to @xmath344 in @xmath339 , and @xmath355 is ambient isotopic to @xmath341 @xmath356 in @xmath339 . let @xmath357 be a regular neighborhood of @xmath358 in @xmath122 such that @xmath359 , and @xmath360 . then @xmath357 is a genus two handlebody , and @xmath295 is an annulus in @xmath361 . note that @xmath362 is a regular neighborhood of @xmath0 . then we denote by @xmath363 the closure of the exterior of this regular neighborhood of @xmath0 . note that @xmath295 is embedded in @xmath364 . then attach @xmath365 to @xmath357 along @xmath366 . it is directly observed ( see figure 4.11 ) that we obtain a solid torus , say @xmath367 , such that @xmath295 wraps around @xmath368 longitudally @xmath69times . then , let @xmath369 . note that @xmath370 is an annulus properly embedded in @xmath363 . assume that @xmath370 is compressible in @xmath363 . then the compressing disk is not contained in @xmath367 since @xmath370 is incompressible in @xmath367 . hence @xmath367 together with a regular neighborhood of this compressing disk produces a punctured lens space with fundamental group a cyclic group of order @xmath69 in @xmath1 , a contradiction . hence @xmath370 is incompressible in @xmath363 . then assume that @xmath370 is boundary parallel , and let @xmath371 be the corresponding parallelism . since @xmath372 , @xmath371 is not @xmath367 . hence @xmath373 , and this shows that @xmath363 is a solid torus , which implies that @xmath0 is a trivial knot , a contradiction . hence @xmath370 is an essential annulus in @xmath363 , and this completes the proof of claim 5 . suppose that @xmath342 contains at least two proper isotopy classes in @xmath339 . we suppose that @xmath341 , @xmath374 @xmath375 belong to mutually different isotopy classes . let @xmath376 , @xmath377 be the components of @xmath378 . since @xmath324 and @xmath323 intersects transversely , we easily see that we may suppose that @xmath379 , and @xmath380 . let @xmath311 be the solid torus obtained by cutting @xmath122 along @xmath297 , and @xmath381 ( , hence , @xmath382 is homeomorphic to ( torus)@xmath383 $ ] ) . here we may regard that @xmath133 is obtained from @xmath384 by adding a 1handle @xmath385 corresponding to @xmath386 , where @xmath387 is a core of @xmath385 . let @xmath388 , @xmath389 be the components of the image of @xmath110 in @xmath382 , where we may regard that @xmath384 is obtained from @xmath390 by adding @xmath391 . * proof*by extending @xmath394 ( @xmath395 resp . ) to the cores of @xmath391 , we obtain either an annulus which contains @xmath388 or @xmath389 ( if @xmath396 ( @xmath397 resp . ) is contained in @xmath376 or @xmath377 ) , or a rectangle two edges of which are @xmath388 and @xmath389 ( if @xmath396 ( @xmath397 resp . ) joins @xmath376 and @xmath377 ) in @xmath382 . in this case , we obtain an annulus @xmath398 by taking the union of the rectangles from @xmath337 and @xmath399 . since @xmath341 and @xmath374 are not ambient isotopic in @xmath339 , @xmath398 is incompressible in @xmath382 . we note that every incompressible annulus in @xmath400 $ ] with one boundary component contained in @xmath401 , the other in @xmath402 is vertical ( for a proof of this , see , for example , @xcite ) . hence @xmath398 is vertical , and this shows that @xmath392 is vertical . in this case , we see that @xmath403 or @xmath389 is vertical by the existence of the annulus from @xmath399 . then the existance of the rectangle from @xmath337 shows that @xmath403 and @xmath404 are parallel , and this implies that @xmath392 is vertical . assertion shows that @xmath3 is isotopic to @xmath110 or there is an essential annulus in @xmath2 , and this together with the conclusions of cases 1 , and 2.1 shows that we have the conclusions of lemma [ firststep ] for all cases . [ induction ] suppose that @xmath211 consists of more than two components . then we can deform @xmath210 by an ambient isotopy in @xmath2 to reduce @xmath405 still with non - empty intersection each component of which is @xmath0essential in @xmath90 , and essential in @xmath210 . * claim 1*suppose that there is an annulus component @xmath86 of @xmath409 ( @xmath410 or @xmath87 ) such that @xmath86 is @xmath0compressible in @xmath23 . then the @xmath0compressing disk is disjoint from @xmath0 . * proof*let @xmath35 be the @xmath0compressing disk for @xmath86 . assume that @xmath411 ie , @xmath412 consists of a point . then , by compressing @xmath86 along @xmath35 , we obtain two disks each of which intersects @xmath0 in one point . but this is impossible , since each component of @xmath214 separates @xmath413 into two disks each intersecting @xmath0 in two points . * claim 2*suppose that there is an annulus component @xmath414 in @xmath415 , and an annulus component @xmath416 in @xmath417 . then either @xmath414 or @xmath416 is @xmath0incompressible in @xmath157 or @xmath158 . * proof*we first suppose that @xmath416 is @xmath0compressible in @xmath158 . then , by claim 1 , the @xmath0compressing disk is disjoint from @xmath0 . hence , by compressing @xmath416 along the disk , we obtain two disks in @xmath158 which are @xmath0essential in @xmath158 and disjoint from @xmath0 . let @xmath418 be one of the disks . assume , moreover , that @xmath414 is also @xmath0compressible . then , by using the same argument , we obtain a @xmath0essential disk @xmath419 in @xmath157 such that @xmath420 . note that @xmath421 and @xmath422 are parallel in @xmath423 . this implies that @xmath0 is a two - component trivial link , a contradiction . * proof*note that there are at most three mutually non - parallel , disjoint essential simple closed curves on @xmath210 . hence if @xmath424 , then there are three components , say @xmath177 , @xmath178 , @xmath184 , of @xmath211 which are mutually parallel on @xmath210 . we may suppose that @xmath177 , @xmath178 , @xmath184 are located on @xmath210 successively in this order . let @xmath425 ( @xmath426 resp . ) be the annulus on @xmath210 bounded by @xmath427 ( @xmath428 resp . ) . without loss of generality , we may suppose that @xmath425 ( @xmath426 resp . ) is properly embedded in @xmath157 ( @xmath158 resp . ) . since @xmath0 is connected , we may suppose , by exchanging suffix if necessary , that each component of @xmath429 separates the boundary points of each component of @xmath217 on @xmath90 . since each component of @xmath217 is an unknotted arc , we see that @xmath425 is an unknotted annulus . hence there is an annulus @xmath430 in @xmath90 such that @xmath431 and @xmath430 and @xmath425 are pairwise ( @xmath0)parallel in @xmath157 . let @xmath27 be the parallelism between @xmath430 and @xmath425 . if @xmath432 , then we can push the components of @xmath433 out of @xmath157 along the parallelism @xmath27 , still with at least two components of intersection @xmath427 . if @xmath434 , then we can push @xmath425 out of @xmath157 along this parallelism to reduce @xmath405 by two . according to claim 3 and its proof , we suppose that @xmath435 or @xmath436 , and no three components of @xmath211 are mutually parallel in @xmath210 . note that the intersection numbers of any simple closed curves on @xmath210 with @xmath211 are even , because @xmath90 is a separating surface . this shows that @xmath211 consists of two ( in case when @xmath437 ) or three ( in case when @xmath438 ) parallel classes in @xmath210 each of which consists of two components . hence , each component of @xmath409 ( @xmath410 or @xmath87 , say 1 ) is an annulus . if a component of @xmath415 is @xmath0incompressible in @xmath157 , then , by the argument in the proof of claim 3 , we have the conclusion of lemma [ induction ] . hence , in the rest of the proof , we suppose that each component of @xmath415 is a @xmath0compressible annulus in @xmath157 . let @xmath439 be the closure of the component of @xmath440 such that @xmath441 . note that @xmath439 is a 3ball such that @xmath442 consists of some components of @xmath415 . then , by the assumptions , we see that @xmath442 consists of either one , two , or three annuli . * proof*let @xmath443 . since @xmath425 is compressible , there is a @xmath0compressing disk @xmath293 for @xmath425 in @xmath157 . note that @xmath444 ( claim 1 ) . we may regard that @xmath293 is properly embedded in @xmath122 and @xmath293 is parallel to @xmath120 and @xmath121 in @xmath122 . since @xmath0 is connected , we see that @xmath293 is non - separating in @xmath122 . by cutting @xmath122 along @xmath293 , we obtain a solid torus @xmath108 such that @xmath0 is a core circle of @xmath108 . recall that @xmath407 are the closures of the components of @xmath212 . note that @xmath445 is properly embedded in @xmath446 . since the 3sphere does not contain a non - separating 2sphere , we see that @xmath445 in incompressible in @xmath108 . since every incompressible surface in ( torus)@xmath447 is either vertical or boundary parallel annulus ( see @xcite ) , @xmath445 is boundary parallel in @xmath108 . let @xmath448 be the parallelism for @xmath445 , and @xmath449 . since @xmath0 is connected , and @xmath0 intersects @xmath120 and @xmath121 , we see that @xmath426 is disjoint from the images of @xmath120 and @xmath121 in @xmath108 . hence we see that @xmath426 is disjoint from the images of @xmath293 in @xmath450 . this shows that the parallelism @xmath448 survives in @xmath122 , and , hence , we have the conclusion of lemma [ induction ] by the argument as in the proof of claim 3 . * claim 5*if @xmath442 consists of two annuli @xmath425 , @xmath426 , then there is a component of @xmath415 which is ( @xmath0)boundary parallel in @xmath157 , and , hence , we have the conclusion of lemma [ induction ] . * proof*by exchanging suffix , if necessary , we may suppose that the annulus @xmath451 is incident to @xmath266 @xmath24 . if @xmath437 , then we have @xmath452 . if @xmath438 , then , by reversing the order of @xmath453 , and changing the suffix of @xmath451 if necessary , we may suppose that @xmath452 . then let @xmath448 be the 3manifold in @xmath157 such that @xmath454 . note that @xmath448 is embedded in @xmath123 and @xmath455 . * proof*assume that both @xmath120 and @xmath121 are separating in @xmath122 . then @xmath120 and @xmath121 are parallel in @xmath122 , but this contradicts the fact that @xmath448 and @xmath0 are connected . since @xmath120 is a non - separating disk in @xmath122 , and @xmath1 does not contain a non - separating 2sphere , we see that @xmath226 is incompressible in @xmath123 . then , since @xmath1 does not contain a punctured lens space with non - trivial fundamental group , we see that @xmath226 is boundary parallel in @xmath123 by lemma c-3 ( see the proof of claim 1 in case 2 of the proof of lemma 4.2 ) . hence @xmath448 is a parallelism between @xmath226 and @xmath425 , and this shows that @xmath425 is @xmath0boundary parallel in @xmath157 along this parallelism to give the conclusion of lemma [ induction ] . * proof*assume that @xmath442 consists of three annuli @xmath425 , @xmath426 , and @xmath456 , where @xmath457 , @xmath458 , and @xmath459 . since @xmath425 , @xmath426 , @xmath456 are @xmath0compressible in @xmath157 , there are mutually disjoint @xmath0compressing disks @xmath419 , @xmath418 , @xmath460 for @xmath425 , @xmath426 , @xmath456 respectively . we may regard that @xmath419 , @xmath418 , @xmath460 are properly embedded in @xmath122 . note that @xmath421 , @xmath422 , @xmath461 are not mutually parallel in @xmath220 . hence we see that @xmath462 cuts @xmath122 into two components @xmath463 , @xmath464 such that one component of @xmath217 is contained in @xmath463 , and the other component is contained in @xmath464 ( see figure 4.14 ) . but this contradicts the fact that @xmath0 is connected . * proof of proposition [ theorem2]*by lemma [ induction ] , we may suppose that @xmath211 consists of two transverse simple closed curves which are @xmath0essential in @xmath90 , and essential in @xmath210 . then , by lemma [ firststep ] , we have the conclusion of proposition [ theorem2 ] . * proof of theorem [ main]*let @xmath3 be an unknotting tunnel for a non - trivial 2bridge knot @xmath0 , and @xmath125 a genus 2 heegaard splitting of @xmath1 obtained from @xmath465 as above . if @xmath125 is weakly @xmath0reducible , then by propositions [ 1to2 ] , and [ weakly reducible ] , we see that @xmath3 is isotopic to @xmath110 , @xmath111 , @xmath112 , @xmath113 , @xmath114 , or @xmath115 . if @xmath125 is strongly @xmath0irreducible , then by corollary [ corollary : essential intersection ] , and proposition [ theorem2 ] , we see that @xmath3 is isotopic to @xmath110 or @xmath111 , or @xmath2 contains an essential annulus . if @xmath2 contains an essential annulus , then @xmath0 is a @xmath466torus knot . then , by @xcite , it is known that every unknotting tunnel for @xmath0 is isotopic to one of @xmath110 or @xmath112 ( and that @xmath110 and @xmath111 are pairwise isotopic , and @xmath112 , @xmath113 , @xmath114 , @xmath115 are mutually isotopic ) . hence we have the conclusion of theorem [ main ] . let @xmath31 be the union of mutually disjoint arcs and simple closed curves properly embedded in a 3manifold @xmath27 such that @xmath27 admits a 2fold branched covering space @xmath468 along @xmath31 . * lemma a-3 * @xmath31incompressible surface @xmath32 is @xmath31boundary parallel if and only if @xmath470 is boundary parallel . in particular , if @xmath472 is irreducible , and @xmath32 is a disk intersecting @xmath31 in one point , and @xmath55 bounds a disk @xmath35 in @xmath473 such that @xmath35 intersects @xmath31 in one point , then @xmath32 is @xmath31boundary parallel ( in fact , @xmath32 and @xmath35 are @xmath31parallel ) . 1 . @xmath32 is a disk disjoint from @xmath79 , and @xmath32 separates the components of @xmath79 . particularly , in this case , @xmath32 is @xmath79essential , 2 . @xmath32 is a @xmath79boundary parallel disk intersecting @xmath79 in at most one point , 3 . @xmath32 is a @xmath79boundary parallel disk intersecting @xmath79 in two points and @xmath32 separates @xmath81 into the parallelism and a rational tangle , or 4 . @xmath32 is a @xmath79boundary parallel annulus such that @xmath474 . * lemma b-2 * let @xmath35 be an @xmath475essential disk in @xmath140 such that @xmath476 consists of two points . then there exists an @xmath475compressing disk @xmath255 for @xmath143 such that @xmath477 and @xmath478 consists of one point . moreover , by cutting @xmath479 along @xmath255 , we obtain a 2string trivial tangle @xmath81 such that @xmath35 is a @xmath79incompressible disk in @xmath81 ( hence , @xmath35 is @xmath79boundary parallel ) . * lemma c-1 * there exists an essential disk @xmath35 in @xmath25 such that @xmath480 . moreover the disk @xmath35 can be taken as a separating disk , or a non - separating disk according as @xmath86 is separating or non - separating . * proof*there exists boundary compressing disk @xmath59 for @xmath86 . apply a boundary compression on @xmath86 along @xmath59 to obtain a disk @xmath255 . by moving @xmath255 by a tiny isotopy , we obtain a desired disk @xmath35 . for a detail , see , for example , @xcite . * proof*let @xmath35 be as in lemma c-1 . by the proof of lemma c-1 , we see that @xmath86 is isotopic to an annulus obtained from @xmath35 by adding a band . by isotopy , we may suppose that @xmath480 . let @xmath140 be the closure of the component of @xmath481 such that @xmath482 . then @xmath140 is a solid torus , and @xmath86 is incompressible in @xmath140 . hence each component of @xmath214 is non - separating in @xmath143 . this implies that each component of @xmath214 is non - separating in @xmath473 . let @xmath120 , @xmath121 be the copies of @xmath35 in @xmath143 . note that @xmath214 separates @xmath143 into two annuli , say @xmath226 , @xmath445 . if @xmath35 is separating in @xmath25 , then @xmath483 is contained in one of @xmath226 or @xmath445 , say @xmath226 . then the components of @xmath214 are mutually parallel in @xmath473 through the annulus @xmath445 . if @xmath35 is non - separating in @xmath25 , then , by exchanging the suffix if necessary , we may suppose that @xmath120 is contained in @xmath226 , and @xmath121 is contained in @xmath445 . this shows that the components of @xmath214 are not parallel in @xmath473 . * lemma c-3 * let @xmath35 be as in lemma c-1 . suppose that @xmath86 is separating in @xmath25 . then each component of @xmath214 does not represent a generator of the fundamental group of the solid torus obtained from @xmath25 by cutting along @xmath35 , which contains @xmath86 . * proof*let @xmath140 be the solid torus obtained from @xmath25 by cutting along @xmath35 such that @xmath482 . then @xmath484 is homeomorphic to @xmath485 as pairs . this shows that the closure of a component of @xmath486 gives a parallelism between @xmath86 and a subsurface of @xmath473 . let @xmath0 be a knot in a genus two handlebody @xmath25 with an essential disk @xmath293 such that @xmath293 cuts @xmath25 into a solid torus , where @xmath0 is a core circle of @xmath140 . note that there exists a two - fold branched cover @xmath487 of @xmath25 along @xmath0 , where @xmath488 is a genus three handlebody . * proof*let @xmath489 be the lift of @xmath35 in @xmath488 . then , by lemmas a-1 and a-3 , we see that @xmath489 is an essential annulus in a genus three handlebody @xmath488 . then @xmath489 is boundary compressible in @xmath488 . hence , by lemma a-2 , we see that @xmath35 is @xmath0boundary compressible in @xmath25 . * lemma d-2 * let @xmath35 , @xmath120 , @xmath121 be as above . suppose , moreover , that @xmath35 is separating in @xmath25 . then @xmath120 and @xmath121 are @xmath0parallel in @xmath25 , and , by cutting @xmath491 along @xmath266 ( @xmath410 or @xmath87 , say 1 ) , we obtain a 1string trivial arc in a solid torus , say @xmath479 . moreover , @xmath121 is @xmath475boundary parallel in @xmath140 . * proof*we note that @xmath35 separates @xmath25 into two solid tori @xmath108 , @xmath109 , where @xmath120 , @xmath121 are properly embedded in @xmath108 . since each @xmath266 intersects @xmath0 in one point , @xmath266 is an essential disk of @xmath108 , and this shows that @xmath120 and @xmath121 are parallel in @xmath108 , and in @xmath25 . then , @xmath469smith conjecture shows that they are actually @xmath0parallel . then , by using @xmath469equivariant loop theorem , we see that we obtain a 1string trivial tangle in a solid torus @xmath479 , by cutting @xmath491 along @xmath120 . since @xmath120 and @xmath121 are @xmath0parallel in @xmath25 , we see that @xmath121 is @xmath475boundary parallel in @xmath140 . * lemma d-3 * let @xmath35 be as in lemma d-1 . suppose , moreover , that @xmath35 is non - separating in @xmath25 . then @xmath483 is non - separating in @xmath25 , and , by cutting @xmath491 along @xmath483 , we obtain a 2string trivial tangle , say @xmath81 . moreover , @xmath35 is @xmath79boundary parallel in @xmath80 . * proof*let @xmath140 be the solid torus obtained from @xmath25 by cutting along @xmath35 . we may suppose that @xmath120 and @xmath121 are properly embedded in @xmath140 . since each @xmath266 intersects @xmath0 in one point , we see that @xmath266 is an essential disk in @xmath140 . by the construction of @xmath120 , @xmath121 , we see that @xmath483 separates the copies of @xmath35 in @xmath140 . this shows that @xmath483 is non - separating in @xmath25 . then , by cutting @xmath491 along @xmath483 , we obtain a 2string tangle in a 3ball , say @xmath81 . since @xmath488 is a genus three handlebody , we see that the 2fold covering space of @xmath80 branched along @xmath79 is a solid torus . hence , @xmath81 is a 2string trivial tangle . by lemma b-1 ( 3 ) , we see that @xmath35 is @xmath79boundary parallel in @xmath80 . * lemma d-4 * let @xmath35 , @xmath255 be pairwise disjoint , pairwise parallel , non @xmath0parallel , @xmath0essential disks in @xmath25 such that @xmath412 , and @xmath492 consist of two points . then there are two @xmath0compressing disks @xmath493 , @xmath494 for @xmath473 such that @xmath495 is non - separating in @xmath25 and is disjoint from @xmath496 , and , by cutting @xmath491 along @xmath495 , we obtain a 2string trivial tangle , say @xmath81 . moreover @xmath35 , @xmath255 are @xmath79boundary parallel in @xmath81 , and , hence , @xmath496 cobounds a 2string trivial tangle in @xmath491 . let @xmath493 be the disk corresponding to @xmath120 or @xmath121 in lemma d-2 , and @xmath479 the 1string trivial arc in a solid torus @xmath140 obtained from @xmath491 by cutting along @xmath493 . then , by lemma b-2 , we see that there exists an @xmath475compressing disk @xmath494 for @xmath143 such that @xmath494 cuts @xmath479 into a 2string trivial tangle . here we may suppose that @xmath494 is disjoint from the images of @xmath493 in @xmath143 , and , hence , we may regard that @xmath494 is properly embedded in @xmath25 . then @xmath495 satisfies the conclusion of lemma d-4 .

nuclear magnetic resonance ( nmr ) , neutron diffraction ( nd ) , x - ray diffraction , magnetic susceptibility and specific heat measurements on the frustrated-site spinel coalo compound reveal a collinear antiferromagnetic ordering below = 9.8(2 ) k. a high quality powder sample characterized by x - ray diffraction that indicates a relatively low co - al inversion parameter in ( coal)[alcoo , shows a broad maximum around 15 k in and a sharp peak at in heat capacity . the average ordered magnetic moment of co ( = 3/2 ) ions at the-site is estimated to be 2.4(1 ) from nmr and 1.9(5 ) from nd which are smaller than the expected value of 3 for and = 2 . antiferromagnetic spin fluctuations and correlations in the paramagnetic state are revealed from the , nmr and nd measurements , which are due to spin frustration and site inversion effects in the system . the nd data also show short - range dynamic magnetic ordering that persists to a temperature that is almost twice .

1308.5238

the investigation of magnetism in the insulating @xmath1-site spinel - structure compound coal@xmath2o@xmath3 has a long history which started with a suggestion of long - range antiferromagnetic ( afm ) ordering below 4 k from magnetic susceptibility @xmath0 and neutron diffraction ( nd ) measurements performed by roth.@xcite the magnetism of this system originates from co@xmath12 ions [ ( @xmath17(@xmath18 with spin @xmath13 = 3/2 at the tetrahedral @xmath1-sites which form a diamond lattice ( fig . [ fig : figure_a - site_structure ] ) . the general formula for a normal oxide spinel compound is @xmath1[@xmath19o@xmath3 where the @xmath1 and @xmath20 atoms occupy the tetrahedral and octahedral holes between nearly cubic - close - packed o layers perpendicular to the [ 111 ] direction . later on , the interest shifted to the spin frustration effects in the @xmath1-site spinel compounds after the ground state of the diamond lattice afm was investigated theoretically.@xcite the diamond lattice is composed of two interpenetrating face - centered cubic ( fcc ) sublattices as shown in fig . [ fig : figure_a - site_structure ] . the frustration is caused by the next - nearest - neighbor afm interaction @xmath21 which couples nearest - neighbor sites of each fcc sublattice of the diamond structure , while the nearest - neighbor interaction @xmath22 between the two fcc sublattices alone does not induce any frustration for afm ordering ( fig . [ fig : figure_a - site_structure ] ) . o@xmath3 , composed of two interpenetrating face - centered cubic ( fcc ) sublattices shown by purple and green spheres . the nearest - neighbor interaction @xmath22 between the two fcc sublattices and the next - nearest - neighbor interaction @xmath21 within the same fcc sublattice are shown.,width=240 ] bergman _ _ et al.__@xcite pointed out theoretically that the ground state of the diamond lattice is a nel - type antiferromagnet for a ratio of @xmath23 . the ground state changes to a spiral spin - liquid state for @xmath24 . @xmath25 and @xmath26 for coal@xmath2o@xmath3 are reported to be @xmath27 mev @xmath28 10.67 k and with @xmath29 mev @xmath30 k , @xmath31 = 0.11.@xcite since the ratio is close to the critical ratio of @xmath32 , coal@xmath2o@xmath3 is considered to be located in the critical region between the afm and the spiral spin - liquid states . experimental observations of the ground state of coal@xmath2o@xmath3 are contradictory . tristan _ _ et al.__@xcite reported from @xmath0 and heat capacity @xmath11 measurements on powder samples that a spin - glass state is realized below 4.8 k with a high frustration parameter@xcite @xmath33 = 22 , where @xmath34 is the curie - weiss temperature in the curie - weiss law and @xmath35 is the spin - glass transition temperature . on the other hand , suzuki _ proposed a possible spin - liquid ground state below 9 k ( ref ) with @xmath36 where @xmath37 is defined as the temperature at which a broad peak of @xmath11 is observed . _ suggested an unconventional magnetically ordered phase in which the spin - liquid correlations were observed from neutron powder diffraction@xcite ( below 5 k ) and single - crystal nd experiments@xcite ( below 8 k ) . _ also carried out a nd experiment on a single crystal of coal@xmath2o@xmath3 which showed a sharp cusp at 6 k in the @xmath38-dependence of @xmath0 and an afm state below 6.5 k.@xcite they observed strong diffusive scattering in the afm state where the phase transition was suggested to be first order nature.@xcite the origin of this ambiguity is probably site inversion , as has been pointed out by zaharko _ et al_.@xcite the degree of site inversion between the @xmath1- and @xmath20-sites is usually defined by an inversion parameter @xmath39 in the formula @xmath40o@xmath3 , which here corresponds to ( co@xmath6al@xmath41)[al@xmath8co@xmath9o@xmath3 . the @xmath39 value depends delicately on the method of sample preparation . the @xmath39 values of the polycrystalline samples reported by suzuki _ _ et al.__@xcite , tristan _ _ et al.__@xcite and zaharko _ _ et al.__@xcite are 0.04(1 ) , 0.080.104 and 0.17 , respectively . the @xmath39 values for the single - crystal samples studied by zaharko@xcite and macdougall@xcite are 0.08 and 0.02(4 ) , respectively . quite recently hanashima _ _ et al.__@xcite carried out a systematic study of the effects of site inversion on the magnetic properties of ( co@xmath6al@xmath41)[al@xmath8co@xmath9o@xmath3 where @xmath39 is controlled from 0.0467 to 0.153 by changing the heat treatment conditions . the @xmath0(@xmath38 ) of the sample with the lowest @xmath39 value @xmath42 shows a broad maximum at 14 k and does not show any signature of magnetic ordering down to 2 k , similar to that reported by suzuki _ et al._,@xcite suggesting a spin - liquid state . with increasing @xmath39 , the broad maximum moves to lower temperature ( 10 k for @xmath43 ) , and for @xmath44 , a clear cusp in the @xmath38-dependence of @xmath0 is observed which is attributed to a spin - glass transition below @xmath45 k. in the intermediate region 0.0791 @xmath46 0.847 , a coexistence of the spin - liquid state and spin - glass state is proposed . _ conclude that with increasing @xmath39 the spin - liquid state collapses and the spin - glass order emerges for @xmath44 . it is important to understand whether the intrinsic ground state of coal@xmath2o@xmath3 for small values of @xmath39 is a spin - glass state , an antiferromagnet or a spin - liquid state due to the spin frustration . to investigate the ground state of coal@xmath2o@xmath3 one needs a high - quality sample with a small value of the inversion parameter @xmath39 . in this paper , we report measurements of crystallography , magnetization , @xmath0 , specific heat , nuclear magnetic resonance ( nmr ) and nd on a well - characterized polycrystalline coal@xmath2o@xmath3 sample with inversion parameter @xmath5 . we conclude that coal@xmath2o@xmath3 exhibits collinear afm ordering below the @xmath4 = 9.8(2 ) k , with an ordered moment of 2.4(1 ) @xmath14 from nmr and 1.9(5 ) @xmath14 from nd measurements . a polycrystalline sample of coal@xmath2o@xmath3 was synthesized by the solid state reaction method using @xmath47 ( 99.9985% ) and @xmath48 ( 99.995% ) from alfa aesar as the starting materials . as a precautionary measure to check the phase - purity of the starting materials , room temperature x - ray diffraction ( xrd ) measurement was performed on the @xmath47 starting material . the results revealed that the material was not single - phase and with a sizable presence of binary coo . the weight fraction of coo obtained by two - phase rietveld refinement of the powder xrd pattern was 1.4(1)% . the molar fraction of coo in @xmath47 was considered while calculating the stoichiometry of cobalt in the starting mixture . stoichiometric mixtures of @xmath47 and @xmath48 were thoroughly mixed and pressed into a pellet . the pressed pellet was placed inside an alumina crucible and heated to 875 @xmath49c in 20 h , held there for 40 h and then cooled to 500 @xmath49c in 100 h and held there for 80 h , and finally the sample was cooled by switching off the furnace . the material was then thoroughly reground and repressed into a pellet . the pellet was placed in the same alumina crucible and heated to 1000 @xmath49c in 20 h , held there for 50 h , then cooled to 700 @xmath49c in 60 h , held there for 2 h and finally slowly cooled to 400 @xmath49c in 100 h and held there for 100 h. at this time the furnace was switched off and the sample cooled to room temperature . the sample was persian blue in color . structural characterization of the sample was performed at room temperature using powder xrd data obtained from a rigaku geigerflex powder x - ray diffractometer utilizing cuk@xmath50 radiation . the fullprof package@xcite was used in the rietveld xrd data refinement . temperature- and applied magnetic field @xmath51-dependent magnetic measurements were performed using a magnetic properties measurement system ( mpms , quantum design , inc . ) . heat capacity @xmath11 was measured by the thermal relaxation method using a physical properties measurement system ( ppms , quantum design , inc . ) . nmr measurements were conducted by probing the @xmath52al nuclei ( @xmath53 = 5/2 , @xmath54/2@xmath55 = 11.09375 mhz / t , @xmath56 = 0.149 barns ) and the @xmath57co nuclei ( @xmath53 = 7/2 , @xmath54/2@xmath55 = 10.03 mhz / t , @xmath56 = 0.4 barns ) by using a phase - coherent spin - echo pulse spectrometer . the nmr spectrum was taken by recording the spin - echo signal intensity while sweeping the external magnetic field at a fixed resonance frequency . the co nmr spectrum under zero magnetic field was obtained by plotting the spin - echo intensity point by point at different fixed frequencies . neutron diffraction measurements were performed using the triple - axis spectrometer ( triax ) at the university of missouri research reactor ( murr ) . approximately 2 grams of coal@xmath2o@xmath3 powder was loaded into a vanadium sample cell and mounted on the cold head of an advanced research systems closed - cycle refrigerator on triax . powder nd patterns were obtained on the spectrometer operated with a cu(220 ) monochromator at a wavelength of 1.348 , a pg(002 ) analyzer set at the elastic position , and collimation of @xmath58 . sapphire and si filters were used upstream from the monochromator and none after the sample . additional scans with greater statistics were also obtained at the key positions in wavevector ( @xmath59 ) space corresponding to the putative spiral or collinear magnetic orderings . the 2 g sample held in the same holder was also measured on the murr high resolution neutron powder diffractometer ( psd ) , which utilizes a double - focusing perfect single si monochromator at the ( 511 ) reflection to illuminate the sample with 1.4805 wavelength neutrons . diffracted neutrons are detected by five linear position sensitive detectors . the sample temperature was adjusted with a leybold closed - cycle refrigerator with a 18 k base temperature . versus site inversion parameter @xmath39 obtained by single - phase rietveld refinement of our polycrystalline coal@xmath2o@xmath3 sample . the red line is shown as eye guides . ( b ) room - temperature powder x - ray diffraction data , two - phase rietveld refinement , bragg positions and difference profile for coal@xmath2o@xmath3 and the al@xmath2o@xmath60 impurity phase.,title="fig:",width=288 ] versus site inversion parameter @xmath39 obtained by single - phase rietveld refinement of our polycrystalline coal@xmath2o@xmath3 sample . the red line is shown as eye guides . ( b ) room - temperature powder x - ray diffraction data , two - phase rietveld refinement , bragg positions and difference profile for coal@xmath2o@xmath3 and the al@xmath2o@xmath60 impurity phase.,title="fig:",width=288 ] as described in the sec . i , it was shown that the magnetic properties of coal@xmath2o@xmath3 depend delicately on the amount of site inversion present in the sample.@xcite the goal in this work is to investigate the magnetic properties of a high - quality polycrystalline sample of coal@xmath2o@xmath3 having minimal site inversion . to achieve this we synthesized a polycrystalline sample of coal@xmath2o@xmath3 using the controlled thermal treatments described in the previous section . however , we could not achieve a completely site - inversion - free sample . rietveld analysis of the room - temperature powder xrd data of our sample showed a minor amount of site inversion corresponding to the formula ( co@xmath6al@xmath41)[al@xmath8co@xmath9o@xmath3 . the goodness of fit parameter @xmath61 obtained by single - phase rietveld refinements of the room - temperature powder xrd data versus site inversion @xmath39 is shown in fig . [ fig : figure_xrd](a ) . from this figure we estimate @xmath5 . this value of @xmath39 is slightly smaller than that reported by tristan _ et al . _ ( 0.0800.104 ) in refs . and and zaharko _ et al . _ ( 0.08 ) in ref . , and is equal , within error bars , to that reported by suzuki _ et al . _ ( 0.04 ) in ref . and hanashima _ et al . _ ( 0.0467).@xcite l l structure & spinel ( @xmath62-type ) + space group & @xmath63 ( no . 227 ) + lattice parameter & 8.1011(2 ) + coal@xmath2o@xmath3 weight fraction & 96.5(2)% + inversion parameter ( @xmath39 ) & 0.057(20 ) + + atom & wyckoff positions + co & 8@xmath64 ( 1/8 , 1/8 , 1/8 ) + al & 16@xmath65 ( 1/2 , 1/2 , 1/2 ) + o & 32@xmath66 ( @xmath67 ) + & where @xmath68 = 0.2630(1 ) + refinement quality parameters + @xmath69 & 2.49 + @xmath70 ( % ) & 4.69 + @xmath71 ( % ) & 6.37 + additionally , powder xrd data showed the presence of a minor secondary phase al@xmath2o@xmath60 in our sample . after estimating the site inversion parameter by single - phase rietveld refinement we performed a two - phase rietveld refinement with a fixed value of @xmath72 . figure [ fig : figure_xrd](b ) shows the results of the two - phase rietveld refinement . we achieved an excellent fit to the data , and the calculated weight fraction of al@xmath2o@xmath60 was 3.5(2)% . the crystal and rietveld refinement quality parameters are listed in table [ table : structure ] . versus applied magnetic field @xmath51 for coal@xmath2o@xmath3 measured at four different temperatures between 1.8 and 300 k.,width=288 ] isothermal magnetization @xmath73 versus @xmath51 data of our polycrystalline coal@xmath2o@xmath3 sample at four different temperatures are shown in fig . [ fig : figure_mh ] . the @xmath74 data at @xmath75 and 300 k are proportional to @xmath51 in the entire field range of measurement . however , the @xmath74 data at 1.8 k shows nonlinear behavior with positive curvature in the entire field range of measurement . this behavior is typical of heisenberg afms with anisotropy as observed from the nd measurements.@xcite the @xmath10 of our sample measured under zero - field - cooled ( zfc ) and field - cooled ( fc ) conditions in the presence of an applied magnetic field @xmath76 t is shown in fig . [ fig : figure_susceptibility ] . the @xmath10 exhibits a broad maximum centered at @xmath38 = 15(1 ) k suggesting the presence of afm correlations in the material . the zfc and fc @xmath10 data nearly overlap in the entire temperature range of measurement [ inset ( a ) , fig . [ fig : figure_susceptibility ] ] , which is consistent with the result on polycrystalline coal@xmath2o@xmath3 reported by suzuki _ et al . _ with a low inversion parameter @xmath77.@xcite our @xmath10 data are also consistent with the results for the @xmath42 sample of hanashima _ et al._,@xcite except below 9.6 k where there is a rather significant bifurcation of zfc and fc data in their sample . as discussed in sec . i , it has been shown in previous reports that the low - temperature magnetic properties of coal@xmath2o@xmath3 are very sensitive to the amount of site inversion between the @xmath1-site and the @xmath20-site in the samples . samples with relatively large inversion parameters ( @xmath78 ) have been reported to show spin - glass - like behaviors,@xcite while samples with @xmath79 have been argued to have spin - liquid ground states.@xcite l c c c c c c c c l @xmath39 & @xmath35 & @xmath37 & @xmath34 & @xmath80 & @xmath81 & @xmath82 & @xmath16 & ground & ref . + ( % ) & ( k ) & ( k ) & ( k ) & & ( @xmath83k / mol ) & ( @xmath14/f.u . ) & & state & + & & & @xmath8473(8 ) & 8(1 ) & 2.1(1 ) & 4.1(1 ) & 2.12(5 ) & & this work + & & & @xmath84104(9 ) & 11(1 ) & 2.52(6 ) & 4.49(5 ) & 2.32(3 ) & & this work + [ 2ex ] 4(2 ) & & 9 & @xmath8489(6 ) & 9.9(7 ) & 2.5(1 ) & 4.45(8 ) & 2.30(4 ) & spin liquid & [ ] + 4.67 & & 9.6 & @xmath8493.0 & 9.7 & 2.38 & 4.36 & 2.25 & spin liquid & [ ] + 2(4 ) & 6.5 & & @xmath84109(1 ) & 18.0(2 ) & 2.99(4 ) & 4.89(3 ) & 2.53(2 ) & afm & [ ] + 8(3 ) & 4.8(2 ) & & @xmath84104(2 ) & 22(1 ) & 2.7(1 ) & 4.65(9 ) & 2.40(4 ) & spin glass & [ ] + 8 & 8 & & @xmath8494(1 ) & 12(1 ) & 2.7(1 ) & 4.63(2 ) & 2.40(1 ) & unconventional & [ , ] + versus temperature @xmath38 for coal@xmath2o@xmath3 . the zfc and fc data nearly overlap each other in the entire @xmath38-range of measurement . inset ( a ) : expanded plot of @xmath10 below 50 k. inset ( b ) : inverse magnetic susceptibility @xmath85 versus @xmath38 . solid line is curie - weiss fit to the data as described in the text.,width=288 ] the @xmath10 data between 100 and 350 k were fitted by the curie - weiss law [ inset ( b ) in fig . [ fig : figure_susceptibility ] ] , @xmath86 where @xmath81 is the curie constant and @xmath87 is the @xmath38-independent orbital contribution to @xmath0 . the fitted value of @xmath88 is @xmath89 @xmath83/mol , which is in good agreement with van vleck susceptibility @xmath90 @xmath83/mol for co@xmath12 ions at the @xmath1-site estimated from nmr measurements as will be described in sec . vi . b. this value is almost one order of magnitude greater than the diamagnetic susceptibility @xmath91 = @xmath92 @xmath83/mol estimated from the reference compound @xmath93 ( ref . ) . the fitted values of @xmath81 , @xmath34 and the effective moment @xmath82 and @xmath16-factors determined from @xmath81 are listed in table [ table : magnetism ] . for comparison , parameters for some other coal@xmath2o@xmath3 samples reported in the literature are also included in table [ table : magnetism ] . the @xmath81 obtained for our sample is larger than the value 1.875 @xmath83k / mol expected for spin @xmath15 and @xmath94 . however , our value of @xmath81 is smaller than those reported in the literature for this material ( table [ table : magnetism ] ) . if we consider that magnetism in coal@xmath2o@xmath3 is entirely due to co@xmath12 spins with @xmath15 then the calculated value of powder - averaged @xmath16-factor is 2.12(5 ) which is again considerably smaller than that reported for coal@xmath2o@xmath3 in the literature ( table [ table : magnetism ] ) and @xmath16 = 2.26 reported from electron spin resonance ( esr ) measurements.@xcite it should be noted , however , that c value strongly depends on @xmath87 . if one assumes @xmath87 = 0 , then @xmath81 = 2.52(6 ) @xmath83k / mol was obtained with @xmath34 = @xmath95 k ( table [ table : magnetism ] ) . using this value of @xmath81 , the @xmath16-factor is estimated to be 2.32(2 ) which is slightly larger than @xmath16 = 2.26 obtained from the esr measurements.@xcite it appears that the value of the frustration parameter @xmath96 depends sensitively on the degree of antisite disorder present in coal@xmath2o@xmath3 samples . the @xmath96 value of our sample is lower than those reported in the literature ( table [ table : magnetism ] ) . the @xmath11 versus @xmath38 for coal@xmath2o@xmath3 is shown in fig . [ fig : figure_hc ] . low - temperature @xmath97 data of coal@xmath2o@xmath3 along with those of the nonmagnetic reference compound @xmath93 ( ref . ) are shown in inset on an expanded scale . the @xmath97 data of coal@xmath2o@xmath3 show a sharp peak centered at @xmath38 = 9.7(2 ) k suggesting a magnetic transition at this temperature . the magnetic contribution @xmath98 to the total heat capacity was obtained by subtracting the @xmath97 data of @xmath93 from that of coal@xmath2o@xmath3 [ fig . [ fig : figure_hc - mag](a ) ] . versus temperature @xmath38 for coal@xmath2o@xmath3 . inset : the @xmath97 of coal@xmath2o@xmath3 at an expanded scale near magnetic transition temperature @xmath4 along with @xmath97 of nonmagnetic analog compound @xmath93 ( ref . ) . the @xmath99 data of @xmath93 provided by tristan _ _ were for @xmath100 k. we extrapolated the data below 10 k using @xmath101 . , width=288 ] over temperature @xmath38 versus @xmath38 . an additional data point @xmath102 at @xmath103 was inserted for better estimation of the magnetic entropy . inset : @xmath104 versus @xmath38 below 10 k. solid line shows @xmath104 @xmath105 @xmath106 dependence . ( b ) magnetic entropy @xmath107 versus @xmath38 . the horizontal dashed line represents the calculated high-@xmath38 limit @xmath107 = 10.9 j / mol k.,title="fig:",width=288 ] over temperature @xmath38 versus @xmath38 . an additional data point @xmath102 at @xmath103 was inserted for better estimation of the magnetic entropy . inset : @xmath104 versus @xmath38 below 10 k. solid line shows @xmath104 @xmath105 @xmath106 dependence . ( b ) magnetic entropy @xmath107 versus @xmath38 . the horizontal dashed line represents the calculated high-@xmath38 limit @xmath107 = 10.9 j / mol k.,title="fig:",width=288 ] the magnetic contribution @xmath108 to the total entropy was calculated from @xmath109 using @xmath110 the @xmath38-dependence of @xmath107 is shown in fig . [ fig : figure_hc - mag](b ) , where we have also indicated the expected high temperature limit @xmath111 where we consider only the contribution from the @xmath1-site co@xmath12 ions with @xmath112 for our sample with @xmath39 = 0.057 for simplicity . from the @xmath108 data presented in the fig . [ fig : figure_hc - mag](b ) we find that ( i ) @xmath107 does not reach the expected saturation limit and its value at 100 k is 9.74 j / mol k which is @xmath113 smaller than 10.9 j / mol k. \(ii ) the value of @xmath107 at @xmath114 k is 3.11 j / mol k which is only @xmath115 of it s high-@xmath38 value . this clearly shows that a major part of the magnetic entropy is released between 10 and 50 k , which indicates the persistence of strong dynamic short - range magnetic order above @xmath4 up to approximately 50 k. there are contradictory reports in the literature about the @xmath38-dependence of @xmath11 below the magnetic ordering temperature . while tristan _ _ et al.__@xcite and macdougall _ _ et al.__@xcite reported that @xmath11 varies as @xmath116 , suzuki _ _ find that this dependence is @xmath117.@xcite in our sample we find that @xmath109 below @xmath4 follows @xmath109 @xmath105 @xmath118 which is approximately mid way between the @xmath116 and @xmath117 power law dependences . our observation is inconsistent with the linear @xmath38-dependence of @xmath11 expected for spin glass systems@xcite and the exponent is slightly lower than the @xmath119 dependence predicted by bergman _ et al . _ for spiral spin - liquid frustrated diamond - lattice antiferromagnets.@xcite figure [ fig:1al - nmr - specfit ] shows the field - swept @xmath52al - nmr spectrum at @xmath38 = 25 k and resonance frequency @xmath120 = 9.3 mhz . since our sample consists of grains with randomly oriented crystal axes , the spectrum is a powder pattern . the spectrum can be explained by a combination of a large zeeman interaction @xmath121 and a small quadrupole interaction @xmath122 . the nuclear spin hamiltonian can therefore be expressed as @xmath123 where @xmath124 and @xmath125 , $ ] @xmath56 is the quadrupole moment of the @xmath52al nucleus and @xmath126 is the nmr shift which is a sum of isotropic ( @xmath127 ) and axial ( @xmath128 ) parts . @xmath129 is the asymmetry parameter for the electric field gradient ( efg ) defined by @xmath130 , @xmath131 , @xmath132 = @xmath133 and the quadrupole frequency @xmath134 = @xmath135 . the @xmath136 and @xmath137 axes are the quantization axes for the zeeman and quadrupole interactions , respectively . the observed spectrum is well reproduced by a simulated spectrum shown by a red line in fig . [ fig:1al - nmr - specfit ] with @xmath138 = 0.55 mhz and @xmath129 = 0 . the value of @xmath138 is independent of @xmath38 in the measured range of 1.8 to 300 k within our experimental uncertainty as shown in the inset of fig . [ fig:1al - nmr - specfit ] . in contrast , @xmath127 and @xmath128 show weak temperature dependences . the weak @xmath38-dependences of the nmr shifts are consistent with previous data reported by miyatani _ _ et al.__@xcite who measured the @xmath52al - nmr spectrum in coal@xmath2o@xmath3 with the continuous wave nmr method above 77 k , although the @xmath128 component was not detected in the previous measurement . al - nmr spectrum at 25 k and @xmath139 = 9.3 mhz . the red line shows the simulated spectrum with @xmath134 = 0.55 mhz and @xmath129 = 0 . the inset shows the temperature @xmath38 independence of the quadrupole frequency @xmath134.,width=355 ] the @xmath38-dependences of @xmath127 and @xmath128 follow the @xmath38-dependence of @xmath0 as shown in fig . [ fig:2al - nmr - k ] where the nmr shifts are plotted as a function of @xmath0 with @xmath38 as an implicit parameter . here the nmr shifts are determined by measuring the nmr spectrum with a higher resonance frequency of 70.9 mhz to improve the accuracy . the nmr shift has contributions from the @xmath38-dependent spin part @xmath140 and a small @xmath38-independent orbital part @xmath141 . @xmath140(@xmath38 ) is related to the spin susceptibility with hyperfine coupling constant @xmath142 as follows : @xmath143 where @xmath144 is the avogadro number . the isotropic ( @xmath145 ) and axial ( @xmath146 ) parts of the hyperfine coupling constants for the al nucleus are evaluated from the slope of the @xmath126-@xmath0 plots ( fig . [ fig:2al - nmr - k ] ) to be 1.25(2 ) koe/@xmath147 and @xmath148 koe/@xmath147 respectively , and @xmath149 is 0.023(4)% and 0.013(3)% for the isoisotropic and axial parts , respectively . classical dipolar field calculations indicate that the hyperfine field can not be due to the classical dipolar field at the al - sites from neighboring @xmath150 ( @xmath13 = 3/2 ) spins , but is rather due to the transferred hyperfine field due to finite spin transfer from @xmath150 spins . the anisotropy in the hyperfine field suggests that anisotropic orbitals such as the 2@xmath151 and 3@xmath151 orbitals on al atoms are polarized by the spin transfer from the neighboring co atoms . al - nmr shifts ( @xmath152 and @xmath153 ) versus @xmath0 with @xmath38 as an implicit parameter . the solid lines show the linear fit using the eq . ( [ eq : k - chi]).,width=355 ] al - nmr spectra at various temperatures with resonance frequency @xmath120 = 9.3 mhz . , width=355 ] the spectrum starts to broaden suddenly below @xmath28 10 k , as shown in fig . [ fig:3al - nmr - spect ] . the broadening indicates a static magnetic ordering of the @xmath150 spins which produce a finite static internal field at the al - sites . the symmetric line broadening indicates an afm state . figure [ fig:4al - nmr - linewidth ] shows the @xmath38 dependence of the line width ( @xmath154 ) determined at 10 % of the peak intensity . since the broadening of the @xmath52al - nmr line width in our polycrystalline sample below @xmath28 10 k is produced by the co@xmath12 spin moments in the magnetically ordered state , the @xmath38 dependence of the line width reflects that of the sublattice magnetization in the afm state . the nel temperature @xmath155 is determined by fitting @xmath154 near @xmath155 in fig [ fig:4al - nmr - linewidth ] by @xmath156 where @xmath157 @xmath28 1669 oe and @xmath158 @xmath28 1145 oe is a constant determined by averaging the nearly @xmath38-independent @xmath154 above @xmath155 . by fitting data in the temperature range of 7.5 k @xmath159 10 k , by eq . ( [ eq : orderparameter ] ) , @xmath155 is estimated to be 9.8(2 ) k and @xmath160 = 0.65(5 ) . in general , one can drive the universality class / spin dimensionality from the critical exponent @xmath160 . the @xmath160 = 0.65(5 ) is much larger than @xmath160 = 0.367 for the 3 dimensional ( d ) heisenberg model , 0.345 for the 3d xy model , and 0.326 for the 3d ising model.@xcite it should be noted that our @xmath160 value strongly depends on the fitting range . by decreasing the minimum temperature of the fit , the @xmath160 was found to decrease with a minimum value @xmath160 = 0.48(5 ) obtained by using all data below 10 k. the @xmath160 value is also estimated to be 0.255(44 ) from nd data for 3 k @xmath161 @xmath38 @xmath161 10 k as will be described in sec . vii . since the @xmath160 value strongly depends on the fitting range , we can not discuss the details of the universality class or spin dimensionality in coal@xmath2o@xmath3 based on our nmr and nd data . -dependence of the line width at 10% of the peak intensity . the solid purple line shows the fit by eq . ( [ eq : orderparameter ] ) using data in the temperature range of 7.5 k @xmath159 10 k.,width=355 ] in the paramagnetic state at @xmath38 = 125 k , we observe two peaks in the @xmath57co - nmr spectrum as shown in fig . [ fig:5co - nmr - specfit ] . with decreasing @xmath38 , the peak observed at @xmath162 6.9 t shifts to higher magnetic field and the other peak at @xmath162 6.8 t moves to slightly lower magnetic field . while the former peak disappeared below 10 k , the latter peak could be observed down to the lowest temperature of 1.5 k. these behaviors are very similar to @xmath57co - nmr data for the magnetic @xmath163 ( @xmath164 ) at the @xmath1-site and nonmagnetic low - spin @xmath165 [ ( @xmath166(@xmath167 , @xmath168 at the @xmath20-site in the spinel - structure antiferromagnet @xmath47 with @xmath155 = 34 k. @xcite thus the former signal , which vanishes due to very large internal field at the @xmath163-site , can be assigned to be the co at the tetrahedral @xmath1-site and the latter signal can be attributed to the co occupying the octahedral @xmath20-site . co - nmr spectrum at @xmath38 = 125 k and resonance frequency @xmath120 = 70.2 mhz . the two peaks correspond to the co ions occupying the @xmath1-site ( magnetic @xmath163 ion ) and the @xmath20-site ( nonmagnetic @xmath165 ion ) , respectively . the transverse relaxation time @xmath169 = 8.2 @xmath170s and @xmath169 = 66 @xmath170s for the @xmath1- and @xmath20-sites , respectively.,width=355 ] from the intensities of the co signals occupying the @xmath1- and the @xmath20-sites , corrected by their respective longitudinal and transverse relaxation times ( @xmath171 and @xmath169 , respectively ) , we have evaluated the inversion parameter to be @xmath172 0.06(2 ) . this value is in good agreement with that from the x - ray analysis . we notice that the @xmath38 dependence and the magnitude of the nmr shift of the co signal from the @xmath20-sites in coal@xmath2o@xmath3 is very close to that observed in the co nmr signal from the @xmath20-sites in @xmath47 where the @xmath165 ions are in a nonmagnetic state with low spin @xmath13 = 0.@xcite from these results , we may consider that the co ions occupying the @xmath20-site in coal@xmath2o@xmath3 could be @xmath165 with low spin state as in the case of co ions at the @xmath20-site in @xmath47 . this could be possible if we assume that the deviation from oxygen stoichiometry exists as in ( co@xmath6al@xmath7)[al@xmath8co@xmath9o@xmath173 to satisfy charge compensation of the compound . however , we can not rule out completely that the observed nonmagnetic co is from an impurity phase , although it seems unlikely . co - nmr shift @xmath126 versus @xmath0 with @xmath38 as an implicit parameter . the solid line is a linear fit.,width=336 ] the hyperfine coupling constant of co at the @xmath1-site is estimated from the @xmath126-@xmath0 plot in fig . [ fig:6co - nmr - k ] to be @xmath8418.0(5 ) koe/@xmath14 which is considerably smaller than the value of @xmath84105 koe/@xmath14 for a free @xmath163 ion . such a small co hyperfine coupling constant at the @xmath1-site has been reported previously to be @xmath8418.3 koe/@xmath14 and @xmath8415.7 koe/@xmath14 in the antiferromagnetic @xmath47 ( @xmath4 = 34 k ) ( ref . ) and in ferrimagnetic @xmath174 ( @xmath175 = 97 k ) ( ref . ) , respectively . the small coupling constant was well explained by taking the second - order orbital moment due to the spin - orbit interaction into consideration . here we present a similar discussion as in the previous papers.@xcite the major contributions to the magnetic hyperfine coupling constant at the @xmath57co nucleus are dominated by the @xmath65 core polarization ( @xmath176 ) , dipolar ( @xmath177 ) and orbital ( @xmath178 ) terms:@xcite @xmath179 due to the crystal field effect , the ground state of the @xmath163(@xmath180 ) ion occupying the tetrahedral @xmath1-site has four electrons in the two lower @xmath181 orbitals and three in the upper three @xmath182 orbitals . this electronic configuration results in @xmath177 = 0 . @xcite for a free @xmath163 ion , the @xmath65 core polarization coupling constant is @xmath84105 koe/@xmath14 . @xcite due to quenching of the orbital angular momentum @xmath183 , there is no finite contribution of the @xmath183 in the first order . but if we consider the second order perturbation of the spin - orbit interaction , the orbital quenching is partially lifted . this contribution due to the spin - orbit coupling has a positive sign and is estimated to be 85.1 koe/@xmath14 for @xmath163 ions at the tetrahedral sites.@xcite from eq . ( [ eq : totalhyperfinecoupling ] ) , we obtain @xmath184 = @xmath8419.9 koe/@xmath14 , which can reasonably explain the experimental value of @xmath8418.0(5 ) koe/@xmath14 . the temperature independent orbital part of the nmr shift for the co@xmath12 at the a - site is estimated to be @xmath185 = 5.46(3)% from the @xmath126-@xmath0 analysis . the @xmath185 relates to van vleck susceptibility @xmath186 for co@xmath12 at the a - site as @xmath185 = @xmath187@xmath186 where @xmath188 is the orbital hyperfine coupling constant given by @xcite @xmath189 here @xmath190 is an average of @xmath191 over 3@xmath65 electrons . using @xmath190 = 6.02 a.u . for co@xmath12 ( ref . ) , @xmath186 is estimated to be @xmath90 @xmath83/mol which is in good agreement with @xmath38 independent magnetic susceptibility discussed in sec . co - nmr spectrum under zero external field at 1.6 k. , width=336 ] below @xmath4 , we observe an nmr signal of co nuclei at the @xmath1-site under zero magnetic field , which again confirms long - range magnetic ordering . figure [ fig:7co - nmr - zerofield ] shows the zero - field @xmath57co - nmr spectrum at 1.6 k. a similar zero - field @xmath57co - nmr spectrum at the @xmath1-site has been previously reported for the spinel compound @xmath47 in the afm state.@xcite the resonance frequency @xmath120 = 43.5(4 ) mhz at the peak intensity point is lower than @xmath120 = 55.7 mhz in @xmath47.@xcite the internal field @xmath192 at the co @xmath1-site is evaluated to be 43.3(4 ) koe at 1.6 k. with the value of @xmath184 = @xmath8418.0 koe/@xmath14 deduced from the @xmath126-@xmath0 plot , the ordered moment @xmath193 of @xmath163 is estimated to be 2.4(1 ) @xmath14/co using the relation @xmath192 = @xmath194@xmath193 . this value is smaller than the expected value of 3 @xmath14/co for @xmath13 = 3/2 of @xmath163 ions with @xmath16 = 2 and is larger than the value of 1.58 @xmath14/co reported from the nd study done previously by zaharko _ _ et al.__@xcite the shoulder - like structures of the spectrum are due to a distribution of @xmath163 spin moments in the afm state . in order to investigate the dynamical behavior of the @xmath163 spins in coal@xmath2o@xmath3 , @xmath52al spin - lattice relaxation rates ( @xmath195 ) were measured at different temperatures . in the case of @xmath52al ( @xmath53 = 5/2 ) , the nuclear magnetization recovery of the central transition after a @xmath55/2 saturation pulse is given by @xcite @xmath196 where @xmath197 is the nuclear magnetization at time @xmath198 after the saturation pulse . the experimental recovery curves above @xmath4 were well fitted by the equation , while below @xmath4 the equation could not fit the experimental data due to distributions of @xmath171 . then we tentatively assumed that the experimental recovery curve is composed of two components , one with a short relaxation time @xmath199 and another with long relaxation time @xmath200 and we carried out a fit using the following equation : @xmath201+m_{\rm l}[(0.029 e^ { -t / t_{\rm{1l}}})\\ & + ( 0.18 e^ { -6t / t_{\rm{1l } } } ) + ( 0.79 e^ { -15t / t_{\rm{1l } } } ) ] \end{split } \label{eq:2componentsfort1fori3/2}\ ] ] where @xmath202 + @xmath203 = 1 . the temperature dependence of @xmath202 is shown in the inset of fig . [ fig:8al - nmr - t1 ] . figure [ fig:8al - nmr - t1 ] shows the temperature dependence of @xmath204 and @xmath205 . with decreasing @xmath38 , @xmath204 decreases and shows a minimum around 50 k and exhibits a sharp peak at @xmath38 = 10.0(5 ) k. the sharp peak of @xmath195 is due to critical slowing down of spin fluctuations expected for a second order phase transition , which again confirms the afm long - range magnetic ordering at @xmath155 . below 10 k , @xmath195 rapidly falls and then decreases slowly where both @xmath204 and @xmath205 show @xmath206 power - law behavior ( shown by the straight lines in the figure ) below 4 k. in the antiferromagnetically ordered state , @xmath195 is mainly driven by scattering of magnons , leading to @xmath207 or @xmath208 power - law temperature dependences due to a two- or three - magnon raman process , respectively . the weak @xmath38 dependence of @xmath195 @xmath209 @xmath206 below 4 k can not be explained by the magnon - scattering , and suggests the presence of other magnetic fluctuations in the magnetically ordered state , which could originate from spin frustrations and/or short - range magnetic correlations revealed by the neutron scattering measurements described below . dependence of the short and long nuclear spin - lattice relaxation rates , @xmath210 and @xmath211 , respectively . below 4 k , both @xmath210 and @xmath211 show @xmath212 power law behavior shown by the solid lines . the inset shows the @xmath38 dependence of @xmath202 which corresponds to the fraction of @xmath210 defined in eq . ( [ eq:2componentsfort1fori3/2]).,width=336 ] in order to see spin fluctuation effects in the paramagnetic state , it is useful to re - plot the data by changing the vertical axis from @xmath195 to @xmath213 as shown in fig . [ fig:9al - nmr - t1tchi ] . @xmath214 can be expressed in terms of the imaginary part of the dynamic susceptibility @xmath215 per mole of electronic spins as:@xcite @xmath216 where the sum is over the wave vectors @xmath217 within the first brillouin zone , @xmath218 is the form factor of the hyperfine interactions and @xmath215 is the imaginary part of the dynamic susceptibility at the larmor frequency @xmath219 . on the other hand , the uniform @xmath0 corresponds to the real component @xmath220 with @xmath221 = 0 and @xmath219 = 0 . thus a plot of @xmath213 versus @xmath38 shows the @xmath38 dependence of @xmath222 with respect to that of the uniform susceptibility @xmath223(0 , 0 ) . for high temperatures above 100 k , @xmath213 is a nearly constant showing that the temperature dependence of @xmath222 scales to that of @xmath223(0 , 0 ) . on the other hand , with decrease in temperature , @xmath213 starts to increase below 100 k which is @xmath209 10 times higher than @xmath4 . the increase of the @xmath213 at @xmath224 can not be simply attributed to the critical slowing down of spin fluctuations near @xmath4 . this implies @xmath222 increases more than @xmath223(0 , 0 ) , which is due to a growth of spin fluctuations with @xmath221 @xmath225 0 , most likely with afm wave vector @xmath221 = @xmath226 , even at @xmath38 much higher than @xmath4 . thus we conclude that strong afm spin fluctuations are realized in a wide temperature region up to @xmath209100 k in the paramagnetic state . @xmath0 versus temperature @xmath38 for a resonance frequency of 70.9 mhz . @xmath214@xmath0 increases with decreasing temperature indicating the growth of afm spin correlations.,width=336 ] analysis of a powder neutron diffraction pattern taken at room temperature ( fig . [ nd - rietveld ] ) is consistent with a single @xmath63 phase of coal@xmath2o@xmath3 . due to the relatively small difference in the neutron coherent scattering lengths between al and co , the rietveld refinement of the powder diffraction pattern is not sensitive to the co - al site inversion discussed in section iii . at room temperature ( red circles ) and the calculated refined pattern ( solid black line ) . the vertical green tick marks indicate the peak positions of bragg reflections and the lower blue line shows the difference between the model calculations and measured data . , width=288 ] a prominent ( 002 ) reflection , not allowed by the chemical structure , emerges below 10 k as shown in fig . [ fig : nrd](a ) and its intensity increases upon cooling to the base temperature of 3 k. measurable intensity increases are also observed for the ( 111 ) , ( 113 ) , ( 222 ) , and ( 133 ) nuclear bragg reflections between scans taken at 15 and 3 k ( not shown ) . whereas the ( 002 ) peak is slightly broader than the bragg peaks in its vicinity , the ( 111 ) magnetic peak is much broader than the ( 111 ) nuclear peak in agreement with similar observations of single crystals.@xcite the unusual broadening along the ( 111 ) reflection , ( reported for a single crystal @xcite ) , makes the observation of the ( 222 ) reflection virtually impossible with the small amount of sample we used in this study . the emergence of the ( 002 ) and the increase in intensities of the bragg reflections , that fall off systematically with the increase in @xmath227 ( due to magnetic form factor ) , are consistent with long - range afm ordering of the co spins at the @xmath1-sites with a magnetic space group @xmath228 . based on the temperature dependence of the integrated intensity of the ( 002 ) ( see fig . [ fig : nrd](b ) ) , @xmath4 is estimated to be 10.3(9 ) k , which is in agreement with the value estimated from nmr and specific heat measurements within experimental uncertainty . the diamond structure of co@xmath12 consists of two interpenetrating fcc sublattices , with origins at ( 0,0,0 ) ( spin up ) and at ( @xmath229 ) ( spin down ) , as shown in the inset of fig . [ fig : nrd](b ) . the four spins in one sublattice are parallel to the [ 001 ] direction and the other four in the other sublattice are antiparallel to those . this afm structure is similar to that observed for the @xmath1-site of co@xmath12 in co@xmath60o@xmath3 , for which an average magnetic moment 3.26 @xmath14/co at 4.2 k was obtained from nd analysis . @xcite our detailed analysis of the magnetic diffraction pattern yields a 1.9(5 ) @xmath14/co average ordered magnetic moment , lower than 3 @xmath14/co expected for co@xmath12 ( @xmath15 , @xmath16 = 2 ) , in agreement with 2.4(1 ) @xmath14/co from the above nmr measurements within errors . ) of the ( 002 ) magnetic reflection at various temperatures demonstrating the long - range magnetic order . temperature dependence of background in the @xmath227 range shown in ( a ) indicates possible short - range magnetic order above @xmath4 . the weak peak above the transition is nonmagnetic in origin ( most likely due to a strong nuclear ( 004 ) reflection by @xmath230 contamination in the beam ) and persists to room temperature . ( b ) integrated intensity of the ( 002 ) as a function of temperature . the solid line is a fit to a power - law with @xmath231 k and @xmath232 . inset : antiferromagnetic structure below @xmath4 . magnetic and crystallographic unit cell are same with @xmath233 . , width=288 ] figure [ sro - mag ] shows the diffraction intensity versus momentum transfer [ @xmath234 at three temperatures above @xmath4 after subtracting a similar data set taken at higher @xmath38 = 150 k. the patterns indicate remnant dynamic short - range order ( sro ) above the transition that disappears between 30 and 50 k. the pattern , shown in fig . [ sro - mag ] at @xmath38 = 18 k , is very similar to that observed by krimmel _ et al . _ @xcite at @xmath38 = 1.6 k for a sample with 68% site inversion which does not show long - range magnetic ordering down to 1.6 k. in order to see the effects of the site inversion on the short - range magnetic ordering , we have calculated the @xmath221-dependence of the intensity by assuming 6% random co@xmath12 vacancies in a unit cell . in the calculation , we assumed the magnetic structure at the base temperature with the 6% random vacancies , and allowed for uniform broadening of all magnetic peaks beyond the spectrometers resolution . the calculated results shown by solid lines in fig . [ sro - mag ] reasonably reproduce the observed patterns . this suggests that the sro above @xmath4 originates mainly from the site inversion effects . it is known that the site inversion can be as high as 22% ( ref . ) and such high inversion can eventually lead to a complete suppression of long - range magnetic ordering . @xcite in this regard , it is interesting to note that the afm transition temperature for co@xmath60o@xmath3 , associated with the ordering of the co@xmath235-site , is at @xmath4 @xmath28 3040 k , @xcite which is very close to the temperature at which the sro disappears in our sample . in view of the site inversion playing a major role in determining @xmath4 , it is likely that reducing the inversion will enhance @xmath4 in coal@xmath2o@xmath3 . of course , @xmath4 will also be controlled by the spin frustration which has been pointed out to play an important role in the system . to investigate the effect of spin frustration due to next - nearest neighbor exchange coupling ( @xmath21 ) in detail , one needs an inversion - free coal@xmath2o@xmath3 sample , although so far no one has obtained such a sample , with the lowest value of inversion parameter being at @xmath39 @xmath28 0.02(4).@xcite another candidate to study spin frustration effects in @xmath1-site spinels would be co@xmath60o@xmath3 where the likelihood for co@xmath12-co@xmath236 inversion is negligible , with the @xmath20-site occupied by the nonmagnetic co@xmath236 . however , since the ratio of the exchange coupling constants for co@xmath60o@xmath3 is @xmath237 = 0.019 ( with @xmath22 = 1.09 mev and @xmath21 = 0.02 mev ) ( ref . ) which is almost 6 times smaller than 0.109 for coal@xmath2o@xmath3 ( ref . ) , one expects much weaker spin frustration effects in co@xmath60o@xmath3 . at various temperatures above @xmath4 , as indicated , after subtraction of a similar scan at @xmath38 = 150 k. the broad peaks at 18 k and 30 k are evidence for short - range afm order that persists up to about @xmath238 k. sharp spikes in the subtracted patterns are due to temperature shifts of prominent nuclear bragg reflections . solid lines are calculated magnetic diffraction patterns assuming the underlying magnetic structure shown in fig . [ fig : nrd](b ) with 6% random co@xmath12 vacancies . , width=288 ] our @xmath97 , nmr and nd data clearly demonstrate that @xmath239 exhibits a collinear afm ground state below @xmath4 = 9.8(2 ) k. the @xmath52al - nmr spectrum exhibits a symmetric broadening below 10 k which indicates the presence of a long - range afm ordered ground state for this compound . furthermore , the line width of the nmr spectrum which is proportional to the order parameter shows a continuous increase below 10 k , as expected for second order phase transition . the presence of spontaneous magnetization below 10 k is confirmed directly by the observation of a @xmath57co nmr signal under zero magnetic field in the magnetically ordered state . using the internal field at the co sites in the magnetically ordered state and the hyperfine coupling constant estimated from the @xmath126-@xmath0 analysis , an ordered moment at the @xmath1-site co@xmath12 ion is estimated to be 2.4(1 ) @xmath14/co . the sharp peak of @xmath195 at around 10 k further confirms a second - order phase transition to a long - range magnetically ordered state at the same temperature . the results from the nd measurements are not only consistent with the nmr results but also indicate a collinear long - range afm - ordered ground state for this compound with an ordered moment of 1.9(5 ) @xmath14/co , consistent with the estimate from nmr . the magnetic entropy @xmath107 at @xmath4 is considerably smaller than its value at high temperature , which indicates that strong afm correlations occur above @xmath4 . the emergence of afm correlations was manifested by the increase of @xmath213 below 100 k , by the presence of broad magnetic peaks observed below 50 k in the nd measurements , and by the broad peak in @xmath0(@xmath38 ) at @xmath28 15 k. there are two factors which govern the magnetic properties of this system , @xmath240 frustration and site inversion . we successfully prepared a polycrystalline sample of @xmath239 with a low inversion parameter @xmath39 = 0.057(20 ) . site disorder creates increased randomness of exchange interactions in the already frustrated compound.@xcite the suppression of the value of magnetic entropy at high temperatures from the value expected for @xmath13 = 3/2 suggests that the effective spin of the co at the @xmath1-site is smaller than 3/2 , which could be a manifestation of both spin frustration and co - al site inversion effects . the presence of dynamic short - range magnetic ordering observed in neutron scattering experiments below @xmath209 50 k can be caused mainly by the site inversion effects . as pointed out previously , although a spin - liquid ground state has been proposed from magnetic susceptibility measurements for @xmath239 with low values of @xmath39,@xcite our nmr and nd measurements conclusively proved the existence of a long - range afm - ordered state below 10 k. since site inversion plays a crucial role in determining the magnetic properties of this spinel compound , it will also be interesting to perform systematic nmr and nd measurements on highly disordered ( high @xmath39 ) @xmath239 samples to investigate how the system changes from the afm state to a spin glass state . we thank dr . vladimir tsurkan for providing us the heat capacity data of the nonmagnetic reference compound @xmath241 in ref . . this research was supported by the u.s . department of energy , office of basic energy sciences , division of materials sciences and engineering . ames laboratory is operated for the u.s . department of energy by iowa state university under contract no . de - ac02 - 07ch11358 . 9 w. l. roth , j. phys . ( paris ) * 25 * , 507 ( 1964 ) . d. bergman , j. alicea , e. gull , s. trebst , and l. balents , nature phys . * 3 * , 487 ( 2007 ) . o. zaharko , n. b. christensen , a. cervellino , v. tsurkan , a. maljuk , u. stuhr , c. niedermayer , f. yokaichiya , d. n. argyriou , m. boehm , and a. loidl , phys . b * 84 * , 094403 ( 2011 ) . n. tristan , j. hemberger , a. krimmel , h .- a . krug von nidda , v. tsurkan , and a. loidl , phys . b * 72 * , 174404 ( 2005 ) . a. p. ramirez , annu . * 24 * , 453 ( 1994 ) . t. suzuki , h. nagai , m. nohara , and h. takagi , j. phys . : condens . matter * 19 * , 145265 ( 2007 ) . o. zaharko , a. cervellino , v. tsurkan , n. b. christensen , and a. loidl , phys . b * 81 * , 064416 ( 2010 ) . g. j. macdougall , d. gout , j. l. zarestky , g. ehlers , a. podlesnyak , m. a. mcguire , d. mandrus , and s. e. nagler , proceedings of the national academy of sciences * 38 * , 15693 ( 2011 ) . n. tristan , v. zestrea , g. behr , r. klingeler , b. bchner , h .- a . krug von nidda , a. loidl , and v. tsurkan , phys . b * 77 * , 094412 ( 2008 ) . k. hanashima , y. kodama , d. akahoshi , c. kanadani , and t. saito , j. phys . . jpn . * 82 * , 024702 ( 2013 ) . j. rodrguez - carvajal , physica b * 192 * , 55 ( 1993 ) ; see also www.illeu / sites / fullprof/. a. maljuk , v. tsurkan , v. zestrea , o. zaharko , a. cerellino , a. loidl , and d. n. argyriou , j. cryst . growth * 311 * , 3997 ( 2009 ) . m. hagiwara , s. kimura , n. nishihagi , t. suzuki , m. nohara , h. takagi , and k. kindo , j. low temp . phys . * 159 * , 11 ( 2010 ) . d. meschede , f. steglich , w. felsch , h. maletta , and w. zinn , phys . lett . * 44 * , 102 ( 1980 ) . l. r. walker and r. e. walstedt , phys . lett . * 38 * , 514 ( 1977 ) . k. miyatani , k. kohn , s. iida , and h. kamimura , j. phys . . jpn . * 20 * , 471 ( 1965 ) . edited by l. j. de jongh ( klewer , dordrecht , 1989 ) . t. fukai , y. furukawa , s. wada , and k. miyatani , j. phys . . jpn . * 65 * , 4067 ( 1996 ) . t. tsuda , a. hirai , and h. abe , phys . 26a * , 463 ( 1968 ) . y. furukawa , s. wada , t. kajitani , and s. hosoya , j. phys . 68 * , 346 ( 1999 ) . a. abragam , j. horowitz , and m. h. l. pryce , proc . ( london ) , ser . a * 230 * , 169 ( 1955 ) . a. narath , _ hyperfine interactions _ edited by a. j. freeman and r. b. frankel ( academic press , new york , 1967 ) . a. j. freeman and r. e. watson , _ magnetism _ edited by g. t. rado and h. shul ( academic press , new york , 1965 ) vol . 2 . a. narath , phys . rev . * 162 * , 320 ( 1967 ) . t. moriya , j. phys . * 18 * , 516 ( 1963 ) . a. v. mahajan , r. sala , e. lee , f. borsa , s. kondo , and d. c. johnston , phys . b * 57 * , 8890 ( 1998 ) . w. l. roth , j. phys . chem . solids * 25 * , 1 ( 1964 ) a. krimmel , v. tsurkan , d. sheptyakov , and a. loidl , physica b * 378 - 380 * , 583 ( 2006 ) . a. nakatsuka y. ikeda , y. yamasaki , n. nakayama and t. mizota , solid state commun . * 128 * , 85 ( 2003 ) .

given a birational normal extension of a two - dimensional local regular ring , we describe all the equisingularity types of the complete-primary ideals in whose blowing - up has some point whose local ring is analytically isomorphic to .

0707.1387

a sandwiched surface singularity @xmath8 is a normal surface singularity that can be projected birationally to a non - singular surface . from a more algebraic point of view , the local ring @xmath0 of any sandwiched singularity is a birational normal extension of a two - dimensional local regular ring @xmath4 . once a sandwiched surface singularity has been fixed , in this paper we address the problem of describing the equisingularity classes of all its birational projections to a plane . the problem of classifying the germs of sandwiched surface singularities was already posed by spivakovsky . as he claims in @xcite this problem has two parts : discrete and continuous . the continuous part is to some extent equivalent to the problem of the moduli of plane curve singularities , while the main result of this paper solves completely the combinatorial part . any birational projection from a sandwiched singularity to a plane is obtained by the morphism of blowing up a complete @xmath9-primary ideal in the local ring of a regular point @xmath10 on the plane . our goal is to give all the equisingularity types of these ideals . namely , fixed a birational normal extension @xmath0 of a local regular ring @xmath11 , we describe the equisingularity type of any complete @xmath9-primary ideal @xmath12 such that its blowing - up @xmath5 has some point @xmath6 whose local ring @xmath7 is analytically isomorphic to @xmath0 . in this case , we will say that the surface @xmath13 contains the singularity @xmath0 for short , making a slight abuse of language . this is done by describing the enriques diagram of the cluster of base points of any such ideal @xmath3 : such a diagram will be called an enriques diagram for the singularity @xmath0 . recall that an enriques diagram is a tree together with a binary relation ( proximity ) representing the topological equivalence classes of clusters of points in the plane ( see [ ed - dualg ] ) . previous works by spivakovsky @xcite and mhring @xcite describe a type of enriques diagram that exists for any given sandwiched surface singularity ( detailed in 2 ) and provide other types mostly in the case of cyclic quotients ( see @xcite 2.7 ) and minimal singularities ( see @xcite 2.5 ) . the organization of the paper is as follows . section 1 is devoted to recalling some definitions concerning the language of infinitely near points , sandwiched surface singularities and graphs . fundamental for our purpose will be the notion of enriques diagram , introduced in @xcite . in section 2 , after some technical results , we introduce the concept of _ contraction _ for a sandwiched surface singularity @xmath0 . by a contraction we mean the resolution graph @xmath14 of @xmath0 ( a sandwiched graph , as introduced in @xcite ) enriched by some proximities between their vertices , these proximities being compatible with the weights of the graph . fixed a sandwiched graph , the problem of finding the whole list of possibilities for such proximities is the hard part of our work . this is achieved in section 3 , by proving that any contraction for @xmath14 may be recovered from some contraction of the graph obtained from @xmath14 by removing one end . this fact is the key result in order to describe a procedure to obtain all the contractions for @xmath0 . finally , in section 4 , we explain how to complete contractions in order to obtain any enriques diagram for @xmath0 . * acknowledgement * the authors thank e. casas - alvero for drawing their attention to the problem addressed in this paper . they also thank m. spivakovsky for the conversations held on this topic . in this section , we fix notation and recall some of the facts concerning sandwiched surface singularities and base points of ideals which will be used throughout this paper , and we focus on our problem . a standard reference for most of the tools and techniques treated here is the book by casas - alvero @xcite . let @xmath15 be a regular local two - dimensional @xmath16-algebra and @xmath17 . a _ cluster _ of points of @xmath18 with origin @xmath10 is a finite set @xmath19 of points infinitely near or equal to @xmath10 such that , for any @xmath20 , @xmath19 contains all points to which @xmath21 is infinitely near . a subset of a cluster @xmath19 is a _ subcluster _ if it is a cluster ( with origin some point of @xmath19 ) . by assigning integral multiplicities @xmath22 to the points of @xmath19 , we obtain a _ weighted cluster _ @xmath23 ; the multiplicities @xmath24 are called the _ virtual multiplicities _ of @xmath25 . a point @xmath21 is said to be _ proximate _ to another point @xmath26 if @xmath21 is infinitely near to @xmath26 and lies on the strict transform of the exceptional divisor of blowing up @xmath26 . we write @xmath27 if @xmath21 is infinitely near or equal to @xmath26 , and @xmath28 if @xmath21 is proximate to @xmath26 . the relation @xmath29 is an ordering of the infinitely near points , and it will be considered as their natural ordering . a point @xmath20 is _ free _ if it is proximate to only one point , which is necessarily the immediate predecessor , and @xmath21 is _ satellite _ if it is proximate to two points ; otherwise , the point is necessarily the origin of the cluster . the number @xmath30 is the _ excess _ at @xmath21 of @xmath31 . _ consistent clusters _ are those weighted clusters with non - negative excesses at all their points . we write @xmath32 for the set of the _ dicritical _ points of @xmath25 , that is , the points with positive excess . if @xmath33 and @xmath34 , define the _ sum _ @xmath35 as the weighted cluster whose set of points is @xmath36 and whose virtual multiplicities are @xmath37 for @xmath38 ( @xcite 8.4 ) . this operation is clearly associative and commutative , thus making the set of all weighted clusters with origin at @xmath10 a semigroup . consider the set @xmath39 of all consistent clusters with origin at @xmath10 with positive virtual multiplicities . again , @xmath39 equipped with the sum is clearly a semigroup . a weighted cluster @xmath40 is called _ irreducible _ if it is so as element of the semigroup @xmath39 , that is , @xmath41 is not the sum of two elements of @xmath39 . to any point @xmath21 , @xmath42 , we associate the irreducible cluster @xmath43 in @xmath39 which has virtual multiplicity one at @xmath21 , which will be called the _ irreducible cluster in _ @xmath39 _ ending at _ @xmath21 . two _ clusters _ @xmath19 and @xmath44 are called _ similar _ if there is a bijection ( _ similarity _ ) @xmath45 so that both @xmath46 and @xmath47 preserve ordering and proximity . two _ weighted clusters _ @xmath33 and @xmath34 are called _ similar _ if there is a similarity between @xmath19 and @xmath44 preserving virtual multiplicities ( see @xcite 8.3 ) . an analytic isomorphism @xmath48 defined in a neighborhood of @xmath10 clearly induces a similarity between each cluster @xmath19 with origin @xmath10 and its image @xmath49 ( @xcite 3.3 ) . furthermore , if @xmath48 is only a homeomorphism , then @xmath19 and @xmath49 are still similar ( @xcite 8.3.12 ) if @xmath50 is the composition of the blowing - ups of all points in @xmath19 , write @xmath51 for the exceptional divisor of @xmath52 and @xmath53 for its irreducible components . we denote by @xmath54 the _ intersection matrix _ of @xmath51 : if @xmath55 , its coefficient is just the self - intersection of @xmath56 , and equals @xmath57 , where @xmath58 is the number of points in @xmath19 proximate to @xmath21 ; if @xmath59 , @xmath60 in case @xmath61 , and @xmath62 otherwise . it can be easily seen that @xmath61 if and only if @xmath21 is maximal among the points of @xmath19 proximate to @xmath26 or vice - versa ( cf . @xcite 4.4.2 ) . notice that @xmath63 is an invariant of the similarity class of @xmath19 . if @xmath31 is a weighted cluster , there is a well established notion for a germ of curve to go through @xmath41 ( which is a linear condition , see @xcite 4.1 ) , and the equations of all curves going through @xmath41 define a complete @xmath9-primary ideal @xmath64 in @xmath4 ( see @xcite 8.3 ) . any complete @xmath9-primary ideal @xmath3 in @xmath4 has a _ weighted cluster of base points _ , denoted by @xmath65 , which consists of the points shared by , and the multiplicities of , the curves defined by generic elements of @xmath3 . moreover , the maps @xmath66 and @xmath67 are reciprocal isomorphisms between the semigroup @xmath68 of complete @xmath9-primary ideals in @xmath4 ( equipped with the product of ideals ) and the semigroup @xmath69 ( see @xcite 8.4.11 for details ) . if @xmath42 , denote by @xmath70 the ideal in @xmath71 corresponding by the preceding isomorphism to the irreducible cluster @xmath72 ending at @xmath21 , that is , @xmath73 . a couple of ideals @xmath74 in @xmath71 are _ equisingular _ if @xmath65 and @xmath75 are similar ( @xcite 8.3 ) . notice that two equisingular complete ideals in @xmath71 have equisingular ( that is , topologically equivalent ) generic germs and equal codimensions ( @xcite 8.3.9 ) . the main references here are @xcite and @xcite . if @xmath76 , we denote by @xmath77 the blowing - up of @xmath78 . the surface @xmath13 is not regular in general , and its singularities are sandwiched singularities . moreover , if @xmath19 is the set of base points of @xmath78 , we have a commutative diagram @xmath79^{f}\ar[rd]_{\pi_k } & { x } \ar[d]^{\pi_i } \\ & { s}}\ ] ] where the morphism @xmath80 , given by the universal property of the blowing - up , is the minimal resolution of the singularities of @xmath13 ( @xcite remark 1.4 ) . let @xmath0 be any singularity of @xmath13 ; then we say that @xmath78 is an _ ideal for _ @xmath0 . it follows that the exceptional divisor @xmath81 associated with the minimal resolution of @xmath0 is a connected subset of the exceptional divisor @xmath51 . there is a bijection between the set of irreducible components of @xmath82 and the set of dicritical points of @xmath83 ( see @xcite ) . this allows to write @xmath84 for the set of these components on @xmath13 . because of this , we may think of @xmath0 as a singularity obtained by contracting a connected curve ( which will be called @xmath81 ) of @xmath51 containing no component with self - intersection @xmath85 ( such a component @xmath56 is necessarily the exceptional divisor of the blowing - up of some maximal point of @xmath19 and thus , a dicritical point ) . for any ideal @xmath86 with positive @xmath87 , we have an analytic isomorphism @xmath88 ( cf . @xcite , corollary i.1.5 ) . since we are interested in sandwiched singularities modulo analytic isomorphism , the relevant information we need to retain about @xmath83 is , on one hand , its set of points @xmath19 and , on the other , knowing which of the points of @xmath19 are dicritical ( the rest being non - dicritical , of excess zero ) . we introduce the enriques diagrams and the weighted dual graphs related to them . the enriques diagrams are combinatorial objects that enclose the topological information of the clusters of infinitely near points in @xmath18 , namely they represent the similarity classes of clusters . a _ tree _ is a finite graph with a partial order relation @xmath89 between the vertices , without loops , which has a single initial vertex , or _ root _ , and every other vertex has a unique immediate predecessor . the vertex @xmath26 is said to be a _ successor _ of @xmath21 if @xmath21 is the immediate predecessor of @xmath26 . if @xmath21 has no successors then it is an _ extremal vertex_. the set of vertices of a graph will be denoted by the same letter as the graph itself . an _ enriques diagram _ @xmath90 ( @xcite enriques iv.i , @xcite casas 3.9 ; see also @xcite and @xcite for a combinatorial presentation ) is a tree with a binary relation between vertices , called _ proximity _ and denoted by @xmath91 , which satisfies : 1 . every vertex but the root is proximate to its immediate predecessor ; the root is proximate to no vertex . 2 . if @xmath92 , then @xmath93 and there is at most one other vertex in @xmath90 proximate to both of them . any vertex is proximate to at most two other vertices . the vertices which are proximate to two points are called _ satellite _ , the other vertices , but the root , are called _ free_. if @xmath26 is the immediate predecessor of @xmath21 , and @xmath94 , then @xmath95 . if @xmath21 is a vertex in @xmath90 , we write @xmath96 for the number of vertices in @xmath90 proximate to @xmath21 . a satellite vertex is said to be _ satellite of _ the last free vertex that precedes it . in order to express graphically the proximity relation , enriques diagrams are drawn according to the following rules : 1 . if @xmath26 is a free successor of @xmath21 then the edge going from @xmath21 to @xmath26 is smooth and curved and , if @xmath21 is not the root , it has at @xmath21 the same tangent as the edge joining @xmath21 to its predecessor . 2 . the sequence of edges connecting a maximal succession of vertices proximate to the same vertex @xmath21 are shaped into a line segment , orthogonal to the edge joining @xmath21 to the first vertex of the sequence . if @xmath19 is a cluster , there is an enriques diagram @xmath97 naturally associated with it by taking one vertex for each point of @xmath19 and the proximity of the cluster as the proximity of @xmath97 ; conversely , for any enriques diagram @xmath90 there is some cluster @xmath19 with origin @xmath10 whose enriques diagram @xmath97 is @xmath90 . if no confusion may arise , we will label the points in @xmath19 and their corresponding vertices in @xmath97 with the same symbol . a connected subtree of an enriques diagram @xmath90 is a _ subdiagram _ if it is an enriques diagram with root some vertex of @xmath90 and whose proximity is the restriction of the proximity of @xmath90 . observe that @xmath44 is a subcluster of @xmath19 if and only if the associated enriques diagram @xmath98 is a subdiagram of @xmath97 . if @xmath90 is the enriques diagram associated with @xmath19 and @xmath99 , we denote by @xmath100 the enriques diagram of the irreducible cluster @xmath43 ending at @xmath21 . if @xmath21 is extremal , @xmath100 is called a _ branch _ of @xmath90 . by assigning to an enriques diagram @xmath90 a _ marking map _ @xmath101 , we obtain a _ marked enriques diagram _ @xmath102 . any consistent cluster @xmath25 induces a marking map @xmath103 by taking @xmath104 if @xmath21 corresponds to a dicritical point of @xmath25 ( in this case , @xmath21 is called a _ dicritical vertex _ ) , and @xmath105 otherwise . a _ marked subdiagram _ @xmath106 of @xmath107 is a marked enriques diagram where @xmath108 is a subdiagram of @xmath90 and @xmath109 is the restriction of @xmath110 to @xmath108 . observe that the extremal vertices of a marked enriques diagram associated with some @xmath111 are always dicritical . if @xmath0 is a sandwiched surface singularity , we say that @xmath112 is an _ enriques diagram for _ @xmath0 if it is the marked enriques diagram of @xmath113 , for some ideal @xmath76 for @xmath0 . under this framework , the goal of this paper is to describe all the enriques diagrams for a given @xmath0 . incidence between the irreducible components of a divisor @xmath114 on a surface is usually represented by means of the _ weighted dual graph _ of @xmath114 . it is defined by taking a vertex for each component of @xmath114 , and by joining two vertices by an edge if and only if the corresponding components of @xmath114 meet ; each vertex is weighted by taking minus the self - intersection of the corresponding component . if @xmath90 is the enriques diagram of a cluster @xmath19 , the _ ( weighted ) dual graph _ of @xmath90 , denoted by @xmath115 , is the weighted dual graph of the exceptional divisor @xmath51 ( which has no loops ) . since the information enclosed in the weighted dual graph is the same as that contained in the intersection matrix of @xmath19 , this definition is consistent . the similarity class of a cluster may be represented either by its enriques diagram or by its weighted dual graph , since from the intersection matrix the ordering @xmath116 ( of being infinitely near ) and the proximity may be inferred . in fact , this is also true for rational surface singularities . from the intersection matrix @xmath117 of a rational surface singularity , the fundamental cycle @xmath118 may be computed ( see @xcite theorem 4.2 ) and from it , the order of the blowing - ups performed to resolve the singularity : the negative entries of @xmath119 correspond to the exceptional components having appeared in the last blowing - up ( cf . theorem 1.14 of @xcite ) . it is worth noticing that the proximity of @xmath90 can not be recovered in general only from its dual graph without weights ( see @xcite 4.4 ) . a _ non - singular graph _ is the weighted dual graph of some enriques diagram ( cf . the vertex in @xmath120 corresponding to @xmath21 in @xmath90 will be denoted by @xmath121 , written in _ roman _ font . by the ( weighted ) dual graph of a marked enriques diagram @xmath122 we mean the dual graph of @xmath90 and it will also be denoted by @xmath123 . the vertices of @xmath123 corresponding to dicritical vertices ( non - dicritical , respectively ) of @xmath107 will be called _ dicritical _ ( _ non - dicritical _ , respectively ) , too . if @xmath78 is a complete @xmath9-primary ideal in @xmath4 , we will write @xmath124 and @xmath125 to mean the marked enriques diagram and the weighted dual graph of its cluster of base points @xmath113 , respectively . [ w=1extrem ] the weighted dual graph @xmath126 can be constructed as follows : take one vertex in @xmath120 for each vertex of @xmath90 , and connect two vertices in @xmath120 by an edge if and only if one of the corresponding vertices in @xmath90 is maximal among the vertices in @xmath90 proximate to the other . moreover , if @xmath121 is a vertex of @xmath120 , its weight @xmath127 is @xmath128 ( cf . 4.4 of @xcite for details ) . a vertex @xmath129 has weight @xmath130 if and only if @xmath121 is extremal in @xmath90 . a _ chain _ @xmath131 of a graph @xmath132 without loops is the subgraph composed of all vertices and edges between the vertices @xmath133 ; it will be described by the ordered sequence of vertices between @xmath134 and @xmath121 , and @xmath135 will denote its length . two vertices @xmath133 are _ adjacent _ if @xmath136 ; a vertex is an _ end _ if it is adjacent to only one vertex . a _ weighted subgraph _ of a weighted graph @xmath132 is a subgraph of @xmath132 whose vertices have the same weights as @xmath132 . the following result describes the proximity relations between the vertices of a chain : [ chain ] let @xmath137 be two vertices of an enriques diagram @xmath90 , and consider the non - singular graph @xmath132 of @xmath90 . * if @xmath138 , then @xmath139 ; if @xmath140 , either @xmath141 or @xmath142 . moreover , all the vertices of @xmath131 correspond to vertices in the same branch of @xmath90 . * write @xmath143 . there exists some @xmath144 satisfying @xmath145 for @xmath146 , and @xmath147 for @xmath148 . furthermore , if @xmath149 , @xmath150 is proximate to some @xmath151 with @xmath152 . the first assertion of ( a ) is just lemma 3.2 of @xcite . now , if @xmath153 , either @xmath154 or @xmath155 ; in any case , either @xmath156 is infinitely near to @xmath157 or viceversa , and hence @xmath156 and @xmath157 can not belong to different branches of @xmath90 . now , we prove ( b ) . first of all , note that for any @xmath158 either @xmath159 is proximate to @xmath160 or viceversa ( cf . [ w=1extrem ] ) . by ( a ) , @xmath161 is necessarily infinitely near to @xmath162 and so , proximate to it . if each @xmath160 is proximate to @xmath159 , the first claim is obvious by taking @xmath163 . assume that there exists some @xmath164 such that @xmath159 is proximate to @xmath160 , and take @xmath165 to be minimal with this property . we claim that @xmath166 for @xmath146 , and @xmath167 for @xmath148 . to show this , assume that there exists some @xmath168 such that @xmath169 and take @xmath170 to be minimal . then , both @xmath171 and @xmath172 are proximate to @xmath173 and , since they are adjacent to it , they are maximal among the vertices of @xmath90 proximate to @xmath173 . however , by ( a ) they are in the same branch of @xmath90 , so they must be equal , which is impossible . note that @xmath174 is the maximal point in @xmath90 among the vertices belonging to @xmath131 . by ( a ) we know that every @xmath150 , @xmath149 is infinitely near to @xmath26 . write @xmath151 for the maximal vertex in @xmath90 among the vertices belonging to @xmath175 such that @xmath150 is infinitely near to it . by ( a ) applied to @xmath176 and the maximality of @xmath151 , necessarily @xmath177 is infinitely near to @xmath150 and , because @xmath177 is proximate to @xmath151 , so is @xmath150 . this completes the proof . the _ resolution graph _ of a sandwiched singularity @xmath0 is the weighted dual graph of the exceptional divisor of the minimal resolution of @xmath0 . these graphs are called _ sandwiched graphs _ and they are characterized as the weighted subgraphs of some non - singular graph containing no vertices of weight @xmath178 ( see @xcite proposition ii.1.11 ; cf . forthcoming [ rem_g ] ) . in particular , the graph obtained from a sandwiched graph by removing an end is still a sandwiched graph . [ rem_g ] if @xmath107 is an enriques diagram for @xmath0 and @xmath179 is the weighted subgraph of @xmath123 comprising only the non - dicritical vertices , then @xmath14 equals one of the connected components of @xmath179 , whose vertices ( and their corresponding vertices in @xmath90 ) will be called _ non - dicritical vertices relative _ to @xmath0 . in remark [ rem_g ] we have observed that if @xmath107 is an enriques diagram for a sandwiched surface singularity @xmath0 , then its dual graph @xmath123 contains the resolution graph @xmath14 as a weighted subgraph . given any enriques diagram @xmath90 , the following proposition shows that this combinatorial condition is sufficient to infer a result of geometrical nature : suitable marking maps @xmath110 can be chosen so that @xmath180 becomes an enriques diagram for @xmath0 . [ graph - ideal ] let @xmath0 be a sandwiched surface singularity and let @xmath181 be a marked enriques diagram . then , @xmath107 is an enriques diagram for @xmath0 if and only if 1 . the dual graph @xmath120 contains @xmath182 as a weighted subgraph . 2 . @xmath183 if @xmath184 ; and @xmath185 if @xmath186 and it is adjacent to some vertex of @xmath182 . the `` only if '' part follows from [ rem_g ] , and , since @xmath187 is an enriques diagram for @xmath0 , the marking map @xmath110 must satisfy the statement in order to assure that the non - dicritical vertices of @xmath90 relative to @xmath0 correspond exactly to the vertices of @xmath182 . for the `` if '' part , consider the reduced exceptional divisor @xmath81 of the minimal resolution of @xmath0 . using plumbing ( see @xcite , @xcite remark i.1.10 ) , we can glue smooth rational curves in order to obtain a configuration @xmath188 containing @xmath81 , having dual graph @xmath120 and being contained on a smooth surface @xmath189 . since @xmath120 is a non - singular graph , by castelnuovo s criterion @xmath188 contracts to a non - singular point @xmath10 ( @xcite ii.1.10 ) on a surface this contraction factors into a composition of point blowing - ups ( @xcite theorem 5.7 ) , so it determines a cluster with origin at @xmath10 and having enriques diagram @xmath90 ( @xcite 4.4 ) . it is clear that @xmath189 is the surface obtained from @xmath18 by blowing up all points in @xmath19 . consider a system of virtual multiplicities @xmath190 for @xmath19 in such a way that all the points have positive excess except for those corresponding to the vertices of the subgraph @xmath191 , which have excess 0 ( such a system exists by @xcite 8.4.1 ) . notice that no maximal point of @xmath19 corresponds to a vertex of @xmath182 , since any vertex of the resolution graph @xmath182 has weight strictly greater than @xmath178 . hence @xmath192 is a consistent cluster with positive virtual multiplicities . let @xmath193 and @xmath194 . the morphism @xmath195 given by the universal property of blowing up is the minimal resolution of the singularities of @xmath13 ( @xcite ii.1.4 ) and its exceptional components correspond to those points of @xmath196 having excess 0 . therefore , @xmath81 is the exceptional divisor of @xmath80 and so , the singularity on @xmath13 given by its contraction is isomorphic to @xmath197 ( @xcite theorem 3.13 ) . given a sandwiched surface singularity @xmath0 , there is not a unique non - singular graph @xmath132 containing @xmath14 as a weighted subgraph ( see example [ ex1 ] ) . in fact , it is possible to construct infinitely many non - singular graphs @xmath132 containing a given sandwiched graph @xmath14 , as example [ ex2 ] shows . [ ex2 ] if @xmath76 is an ideal for @xmath0 and @xmath194 , by choosing any non - singular point in the exceptional locus of @xmath13 , and blowing up this point , we obtain a new surface @xmath198 containing @xmath0 , as well . this @xmath198 is the blowing - up of a complete @xmath9-primary ideal @xmath199 , where @xmath200 has codimension one in @xmath78 ( theorem 3.5 of @xcite ) , and the dual graph @xmath201 contains a vertex more than @xmath125 . in this way , an infinite chain of ideals in @xmath202 @xmath203 for @xmath0 can be constructed , and each @xmath204 contains @xmath14 as a weighted subgraph . moreover , for any @xmath205 the enriques diagram @xmath206 is a marked subdiagram of @xmath207 . [ l.contraction ] let @xmath107 be an enriques diagram for @xmath0 . consider the set @xmath208 of non - dicritical vertices of @xmath107 relative to @xmath0 . * there is a tree structure on @xmath208 induced by the natural ordering @xmath89 of @xmath90 . * for any @xmath209 define @xmath210 if and only if @xmath211 . then,@xmath212 is a proximity , which turns @xmath208 into an enriques diagram . to exhibit the tree structure of @xmath208 , we will prove that 1 . there is a unique minimal element of @xmath208 by @xmath89 , which is taken as the root of @xmath208 ; 2 . for any @xmath213 , its immediate predecessor in @xmath208 is the maximal element of @xmath214 . suppose that @xmath21 and @xmath26 are two different minimal vertices in @xmath208 , and write @xmath215 for the maximal vertex in @xmath216 ( this is , the maximal vertex which both @xmath26 and @xmath21 are infinitely near or equal to ) . then , as @xmath123 contains no loops , @xmath217 and , by the connectivity of @xmath218 , we infer that @xmath219 , contradicting the minimality of @xmath26 and @xmath21 . we denote by @xmath220 the minimal vertex of @xmath208 , which is set as the root of @xmath208 . on the other hand , if @xmath221 , @xmath222 , the vertices in @xmath100 are totally ordered by the natural ordering @xmath89 of @xmath90 . hence , there exists a unique immediate predecessor of @xmath21 , which is the maximal element of @xmath223 , and this proves ( a ) . now , to prove ( b ) , we show that @xmath224 defined as above is a proximity relation for @xmath208 . since the root is the minimal vertex of @xmath208 , it is clear that it is proximate to no other vertex of @xmath208 . if @xmath222 , its immediate predecessor @xmath225 in @xmath208 is the maximal element of @xmath223 ; hence @xmath226 and @xmath227 . then ( b ) of [ chain ] says that @xmath21 is proximate to some vertex @xmath228 of @xmath229 , and ( a ) of [ chain ] says that @xmath215 is infinitely near or equal to @xmath225 . since @xmath230 , this leads to contradiction . this proves the first condition of the proximity ( see 1.3 ) . the conditions 2 and 3 are clearly satisfied . an enriques diagram @xmath208 obtained as in lemma [ l.contraction ] will be called _ a contraction for _ @xmath0 ( or _ for _ @xmath14 ) _ associated with _ reciprocally , we will also say that @xmath107 is _ associated with the contraction _ @xmath208 . [ ] notice that , by virtue of [ graph - ideal ] , any enriques diagram ( respectively , any contraction ) for @xmath0 is in fact an enriques diagram ( respectively , a contraction ) for any sandwiched surface singularity whose resolution graph is @xmath14 . a contraction for @xmath0 may also be regarded as an enrichment of the resolution graph @xmath14 by some proximities between their vertices , these proximities being compatible with the weights of @xmath14 in the sense of lemma [ l.ineq-weights ] below . [ ex1 ] figure [ eds ] provides three distinct enriques diagrams for the same sandwiched singularity : they are not apparently related , namely one is not a subdiagram of the other , as was the case in example [ ex2 ] . notice that @xmath231 and @xmath232 give rise to the same contraction , which is the enriques subdiagram of @xmath233 comprising the black dots . three different marked enriques diagrams for the same singularity @xmath0 and their corresponding dual graphs . dicritical vertices are represented with white dots . ] [ r.contraction ] if @xmath107 is an enriques diagram for @xmath0 and @xmath208 is the associated contraction , @xmath208 is not , in general , an enriques subdiagram of @xmath90 ( see enriques diagram @xmath232 of figure [ eds ] ) . in particular , if @xmath76 is an ideal for @xmath0 with enriques diagram @xmath107 , the set of points of @xmath83 corresponding to the vertices of @xmath208 does not constitute , in general , a subcluster of @xmath25 . [ l.ineq-weights ] let @xmath208 be a contraction for @xmath14 associated with an enriques diagram @xmath107 . then for any vertex @xmath213 , @xmath234 and the inequality is strict at the extremal vertices of @xmath208 . in particular , @xmath218 is not a weighted subgraph of @xmath126 . the inequality comes from the definition of contraction , since the number of vertices proximate to @xmath21 in @xmath208 is less or equal than in @xmath90 . if @xmath21 is extremal in @xmath208 , @xmath235 , while @xmath236 , since @xmath21 is a non - dicritical vertex of @xmath107 and hence necessarily non - extremal in @xmath90 ( see [ w=1extrem ] ) . the last assertion follows by considering the weights at the extremal vertices of @xmath208 . in @xcite corollary ii.1.14 , spivakovsky introduced a type of birational projection into a plane that could be achieved for any sandwiched singularity . namely he showed that , once a sandwiched surface singularity @xmath0 is fixed , an ideal @xmath76 can be chosen in such a way that : * @xmath0 is the only singularity of @xmath194 ; * the strict transform ( by the minimal resolution of @xmath13 ) of any exceptional component of @xmath237 is a curve of the first kind , that is , the strict transform by @xmath80 ( see diagram [ commdiagr ] ) of any @xmath238 with @xmath239 has self - intersection equal to @xmath85 . an ideal satisfying the above conditions ( i ) and ( ii ) ( cf . @xcite 2.3 ) will be called an _ s - ideal for _ @xmath0 . a marked enriques diagram associated with an s - ideal for @xmath0 will be called an _ s - enriques diagram for _ @xmath0 . the following result describes what the equisingularity classes of s - enriques diagrams look like : [ idspivak ] an ideal @xmath240 is an s - ideal if and only if the dicritical vertices of @xmath124 are free and extremal . write @xmath107 for the enriques diagram of @xmath83 . first of all , note that @xmath194 has only one singularity @xmath0 if and only if any non - dicritical vertex of @xmath123 belongs to @xmath14 . let @xmath241 and assume that there exists some @xmath242 infinitely near to @xmath21 . we may assume that @xmath26 is an immediate successor of @xmath21 . then , @xmath243 against condition @xmath244 . therefore , @xmath21 must be maximal in @xmath19 . now , assume that @xmath21 is satellite , proximate to @xmath161 and @xmath245 . then , @xmath246 . necessarily , @xmath161 and @xmath245 are not dicritical points of @xmath19 and thus , @xmath247 . if follows that @xmath248 against the assumption @xmath241 . conversely , if the dicritical vertices of @xmath107 are free and extremal , the union of the non - dicritical vertices of @xmath123 is connected and hence @xmath13 has only one singularity . moreover , as above , the self - intersection of the strict transform on @xmath249 of any component @xmath238 with @xmath241 is @xmath85 . a contraction @xmath208 associated with an s - enriques diagram will be called an _ s - contraction_. contrary to what happened for general contractions ( recall [ r.contraction ] ) , an s - contraction @xmath208 is a subdiagram of its associated s - enriques diagram @xmath107 ; furthermore , any s - contraction is associated with a unique s - enriques diagram : [ p.s - contraction ] if @xmath107 is an s - enriques diagram for @xmath0 , then an s - contraction @xmath208 associated with @xmath107 satisfies : 1 . @xmath208 is a subdiagram of @xmath90 ; 2 . @xmath90 can be recovered from @xmath208 by adding at any vertex @xmath213 a number of @xmath250 free successors ; @xmath187 is recovered by defining the marking map @xmath110 as @xmath251 if @xmath213 , and @xmath104 otherwise . by virtue of [ idspivak ] , any dicritical point of @xmath90 is free and extremal . therefore , for any @xmath221 , there are exactly @xmath250 of these vertices in the first neighborhood of @xmath21 . this gives both claims . next result is a sort of converse of lemma [ l.ineq-weights ] : [ lem_2.2 ] let @xmath208 be an enriques diagram and assume that the dual graph @xmath218 equals @xmath14 and satisfies @xmath252 at each vertex @xmath213 . then @xmath208 is an s - contraction for @xmath14 . to any @xmath221 add exactly @xmath253 free successors to obtain from @xmath208 a marked enriques diagram @xmath107 whose dicritical vertices are the extremal ones . by construction the dicritical vertices are also free , @xmath14 is a weighted subgraph of @xmath126 . then invoking [ graph - ideal ] and [ idspivak ] we are done . let us end this section by showing that the family of contractions for a sandwiched surface singularity equals the family of s - contractions : [ generalityspivideals ] any contraction for @xmath14 is an s - contraction . let @xmath107 be an enriques diagram for @xmath0 and let @xmath208 be a contraction for @xmath14 associated with @xmath107 . define a new enriques diagram @xmath108 from @xmath208 by adding to each vertex @xmath221 as many free successors as the number of vertices in @xmath254 that are proximate to @xmath21 . taking @xmath105 if @xmath221 and @xmath104 otherwise , the resulting enriques diagram @xmath255 for @xmath0 is an s - enriques diagram associated with @xmath208 ( see [ graph - ideal ] ) . in this section we describe all the contractions for a given sandwiched surface singularity @xmath0 . observe that , by virtue of [ p.s - contraction ] and [ generalityspivideals ] , this is equivalent to listing all the equisingularity classes of the s - ideals for @xmath0 . suppose that the resolution graph @xmath256 has @xmath205 vertices and that @xmath257 is an end of @xmath258 . the weighted graph obtained by removing @xmath257 is again a sandwiched graph and will be denoted by @xmath259 ( see [ ed - dualg ] ) . we want to detail a procedure to obtain all the contractions for @xmath260 from the contractions for @xmath259 . then , by induction on @xmath205 , the whole list of contractions for a given sandwiched singularity will be inferred just from its resolution graph . the first result of this section describes how the vertex @xmath157 looks like in any contraction : [ lem_2.1 ] the vertex @xmath157 ( corresponding to the end @xmath257 of @xmath14 ) in any contraction @xmath208 for @xmath0 is either the root or free . furthermore , if @xmath157 is not the root of @xmath208 , then either @xmath157 is extremal or @xmath157 has a unique successor , which is satellite of @xmath157 . assume that @xmath157 is satellite , proximate to the vertices @xmath161 and @xmath245 in @xmath208 , and suppose that @xmath245 is proximate to @xmath161 . we will show that @xmath261 , thus contradicting that @xmath257 is an end , and proving the first claim . consider the enriques subdiagram @xmath262 of @xmath208 comprising all the points preceding or equal to @xmath157 . in particular , @xmath263 are both in @xmath262 , and @xmath157 is maximal among the points of @xmath262 proximate to @xmath161 , and also to @xmath245 . hence , as vertices of @xmath264 , @xmath257 is adjacent to both @xmath265 and @xmath266 and so , @xmath267 . now , the rest of vertices of @xmath268 all lie after some vertex of @xmath262 , giving rise to blowing - ups of extra points . the combinatorial effect of these blowing - ups is translated in the dual graph by the elementary modifications introduced in i.1.5 of @xcite , those of the first kind corresponding to the blowing - ups of free points while those of the second kind to the blowing - ups of satellite points . from their definition , it is immediate that these modifications respect the property of being in the chain determined by two vertices already in the graph . for the second claim , let @xmath134 be the only vertex in @xmath269 to which @xmath257 is adjacent , and assume that @xmath157 is not the root of @xmath208 . we distinguish two cases . the first one is when @xmath157 is maximal among the points in @xmath208 proximate to @xmath26 . since @xmath157 is an end , there are no vertices in @xmath208 proximate to @xmath157 and @xmath157 is an extremal vertex of @xmath208 . the second case is when @xmath26 is maximal among the points in @xmath208 proximate to @xmath157 . since @xmath157 is and end , @xmath157 must have a unique successor , say @xmath215 , preceding @xmath26 . denote by @xmath156 the immediate predecessor of @xmath157 . since @xmath157 is free and @xmath257 is adjacent only to @xmath134 , @xmath157 can not be the last point in @xmath208 proximate to @xmath156 . hence @xmath215 must be proximate to @xmath156 and so @xmath215 is satellite of @xmath157 . this completes the proof . the following result shows how to construct a contraction for @xmath270 from a contraction for @xmath271 . moreover , any contraction for @xmath270 can be obtained in this way . [ rules ] suppose @xmath271 is obtained from a sandwiched graph @xmath270 by removing an end @xmath257 . let @xmath272 be the unique vertex in @xmath270 to which @xmath257 is adjacent , and let @xmath273 be a contraction for @xmath271 . define a new enriques diagram by taking @xmath274 and adding to @xmath273 extra proximities relating @xmath157 according to one of the following rules : 1 . if @xmath275 , set @xmath157 in @xmath208 as a free successor of @xmath156 . 2 . if @xmath276 are free vertices in @xmath273 with @xmath277 and @xmath278 , then set @xmath279 for all @xmath280 , and either set @xmath281 , provided @xmath225 is the root of @xmath273 and @xmath282 ( with @xmath157 becoming the root of @xmath208 ) , or set @xmath283 , provided @xmath284 . then @xmath208 is a contraction for @xmath285 . moreover , any contraction @xmath208 for @xmath270 can be constructed from some contraction @xmath273 for @xmath271 as above . clearly @xmath208 defined as above satisfies @xmath286 and @xmath287 at each vertex @xmath288 . thus , invoking [ lem_2.2 ] and [ generalityspivideals ] , the first claim follows . by virtue of [ lem_2.1 ] , the vertex @xmath157 of @xmath208 corresponding to @xmath257 is either free or the root of @xmath208 . since @xmath257 is adjacent to @xmath272 in @xmath289 , there are only two possibilities for their corresponding vertices in @xmath208 : case 1 : : : @xmath157 is maximal among the vertices in @xmath208 proximate to @xmath156 . hence @xmath157 can not be the root of @xmath208 , and by [ lem_2.1 ] @xmath157 is free . moreover , @xmath157 is an extremal vertex of @xmath208 : otherwise , [ lem_2.1 ] implies that @xmath157 has a unique successor , which is satellite of @xmath157 and thus proximate to @xmath156 , contradicting the maximality of @xmath157 among the vertices proximate to @xmath156 . therefore the set of vertices of @xmath290 is connected and has a tree structure . by considering the restriction of the proximity of @xmath208 to @xmath273 , @xmath273 becomes an enriques diagram . clearly the graphs @xmath291 and @xmath292 are equal ( disregarding weights ) . observe that @xmath293 and that @xmath294 if @xmath295 , @xmath296 . applying [ w=1extrem ] , we have @xmath297 for any @xmath298 and @xmath299 thus , invoking [ lem_2.2 ] and [ generalityspivideals ] , @xmath273 is a contraction for @xmath271 . finally , notice that the enriques diagram @xmath208 is obtained from @xmath273 by the procedure of the first rule of the statement , and we are done in this case . case 2 : : : @xmath156 is maximal among the vertices in @xmath208 proximate to @xmath157 . let @xmath300 be the vertices in @xmath208 preceding or equal to @xmath156 which are proximate to @xmath157 . first of all , we define on the set of vertices of @xmath301 a tree structure . we distinguish two cases : + * 2.1 . * if @xmath157 is the root of @xmath208 , then take @xmath302 as the root of @xmath273 , and for any @xmath303 declare that @xmath21 is the immediate predecessor of @xmath26 in @xmath273 if and only if @xmath21 is the immediate predecessor of @xmath26 in @xmath208 . + * 2.2 . * otherwise , take the root of @xmath208 as the root of @xmath273 ; for any @xmath303 declare that @xmath21 is the immediate predecessor of @xmath26 in @xmath273 if and only if @xmath21 is the immediate predecessor of @xmath26 in @xmath208 ; declare that the immediate predecessor of @xmath302 in @xmath273 is the immediate predecessor @xmath304 of @xmath157 in @xmath208 . + restrict the proximity of @xmath208 to @xmath273 , namely , for any @xmath305 set @xmath306 if and only if @xmath307 . let us check that it satisfies the properties 1 to 3 of a proximity ( see [ ed - dualg ] ) . in the first case ( where @xmath157 is the root of @xmath208 ) these properties are clearly satisfied . in the second case , the only condition that must be checked is property 1 for the vertex @xmath302 , namely , that @xmath308 . since @xmath302 is a successor of @xmath157 in @xmath208 , by [ lem_2.1 ] we infer that @xmath302 is satellite of @xmath157 . thus @xmath302 is satellite in @xmath208 : @xmath302 is proximate to its immediate predecessor in @xmath208 , which is @xmath157 , and to some point of @xmath208 , say @xmath21 ; moreover , @xmath157 must be proximate to @xmath21 , as well . since @xmath157 is proximate to @xmath304 , we infer that @xmath309 and @xmath308 , as desired . therefore @xmath273 is an enriques diagram , whose dual graph @xmath291 equals @xmath292 disregarding weights . observe that @xmath310 and @xmath311 if @xmath312 , @xmath313 . applying [ w=1extrem ] , we have @xmath297 for any @xmath314 and @xmath315 thus , invoking [ lem_2.2 ] and [ generalityspivideals ] , @xmath273 is a contraction for @xmath271 . finally , it remains to show that the enriques diagram @xmath208 may be obtained from @xmath273 by the procedure of the second rule of the statement . indeed , according to the proximity defined in @xmath273 , notice first that @xmath316 is a chain of free vertices in @xmath273 preceding or equal to @xmath156 , and that @xmath302 is the root of @xmath273 if and only if @xmath157 is the root of @xmath208 . on the other hand , recall that the proximity relations in @xmath208 involving the vertex @xmath157 are @xmath317 and the further proximity relation @xmath318 must be added in case @xmath302 is not the root of @xmath273 . this is exactly what performs the operation of the second rule , and we are done . the whole list of contractions for @xmath14 are obtained by applying recursively the rules of [ rules ] . let us just sketch the main steps of an implementation of this procedure . each step of this procedure adds a new vertex keeping the proximities already defined . the idea is that the bigger the weights of @xmath14 are , the more enriques diagrams for @xmath0 can be found . * choose any vertex of @xmath14 , say @xmath319 , and take @xmath320 and @xmath321 . * step i. * assume that @xmath322 , @xmath323 have been obtained , where @xmath322 is a subgraph of @xmath14 . choose a vertex @xmath324 adjacent to some @xmath325 . the graph @xmath326 is obtained by adding @xmath324 with weight @xmath327 to @xmath322 , adjacent to @xmath134 ; the new enriques diagram @xmath328 is obtained from @xmath323 by adding @xmath329 according to one of the rules of [ rules ] : 1 . if @xmath330 , @xmath329 can be added to @xmath323 as a free successor of @xmath26 , @xmath331 . if @xmath332 are free vertices in @xmath323 with @xmath333 and @xmath334 , then set @xmath335 for all @xmath336 , and either set @xmath337 , provided @xmath225 is the root of @xmath323 and @xmath338 ( with @xmath329 becoming the root of @xmath328 ) , or set @xmath339 , provided @xmath340 . the procedure stops at step @xmath205 , the number of vertices of @xmath14 . at this point , the obtained weighted graph @xmath270 equals @xmath14 , and the enriques diagram @xmath341 is just a contraction for @xmath14 . at any step of the procedure , there may be several choices to add a fixed new vertex ( for example , we may apply either rule 1 or 2 to add the vertex @xmath329 to @xmath323 ) . in order to obtain the whole list of all the contractions for @xmath14 , all these possibilities must be performed . it might also happen that an enriques diagram @xmath323 to which the new vertex can not be added is reached . this means that no enriques diagram for @xmath0 with the subset of proximities of @xmath323 exists . minimal singularities are rational surface singularities whose fundamental cycle is reduced . they are characterized as those sandwiched singularities that can be obtained by blowing up a complete ideal all whose base points are free ( see @xcite 2.5 ; cf . @xcite ) . as a consequence of our results , a sandwiched surface singularity @xmath0 is minimal if and only if there exists a contraction for @xmath0 that is obtained by applying the first rule at each step of the above procedure . [ exsp ] let @xmath0 be a singularity whose resolution graph is shown at the bottom of figure [ exsp ] . by applying the procedure just described , we obtain the whole list of contractions for @xmath0 . the s - enriques diagrams shown in figure [ llistated ] are obtained by adding free successors to them as explained in ( b ) of [ p.s - contraction ] . the complete list of s - enriques diagrams for a singularity @xmath0 with resolution graph @xmath14 in example [ exsp ] . the white - filled dots represent the dicritical points , added to the contractions . ] in this section we address the problem of describing the equisingularity classes of the ideals for a given sandwiched surface singularity @xmath0 , that is , of describing all the possible enriques diagrams for @xmath0 . the ( finite ) family of contractions for @xmath0 was inferred from the resolution graph @xmath14 of @xmath0 by the procedure explained in the preceding section . it remains to find out all the different enriques diagrams for @xmath0 giving rise to the same contraction ( an infinite family ) . here we will show how to complete contractions in order to describe all the different enriques diagrams for @xmath0 , thus solving completely the problem we are concerned with . given a contraction @xmath208 for @xmath0 , our aim is to describe all the enriques diagrams for @xmath0 associated with @xmath208 . consider the marked enriques diagram @xmath342 with @xmath343 for any @xmath213 , and the number @xmath344 . by [ l.ineq-weights ] , @xmath345 . let us describe a procedure to add vertices to @xmath346 in order to reach an enriques diagram for @xmath0 . write @xmath347 . for @xmath348 , choose a vertex @xmath329 in @xmath208 such that @xmath349 and then define inductively @xmath350 by taking @xmath351 and @xmath352 , where the new vertex @xmath353 is set as a successor of @xmath329 either * as a successor preceding @xmath356 , namely @xmath354 and @xmath357 are the only proximities relating @xmath353 , and then set @xmath358 in case @xmath359 ( otherwise @xmath360 can be chosen no matter @xmath361 or @xmath362 ) . notice that at step @xmath363 the operation of type a may always be performed , independently of the existence of a free successor of @xmath329 , which would offer the possibility to choose also an operation of type b. observe that @xmath364 . thus the procedure performs effectively the @xmath365 steps . any of such marked enriques diagram @xmath366 , obtained from @xmath208 by the above procedure , will be called an _ extension _ of the contraction @xmath208 . clearly any extension of @xmath208 is an enriques diagram for @xmath0 associated with @xmath208 . notice that any extension of @xmath208 all whose vertices have been added performing operation a at each step is an s - enriques diagram for @xmath0 ( in fact , the unique s - enriques diagram for @xmath0 associated with @xmath208 ) . conversely , if a marked enriques diagram @xmath107 contains , as a marked subdiagram , an extension @xmath367 of some contraction @xmath208 for @xmath0 and satisfies that any vertex of @xmath368 is proximate to no vertex of @xmath208 , then @xmath107 is an enriques diagram for @xmath0 . for the first assertion , we need to find a marked subdiagram @xmath367 of @xmath107 which is an extension of @xmath208 . take @xmath369 , and define @xmath370 , where @xmath371 and @xmath372 is the restriction of @xmath373 to @xmath114 . notice that @xmath114 is a connected subtree of @xmath90 since , if @xmath21 is proximate to some @xmath295 , then any vertex in @xmath100 infinitely near to @xmath26 is also proximate to @xmath26 . hence , @xmath114 together with the proximities inherited from the proximity of @xmath90 is an enriques subdiagram of @xmath90 . furthermore , @xmath367 is a marked subdiagram of @xmath107 . moreover , by [ w=1extrem ] , the cardinality of @xmath374 equals @xmath375 . denote the vertices of @xmath374 by @xmath376 so that @xmath329 is not infinitely near to @xmath377 if @xmath378 . write @xmath379 and for @xmath380 , define , recursively @xmath381 as the marked enriques diagram obtained from @xmath382 by deleting @xmath329 ( and keeping the restricted proximity and marking map ; the successors of @xmath329 become successors of the immediate predecessor of @xmath329 ) . notice that the @xmath381 are the marked enriques diagrams generated by the procedure detailed above to reach @xmath367 , proving that @xmath367 is an extension of @xmath208 , as wanted . for the converse , let @xmath370 be an extension of a contraction @xmath208 for @xmath0 . thus , by [ graph - ideal ] , @xmath383 as weighted graphs , @xmath384 for any @xmath248 and @xmath385 for any @xmath386 being adjacent to some vertex of @xmath14 . if @xmath387 contains @xmath367 as a marked subdiagram and there any vertex of @xmath388 is proximate to no vertex of @xmath389 , then @xmath390 as weighted graphs , and @xmath373 satisfies the marking map hypothesis of [ graph - ideal ] , 2 : @xmath391 for any @xmath248 and @xmath392 for any @xmath393 adjacent to some vertex of @xmath14 . hence , applying [ graph - ideal ] to @xmath107 we are done . we have already pointed out that sandwiched singularities are normal birational extensions of the regular ring @xmath4 . if @xmath394 is such an extension , there exists a complete ideal @xmath395 such that @xmath396_{n_q}$ ] , where @xmath397 is a height two maximal ideal in @xmath398 $ ] containing @xmath399 ( the maximal ideal of @xmath4 ) , and @xmath400 is a generic element of @xmath78 ( see @xcite ) . @xmath4 is said to be _ maximally regular _ in @xmath0 if there is no other regular ring @xmath401 such that @xmath402 write @xmath124 for the marked enriques diagram of the base points of @xmath78 . let @xmath367 and @xmath208 be the extension and the contraction for @xmath0 associated with @xmath124 . then , by virtue of [ final ] , @xmath124 can be thought as being constructed from @xmath367 by adding new vertices which are infinitely near to some dicritical vertex of @xmath367 and not proximate to any vertex of @xmath208 , or preceding the root of @xmath208 ( notice that in any case , the proximities of @xmath367 , and hence also the proximities of @xmath208 , are preserved ) . moreover , @xmath4 is maximally regular in @xmath0 if and only if the root of @xmath124 equals the root of @xmath367 , i.e. no vertices have been added to @xmath367 preceding the root . let @xmath0 be a sandwiched singularity whose resolution graph is shown in the top left corner of figure [ llistated2 ] . the contractions for @xmath0 are shown at the top of the figure , and below each one of them , a complete list of the associated extensions is drawn . any enriques diagram for @xmath0 contains one of these extensions as a marked subdiagram . a complete list of extensions for a sandwiched singularity whose resolution graph is drawn in the top left corner . at the top of the figure the contractions of @xmath0 are shown . white - filled dots represent dicritical vertices . ] f. enriques and o. chisini , _ lezioni sulla teoria geometrica delle equazioni e delle funzioni algebriche . _ , collana di matematica [ mathematics collection ] , vol . 5 , nicola zanichelli editore s.p.a . , bologna , 1985 , reprint of the 1915 , 1918 , 1924 and 1934 editions , in 2 volumes .

we examine a dynamic model for the disruption of information flow in hierarchical social networks by considering the vertex - pursuit game seepage played in directed acyclic graphs ( dags ) . in seepage , agents attempt to block the movement of an intruder who moves downward from the source node to a sink . the minimum number of such agents required to block the intruder is called the green number . we propose a generalized stochastic model for dags with given expected total degree sequence . seepage and the green number is analyzed in stochastic dags in both the cases of a regular and power law degree sequence . for each such sequence , we give asymptotic bounds ( and in certain instances , precise values ) for the green number . [ section ] [ theorem]lemma [ theorem]definition [ theorem]remark [ theorem]conjecture [ theorem]proposition [ theorem]algorithm [ theorem]corollary [ theorem]observation [ theorem]open problem

1308.3380

the on - line social network twitter is a well known example of a complex real - world network with over @xmath0 million users . the topology of twitter network is highly directed , with each user following another ( with no requirement of reciprocity ) . by focusing on a popular user as a source ( such as lady gaga or justin bieber , each of whom have over 11 million followers @xcite ) , we may view the followers of the user as a certain large - scale _ hierarchical social network_. in such networks , users are organized on ranked levels below the source , with links ( and as such , information ) flowing from the source downwards to sinks . we may view hierarchical social networks as _ directed acyclic graphs _ , or _ dags _ for short . hierarchical social networks appear in a wide range of contexts in real - world networks , ranging from terrorist cells to the social organization in companies ; see , for example @xcite . in hierarchical social networks , information flows downwards from the source to sinks . disrupting the flow of information may correspond to halting the spread of news or gossip in on - line social network , or intercepting a message sent in a terrorist network . how do we disrupt this flow of information while minimizing the resources used ? we consider a simple model in the form of a vertex - pursuit game called seepage introduced in @xcite . seepage is motivated by the 1973 eruption of the eldfell volcano in iceland . in order to protect the harbour , the inhabitants poured water on the lava in order to solidify it and thus , halt its progress . the game has two players , the _ sludge _ and a set of _ greens _ ( note that one player controls all the greens ) , a dag with one source ( corresponding to the top of the volcano ) and many sinks ( representing the lake ) . the players take turns , with the sludge going first by contaminating the top node ( source ) . then it is the greens turn , and they choose some non - protected , non - contaminated nodes to protect . on subsequent rounds the sludge moves a non - protected node that is adjacent ( that is , downhill ) to the node the sludge is currently occupying and contaminates it ; note that the sludge is located at a single node in each turn . the greens , on their turn , proceed as before ; that is , choose some non - protected , non - contaminated nodes to protect . once protected or contaminated , a node stays in that state to the end of the game . the sludge wins if some sink is contaminated ; otherwise the greens win , that is , if they erect a cutset of nodes which separates the contaminated nodes from the sinks . the name `` seepage '' is used because the rate of contamination is slow . the game is related to vertex - pursuit games such as cops and robbers ( for an introduction and further reading on such games , see @xcite ) , although the greens in our case need not move to neighbouring nodes . for an example , see the dag in figure [ seepageex1 ] . ( we omit orientations of directed edges in the figure , and assume all edges point from higher nodes to lower ones . ) [ h ] to obtain the results in this paper , a number of different winning strategies are employed by the two players . in some cases one of the two players can play arbitrarily ( at least up to some point ) , whereas in other cases the optimal strategy is simply a `` greedy '' one ( for example , when the greens protect neighbours as close as possible to the current position of the sludge ) . in some other cases , much more sophisticated strategies have to be applied . to date the only analysis of seepage was in @xcite , which presented results for dags . seepage may be extended to certain directed graphs with cycles , although we do not consider this variation here ( see also section [ directions ] ) . in @xcite , a characterization was given of directed trees where one green has a winning strategy , and bounds were given on the number of greens needed to win in truncated products of paths . see also chapter 9 of @xcite . seepage displays some interesting similarities to an approach used in mathematical counterterrorism , where cut sets in partially ordered sets ( which are just a special kind of dag ) are used to model the disruption of terrorist cells . as described in farley @xcite , the maximal elements of the poset are viewed as the leaders of the terrorist organization , who submit plans down via the edges to the nodes at the bottom ( the foot soldiers or minimal nodes ) . only one messenger needs to receive the message for the plan to be executed . farley considered finding minimum - order sets of elements in the poset , which when deleted , disconnect the minimal elements from the maximal one ( that is , find a _ minimum cut _ ) . we were struck by the similarities in the underlying approaches in @xcite and @xcite ; for example , in seepage the greens are trying to prevent the sludge from moving to the sinks by blocking nodes . the main difference is that seepage is `` dynamic '' ( that is , the greens can move , or choose new sets of nodes each time - step ) , while the min - cut - set approach is `` static '' ( that is , find a cutset in one time - step ) . seepage is perhaps a more realistic model of counterterrorism , as the agents do not necessarily act all at once but over time . however , in both approaches deterministic graphs are used . we note that a stochastic model was presented for so - called _ network interdiction _ in @xcite , where the task of the interdictor is to find a set of edges in a weighted network such that the removal of those edges would maximally increase the cost to an evader of traveling on a path through the network . a stochastic model for complex dags was given in @xcite . for more on models of on - line social networks and other complex networks , see @xcite . our goal in the present article is to analyze seepage and the green number when played on a random dag as a model of disrupting a given hierarchical social network . we focus on mathematical results , and give a precise formulation of our random dag model in section [ sec : definitions ] . our model includes as a parameter the total degree distribution of nodes in the dag . this has some similarities to the @xmath1 model of random graphs with expected degree sequences ( see @xcite ) or the pairing model ( see @xcite ) . we study two cases : regular dags ( where we would expect each level of the dag to have nodes with about the same out - degree ) , and power law dags ( where the degree distribution is heavy tailed , with many more low degree nodes but a few which have a high degree ) . rigorous results are presented for regular dags in theorem [ thm : main_d - reg ] , and for power law dags in theorem [ thm : main_power - law ] . an overview of the main results is given in section [ mainr ] . throughout , @xmath2 will represent a finite dag . for background on graph theory , the reader is directed to @xcite . the total degree of a vertex is the sum of its in- and out - degrees . additional background on seepage and other vertex - pursuit games may be found in @xcite . we denote the natural numbers ( including @xmath3 ) by @xmath4 , and the positive integers and real numbers by @xmath5 and @xmath6 , respectively . for an event @xmath7 on a probability space , we let @xmath8 denote the probability of @xmath7 . given a random variable @xmath9 , we let @xmath10 and @xmath11 be the expectation and the variance of @xmath9 , respectively . we now give a formal definition of our vertex - pursuit game . fix @xmath12 a node of @xmath2 . we will call @xmath13 the _ source_. for @xmath14 let @xmath15 where @xmath16 is the distance between @xmath17 and @xmath13 in @xmath18 in particular , @xmath19 . for a given @xmath20 and @xmath21 , let @xmath22 be the game played on graph @xmath2 with the source @xmath13 and the _ sinks _ the game proceeds over a sequence of discrete time - steps . exactly @xmath24 new nodes are protected at time - step @xmath25 . ( in particular , at most @xmath26 nodes are protected by the time @xmath25 . ) note that if @xmath27 is an integer , then exactly @xmath27 nodes are protected at each time - step , so this is a natural generalization of seepage . to avoid trivialities , we assume that @xmath28 . the _ sludge _ starts the game on the node @xmath29 . the second player , the _ greens _ , can protect @xmath30 nodes of @xmath31 . once nodes are protected they will stay protected to the end of the game . at time @xmath32 , the sludge makes the first move by sliding along a directed edge from @xmath33 to @xmath34 , which is an out - neighbour of @xmath33 . after that the greens have a chance to protect another @xmath35 nodes . since the graph is finite and acyclic , the sludge will be forced to stop moving , and so the game will eventually terminate . if he reaches any node of @xmath23 , then the sludge wins ; otherwise , the greens win . if @xmath36 ( the maximum out - degree of @xmath2 ) , then the game @xmath22 can be easily won by the greens by protecting of all neighbours of the source . therefore , the following graph parameter , _ the green number _ , is well defined : @xmath37 it is clear that for any @xmath38 we have @xmath39 . there are two parameters of the model : @xmath40 and an infinite sequence @xmath41 of non - negative integers . note that the @xmath42 s may be functions of @xmath43 . the first layer ( that is , the source ) consists of one node : @xmath44 . the next layers are recursively defined . for the inductive hypothesis , suppose that all layers up to and including the layer @xmath45 are created , and let us label all nodes of those layers . in particular , @xmath46 where @xmath47 we would like the nodes of @xmath48 to have a total degree with the following distribution @xmath49 . however , it can happen that some node @xmath50 has an in - degree @xmath51 already larger than @xmath42 , and so there is no hope for the total degree of @xmath42 . if this is not the case , then the requirement can be easily fulfilled . as a result , @xmath52 , the desired degree distribution , will serve as a ( deterministic ) lower bound for the actual degree distribution we obtain during the ( random ) process . let @xmath53 be a new set of nodes of cardinality @xmath43 . all directed edges that are created at this time - step will be from the layer @xmath48 to a random subset of @xmath53 that will form a new layer @xmath54 . each node @xmath50 generates @xmath55 random directed edges from @xmath56 to @xmath53 . therefore , we generate @xmath57 random edges at this time - step . the destination of each edge is chosen uniformly at random from @xmath53 . all edges are generated independently , and so we perform @xmath58 independent experiments . the set of nodes of @xmath53 that were chosen at least once forms a new layer @xmath54 . note that it can happen that two parallel edges are created during this process however , this is a rare situation for sparse random graphs we are going to investigate in this paper . hence , our results on the green number will also hold for a slightly modified process which excludes parallel edges . in this paper , we focus on two specific sequences : regular and power law . we will describe them both and state the main results in the next two subsections . we consider asymptotic properties of the model as @xmath59 . we say that an event in a probability space holds _ asymptotically almost surely _ ( _ a.a.s . _ ) if its probability tends to one as @xmath43 goes to infinity . we consider a constant sequence ; that is , for @xmath60 we set @xmath61 , where @xmath62 is a constant . in this case , we refer to the stochastic model as _ random @xmath63-regular dags_. since @xmath64 , observe that @xmath65 ( deterministically ) for any @xmath66 , since at most @xmath67 random edges are generated when @xmath23 is created . we will write @xmath68 for @xmath69 since the graph @xmath2 is understood to be a @xmath63-regular random graph , and @xmath70 . [ thm : main_d - reg ] let @xmath71 be any function that grows ( arbitrarily slowly ) as @xmath43 tends to infinity . for the random @xmath63-regular dags , we have the following . * a.a.s . @xmath72 . * if @xmath73 , then a.a.s.@xmath74 * if @xmath75 , then a.a.s.@xmath76 * if @xmath77 for some @xmath78 , then a.a.s.@xmath79 * let @xmath78 , @xmath80 . there exists a constant @xmath81 such that if @xmath82 , then a.a.s.@xmath83 the whole section [ sec : d - reg ] is devoted to prove this theorem . theorem [ thm : main_d - reg ] tells us that the green number is slightly bigger than @xmath84 if the sinks are located near the source , and then it is @xmath84 for a large interval of @xmath66 . later , it might decrease slightly since an increasing number of vertices have already in - degree @xmath85 or more , but only for large @xmath66 ( part ( v ) ) we can prove better upper bounds than @xmath86 one interpretation of this fact is that the resources needed to disrupt the flow of information is in a typical regular dag is ( almost ) independent of @xmath66 , and relatively low ( as a function of @xmath66 ) . we have three parameters in this model : @xmath87 , @xmath88 , and @xmath89 . for a given set of parameters , let @xmath90 and @xmath91 finally , for @xmath92 let @xmath93 in this case , we refer to the model as _ random power law dags_. we note that the sequence @xmath52 is decreasing ( in particular , the source has the largest expected degree ) . moreover , the number of coordinates that are at least @xmath45 is equal to @xmath94 and hence the sequence follows a power - law with exponent @xmath95 . from the same observation it follows that the maximum value is @xmath96 finally , the average of the first @xmath43 values is @xmath97 since @xmath98 . our main result on the green number @xmath99 in the case of power law sequences is the following . [ thm : main_power - law ] let @xmath100 if @xmath101 , and @xmath102 otherwise . let @xmath103 be the largest integer satisfying @xmath104 . let @xmath105 be the largest integer such that @xmath106 finally , let @xmath107 then , for @xmath108 we have that a.a.s . @xmath109 where @xmath110 , for @xmath111 , @xmath112 and for @xmath113 , @xmath114 in the power law case , theorem [ thm : main_power - law ] tells us that the green number is smaller for large @xmath66 . this reinforces the view that intercepting a message in a hierarchical social network following a power law is more difficult close to levels near the source . before analyzing the game on random @xmath63-regular dags , we need a few lemmas . we will be using the following version of a well - known chernoff bound . [ lem : chernoff ] let @xmath9 be a random variable that can be expressed as a sum @xmath115 of independent random indicator variables where @xmath116 is a bernoulli random variable with success probability @xmath117 with ( possibly ) different @xmath118 . then the following holds for @xmath119 : @xmath120 in particular , if @xmath121 , then @xmath122 we will start by proving the threshold for appearance of vertices of in - degree @xmath45 . [ lem : layersize ] let @xmath71 be any function that grows ( arbitrarily slowly ) as @xmath43 tends to infinity . then a.a.s . the following properties hold . * @xmath123 for any @xmath124 . * for all @xmath125 , let @xmath126 for every @xmath127 , we have that @xmath128 if @xmath129 , and @xmath130 for some @xmath131 . in particular , the threshold for the appearance of vertices of in - degree @xmath45 is @xmath132 . * @xmath133 for @xmath134 . for ( i ) note that the probability that a given vertex @xmath135 has in - degree @xmath136 at level @xmath137 is at most @xmath138 thus , the expected number of vertices of in - degree @xmath45 at level @xmath66 is @xmath139 and , in particular , the expected number of vertices of in - degree @xmath85 or more at @xmath23 is @xmath140 . set @xmath141 . by markov s inequality , with probability at least @xmath142 we derive that @xmath143 since @xmath144 we obtain that a.a.s . holds for all values of @xmath145 . since we aim for a statement that holds a.a.s . , we can assume for @xmath145 that @xmath146 , so ( i ) holds for @xmath147 . suppose that ( i ) holds for all @xmath148 ; that is , @xmath149 . by and the inductive hypothesis ( used recursively ) , we obtain that @xmath150 where @xmath151 note that @xmath152 we now prove ( ii ) and ( iii ) . by part ( i ) , the number of random edges @xmath153 emanating from @xmath154 is @xmath155 when layer @xmath66 is created , these edges are joined to random vertices in the set @xmath156 of cardinality @xmath43 . for any fixed @xmath125 and any fixed layer @xmath124 , we define the indicator variable @xmath157 to be @xmath158 if @xmath159 has in - degree @xmath45 and @xmath3 otherwise , for @xmath160 . let @xmath161 . as observed in part ( i ) of this proof , @xmath162 and thus , @xmath163 . by markov s inequality , a.a.s . for @xmath164 , no vertices of in - degree @xmath45 are present . by considering the case @xmath165 , this shows that for @xmath166 , all vertices have in - degree @xmath158 a.a.s . , and thus , for such @xmath66 we have @xmath167 a.a.s . hence , ( iii ) holds . we find that @xmath168 thus , @xmath169 ^ 2 ( 1+o(1 ) ) + { \mathbb e}(x ) \\ & = & ( { \mathbb e}(x))^2(1+o(1))+{\mathbb e}(x).\end{aligned}\ ] ] hence , for @xmath170 , it follows from chebyshev s inequality that @xmath171 and hence , a.a.s . there are vertices of in - degree @xmath45 , proving part ( ii ) of the lemma . the lemma is enough to prove the first two parts of the main theorem . by lemma [ lem : layersize ] ( iii ) for @xmath172 the game is played on a tree . part ( i ) is trivial , since the greens have to protect all vertices in @xmath173 , or they lose . to derive the upper bound of ( ii ) , note that for @xmath174 we have that @xmath175 ( @xmath176 for @xmath177 ) . the greens can play arbitrarily during the first @xmath178 steps , and then block the sludge on level @xmath66 . if @xmath179 , then we have that @xmath84 is an upper bound of @xmath68 , and the upper bound of ( iii ) holds . to derive the lower bound of ( ii ) , note that if @xmath180 , then exactly @xmath84 new vertices are protected at each time - step . without loss of generality , we may assume that the greens always protect vertices adjacent to the sludge ( since the game is played on the tree , there is no advantage to play differently ) . no matter how the greens play , there is always at least one vertex not protected and the sludge can reach @xmath23 . for a given vertex @xmath181 and integer @xmath66 , let us denote by @xmath182 the subset of @xmath183 consisting of vertices at distance @xmath66 from @xmath13 ( that is , those that are in the @xmath66-th level of the subgraph whose root is @xmath13 ) . let @xmath184 be the subgraph of all vertices of depth @xmath66 pending at @xmath13 . call a vertex @xmath185 _ bad _ if @xmath186 and @xmath17 has in - degree at least @xmath85 ( recall , that @xmath53 is a set of @xmath43 vertices used in the process of generating a random graph ) . let @xmath187 be the total number of bad vertices in @xmath188 . in the next lemma , we estimate @xmath187 . [ lem : badvertices ] a.a.s . the following holds for some large enough constant @xmath189 for any @xmath181 , where @xmath190 , and any @xmath66 such that @xmath191 , @xmath192 fix @xmath181 and let @xmath66 be the maximum integer satisfying @xmath191 . since there are @xmath193 possible vertices to consider , it is enough to show that the bound holds with probability @xmath194 . for @xmath185 , let @xmath195 ( @xmath196 ) be the event that @xmath197 and @xmath17 is bad . in order for @xmath17 to be in @xmath198 , @xmath17 must receive at least one edge from a vertex in @xmath199 , and in order to be bad it must have at least one more edge from either @xmath199 or from another vertex at layer @xmath200 . thus , @xmath201 since there are @xmath202 edges emanating from @xmath199 , and there are @xmath203 edges emanating from @xmath200 . letting @xmath204 being the corresponding indicator variable , we have that @xmath205 note that @xmath206 , since @xmath207 ( for a fixed total number of edges , the probability for @xmath17 to be bad is smaller if another vertex @xmath208 is bad ) and , by the law of total probability , at least one of the two conditional probabilities has to be at most @xmath209 . thus , @xmath210 is bounded from above by @xmath211 , where the @xmath212 are independent indicator random variables with @xmath213 for some sufficiently large @xmath214 . the total number of bad vertices in the subgraph of depth @xmath66 pending at @xmath13 is @xmath215 since @xmath216 @xmath217 and by the chernoff bound given by ( [ eq : ch1 ] ) , @xmath218 with probability @xmath219 for some @xmath220 large enough . we need one more lemma . for a given vertex @xmath181 and integer @xmath66 , a vertex @xmath185 is called _ very bad _ if it has at least two incoming edges from vertices in @xmath199 . in particular , every very bad vertex is bad . let @xmath221 be the number of very bad vertices in @xmath188 . for a given @xmath222 , and any @xmath223 such that @xmath224 , we will consider the subgraph @xmath225 consisting of all vertices to which there is a directed path from some vertex in @xmath226 . for any @xmath227 , let @xmath228 be a subset of @xmath229 that is in @xmath225 . [ lem : notworoots ] let @xmath223 for some @xmath222 be such that @xmath230 . then a.a.s . for any @xmath231 , where @xmath232 , and any integer @xmath66 with @xmath191 , we have that @xmath233 . fix any @xmath231 for some @xmath232 . as in lemma [ lem : badvertices ] , by letting @xmath234 be the event that @xmath185 is very bad , we have @xmath235 letting @xmath236 be the corresponding indicator variable , we have that @xmath237 analogously as in the previous proof , define independent indicator random variables @xmath238 with @xmath239 . we have @xmath240 and so @xmath241 since @xmath191 . as @xmath224 , we have that @xmath242 and so the expected number of very bad vertices found in @xmath228 is @xmath243 . finally , the expected number of very bad vertices in any sublayer @xmath228 ( @xmath244 ) is @xmath245 , and the result holds by markov s inequality . we now come back to the proof of the main theorem for random regular dags . note that we already proved an upper bound of ( iii ) ( see the proof of parts ( i ) and ( ii ) ) . since @xmath68 is non - increasing as a function of @xmath66 , an upper bound of ( iv ) also holds . we will prove a lower bound of ( iv ) first . the lower bound of ( iii ) will follow easily from there . let @xmath78 and suppose that we play the game with parameter @xmath246 . if @xmath247 , then for every @xmath14 , we have that @xmath248 and @xmath249 otherwise . ( for @xmath250 we find that @xmath251 for any @xmath25 . ) suppose that the greens play _ greedily _ ( that is , they always protect vertices adjacent to the sludge ) and the graph is locally a tree . note that during the time between @xmath252 and @xmath253 , they can direct the sludge leaving him exactly one vertex to choose from at each time - step . however , at time - step @xmath254 , the sludge has 2 vertices to choose from . the sludge has to use this opportunity wisely , since arriving at a bad vertex ( see definition above ) when the greens can protect @xmath84 vertices would result in him losing the game . our goal is to show that the sludge can avoid bad vertices and , as a result , he has a strategy to reach the sink @xmath23 . since we aim for a statement that holds a.a.s . we can assume that all properties mentioned in lemmas [ lem : layersize ] , [ lem : badvertices ] , and [ lem : notworoots ] hold . before we describe a winning strategy for the sludge , let us discuss the following useful observation . while it is evident that the greens should use a greedy strategy to play on the tree , it is less evident in our situation . perhaps instead of playing greedily , the greens should protect a vertex far away from the sludge , provided that there are at least two paths from the sludge to this vertex . however , this implies that the vertex is very bad and we know that very bad vertices are rare . it follows from lemma [ lem : notworoots ] that there is no very bad vertex within distance @xmath66 , provided that the sludge is at a vertex in @xmath229 , @xmath255 and @xmath256 . ( for early steps we know that the graph is locally a tree so there are no bad vertices at all . ) therefore , without loss of generality , we can assume that at any time - step @xmath25 of the game , the greens protect vertices greedily or protect vertices at distance at least @xmath66 where @xmath66 is the smallest value such that @xmath257 . we call the latter protected vertices _ dangerous_. the sludge has to make sure that there are no nearby bad nor dangerous vertices . let @xmath258 where the constant @xmath214 will be determined soon and is sufficiently large such that the sludge is guaranteed to escape from all bad or dangerous vertices which are close to him . let @xmath259 during the first @xmath260 time - steps , the sludge chooses any arbitrary branch . since he is given this opportunity at least @xmath261 times and each time he cuts the number of possible destinations by a factor of @xmath262 , the number of possible vertices the sludge can reach at time @xmath260 is @xmath263 . from that point on , it follows from lemma [ lem : notworoots ] that there are no nearby very bad vertices . at time @xmath264 , by lemma [ lem : layersize ] , there are no bad vertices at distance @xmath265 from the sludge , and hence , no dangerous vertices within this distance . it follows from lemma [ lem : badvertices ] that there are @xmath266 bad vertices at distance @xmath267 there are @xmath266 dangerous vertices within this distance ( since the total number of protected vertices during the whole game is of this order ) . thus , there are @xmath266 bad or dangerous vertices at a distance between @xmath268 and @xmath269 from the sludge . to derive a lower bound on the length of the game , we provide a strategy for the sludge that allows him to play for at least a certain number of steps , independently of the greens behaviour . in particular , his goal is to avoid these bad or dangerous vertices : as long as the sludge is occupying a vertex that is not bad , there is at least one vertex on the next layer available to choose from . more precisely , it follows from lemma [ lem : notworoots ] , that from time @xmath260 onwards , locally there are no very bad vertices . let us call a _ round _ a sequence of @xmath270 time - steps . since all bad vertices are in distinct branches , in every @xmath271-th time - step the sludge can half the number of bad vertices . therefore , after one round the sludge can escape from all @xmath272 bad or dangerous vertices that are under consideration in a given round , provided that @xmath214 is large enough constant . ( recall the constant @xmath273 is defined in lemma [ lem : badvertices ] . ) using this strategy , at time @xmath274 there are no bad or dangerous vertices at distance @xmath275 to see this , note that since the sludge escaped from all bad or dangerous vertices , which at time @xmath276 were at distance @xmath277 , and he has advanced @xmath270 steps by now . using lemma [ lem : badvertices ] again , we find that there are @xmath266 bad or dangerous vertices at distance @xmath278 arguing as before , we find that it takes another @xmath270 steps to escape from them . in general , at time @xmath279 , there are @xmath266 bad or dangerous vertices at a distance between @xmath280 and @xmath281 thus , as long as @xmath282 , the strategy of escaping from bad or dangerous vertices before actually arriving at that level is feasible . moreover , we can finish this round , and so the sludge is guaranteed to use this strategy until time @xmath283 , where @xmath284 is the smallest value such that @xmath285 . solving this for @xmath284 we obtain that @xmath286 and so @xmath287 hence , @xmath288 finally , note that if @xmath284 is the smallest value such that @xmath289 , we get that @xmath290 and so @xmath291 . hence , another @xmath292 steps can be played , and the constant of the second order term can be improved from @xmath293 to @xmath294 , yielding part ( iv ) . part ( iii ) follows by taking @xmath271 to be a function of @xmath43 slowly growing to infinity . our next goal is to show that when @xmath295 , the value of @xmath68 is slightly smaller than @xmath84 , provided that @xmath296 is a sufficiently large constant . however , before we do it , we need one more observation . it follows from lemma [ lem : layersize](i ) that a.a.s . @xmath297 for @xmath298 ( @xmath299 is any function tending to infinity with @xmath43 , as usual ) . however , this is not the case when @xmath300 . at this point of the process , a positive fraction of vertices of @xmath229 are bad . this , of course , affects the number of edges from @xmath229 to @xmath301 . in fact , the number of edges between two consecutive layers converges to @xmath302 as shown in the next lemma . [ lem : edges_conv ] let @xmath303 be the constant satisfying @xmath304 for every @xmath305 , there exists a constant @xmath306 such that a.a.s . for every @xmath307 , @xmath308 and the number of edges between @xmath229 and @xmath301 is at least @xmath309 and at most @xmath310 . suppose that the layer @xmath229 has in total @xmath311 random incoming edges , for some @xmath312 $ ] . then the probability that a vertex @xmath135 ( recall that @xmath53 is the set of cardinality @xmath43 used to create layer @xmath229 ) has in - degree @xmath313 ( that is , absorbs @xmath45 incoming edges , or attracts no edge if @xmath314 ) is @xmath315 note that each vertex of in - degree @xmath316 generates @xmath317 edges to the next layer . further , vertices of in - degrees @xmath45 or more do not have any offspring , and vertices of @xmath53 of in - degree zero are not in @xmath229 . therefore , the expected number of outgoing edges produced by all vertices in layer @xmath229 is @xmath318 the events considered here are almost independent ( one can compute higher moments and see that the @xmath45th moment is asymptotically equal to the @xmath45th power of the first moment ) , so for any @xmath319 it follows from chernoff bounds that with probability @xmath320 the number of vertices of degree @xmath45 is @xmath321 . thus , with the same probability , strong concentration also follows for the number of edges . if the number of incoming edges equals @xmath302 , then the expected number of outgoing edges equals @xmath322 if less than @xmath302 edges are incoming , then more will be going out , and vice versa . the process converges and so there exists a constant @xmath306 such that a.a.s . the number of edges between two consecutive layers @xmath229 and @xmath301 is between @xmath309 and @xmath310 for any @xmath25 such that @xmath307 . finally , let us recall that the layer @xmath229 consists of vertices of @xmath53 with in - degree at least one . the number of in - degree @xmath3 vertices is concentrated around its expectation , and thus we have that a.a.s.@xmath323 the lemma is proved . the value of @xmath303 ( and so @xmath324 as well ) can be numerically approximated . it is straightforward to see that @xmath303 tends to @xmath325 ( hence , @xmath324 tends to @xmath158 ) when @xmath326 . we present below a few approximate values . .approximate values of @xmath303 and @xmath324 . [ cols="^,^,^,^,^,^",options="header " , ] finally , we are ready to finish the last part of theorem [ thm : main_d - reg ] . _ proof of theorem [ thm : main_d - reg](v)_. we assume that the game is played with parameter @xmath246 for some @xmath327 . for every @xmath14 , we have that @xmath248 , and @xmath328 , otherwise . to derive an upper bound of @xmath68 that holds a.a.s . , we need to prove that a.a.s . there exists no winning strategy for the sludge . we will use a combinatorial game - type argument . the greens will play _ greedily _ ( that is , they will always protect nodes adjacent to the sludge ) . suppose that the sludge occupies node @xmath329 for some @xmath14 ( at time @xmath330 he moves from @xmath13 to some node in @xmath229 ) and he has a strategy to win from this node , provided that no node in the next layers is protected by the greens . we will call such a node _ sludge - win_. note that during the time period between @xmath252 and @xmath253 , the greens can protect @xmath84 nodes at a time , so they can direct the sludge leaving him exactly one node to choose from at each time - step . therefore , if there is a node of in - degree at least 2 in any of these layers , the greens can force the sludge to go there and finish the game in the next time - step . this implies that all nodes within distance @xmath331 from @xmath13 ( including @xmath13 itself ) must have in - degree 1 and so the graph is locally a tree . however , at time - step @xmath254 , the greens can protect @xmath332 nodes , one less than in earlier steps . if the in - degree of a node reached at this layer is at least 3 , then the greens can protect all out - neighbours and win . further , if the in - degree is 2 and there is at least one out - neighbour that is not sludge - win , the greens can force the sludge to go there and win by definition of not being sludge - win . finally , if the in - degree is 1 , the sludge will be given 2 nodes to choose from . however , if there are at least two out - neighbours that are not sludge - win , the greens can `` present '' them to the sludge and regardless of the choice made by the sludge , the greens win . we summarize now the implications of the fact that @xmath329 is sludge - win . first of all , all nodes within distance @xmath331 are of in - degree 1 . nodes at the layer @xmath333 below @xmath13 have in - degree at most 2 . if @xmath334 has in - degree 2 , then all of the @xmath84 out - neighbours are sludge - win . if @xmath334 has in - degree 1 , then all out - neighbours except perhaps one node are sludge - win . using this observation , we characterize a necessary condition for a node @xmath335 to be sludge - win . for a given @xmath335 that can be reached at time 1 , we define a _ sludge - cut _ to be the following cut : examine each node of @xmath336 , and proceed inductively for @xmath60 . if @xmath337 has out - degree @xmath338 , then we cut away any out - neighbour and all nodes that are not reachable from @xmath13 ( after the out - neighbour is removed ) . the node that is cut away is called an _ avoided node_. after the whole layer @xmath336 is examined , we skip @xmath339 layers and move to the layer @xmath333 . we continue until we reach the sink , the layer @xmath340 for some @xmath341 ( we stop at @xmath23 without cutting any further ) . the main observation is that if the sludge can win the game , then the following claim holds . _ claim_. there exists a node @xmath335 and a sludge - cut such that the graph left after cutting is a @xmath342-regular graph , where each node at layer @xmath336 , @xmath343 has out - degree @xmath84 , and all other nodes have out - degree @xmath338 . in particular , for any @xmath343 the graph induced by the set @xmath344 is a tree . it remains to show that a.a.s . the claim does not hold . ( since there are at most @xmath63 nodes in @xmath173 it is enough to show that a.a.s . the claim does not hold for a given node in @xmath173 . ) fix @xmath335 . the number of avoided nodes at layer @xmath345 is at most the number of nodes in @xmath336 ( after cutting earlier layers ) , which is at most @xmath346 in particular , @xmath347 , the number of nodes in the sink after cutting , is at most @xmath348 . it can be shown that a.a.s . @xmath349 for some @xmath350 . fix @xmath351 . we need to show that for this given @xmath347 the claim does not hold with probability @xmath194 . since each node in @xmath352 has in - degree at most 2 , the number of nodes in @xmath353 is at most @xmath354 ( as before , after cutting ) . since the graph between layer @xmath355 and @xmath353 is a tree , the number of nodes in @xmath352 is at most @xmath356 , which is an upper bound for the number of avoided nodes at the next layer @xmath357 . applying this observation recursively we obtain that the total number of avoided nodes up to layer @xmath358 is at most @xmath359 to count the total number of sludge - cuts of a given graph , observe that each avoided node corresponds to one out of @xmath338 choices . hence , the total number of sludge - cuts is at most @xmath360 we now estimate the probability that the claim holds for a given @xmath335 and a sludge - cut . to obtain an upper bound , we estimate the probability that all nodes in the layer @xmath353 are of in - degree 1 . conditioning on the fact that we have @xmath347 nodes in the last layer , we find that the number of nodes in @xmath353 is at least @xmath361 let @xmath341 be large enough such that we are guaranteed by lemma [ lem : edges_conv ] that the number of edges between the two consecutive layers is at least @xmath362 . hence , the probability that a node in @xmath353 has in - degree 1 is at most @xmath363 where @xmath305 can be arbitrarily small by taking @xmath341 large enough . let @xmath364 be the probability in ( [ woo ] ) . we derive that @xmath365 , where @xmath366 is a large enough constant . conditioning under @xmath367 having in - degree @xmath158 , it is harder for @xmath368 to have in - degree @xmath158 than without this condition , as more edges remain to be distributed . thus , the probability that all nodes in @xmath353 have the desired in - degree is at most @xmath369 thus , by taking a union bound over all possible sludge - cuts ( the upper bound for the number of them is given by ) , the probability that the claim holds is at most @xmath370 which can be made @xmath371 by taking @xmath372 small enough , provided that @xmath271 is large enough so that @xmath373 by considering the extreme case for the probability of having in - degree one when @xmath374 we obtain that @xmath375 for @xmath62 ( see table [ tab : small_values ] ) . it is straightforward to see that @xmath80 will work for any @xmath62 , and @xmath376 for @xmath377 . @xmath378 let us recall that we have three parameters in this model : @xmath87 , @xmath88 , and @xmath89 . for a given set of parameters , we defined @xmath379 finally , for @xmath92 we have that @xmath93 before we analyze the game for this model , let us focus on investigating some properties of the random graph we play on . we already mentioned that the sequence @xmath380 is decreasing but it is not obvious which weights we obtain for a given level @xmath23 . we start by providing a lower bound for the weight of vertices in each layer @xmath66 ( which will imply an upper bound for the previous layer @xmath178 ) . since the weight @xmath42 is a function of the index @xmath284 , it is enough to focus on the latter . for @xmath381 , let @xmath382 be the smallest index among the vertices of layer @xmath23 . using the notation introduced in section [ sec : definitions ] , @xmath383 . the maximum weight at @xmath23 is @xmath384 , the minimum one is @xmath385 . the first lemma investigates the behaviour of @xmath382 . [ lem : max_deg ] let @xmath100 if @xmath101 , and @xmath386 otherwise . let @xmath107 the following holds a.a.s . * @xmath387 , @xmath388 , and @xmath389 . + in particular , @xmath390 , @xmath391 , and @xmath392 * for @xmath393 we have that @xmath394 in particular , @xmath395 * if @xmath396 , then , @xmath397 in particular , @xmath398 * let @xmath103 be the largest integer satisfying @xmath104 . let @xmath105 be the largest integer such that @xmath399 then for @xmath400 we have that @xmath401 in particular , for @xmath400 we have that @xmath402 * for any @xmath403 , the number of edges between @xmath154 and @xmath404 is @xmath405. clearly , we have @xmath406 , @xmath407 ; ( i ) holds deterministically for @xmath408 . the number of vertices on levels @xmath3 and @xmath158 is at most @xmath409 but can be slightly smaller if there are some parallel edges ( which happens a.a.s . if @xmath410 ) . we derive a deterministic upper bound for @xmath411 of @xmath412 but in fact , using the first moment method , we can show that a.a.s . @xmath413 . indeed , the probability that a given vertex from @xmath53 has in - degree at least 2 is @xmath414 so we expect @xmath415 vertices of in - degree at least 2 . with probability @xmath416 we have @xmath417 of such vertices and so @xmath418 . the statement for @xmath419 holds , and hence , part ( i ) follows . now , let us generalize this observation . let @xmath420 and suppose that @xmath421 is already estimated . note that @xmath422 so it remains to estimate the size of @xmath154 . we obtain that @xmath423 note that @xmath424 is an upper bound for the number of edges between layer @xmath425 and @xmath154 ( and so an upper bound for @xmath426 ) , and we derive the equality @xmath427 if all vertices in @xmath154 and @xmath425 have in - degree 1 . arguing as before , we deduce that with probability @xmath416 the number of edges going to vertices in @xmath425 ( in @xmath154 ) that are of degree at least 2 is @xmath428 ( @xmath429 , respectively ) . each edge of this type directed to a vertex in @xmath425 affects its out - degree , and so decreases the number of vertices in @xmath154 by at most one . similarly , one edge going to a vertex in @xmath154 of in - degree at least 2 decreases by at most one the number of vertices of in - degree 1 . thus , with probability @xmath416 , by considering vertices of in - degree 1 only , we obtain that @xmath430 ( the last equality follows from the fact that @xmath431 , provided that @xmath432 . in fact , we consider values of @xmath66 at most @xmath433 for which it will be shown that @xmath434 and so @xmath435 . ) this , together with the fact that @xmath422 , implies that @xmath436 if @xmath437 , then @xmath438 note that @xmath439 and @xmath440 . therefore this recursive formula is to be applied @xmath441 times only before the condition @xmath442 fails ( it may , of course , happen that it fails for @xmath443 so we do not apply it at all ) . we have that a.a.s . the statement holds for any value of @xmath66 such that @xmath444 ; that is , @xmath445 moreover , the error term can be estimated much better ; it is , in fact , @xmath446 for some @xmath305 . let us note one more time that it may happen that @xmath447 and so the condition fails for @xmath443 but then ( ii ) trivially holds . thus , part ( ii ) is finished . for part ( iii ) , suppose that @xmath396 . since our goal is to show that the statement holds a.a.s . , we may assume that @xmath448 for some @xmath305 . from the assumption it follows that @xmath449 and @xmath450 are of the same order . by the relations between @xmath450 and @xmath451 , we have @xmath452 . thus , @xmath453 , and hence , @xmath454 . it follows from ( [ ar : equations ] ) that a.a.s . @xmath455 so ( iii ) holds . for part ( iv ) , let @xmath103 be the largest integer satisfying @xmath104 . based on earlier parts , we may assume that @xmath456 . note that @xmath457 for some @xmath305 , and so @xmath458 for any @xmath459 , since @xmath382 is monotonic as a function of @xmath66 . fix @xmath460 . this time we derive from ( [ ar : equations ] ) that with probability @xmath416@xmath461 provided that @xmath462 ( note that we have @xmath463 , and thus we obtain the first condition , coming from the term @xmath429 . ) the first condition is equivalent to @xmath464 , and so both conditions combined together are equivalent to @xmath465 if this condition is satisfied , then we obtain that @xmath466 , or equivalently @xmath467 . by using the condition on @xmath468 , we obtain the following slightly stronger condition ( where we ignore the factor of @xmath85 ) for @xmath382 : @xmath469 now , suppose that ( [ eq : cond ] ) is satisfied , and we rewrite the relation between @xmath382 and @xmath421 using the fact that @xmath470 : @xmath471 applying this argument recursively , we obtain that @xmath472 finally , since we will soon show that @xmath473 we derive by the previous cases that @xmath474 indeed , since @xmath475 the condition ( [ eq : cond ] ) fails for @xmath476 ( by taking @xmath214 large enough ) , and item ( iv ) follows . finally , the proof of part ( v ) follows now by inspecting closely the parts ( ii ) , ( iii ) , and ( iv ) . in each case , @xmath477 is the size of @xmath23 and it follows from earlier parts that @xmath478 . this is clearly a lower bound for the number of edges we try to estimate . on the other hand , @xmath479 serves as an upper bound . hence , it remains to show that @xmath480 . if @xmath481 , then @xmath482 . by looking at part ( ii ) , we see that the leading term of both @xmath483 and @xmath484 is @xmath485 , and the result follows for this case ( alternatively , in this case we can also observe @xmath486 , @xmath487 , and by the trivial bound on the degree , @xmath488 holds ) . next , if @xmath489 , then @xmath490 , and by the calculations of part ( iii ) , in both @xmath484 and in @xmath479 the leading term is of order @xmath491 , and the result follows also for this case . finally , if @xmath492 , then @xmath493 , @xmath494 , and as observed in part ( iv ) , @xmath495 , and thus @xmath496 , and part ( v ) follows . we are now ready to come back to investigating the green number . we provide some obvious bounds for the green number and after that we sketch the idea that could be used to estimate it precisely . however , we do not perform these calculations rigorously , since the approach is rather delicate . [ lem : green_power_law ] let @xmath497 and @xmath103 be defined as in lemma [ lem : max_deg ] . that is , let @xmath100 if @xmath101 , and @xmath102 otherwise . let @xmath103 be the largest integer satisfying @xmath498 . moreover , let @xmath105 be the largest integer such that @xmath106 then , for @xmath499 we have that a.a.s . @xmath500 note that the definition of @xmath433 in lemma [ lem : green_power_law ] is slightly modified compared to the one from lemma [ lem : max_deg ] . however there is only a @xmath441 difference between these values ; in this case it is smaller . fix @xmath108 and suppose that the game is played with the sink @xmath23 . since the maximum total degree ( and thus , the maximum out - degree ) of vertices in @xmath154 is at most @xmath501 , the greens can easily win when the game is played with parameter @xmath502 . they can play arbitrarily at the beginning of the game when the sludge is moving towards the sink . once he reaches a vertex @xmath503 , the greens can block all out - neighbours and the game ends . we obtain that @xmath504 . in order to derive a lower bound , we will need the following property that follows directly from the proof of lemma [ lem : max_deg ] and holds a.a.s . let us note that for @xmath400 we have that @xmath505 for @xmath506 we have @xmath507 so in fact ( [ eq : cond_ell ] ) holds for any @xmath508 . now let us play the game with parameter @xmath509 for some @xmath305 . we will show that a.a.s . the sludge can win the game , independently of the strategy of the greens . this will prove that @xmath510 a.a.s . and the result will hold after taking @xmath511 . if @xmath147 , then for any @xmath512 , a.a.s . @xmath513 , and thus , the greens clearly can not win the game by protecting @xmath514 vertices . hence , we may assume that @xmath515 . suppose first that @xmath516 and @xmath517 . since in this case @xmath518 , by the formulas for @xmath382 given by lemma [ lem : max_deg ] , for any @xmath517 , @xmath519 . moreover , by part ( v ) of lemma [ lem : max_deg ] , the number of edges between @xmath154 and @xmath23 is at most @xmath520 , with @xmath521 . hence , for any vertex @xmath522 , @xmath523 where @xmath524 denoting by @xmath525 the number of out - neighbours of @xmath13 with in - degree @xmath85 or more , we have that @xmath526 . since @xmath527 is a fixed upper bound on the number of edges between two consecutive layers , for any two vertices @xmath528 , @xmath529 . consider now another stochastic process in which each vertex @xmath522 , has exactly @xmath530 out - neighbours , and for each out - neighbour @xmath531 of @xmath13 , independently of all other vertices , @xmath532 . denote by @xmath533 the number of out - neighbours of @xmath13 with in - degree @xmath85 or more in this new stochastic process . clearly , @xmath534 for any @xmath13 . furthermore , by the previous observation of negative correlation between vertices of in - degree @xmath85 or more ( in the original process ) , @xmath535 for any @xmath536 . set @xmath537 to be a sufficiently small constant . then , by lemma [ lem : chernoff ] , for some @xmath538 we have that @xmath539 by taking a union bound over all @xmath193 vertices of the first @xmath540 layers , a.a.s . all vertices have at least a @xmath541 fraction of out - neighbours with in - degree @xmath158 . now , in order to show a lower bound on the green number , we can assume that all vertices with in - degree @xmath85 or more are already protected by the greens in the very beginning ( they are cut away from top to bottom together with the subgraphs pending at them ) , and thus , the sludge is playing on the remaining graph that is a tree . we showed that a.a.s . the minimum degree in the remaining tree is at least @xmath542 observe that in a tree the best strategy for the greens is always to protect neighbours of the vertex currently occupied by the sludge . indeed , if they protect a vertex at distance @xmath85 or more from the vertex occupied by the sludge , they can consider the path between the vertex occupied by the sludge and the vertex originally protected , and instead protect the unique out - neighbour of the vertex occupied by the sludge . clearly , this is at least as good move as the original one . since the greens have only @xmath543 at their disposal , in each round at least @xmath544 neighbours remain unprotected , and the sludge can go to any of these , and finally reach the sink . suppose now that @xmath545 or @xmath546 . for any @xmath547 we have @xmath548 moreover , note that the formula is true if @xmath549 , but @xmath550 , and also in the case @xmath545 we have @xmath551 , and @xmath552 for some @xmath553 . thus , combining these statements , for any @xmath554 , or any @xmath555 in the case @xmath545 , @xmath556 . since @xmath557 is a monotonically decreasing sequence , any vertex up to ( and including ) layer @xmath425 has weight at least @xmath558 we will provide a strategy for the sludge and show that it guarantees him win a.a.s . , provided that the game is played with parameter @xmath559 , as before . the strategy is straightforward ; in particular , he always goes to any non - protected vertex @xmath17 with the property that no out - neighbour of @xmath17 is protected . note that the total number of vertices protected at the end of the game is @xmath560 . moreover , each protected vertex can only eliminate this vertex or its parents . the number of parents of a given vertex @xmath17 , the in - degree of @xmath17 , can be large but these parents are `` scattered '' across the whole layer , as we will show in the following claim . _ claim . _ the following holds a.a.s . the number of paths from any vertex @xmath13 ( a vertex possibly occupied by the sludge ) and vertex @xmath17 two layers below ( a vertex possibly protected by the greens ) is bounded by some universal constant @xmath561 . _ proof of the claim . _ fix @xmath562 to be an arbitrarily small constant . suppose first that @xmath66 is such that @xmath563 . then , by monotonicity of @xmath384 , the number of directed paths of length two starting at @xmath564 is at most @xmath565 . thus , the probability that for a given vertex @xmath566 there are @xmath561 paths of length two starting from @xmath13 , is at most @xmath567 . by taking a union bound over all @xmath43 vertices @xmath17 from @xmath568 and all starting vertices @xmath13 ( there are at most @xmath43 of them ) , we see that @xmath569 for a sufficiently large constant @xmath561 . the claim then holds for this range of values of @xmath66 , by taking a union bound over @xmath570 possible values of @xmath66 . on the other hand , if @xmath66 is such that @xmath571 , then by the formulas for @xmath384 given in lemma [ lem : max_deg ] we see that @xmath572 and so @xmath573 as well . ( indeed , the exponent of @xmath43 in the formula for @xmath574 in lemma [ lem : max_deg](iv ) can be made arbitrarily small by taking a sufficiently large constant @xmath575 . ) since @xmath573 , it follows from lemma [ lem : max_deg ] that @xmath576 for some @xmath577 , and as shown in part ( v ) of lemma [ lem : max_deg ] , the number of edges between @xmath154 and @xmath23 is at most @xmath578 thus , the number of paths of length two starting at @xmath564 is at most @xmath527 . hence , as before , the probability that for a given vertex @xmath566 there are @xmath561 paths of length two starting from @xmath13 , is at most @xmath579 , and as before , by taking a union bound over all @xmath580 pairs of vertices @xmath581 and over all @xmath540 , for @xmath561 sufficiently large , @xmath582 , and the claim follows . hence , by the claim , we obtain that the number of eliminated vertices is still @xmath583 . finally , since the degree of each vertex in layers up to and including the layer @xmath425 is @xmath584 , the sludge can easily reach the layer @xmath154 . since the minimum degree in this layer is @xmath585 and the game is played with parameter @xmath509 , no matter what the greens do in this very last move , the sludge reaches the sink . thus , @xmath510 a.a.s . as we already mentioned , the result follows by taking @xmath372 tending to zero . theorem [ thm : main_power - law ] follows immediately from lemma [ lem : max_deg ] and lemma [ lem : green_power_law ] . we finish by remarking on how we can try to close the gap in the previous lemma . it follows from lemma [ lem : green_power_law ] that for @xmath499 we have that a.a.s . suppose then that the game is played with parameter @xmath27 such that @xmath587 for some @xmath562 . clearly , the sludge tries to stay on vertices with as small label as possible ( that is , the largest possible total degree ) . the greens aim for the opposite , they want the sludge to go to large labels ( the smallest total degree ) . in the first round , the sludge is guaranteed to be able to go to a vertex with label at most @xmath588 and , in fact , the greens can force him to go to @xmath589 . in the next round , it might be the case that there are some `` shortcuts '' to vertices of degree at least 2 with small labels but , since the number of such edges is very small , the greens can easily prevent the sludge from using these edges . after securing these edges , the greens should protect the remaining neighbours of @xmath589 with small labels . however , this time this does not help much . the sludge is forced to ( but also is able to ) go to a vertex whose label is @xmath590 repeating this argument , and the calculations performed in the proof of lemma [ lem : max_deg ] , we can compute the position of the sludge at time @xmath178 and based on that we can decide if he wins or looses this game . optimizing this with respect to the parameter @xmath27 would yield the asymptotic value of @xmath68 . we considered seepage played on regular dags in theorem [ thm : main_d - reg ] , and in power law dags in theorem [ thm : main_power - law ] . it would be interesting to analyze the game on random dags with other degree sequences ; for example , where the degree distribution remains the same at each level , or there are the same number of vertices at each level . while our emphasis was on asymptotic results for the green number in random dags , our results could be complemented by an analysis ( via simulations ) of the green number on small dags , say up to 100 vertices . we will consider such an approach in future work . finally , hierarchical social networks are not usually strictly acyclic ; for example , on twitter , directed cycles of followers may occur . seepage was defined in @xcite for dags , but it naturally extends to the setting with directed cycles ( here , the directed graphs considered must have at least one source and a set of sinks ; the game is then played analogously as before ) . a next step would be to extend our results , if possible , to a setting where such cycles occur , and analyze the green number on , say , their strongly connected components . one question is to determine if the green number change as a function of the number of backward edges . we would like to thank the anonymous referees for suggestions which improved the paper . a. gutfraind , a. hagberg , and f. pan , optimal interdiction of unreactive markovian evaders , in : _ integration of ai and or techniques in constraint programming for combinatorial optimization problems _ , hoeve , willem - jan van ; hooker , john n. ( eds ) , ( springer berlin / heidelberg ) 2009 . wormald , models of random regular graphs , _ surveys in combinatorics _ , 1999 , j.d . lamb and d.a . preece , eds . london mathematical society lecture note series , vol 276 , pp . 239298 . cambridge university press , cambridge , 1999

from a recent re - analysis of the available data , be and b galactic evolution appears to show evidence for a two - slope behaviour with respect to o. the inferred be / o abundance ratio in halo stars is constant at very low metallicity ( primary behaviour ) and increases proportionally to o / h at high metallicity ( secondary behaviour ) . we show that this can be explained in the framework of one single model , the ` superbubble model ' , in which li , be and b are produced by spallation reactions induced by energetic particles accelerated out of a mixture of supernova ( sn ) ejecta and ambient interstellar medium inside superbubbles ( sbs ) . all the qualitative and quantitative constraints , including the energetics , the value of the transition metallicity and theli/be isotopic ratio , are satisfied provided that the energetic particles have a spectrum flattened at low energy ( in ) and that the proportion of the sn ejecta inside sbs is of the order of a few percent . this lends support to bykov s acceleration mechanism inside sbs and to the sb dynamical evolution model of mac low & mccray ( 1988 ) .

astro-ph0006099

the issue of galactic chemical evolution is one of the most fundamental in astrophysics . how does one pass from the state of the universe following primordial nucleosynthesis , where virtually all the baryonic matter is composed of h and he nuclei ( plus about ten percent of the current amount of @xmath3li ) , to a state in which the heavier chemical elements ( i.e. the ` metals ' ) exist in a sufficient proportion to form planets , rocks , crystals , plants , animals and the bodies of human beings ? most of these elements are synthesized in stellar cores or during the explosion of massive stars , and are released progressively in the interstellar medium ( ism ) , increasing the galactic content in metals , i.e. the _ metallicity _ , @xmath4 . the case of li , be and b ( libeb ) , however , is different . these _ light elements _ are believed to be produced through spallation reactions induced by the interaction of energetic particles ( eps ) in the ism . in these reactions , an heavier nucleus ( most significantly c , n or o ) is ` broken into pieces ' , or _ spalled _ , and transmuted into one of the lighter @xmath0li , @xmath3li , @xmath1be , @xmath5b or @xmath6b nuclei . except for @xmath3li , spallative nucleosynthesis is thought to be the main ( if not the only ) light element production mechanism . the case of @xmath6b is also slightly more complicated , as neutrino - induced spallation in supernovae ( the so - called @xmath7-process ) is sometimes invoked to increase the b / be and @xmath8 ratios which one would expect should the light elements be produced by nucleo - spallation alone . because of the spallative origin of libeb , the galactic evolution of light elements is also relevant to nuclear and high energy astrophysics , as it provides unique constraints about the history of ep interactions in our galaxy . while constraints about the _ current _ ep content of the galaxy ( i.e. cosmic rays as well as possibly other components ) can be derived from gamma - ray astronomy and the study of ism heating and ionization , information about the _ past _ ep populations can only be derived from integrated observables , such as the abundance of libeb in stars of various metallicities . these abundances are the outcome of ongoing libeb production from the ` birth of the galaxy ' to the time when the observed stars formed . therefore , one can regard low - metallicity stars as astrophysical fossils giving evidence for the efficiency of nucleosynthetical processes over time , or in other words for the pace at which the chemical evolution of the various elements took place . assuming that the composition observed today at the surface of the stars reflects that of the gas from which they formed , one can follow the increase of the be and b content of the galaxy , say , as a function of metallicity . the inferred chemical evolution is thus not expressed with respect to the usual physical time , but with respect to what we can call a _ chemical time _ , defined as the abundance of metals in the ism . in the approximation of a homogeneous galaxy ( without infall of primordial gas ) , the chemical time is a mere redefinition and a monotonically increasing function of the physical time . if the galaxy is not chemically homogeneous , which is both expected and observed at early stages , then the chemical time has to be defined locally and can be a non monotonic function of ordinary time ( e.g. if dilution of high metallicity gas by low metallicity gas occurs ; see parizot & drury 1999b ) . a further complication , however , arises from the fact that the ism metallicity proves to be a rather ambiguous measure of the chemical time . this is because the abundances of various elements are not found to increase proportionally to one another , and in particular the ( most often used ) o and fe clocks do not appear to mark the same unambiguous chemical time . indeed , according to some recent measurements , the o / fe abundance ratio is _ not _ constant in the galaxy , even at low metallicity , but decreases as a function of o / h ( israelian , et al . 1998 ; boesgaard , et al . 1999 ; mishenina , et al . this conclusion , however , is still controversial and there is no general consensus about which of the available methods ( oi ir triplet , oi forbidden line , or oh molecular line ) should be used to derive the o abundance whenever they are conflicting ( see however israelian et al . 2000 , for a reconciliation between the various methods ) . it might therefore seem safer , when studying be and b evolution from stellar abundances , to use the fe abundance as the chemical time of reference , since it is in principle easier to measure . however it is o , not fe , which is involved in the spallation reactions producing libeb , hence in a way fe is totally irrelevant to the problem . in the absence of a chemical clock both reliable and relevant to light element galactic evolution , we shall use here o as the reference element , and discuss libeb galactic evolution as a function of o / h , keeping in mind that the exact numerical values might be reconsidered ( or confirmed ! ) in the future , when a consensus about the various observational methods is reached . among light elements the @xmath3li isotope is characterized by a significant production by primordial nucleosynthesis , which accounts for about 10% of the total li content of the galaxy . in comparison , the contribution of galactic nucleosynthesis ( whether spallative or from agb stars or novae ) is negligible in the early galaxy , and the li stellar abundances are found to be almost constant at stellar metallicities lower than about 1/20th solar ( e.g. spite & spite 1982 ) . this implies that the early galactic chemical evolution of li is not constrained by the data , apart from being required to preserve the observed spite plateau . in principle , this does not apply to @xmath0li , which is not produced by primordial nucleosynthesis any significantly . however , @xmath0li represents such a tiny fraction of the total li present in a low - metallicity star that the measurement of its specific abundance is a considerable observational challenge , which has been taken up for only two halo stars up to now ( hobbs & thorburn 1994 , 1997 ; smith et al . 1998 ; cayrel et al . 1999 ) . both of these stars have a metallicity around @xmath9 = -2.3 $ ] , and show a @xmath0li abundance compatible with a @xmath0li/@xmath1be ratio between 20 and 80 , in contrast with the solar value of @xmath10 . as far as be and b ( beb ) are concerned , the contribution of primordial nucleosynthesis is entirely negligible so that the abundances measured in any low - metallicity star provide direct evidence about galactic nucleosynthesis itself . from the phenomenological point of view , two specific behaviours can be identified for be and b galactic evolution : * a _ primary _ behaviour , in which the be and b abundances increase proportionally to o / h , so that the be / o and b / o abundance ratios are constant , and * a _ secondary _ behaviour , in which the be and b abundances increase proportionally to the _ square _ of the o abundance , so that the be / o and b / o ratios are proportional to o / h . the first case is expected if the be and b production rate is proportional to the rate of o release in the ism , i.e. more or less to the sn explosion rate , while the second case corresponds to the standard scenario of galactic cosmic ray nucleosynthesis ( gcrn ) . indeed , according to gcrn , be and b are produced by spallation reactions induced by the interaction of eps accelerated out of the ism ( namely , the cosmic rays ) with the ism itself ( reeves et al . 1970 ; meneguzzi et al . the energy source of these eps is the kinetic energy of the sne , so that the power available for spallation is proportional to the sn explosion rate . now the effective production of be and b through spallation reactions depends on the abundance of o among the eps and the ism ( since o is the main progenitor of be and b ) . the total production rate is thus proportional not only to the rate of o release in the ism , as in the first case above , but also to the o abundance itself , i.e. our chemical clock , o / h . this results in the secondary behaviour first described by vangioni - flam et al . ( 1990 ) . it has to be noted that the gcrn would also lead to a primary behaviour , _ if the ep composition were constant and o - rich_. indeed , most of the be and b production would then be due to interactions of energetic o nuclei with the ism , and thus be independent of the ambient o abundance . the be and b production rates would then be simply proportional to the sn explosion rate , i.e. to the rate of o release in the galaxy . one can thus conclude that , phenomenologically , a primary process is expected when the o abundance among the eps is independent of ( and higher than ) that in the ism , and a secondary process is expected when the ep composition reflects that of the ism . now let us suppose that both a primary and a secondary process for beb production exist in the galaxy . then for obvious reasons , and whatever their respective weight in the global galactic ecology , the primary process is bound to dominated at very low metallicity , while the secondary process must dominate in the high metallicity limit . this is simply because the efficiency of the ( idealized ) secondary process is zero at @xmath11 and infinite at @xmath12 . the abundance diagram showing be / o and b / o as a function of o / h should then look like in fig . [ phenomenology ] , with a constant ratio below some _ transition metallicity _ , @xmath13 , and a linearly increasing ratio above . clearly , the precise determination of @xmath14 is an important goal for observational studies and would provide a strong constraint on the theoretical models of galactic be and b evolution . in a recent study , fields et al . ( 2000 ) have re - analyzed the available data about be and b evolution as a function of o / h , discussing the uncertainties associated with the various methods used to derive the o abundance , the stellar parameters and the incompleteness of the samples . their results are in conformity with the picture proposed above . in particular they have looked for evidence of a transition metallicity , using statistical analysis , and found a range of probable values of @xmath15 between @xmath16 and @xmath17 ( see also olive 2000 ) . above this metallicity , the be and b evolution shows a secondary behaviour and seems compatible with the standard gcrn scenario . unfortunately , very few data points have yet been reported below @xmath14 , and its exact value is rather uncertain . however , energetics arguments show that a primary process _ is _ indeed required below , say , @xmath18 ( e.g. parizot & drury 1999a , b ; ramaty et al . , 2000a , b ) . this provides further support to the general ` two - slopes ' scheme sketched out above , and in particular to the very existence of a transition metallicity ( which is also certified up to a 99% confidence level by fields et al . since the energetics argument has been somewhat controversial , we shall review in detail how it works in sect . [ sec : energetics ] . beforehand , let us summarize the available observational evidence as follows : 1 . the be / o ratio observed in halo stars is constant up to a metallicity @xmath14 , 2 . below @xmath14 , @xmath19 ( see sect . [ sec : energetics ] ) , 3 . @xmath14 is between @xmath20 and @xmath21 , 4 . above @xmath14 , be / o @xmath22 o / h , 5 . at @xmath23 , be / o @xmath24 ( anders & grevesse 1989 ) to these must be added the constraints relating to li : 1 . li / be @xmath25 for @xmath26 ( otherwise spallative nucleosynthesis breaks the spite plateau ) , 2 . @xmath0li/@xmath1be is between 20 and 80 at @xmath9 \simeq -2.3 $ ] , i.e. @xmath27 \sim -1.5 $ ] ( with the observed o / fe trend ; israelian et al . , 2000 ) , and 3 . @xmath28 at solar birth . finally , the constraints relating to b read : 1 . @xmath29 ; 2 . @xmath30 at solar birth ; in the following , we show that the so - called superbubble model for libeb production satisfies the whole of the above constraints with only one free parameter plus an additional mechanism for b production , e.g. the @xmath7-process or nucleo - spallation induced by a component of low - energy cosmic rays ( lecr ) . this model has been shown by parizot & drury ( 1999b ) to offer a natural and efficient primary process for be and b production in the very early galaxy . here , we argue that it also predicts the required change of behaviour ( from primary to secondary ) at a transition metallicity in the range derived by fields et al . ( 2000 ) from the observational data ( this prediction was actually implied by fig . 3 of parizot & drury 1999b , and the related discussion , although not stated explicitly ) . in other words , the full range of @xmath0libeb evolution ( i.e. below _ and _ above the primary - to - secondary transition ) can be accounted for by one single mechanism , namely the acceleration of particles inside a superbubble ( sb ) out of a mixture of sn ejecta and ambient ism . in the following section , we review the gcrn energetics . the sb model is presented in sect . [ sec : sbmodel ] , and the ingredients of the calculation are derived in sect . [ sec : ingredients ] . these ingredients are the composition and the energy spectrum of the superbubble energetic particles ( sbeps ) . in sect . [ sec : results ] , we present the results relating to the evolution of the libeb abundances as a function of o / h . we discuss their implications in sect . [ sec : conclusion ] . in this section , we intend to show that , whatever the assumption about the o / fe trend in halo stars , the standard gcr nucleosynthesis can not account for the be abundance in low metallicity stars , because of energetics problems . in their analysis , fields et al . ( 2000 ) consider three sets of be , b and o data ( which they call the balmer data , the irfm1 data , and the irfm2 data ) , differing by the choice of stellar parameters used to derive the abundances ( see also fields & olive 1999 ) . for the first two data sets , the resulting be abundance found at @xmath31 = -2 $ ] ( i.e. @xmath32 is consistent and equal to @xmath33 . for the third data set , namely irfm2 , the be abundance found at the same metallicity is somewhat higher : @xmath34 as derived from fields best model , and @xmath35 as derived directly from the data . in order to obtain a compelling conclusion , we use here the most conservative ( i.e. lowest ) value of @xmath33 at @xmath31 = -2 $ ] , i.e. @xmath36 ( anders & grevesse , 1989 ) . the be / o ratio is then @xmath37 , which means that , on average , once integrated over the galactic evolution up to the time when @xmath31 $ ] reaches @xmath38 , the production of each nucleus of o must has been accompanied with the production of @xmath37 nuclei of be . considering that a sn produces on average @xmath39 of oxygen , i.e. @xmath40 o nuclei , we end up with an average required production of be of @xmath41 nuclei per supernova ( to be scaled linearly with the actual o yield one wishes to adopt for the sne ) . according to gcrn ( and most models of be production ) , the eps responsible for the cno spallation draw their energy from sn explosions . therefore , assuming that about 10% of the sn energy is eventually imparted to the eps , the total energy available for spallation is of order @xmath42 erg , which implies a be production efficiency of @xmath43 nuclei synthesized per erg of eps injected in the ism . this efficiency has to be compared with the theoretical expectations . we have calculated the be production efficiency as a function of the target metallicity for different ep energy spectra . the ep composition has been taken identical to the ism composition , as appropriate for the gcrn scenario . the results are shown in fig . [ beprodeff ] . it should be noted that an increase of the metal abundances in the eps of about a factor of 10 , as could be invoked to conform to the current over - metallicity of the gcrs as compared to the mean ism composition , does not change the results by more than a factor of 1.7 ( upwards ) . as can be seen from fig . [ beprodeff ] , the be production efficiency obtained for the gcrn , i.e. with the standard cosmic ray source spectrum in @xmath44 , is only @xmath45 be / erg , namely a factor 60 below the required value ( or even more if one considers that the value of @xmath46 derived above is actually the mean value of the be production efficiency for @xmath47 ) . allowing for a steeper slope of the cosmic ray source spectrum , namely @xmath48 or even @xmath49 , does improve the situation , but not enough to reconcile the gcrn expectations with the observed amount of be in low metallicity stars . even with the most efficient spectrum , @xmath50 $ ] ( see sect . [ sbepenergyspectrum ] ) , the be production efficiency obtained is still a factor 7 below the required value at @xmath31 = -2 $ ] , not mentioning the values at lower metallicity . we conclude that the standard gcrn can not account for the production of be and b in the early galaxy , say for @xmath51 . as emphasized in parizot & drury ( 1999a , b ) , a simple solution to the above energetics problem consists in letting more metal - rich eps interact with the ism , so that the energetic carbon , nitrogen and oxygen nuclei ( cno ) can be spalled in flight while propagating through the ism and thereby increase the beb production usually resulting mainly from energetic protons and @xmath52 particles interacting with the ambient cno . if the abundance of cno among the eps is large enough , these _ reverse spallation _ reactions can actually dominate the beb production in the ism and make the production efficiency independent of the ambient metallicity , in contrast with the situation shown in fig . [ beprodeff ] . this is typical of a primary production mechanism , which has been shown to be bound to dominate at very low metallicity . since the cno nuclei in the ism originate from sn explosions , eps with a high abundance of cno can be obtained if particles are accelerated from a material contaminated by large amounts of sn ejecta . this is the case inside the superbubbles ( sbs ) which form consequently to the explosion of many sne in an ob associations . indeed , the dynamical effect of repeated sn explosions in a small region of the galaxy is to blow large ( super)bubbles of hot , rarefied material ( @xmath53k , @xmath54 ) , surrounded by shells of swept - up and compressed ism . the interior of superbubbles consists of the ejecta and stellar winds of evolved massive stars _ plus _ a given amount of ambient ism evaporated off the shell and dense clumps passing through the bubble . the exact composition of the particles to be accelerated inside the bubble depends on i ) how much ambient material has evaporated towards the sb interior , ii ) how well it is mixed with the sn ejecta , and iii ) where the acceleration occurs , whether over the whole sb or more locally around the most recent sn explosion . if the material inside the superbubble is not well mixed , it is possible that the sbeps be richer ( or poorer ) in cno than the average material , depending on where exactly the acceleration takes place . none of the three questions above has yet received a conclusive answer , neither from theoretical nor observational studies . this explains why the superbubble models proposed so far to account for be and b galactic evolution ( parizot & drury 1999b,2000 ; ramaty , et al . , 2000a , b ; bykov , et al . , 2000 ) make different assumptions relative to the sb dynamics and to the acceleration process . in particular , the mass fraction of the sn ejecta inside the superbubble is of the order of a few percent if a thermal conduction model is used to evaluate the evaporation of the material from the supershell , while it would be much higher if the so - called ` magnetic suppression ' mechanism occurred ( higdon et al . 1998 ) . likewise , the mixing of the ejecta with the ambient gas inside the bubble has been found to be efficient by parizot & drury ( 1999b ) , who compared the turbulent mixing time with the age of the sb , while ramaty et al . implicitly assumed a relatively poor mixing . finally , the acceleration mechanism has been assumed by parizot & drury to be of the bykov type ( bykov & fleishman 1992 ; bykov 1995,1999 ) , and thus more or less distributed over the whole sb , while ramaty et al . use the standard shock acceleration model , and thus restrict the acceleration process to the vicinity of the latest sn explosion . note that different ep energy spectra also result from these different assumptions ( see below ) . as indicated above , the current knowledge about superbubble dynamics does not allow one to decide between the models and to determine unequivocally the composition and spectrum of the sbeps . from the point of view of nucleosynthesis , though , the sbep composition and spectrum are the only ingredients we need to calculate the production rates of li , be and b and deduce their galactic chemical evolution . therefore , we propose here to proceed the other way round : first investigate the validity of the superbubble scheme for the libeb evolution , and second , use the available libeb data to constrain the sb models . to achieve the first point , we must show that the sb models naturally produce a be evolution scheme in agreement with the available data , which applies to the qualitative _ and _ quantitative features . in other words , the sb models have to account for the total amount of be found in the halo stars _ and _ the change of behaviour from a primary to a secondary evolution ( with respect to oxygen ) around a transition metallicity @xmath14 in the range currently allowed by the data . then we shall analyze more precisely the link between @xmath14 and the parameters of the sb so as to be able to draw sensible conclusions once @xmath14 is obtained with a greater precision from the data , hopefully in the near future . to this purpose , we now discuss the parameterization of the sb model and derive the relevant physical ingredients ( for more details about the sb model itself , see parizot & drury 1999b,2000 ) . concerning the sbep composition , we assume that the material accelerated inside the superbubble consists of a fraction @xmath55 of pure sn ejecta , and a fraction @xmath56 of the ambient material , i.e. gas with the ism metallicity . as the general ( or average ) metallicity of the galaxy increases , this ambient gas also gets richer and richer in heavy elements . noting @xmath57 and @xmath58 the abundances of element x among the sn ejecta and in the ism , respectively , we can write the abundance of x among the sbeps as : @xmath59 physically , the single parameter @xmath55 gathers all the information relative to the amount of evaporation off the sb shell and the dense clumps inside the sb ( which governs the mass load of the sb by interstellar gas ) and the effective mixing of the evaporated gas with the sn ejecta , before acceleration . the model obtained with @xmath60 corresponds to the acceleration of pure ejecta , while @xmath61 corresponds to the standard gcrn ( acceleration of uncontaminated ism ) . in principle , the decomposition of the sbeps into these two components ( ejecta and ism ) can always be conceived in the abstract . therefore the above parameterization does not introduce any simplification , as long as the acceleration mechanism is assumed to be non - selective ( i.e. each chemical element is accelerated proportionally to its abundance ) . however , the value of the parameter @xmath55 should be expected to vary from one sb to another , and to vary with time for each given sb , due to its dynamical evolution ( parizot & drury , 1999b ) . as we show below , different values of @xmath55 result in different sbep composition and thus different libeb yields . the fact that @xmath55 can be different for different sbs in the galaxy therefore results in some scatter in the @xmath0libeb data . additional causes for scatter in the data have been analyzed in detail in parizot & drury ( 2000 ) and we shall not consider these features here . instead , we concentrate on the mean evolution of be and b in the galaxy and average our results over all the sbs participating to the @xmath0libeb enrichment . this amounts to using a single value of @xmath55 in our models , thought of as the average of the individual @xmath55 s of individual sbs . in the same manner , we neglect the time - dependence of @xmath55 resulting from the sb dynamical evolution ( short time scale , @xmath62 yr ) . while the weight of the swept - up , evaporated ism increases inside the bubble as time passes and more and more sne explode , we use here an average @xmath55 over the whole sb evolution . for results and discussion of the full time - dependent treatment , see parizot & drury ( 1999b ) . we also average over possible changes in the galactic physical conditions , like for instance the mean ism density and magnetic field ( longer time scale , @xmath63@xmath64 yr ) . with the above parameterization , all we need to know to calculate the sbep composition as a function of time ( or mean ism metallicity ) is the composition of the ejecta and that of the ism . the latter is assumed to be the solar composition ( taken from anders & grevesse , 1989 ) , with all the metals scaled in the same way . concerning helium , we follow ramaty et al . ( 1997 ) and adopt a slightly decreasing abundance with decreasing metallicity , to account for the he galactic enrichment , from the primordial to the solar value . the function which we use is slightly different from that of ramaty et al . , and is given by : @xmath65 note also that we scale the cno abundances with o / h , rather than fe / h . according to the recent o / fe observations discussed in sect . [ introduction ] , the fe abundance should not be scaled in the same way . however , this is not relevant to our calculations here because fe does not significantly contribute to the libeb production . table [ ism(z ) ] summarizes our prescription for the ism metallicity as a function of @xmath31 $ ] . .[ism(z ) ] ism abundances relevant to be and b production , scaled with @xmath66 o / h ( @xmath67 is given by eq . [ fhe ] ) . [ cols="^,^",options="header " , ] from the sn yields of table [ snyields ] , we obtain the elemental abundances @xmath57 to be used in eq . ( [ sbepabundances ] ) , and derive the abundances at intermediate metallicities by fitting the ` data points ' linearly in logarithmic scale ( power law ) . for completeness , we now give these fits for the number abundances of h , he , c and n , normalized to @xmath68 . writing the abundance of isotope @xmath69 as @xmath70 , we obtain @xmath71 for @xmath72h , @xmath73 for @xmath74he , @xmath75 for @xmath76c , and @xmath77 for @xmath78n , from which we see that the abundances of h , he , c and o are almost constant in the sn ejecta , while the abundance of @xmath78n increases almost linearly with metallicity , as required for a secondary nucleus . concerning the sbep energy spectrum , we investigate different shapes which we discuss below : 1 . the standard cosmic ray source spectrum , @xmath79 , with @xmath80 , referred to as the ` crs2.0 spectrum ' ; 2 . a modified ( steeper ) cosmic ray source spectrum , @xmath79 , with @xmath81 or even @xmath82 , referred to as ` crs2.2 ' or ` crs2.5 ' ; 3 . the so - called ` weak sb spectra ' , @xmath83 , designed to mimic the results of bykov s acceleration mechanism in the case when most of the shocks responsible for particle acceleration inside the sb are weak shocks . 4 . the so - called ` strong sb spectrum ' , matching the weak sb spectra below some break energy , @xmath84 , @xmath85 or equivalently @xmath86 , and the crs2.0 spectrum above @xmath84 , @xmath87 . the first spectrum , ` crs2.0 ' , is motivated by the results of single shock acceleration models . we assume that it extends from 10 kev / n to @xmath88 gev / n . although a higher cut - off is possible , our results only depend logarithmically on its actual value , through the amount of energy stored ( and thus lost ) in essentially unproductive high energy particles . the second spectrum , ` crs2.2 ' , is sometimes inferred for the cosmic - ray source spectrum by propagating backwards the observed gcr spectrum , assuming a plausible propagation model for the crs in the ism ( e.g. engelmann et al . 1990 ) . as for the spectrum ` crs2.5 ' , it may be regarded as an _ extreme _ cosmic - ray source spectrum , and serves mainly as a ` toy spectrum ' here , for investigation purposes . the weak sb spectra have been used in previous works without clear physical justification . their shape can be understood as follows . considering that from the point of view of acceleration , the interior of a superbubble behaves as an ensemble of thousands of shocks with various velocities stochastically distributed over the sb volume ( bykov & fleishman 1992 ) , one can regard the sb acceleration model as a variant of multiple shock acceleration already investigated by several authors ( see e.g. marcowith & kirk 1999 , and references therein ) . in these models , the low - energy particles are accelerated very efficiently because their probability to escape from the shocks is very low : whenever a particle is advected away from one particular shock front , it can be ` caught ' again by and flow through another shock in a different place . remembering the general relation between the power - law index of particles accelerated at a single shock front and the escape and acceleration timescales : @xmath89 , with @xmath90 ( kirk , et al . 1994 ) , one sees that as the effective escape time tends to infinity in a multiple shock configuration , the power - law index of the ep distribution function tends towards @xmath91 . now a distribution function @xmath92 is equivalent to an ep source spectrum in @xmath93 in our notations , or @xmath94 in the non relativistic regime . this explains the shape of the weak sb spectra below the cut - off energy @xmath95 . clearly this spectral shape can not hold up to the highest energies , since the total energy involved in eps would then be strongly divergent above @xmath96 . in fact , a cut - off is expected in multiple shock acceleration models around an energy @xmath95 such that the diffusion length of the particles having this energy is comparable to the inter - shock distance . for energies greater than @xmath95 , the assumptions normally used to derive the basic transport equation are violated , and the above arguments fail . a comparable change in the acceleration regime around this energy is also discussed by bykov et al . , and found to be responsible for the steepening of the particle distribution function . following very crude arguments , one could expect the spectrum above @xmath95 to resemble the single - shock spectrum , @xmath89 , where @xmath97 and @xmath98 is the compression ratio across the shock discontinuity . now most of the shocks are expected to be quite _ weak _ inside a superbubble ( hence the name of the spectra ) , since they are mostly secondary shock , created by the reflection of a few primary shocks ( from the sne themselves or the stellar winds ) over clumps of denser material or other strong or weak shocks . the compression ratios across these weak shocks are thus probably less than 2 and in fact possibly close to 1 , which implies that the power - law index @xmath99 is quite high . interestingly enough , as long as @xmath99 is greater than 4 , which corresponds to the standard crs2.0 spectrum and is obtained only for strong shock ( @xmath100 ) , its exact value is not relevant to the calculations of libeb production because only a negligible amount of particles have energies above the energy @xmath95 around which the spectral shape changes . the bulk of the spallative libeb production inside the superbubble or in the surrounding shell is thus due to the interactions of sbeps with energies below @xmath95 , so that the actual shape of the cut - off is unimportant . this is the reason why we adopt an exponential cut - off for the weak sb spectra instead of a power - law cut - off , in order to avoid an extra free parameter , and keep only @xmath95 as the relevant free parameter . according to bykov & fleishman ( 1992 ) , @xmath95 is of the order of a few hundreds or a few thousands of mev / n . we use here 500 mev / n as a ` canonical ' value for the cut - off energy , but we also explored other values and found no significant changes ( except for @xmath101 mev / n ) . finally , the ` strong sb spectrum ' can be thought of as a sb spectrum in which the high energy particle distribution would match the standard cosmic - ray source spectrum . according to the above discussion , this would occur if most of the shocks accelerating the particles inside superbubbles were actually _ strong _ shocks ( hence the name of the spectrum ) , with a compression ratio close to 4 . in that case , the sbeps could be but the gcrs observed at earth . this ` unifying scheme ' actually corresponds to the original ideas of bykov & fleishman ( 1992 ) , and has already been proposed in the context of cr acceleration and libeb galactic evolution ( higdon , et al . 1998 ; ramaty & lingenfelter 1999 ; ramaty , et al . , in addition , it does not violate any available observation as the cosmic ray spectrum at low energy is not observable at earth because of the solar modulation . we would then simply be able to predict a flattening of the gcr spectrum at low energy , approaching @xmath85 . interestingly enough , we find that the ` weak ' and ` strong ' sb spectra give very similar results as far as the libeb production is concerned . for this reason , we may loosely refer below to any of the sb spectra as ` the sb spectrum ' . and since the libeb data do not constrain the sbep spectrum at high energy ( say above 1 gev / n ) , both of the following views are possible : 1 ) the sbeps have a spectrum extending up to energies of , say , @xmath88 gev / n , and are just the cosmic rays observed at earth ; or 2 ) the sbep spectrum is cut at relatively low energy , so that the sbeps can not be the same ep component as the gcrs . it is worth noting , however , that all the spectra described above have different implications for gamma - ray line astronomy as well as for the ionization and heating of the gas in the supershells and the adjacent molecular clouds . this means that additional constraints on the sbep energy spectrum can be derived in principle from the calculation of the nuclear excitation rates and the ionization rate resulting from the interaction of the sbeps with the ambient matter . these questions will be addressed elsewhere . in this paper , we focus on the implications of the different spectra for the libeb nucleosynthesis only , without any prejudice coming from other fields of astrophysics , to see whether the sb model can indeed account for the libeb observational data and what constraints can be derived from them . the ultimate goal , of course , will be to gather and combine all the possible constraints from various fields and improve our understanding of the acceleration mechanisms inside a superbubble and of the sbep energy distribution . we now have all the ingredients necessary to calculate the li , be and b production induced by the interaction of the sbeps with the ambient matter . while some of these interactions occur inside the superbubbles themselves , the density there is so low that the corresponding libeb production is negligible as compared to that occurring in the dense regions surrounding the bubbles , namely the supershells and part of the molecular clouds from which the parent ob association formed . dense clumps trapped inside the sbs can also provide efficient targets for the sbeps . in all cases , the chemical composition of the target material is very close to the mean ism composition at the time when the sbep acceleration occurs . this allows us to close our problem , since we have already calculated the ism composition as a function of @xmath102 ( see table [ ism(z ) ] ) . knowing the composition of the projectiles , their energy spectrum and the composition of the target , we can indeed calculate the various reaction rates by using the model described in parizot & lehoucq ( 1999 ) . the relevant physical ingredients here are : 1 . the spallation cross sections , taken from read & viola ( 1984 ) and various updates summarized in ramaty et al . ( 1997 ) ; 2 . the total inelastic cross sections from which we determine the probability for the projectiles to be destroyed before they give rise to a spallation reaction , and for the daughter nuclei ( li , be or b ) to be destroyed before they can thermalize and be ` integrated ' to the ism , so as to participate to the next episode of star formation ( we use the semi - empirical values given by silberberg & tsao 1990 ) ; 3 . the energy loss rates for the various nuclei in the medium of propagation , which compete with the nuclear reactions producing libeb . they have been kindly provided by jrgen kiener . in order to compare the libeb production obtained from our sb models with the observational data , we first study the absolute be production , as a function of the ism metallicity , and then discuss the elemental and isotopic ratios of the three light elements . to derive the galactic evolution of beryllium , we compute the be yield corresponding to the explosion of one supernova , and divide it by the yield of oxygen at the same metallicity , as interpolated logarithmically from table [ snyields ] . this allows us to obtain a be / o yield ratio , which indicates how the be content of the galaxy evolves as a function of the metallicity , @xmath31 $ ] . we have assumed that a total of @xmath42 erg per sn is imparted to the sbeps , which amounts to approximately @xmath103 of the sn kinetic energy . such a value for the particle acceleration efficiency is quite typical of most acceleration models , but a simple scaling of our results is straightforward for more ( or less ) efficient acceleration mechanisms . in fig . [ be / o(crs2.0 ) ] , we show the be / o yield ratio obtained for the crs2.0 spectrum and various values of the dilution parameter , @xmath55 , as a function of the ism metallicity . it is obvious from these curves that each particular sb model , i.e. each particular value of @xmath55 , leads to a be evolution which is primary ( @xmath104 constant ) below some transition metallicity , @xmath14 , and secondary ( @xmath105 ) above this metallicity , as required by the data . these qualitative features can be understood as follows . when the ambient metallicity is low , the metallicity of the sbeps is dominated by the sn ejecta , except for very low values of the mixing parameter , @xmath55 . this is easily seen from eq . ( [ sbepabundances ] ) : when @xmath106 , say , the abundance of c , n and o is @xmath107 times lower in the ism than in the ejecta ; therefore , most of the cno present among the sbeps come from the sn ejecta if @xmath108 . as a result , there is more cno in the sbeps than in the ism and the be production is dominated by reverse spallation reactions . for relatively large values of @xmath55 , thus , the be production efficiency is independent of the ambient metallicity , and the be / o yield ratio is approximately constant ( we even find slightly decreasing values because the relative weight of h and he in the sbeps varies a bit and because the o yield itself is slightly increasing with @xmath109 ) . when the ism metallicity increases , however , the fraction of the cno present in the sbeps coming from the ejecta decreases , because @xmath110 in eq . ( [ sbepabundances ] ) scales with @xmath109 . the contribution of the fresh sn ejecta to the total be yield then becomes marginal for sufficiently large values of @xmath109 , and the be production efficiency is proportional to the ambient metallicity , as in the standard gcrn scenario . indeed , the sb model is then equivalent to the gcrn since the sbep composition is very close to that of the ism . unsurprisingly , thus , we find that the be production efficiency and the be / o yield ratio is proportional to @xmath109 for a sufficiently high ambient metallicity . most importantly , the transition metallicity at which the be evolution changes from a primary behaviour ( be / o @xmath111 constant ) to a secondary behaviour ( be / o @xmath22 o ) depends on the mixing coefficient , @xmath55 . this again is evident from eq . ( [ sbepabundances ] ) . the more we put ejecta in the sbeps ( i.e. the higher the parameter @xmath55 ) , the later the secondary behaviour overcomes the primary one . therefore , we can hope to predict the transition metallicity , @xmath14 , from a sb dynamical model ( which would give the parameter @xmath55 self - consistently ) or alternatively we can constrain the sb evolution models from the value of @xmath14 derived from the observations . for example , fig . [ be / o(crs2.0 ) ] indicates that a transition metallicity between @xmath31 = -2 $ ] and @xmath31 = -1.5 $ ] implies that the sn ejecta represent between 1% and 10% of the sbeps , which sets constraints on the sb evolution models and/or the mixing of the gas inside sbs ( see sect . [ sec : sbmodel ] and below ) . however , the transition metallicity is not the only observable which one has to reproduce . as discussed in sect . [ sec : energetics ] , the data obtained from halo stars imply that the be / o yield ratio in the early galaxy must be of order @xmath37 . this is what we get from our model with the crs2.0 spectrum , as shown in fig . [ be / o(crs2.0 ) ] , if the mixing coefficient , @xmath55 , is about @xmath112 . but this is out of the range derived above . in other words , the two constraints ( @xmath14 and be / o ) are in contradiction with one another if we assume for the sbeps the crs2.0 energy spectrum . we should also note that a value of @xmath55 as high as @xmath113 is very hard to justify since it implies that the particles accelerated inside the superbubbles are almost pure ejecta , in contradiction with most sb evolution models which find a total mass inside the superbubble much larger than the mass of the ejecta ( see e.g. parizot & drury 1999b , and below ) . the situation is different , however , if one assumes that the particle acceleration does not occur over the whole volume of the superbubble , but only in those places where the ejecta of previous sne have been released . but a theoretical justification and/or observational evidence for this is still lacking . moreover , fig . [ be / o(crs2.0 ) ] shows that the transition metallicity corresponding to @xmath114 is above @xmath31 \sim -1 $ ] , which also appears in contradiction with the data . of course , if one allows for more than @xmath42 ergs to be imparted to the sbeps per sn , all the curves on fig . [ be / o(crs2.0 ) ] can be shifted upwards and an agreement may be recovered between the range for @xmath55 allowed by the be / o constraint and that allowed by the preliminary results relating to @xmath14 . this , however , would not occur unless we increase the be yield by a factor of @xmath115 , which amounts to put @xmath116 ergs into sbeps , a rather unreasonable solution . much better appears to be the situation when the sb spectrum ( either weak or strong ) are used instead of the crs2.0 spectrum . in fig . [ be / o(sb500 ) ] , we show the results corresponding to the weak sb spectrum with the cut - off energy @xmath117 mev / n . it can be seen that the correct be / o yield ratio below @xmath31 = -2 $ ] is obtained for a mixing coefficient of a few percent . this is particularly interesting since the ` standard ' sb evolution models ( in the smoothly distributed ism ) predict a very similar value for the fraction of the mass inside the bubble consisting of sn ejecta . indeed , as discussed in parizot & drury ( 1999b ) , the sb mass derived from the non - magnetic evolution model ( weaver et al . 1977 ; mac low & mccray 1988 ) is given by : @xmath118 where @xmath119 is the mechanical luminosity of the ob association responsible for the growth of the superbubble , in units of @xmath120 , @xmath121 is the ambient density in @xmath122 , and @xmath123 is the time elapsed since the first sn explosion ( in myr ) . as a consequence , when 50 sne have exploded , say , the time elapsed is about @xmath124 ( if @xmath119 = 1 ) and the total mass of the gas inside the sb is @xmath125 . considering that each supernova ejects , on average , @xmath126 of material , the total mass of the ejecta reaches about @xmath127 , i.e. @xmath128 of @xmath129 . that this value is precisely in the range required to account for the be / o ratio observed in halo stars is very remarkable , since we have not made any particular assumption to derive it . in other words , by using the standard sb evolution model and the standard efficiency for a particle acceleration model ( i.e. @xmath130 ) , one gets exactly the amount of be production needed to explain the be abundances observed in stars of metallicity lower than @xmath131 solar . this should be considered as a strong argument in favour of both the sb model and the sb spectra . in addition , we see from fig . [ be / o(sb500 ) ] that the very same set of parameters also leads to a transition metallicity between @xmath31 = -2 $ ] and @xmath31 = -1.5 $ ] , i.e. again exactly in the range suggested by the observational data . both constraints are thus found to be consistent if one uses one of the sb spectra . and finally , it is possible to calculate the be / o ratio predicted by this model at solar metallicity . to do so , we first need to recognize that the quantity plotted on the figures is not the stellar be / o abundance ratio itself , but the be / o yield ratio ( i.e. ` at production ' ) . however , it is easy to derive one ratio from the other : focusing on the part of the curve above @xmath14 , we can write @xmath132 for the be / o production ratio , where @xmath133 is a constant which we do not need to specify here , and @xmath134 is the total number of o nuclei in the ism . now the total number of be nuclei in the ism is given as a function of time by : @xmath135 , \label{eq : be(t ) } \end{split}\ ] ] where @xmath136 is the time corresponding to the transition towards the secondary behaviour of be . now acknowledging that @xmath137 and @xmath138 are negligible as compared to the solar values , we obtain @xmath139 from which we deduce the elemental abundance ratio at solar metallicity : @xmath140 from fig . [ be / o(sb500 ) ] we see that the be / o production ratio at solar metallicity is @xmath141 . the ` predicted ' value for the solar be / o abundance ratio for this model is then @xmath142 , remarkably close to the measured value , namely @xmath143 ( anders & grevesse 1989 ) ! we can thus report , in conclusion , that the sb model described above is fully consistent with the available data , and reproduces the whole scheme of be evolution , both qualitatively and quantitatively , namely 1 . the primary behaviour of be up to @xmath14 , 2 . at the correct level of be / o @xmath144 , 3 . the secondary behaviour above @xmath14 , 4 . with a value of @xmath14 consistent with the data , 5 . and the correct value of the be / o abundance ratio at solar metallicity . having shown that the sb model successfully describes the galactic evolution of be over the whole range of halo star metallicities , we now consider the case of the other light elements , li and b , and see whether the libeb data can actually be used to set constraints on the sb dynamical evolution . the observational constraints on li production by sbep - induced spallation have been summarized in sect . [ sec : phenomenology ] . in fig . [ li / be ] , we show the li / be yield ratios obtained with the crs2.0 and the sb spectra , for different values of the mixing parameter , @xmath55 . the general shape of the curves is easily understood if one realizes that be is produced by cno spallation only , while li is produced by cno spallation _ and _ @xmath145 reactions . as a consequence , higher li / be yield ratios are obtained for ep compositions richer in he . at a given metallicity , the latter correspond to lower values of @xmath55 , i.e. smaller proportions of cno - rich ejecta among the eps . however , as the ambient metallicity increases , the weight of @xmath145 reactions decreases and the li production becomes dominated by the cno - spallation anyway , so that the spread in the li / be production ratios for different values of @xmath55 decreases . as can be seen from the figures , the constraint @xmath146 is respected for any sb model with a mixing parameter larger than @xmath147 . this is in very good agreement with the range derived from the energetics constraint and the value of @xmath148 in the case of the sb spectrum . concerning @xmath0li , we found a @xmath3li/@xmath0li isotopic production ratio in the range 1.31.7 for any value of @xmath55 , any ambient metallicity and any sbep spectrum . such a ` constancy ' of the isotopic production ratio is due to the fact that @xmath0li and @xmath3li are produced by the same nuclear reactions , with only slight variations in the reaction thresholds and resonances . the curves showing the @xmath0li/@xmath1be production ratio as a function of metallicity can thus be roughly derived from fig . [ li / be ] , and we only show the exact results for the case of the sb spectrum with @xmath149 mev / n , in fig . [ 6li/9be(sb500 ) ] . interestingly enough , the @xmath0li/@xmath1be production ratio at solar metallicity is virtually independent of the mixing parameter inside superbubbles , and is about @xmath150 . in order to compare this value with the solar value , @xmath10 , one has to average the @xmath0li/@xmath1be production ratio over the whole galactic evolution . proceeding in the same spirit as in eq . ( [ eq : be(t ) ] ) , we have : @xmath151 performing these integrations , we obtain @xmath152 , in reasonable agreement with the observed value . however , as we noted above , this constraint can not be used to distinguish between the sb models , since different values of @xmath55 and different ep spectra give roughly similar results , namely @xmath153 ratios between 3.5 and 5 . on the other hand , [ 6li/9be(sb500 ) ] shows that the value of the @xmath154 ratio at lower metallicity is a very important observable , as it allows one to constrain the effective mixing parameter , @xmath55 , very efficiently . very remarkably , the first few data available at @xmath155 give @xmath154 between 20 and 80 , which corresponds to a value of @xmath55 in a range around one percent , in very good agreement with the values derived from the energetics and @xmath14 constraints . and once again , this agreement is achieved very naturally with the sb spectrum , while the higher value of @xmath55 required for the crs2.0 spectrum ( see above ) would imply a value of the @xmath154 ratio at low z in contradiction with the observation . in this respect , we confirm that the model proposed by ramaty et al . ( 2000a ) in which the eps have a constant metallicity all over the galactic chemical evolution ( which corresponds to a high @xmath55 in our parameterization ) can not account for the @xmath154 ratio observed in halo stars . more data are expected to make this argument even stronger in the near future ( or invalidate it , if in contradiction with the current data ) . in any case , it is worth stressing again that the measurement of the @xmath154 ratio in low metallicity stars is very important , as it allows one to constrain the sb models very efficiently , in a way completely independent of the energetics . and as it stands , it is a strong argument in favour our model that all the available constraints , although independent from one another , namely energetics , @xmath14 and @xmath154 , are in excellent agreement . moreover , they indicate that the proportion of the ejecta inside sbs , @xmath55 , is of the order of a few percent , and that the sbep spectrum is close to the idealized sb spectrum discussed in sect . [ sbepenergyspectrum ] . both results have strong theoretical support from unrelated fields , namely sb dynamical evolution and sb particle acceleration ( see above ) . finally , we turn to the case of boron . for all the sbep spectra and compositions we have studied ( i.e. for any value of @xmath55 ) , we found b / be production ratios between 10 and 12 , for all ambient metallicities . as in the case of the @xmath3li/@xmath0li production ratio , this is due to the fact that be and b are produced by the same nuclear reactions , namely cno spallation . the same is also true for the @xmath6b/@xmath5b isotopic ratio , which is found between 2.0 and 2.3 for any value of the parameters . comparing these values with the observational constraints recalled in sect . [ sec : phenomenology ] , we find again the well - known result that spallation processes underproduce b , and in particular the @xmath6b isotope , with respect to other light elements . as noted above , this should not be considered as a failure of the model since other processes are known for b production ( @xmath7-process and lecr spallation ) and have not been included here . in fact , the present results might be considered as indications that these other processes _ must _ be efficient in the galaxy . this issue will be addressed elsewhere . the sb model described above has been shown to be fully consistent with the qualitative and quantitative constraints of libeb galactic evolution . notably : 1 ) it explains the inferred two - slope behaviour in the framework of one single model ; 2 ) it provides the correct value of be / o at low metallicity ; 3 ) it predicts the correct value of the transition metallicity ; 4 ) it does not break the spite plateau ; 5 ) it is consistent with the @xmath0li/@xmath1be ratio at all metallicity . most importantly , these successes rely on the value of only one free parameter , namely the proportion of the sn ejecta inside a sb . our calculations allow one to derive the value which best reproduces the data , namely a few percent , from the constraints of libeb evolution alone , i.e. without any external prejudice . however , it is remarkable that this value is exactly in the range expected from standard sb dynamical evolution models ( mac low & mccray 1988 ) . likewise , the sb model is found to be successful only if the eps accelerated inside a superbubble have an energy spectrum flattened at low energy ( in @xmath2 ) . now this is just what bykov s sb acceleration model predicts . interestingly enough , the consistency of the sb model is such that one may be tempted to reverse the argument and consider that the results of libeb galactic evolution lend support to the current ideas about sb dynamical evolution and particle acceleration . silich et al . ( 1996 ) have studied in detail the effect of the evaporation of the clouds engulfed inside superbubbles on their dynamical evolution . they assume that the ambient ism is composed of small diffuse clouds with typical internal density of @xmath156 , and consider both cloud evaporation and dynamical disruption through internal flows inside the bubble . in the present context , these mechanisms would imply a modification of the mass loading of the superbubble and thus affect the expected proportion of the ejecta among the sbeps . although our results indirectly confirm the relevance of superbubble simulations in a smoothly distributed ism , with only large - scale density gradients , the models taking into account a cloudy ism can not be rejected . in particular , bykov ( 1999 ) has shown that in time - dependent models most of the acceleration occurs at relatively early times , when the contribution of cloud disruption and evaporation to the superbubble mass loading is small , according to the calculations of silich et al . ( 1996 ) . as far as the particle acceleration inside superbubbles is concerned , we find that only an energy spectrum showing the characteristic shape of multiple shock acceleration at low energy ( @xmath85 ) is compatible with all the data . on the other hand , any sb spectrum is found to give very similar results . this means that libeb evolution alone can not be used to determine whether the sbeps extend to high energy or not . above the so - called break energy , their spectrum could either match the standard crs spectrum or decrease more steeply around an energy of a few hundreds or thousands of mev / n . in the latter case , the sbeps will have to be regarded as a second component of eps in the galaxy , in addition to cosmic - rays . in the opposite case , however , the sbeps could be nothing but the gcrs and the sb model could then be thought of as a mere correction of standard gcrn , taking into account the chemical inhomogeneity of the early galaxy . indeed , gcrn assumes that the libeb - producing energetic particles are accelerated out of the average ism . in a more detailed analysis , however , one has to acknowledge that the metals are released in localized regions of the galaxy and abandon the homogeneous , one - zone representation . for it turns out that particle acceleration occurs precisely where the metals are more abundant this is the essence of the sb model . it makes a crucial difference in the early galaxy , when the chemical inhomogeneities are sharper ( typically for @xmath157 ) . but the value of the parameter @xmath55 which we derived ( a few percent ) is small enough for the chemical composition to be roughly identical inside and outside superbubbles today . this shows that the sb model and the gcrn model are essentially the same at metallicities @xmath158 , i.e. for most of the lifetime of the galaxy . if the bottom line of the sb model is to account for chemical inhomogeneities , then it could actually apply to a wider context than sbs themselves . localized star bursts would behave in about the same way , and globular clusters might also be considered as large ob associations giving rise to the same kind of phenomena , and thus participating to the libeb evolution in about the same way . the framework of libeb evolution described here would thus be quite general and only slight variations of the relevant parameters would occur from one environment to an other , or from one place to another . these variations would cause possibly strong variation of the light elements to metals ratios from star to star , but all fitting in the same general pattern , namely a primary behaviour of @xmath0libeb up to a transition metallicity , @xmath14 , around @xmath131 to @xmath159 , and a secondary behaviour afterwards . the resulting scatter in the libeb data has been discussed in parizot & drury ( 2000 ) ( see also cunha et al . , 2000 ) . in the spirit of the present paper , the scatter in the be and b data can be evaluated by letting @xmath55 vary from one sb to another , or as a function of time . this amounts to allow for some ` diffusion ' across the curves represented in figs . [ be / o(crs2.0 ) ] , [ be / o(sb500 ) ] and [ 6li/9be(sb500 ) ] . as an example , a variation of @xmath55 by a factor of 3 ( from 1% to 3% , say ) would result in a variation of be / o by the same factor at very low @xmath4 , while only by a factor of 2 at @xmath160 } = -2 $ ] , and no variation at @xmath161 . hopefully , improvements of the theoretical modeling of stellar atmospheres and the line formation will allow one to shorten the errors bars and measure the scatter in the data as a function of metallicity . stronger constraints will thus be set on the libeb evolution models . likewise , an accurate determination of the mass fraction of ejecta inside the sb can only be achieved if the transition metallicity , @xmath14 , is measured with enough precision . this requires data points at metallicities much lower than @xmath160 } = -2 $ ] . as for the determination of the sbep energy spectrum at high energy , it shall be more efficiently constrained by gamma - ray astronomy . an important challenge for the satellite integral , to be launched in 2002 , will be to detect nuclear de - excitation lines from ep interactions in nearby superbubbles ( e.g. orion or vela ) . the expected gamma - ray fluxes have been estimated to be comparable with the integral thresholds ( parizot & kndlseder , 1999a , b ) . finally , we wish to stress one of the important differences between the sb model , which accounts for the ` two - slope behaviour ' of be and b evolution within a single model , and an alternative scenario in which the low - metallicity ( primary ) part of the correlation would be attributed to one mechanism , and the secondary part to another mechanism ( e.g. gcrn ) . in our model , the transition between the slope 1 and slope 2 be - o correlation is continuous . any intermediate value is reached over a given range of stellar metallicity ( depending on the value of @xmath55 ) . in the other case , by contrast , one would have a sharp change of slope at the precise metallicity where the secondary process becomes dominant , with no intermediate values . of course , expected physical fluctuations of the parameters would weaken this effect , and current observational error bars prevent us from distinguishing conclusively between the two pictures . but we argue that the observed ` slope 1.5 ' behaviour reported by some authors ( e.g. boesgaard & ryan , 2000 ) can be explained ( in principle ) only if there is a continuous _ transition _ from slope 1 to slope 2 within a _ unique _ model , rather than two unrelated models with a slope 2 eventually superseding a slope 1 . i wish to thank the international astronomical union and the organizers of the iau 198 symposium on libeb evolution in natal ( brazil ) , which stimulated this work . i was supported by the tmr programme of the european union under contract fmrx - ct98 - 0168 .

it is useful to study the space of all cosmological models from a dynamical systems perspective , that is , by formulating the einstein field equations as a dynamical system using appropriately normalized variables . we will discuss various aspects of this work , the choices of normalization factor , multiple representations of models , the past attractor , nonlinear dynamics in close - to - friedmann - lematre models , weyl curvature dominance , and numerical simulations . 1 * cosmological models from a dynamical systems perspective * + j. wainwright and w.c . lim + department of applied mathematics , university of waterloo , + waterloo , ontario , canada n2l 3g1 + department of mathematics and statistics , dalhousie university , + halifax , nova scotia , canada b3h 3j5 + [ 2 mm ] email : [email protected] , [email protected] +

gr-qc0409082

the hot big - bang model of modern cosmology is based on the assumption that the large scale geometry and dynamics of the universe can be described by an exact flrw model , with more detailed and smaller scale physical phenomena , such as density fluctuations or gravitational waves , being described by perturbed fl models , i.e. by solutions of the efe linearized about an exact fl model . we begin with the premise that it is important to study cosmological models more general than perturbed fl models . in particular one is interested in the set @xmath3 , namely , the set of all universes that are compatible with current observations ( ellis 2004 ) . a member of this set must have an epoch , possibly finite , during which the model is close to fl ; at early or later times it may deviate significantly from fl . more generally we are interested in the set @xmath4 of _ all _ cosmological models , and in the relationship between @xmath4 , @xmath3 and @xmath5 , the set of all fl models . for example , we want to study the relation between linearly perturbed fl models and models that are close to fl in some well - defined , gauge - invariant sense , and satisfy the exact efe . are nonlinear effects perhaps significant in close - to - fl models ? another question that takes one outside the realm of linearly perturbed fl models is the detailed nature of the generic cosmological singularity . with the preceding discussion as motivation we consider the class @xmath4 of all cosmological models , whose state at time @xmath6 can be represented by a vector @xmath7 in a state space @xmath8 . the evolution of a model universe is then described by a curve @xmath9 in @xmath8 , called an orbit , that will be a solution of a system of first order autonomous evolution equations and constraints of the form @xmath10 where @xmath11 denotes partial differentiation with respect to the spatial coordinates @xmath12 , and @xmath13 denotes possible higher order spatial derivatives . the goal is to choose state variables that remain bounded during the evolution of the model and a time variable @xmath6 such that @xmath14 at the initial singularity and @xmath15 at late times . within this framework , classes of models with symmetries will be represented by invariant subsets of @xmath8 , the most important being the subset @xmath16 that describes the fl cosmologies . one expects that the orbits of close - to - fl models will shadow orbits in @xmath16 . the appropriate mathematical vehicle for implementing the above program is the orthonormal frame formalism , since it expresses the efe directly as first order ( in time ) autonomous evolution equations . it is also necessary to introduce a process of normalization in order to create bounded variables . a choice that has proven effective is the so - called _ hubble normalization_. this approach to the study of cosmological dynamics has been discussed in some detail in wainwright & ellis 1997 , and more recently by coley 2003 , with emphasis on spatially homogeneous models . since then the approach has been extended to models without symmetry by uggla 2003 . the plan of this paper is as follows . in section [ sec_hub ] we describe the principal features of the hubble - normalized state space , and in section [ sec_c_i ] we discuss close - to - fl models and the related notion of isotropization . in section [ sec_sh ] we give an overview of the dynamics of a special class of models in the cosmological hierarchy , the nontilted spatially homogeneous cosmologies . in section [ sec_vac_num ] we describe a candidate for the past attractor in the hubble - normalized state space , and mention some recent numerical simulations which support this proposal . in section [ sec_col ] we briefly introduce a modification of hubble normalization which provides a description of nontilted spatially homogeneous models that undergo recollapse . section [ sec_conclusion ] contains our concluding remarks . in this section we give the motivation for introducing hubble - normalized variables , and describe the principal features of the resulting hubble - normalized state space and evolution equations . one of the rationals for using dynamical systems methods in cosmology is the hope that one can describe the evolution of cosmological models near the initial singularity by means of a _ past attractor _ of a cosmological dynamical system , and the dynamics at late times in a particular epoch by means of a _ future attractor_. in order to formulate the einstein field equations as a dynamical system it is clear that one has to normalize the variables , since near the initial singularity physical variables typically diverge and at late times typically tend to zero . physical considerations suggest that in a cosmological setting , normalization with the hubble scalar of the fundamental congruence is an appropriate choice . firstly , consider the density parameter @xmath17 , which plays a fundamental role in cosmology , in that it measures the influence of the matter on the dynamics . in geometrized units , it is defined to be the matter density @xmath18 divided by @xmath19 , where @xmath20 is the hubble scalar . secondly , the extent to which the overall expansion of the universe is close to isotropy is measured by the ratio of the shear ( the trace - free part of the expansion tensor ) to the hubble scalar . the importance of hubble normalization appears to have first been mentioned in the literature by kristian & sachs 1966 ( see page 398 ) in connection with their general analysis of the potential constraints that observations of distant galaxies can place on the geometry of spacetime . the role of hubble - normalized variables emerges again when one analyzes the potential constraints that arise from observations of the cmb ( maartens 1995 , 1996 ) . in both analyses one finds that bounds are placed on hubble - normalized physical and geometrical quantities . our principal interest , however , lies in using hubble - normalized variables as dynamical variables , which dates back to collins 1971 , who gave a qualitative analysis of the dynamics of some special classes of cosmological models . working within the orthonormal frame formalism we introduce hubble - normalized variables and write the efe as first order evolution equations in the hubble - normalized state space . for simple classes of ever - expanding models , for example open fl models , and spatially homogeneous models of bianchi type i , the hubble - normalized state space is bounded . near the singularity physical variables such as the matter density diverge , but the hubble scalar also diverges and at such a rate that the corresponding hubble - normalized quantity , the density parameter @xmath21 , remains bounded . in contrast , if @xmath20 is close to zero , then the other physical variables are equally close to zero , so that the hubble - normalized quantities are bounded . for more general models , however , it turns out that this simple picture is not valid _ the hubble - normalized state space is unbounded_. the primary reason is that the hubble - normalized weyl curvature tensor can assume arbitrarily large values . all available evidence , however , suggests that _ the hubble - normalized variables are bounded into the past _ , i.e. on approach to the initial singularity . in addition , if there is a positive cosmological constant there is strong evidence that the hubble - normalized variables are also bounded into the future . thus , even though the hubble - normalized state space is unbounded , one expects that the evolution equations will admit a past attractor and a future attractor . when one formulates the efe as a dynamical system , one uses a time variable @xmath6 that potentially assumes all real values . asymptotic regimes _ are then defined by the limits @xmath22 and @xmath23 . cosmologists model the physical universe as a sequence of epochs in time , an epoch being identified by which source term is dynamically dominant . a typical succession of epochs is i ) inflationary , dominated by a scalar field @xmath24 , \ii ) radiation - dominated , @xmath25 , \iii ) matter - dominated , @xmath26 , \iv ) accelerating , dominated by a cosmological constant @xmath27 . with each source term is associated a hubble - normalized energy density , @xmath28 and a particular epoch is defined as the time interval during which a particular @xmath29 is dominant . in analytic work one typically assumes the presence of one or two source terms and one investigates the asymptotic regimes , which are described mathematically by the past attractor and future attractor of the dynamical system . thus , for example , the past asymptotic regime of the radiation - dominated epoch would physically coincide with the end of the inflationary epoch and the future asymptotic regime would coincide with the beginning of the matter - dominated epoch . a cosmological model is a spacetime with a preferred timelike vector field @xmath30 whose metric is a solution of the efe with appropriate matter / energy content . it is assumed that there is an epoch during which the model is expanding , i.e. the preferred vector field satisfies @xmath31 where @xmath20 is referred to as the hubble scalar . we assume furthermore that @xmath30 is the normal vector field to a family of spacelike hypersurfaces @xmath32 , thereby defining a time variable . we introduce an orthonormal frame @xmath33 , with @xmath34 , and choose the spatial frame vectors @xmath35 to be fermi - propagated . within this framework , the gravitational field variables are the commutation functions @xmath36 associated with the frame , defined by @xmath37 = \gam^c{}_{ab}\ , \mathbf{e}_c.\ ] ] they may conveniently be expressed in terms of geometric quantities \ { h , _ , _ , a_,n _ } , according to @xmath38 ( see for example we , page 32 ) . here @xmath39 is the acceleration and @xmath40 is the shear of the congruence @xmath30 , while @xmath41 and @xmath42 determine the curvature of the spacelike hypersurfaces ( we , pages 18 , 19 , 34 ) . as regards matter / energy content we consider a cosmological constant @xmath43 and a perfect fluid with 4-velocity vector field @xmath44 and linear barotropic equation of state @xmath45 , with @xmath46 . in the orthonormal frame formalism the perfect fluid is described by its energy density @xmath18 relative to @xmath47 and its velocity @xmath48 relative to @xmath47 , i.e. the projection of @xmath44 orthogonal to @xmath47 ( see uggla 2003 , equations ( 2.1)(2.6 ) for details ) . so the gravitational variables ( [ geo ] ) are augmented by the matter variables \ { , v_,}. the velocity @xmath48 is dimensionless and satisfies @xmath49 we now introduce hubble - normalized variables and differential operators according to @xmath50 it is also necessary to introduce the derivatives of the normalization factor , namely the _ deceleration parameter _ @xmath51 and the _ spatial hubble gradient _ @xmath52 , defined by q+1 : = - 1h _ 0 h , r_:= - 1h _ h. at this stage we introduce the separable volume gauge ( uggla 2003 ) , which is characterized by @xmath53 and the existence of local coordinates @xmath6 , @xmath12 such that @xmath54 with the preferred hypersurfaces being given by @xmath55 . here @xmath56 and @xmath11 denote partial differentiation with respect to @xmath6 and @xmath12 . the deceleration parameter @xmath51 can then be expressed in terms of the variables in ( [ hubnormvar ] ) and the differential operator @xmath57 , using the raychaudhuri equation . with the above choices of gauge the hubble - normalized state vector becomes @xmath58 the efe , the jacobi identities and the commutators lead to a system of evolution equations and a system of constraints for the components of @xmath7 . in order to describe the structure of these equations , it is convenient to decompose the state vector @xmath7 as follows : = ( e_^i , r_)^t , where = ( _ , a_,n _ , _ m,_,v_)^t . the evolution equations have the following form : @xmath59 where the deceleration parameter @xmath51 is given by q = 2 ^2 + 12 _ m -_- 13(e_^i _ i - 2a_)r^. there is also a set of constraints that can be written symbolically as 0 = ( , _ i ) . one of the constraints , the so - called gauss constraint , is of particular importance , and we thus give it specifically : _ m + _ + _ k + ^2 = 1 , where @xmath60 is the hubble - normalized shear scalar given by ^2 = 16 _ ^ , and @xmath61 is the hubble - normalized spatial curvature is defined by @xmath62 , where @xmath63 is the curvature scalar of the metric induced on the hypersurfaces @xmath55 . ] of the hypersurfaces @xmath55 , given by _ k = -23(_-r _ ) a^ + 16 n _ n^ -112(n_^)^2 + a_a^. the detailed form of equations ( [ bsys])([esys ] ) is given in lim 2004 ( equations ( 2.18)(2.30 ) with @xmath64 ) . what concerns us here is the overall structure of this system . we make a number of observations . * the equations are first order evolution equations in time , but do contain second order spatial derivatives , namely the second derivatives of the spatial hubble gradient @xmath65 . these derivatives appear in ( [ evor ] ) , on account of ( [ q ] ) . in this respect , the equations are reminiscent of a system of quasi - linear diffusion equations . we shall discuss this matter further in section [ sec_vac_num ] . * the frame variables @xmath66 enter into the remaining equations only through the spatial differential operator @xmath57 . recalling that @xmath52 is the spatial hubble gradient we thus regard the variables @xmath66 and @xmath52 as controlling the spatial inhomogeneity in the cosmological model , while the variables in @xmath67 directly determine the spacetime geometry ( i.e. the gravitational field ) and the matter content . we consider three classes of ever - expanding cosmological models , namely , friedmann - lematre ( fl ) models , spatially homogeneous ( sh ) models and general models , i.e. models without symmetry . the sh models admit a three - parameter local group @xmath68 of isometries acting on spacelike hypersurfaces , while in the fl models the isometry group is a @xmath69 . for brevity we shall refer to the models without symmetry as _ @xmath70 cosmologies_. so we have the following cosmological hierarchy : @xmath71 we will use @xmath72 to denote the corresponding hubble - normalized state spaces : @xmath73 one can construct a more detailed hierarchy by including models with two or one spacelike killing vectors . classes of models with two killing vectors are variously referred to as gowdy , @xmath74-symmetric or @xmath75 cosmologies . models with one killing vector are referred to as @xmath76-symmetric or @xmath77 cosmologies . the sh cosmologies are obtained by requiring that the spatial frame derivatives of the gravitational field and matter variables @xmath67 , and of the normalization factor @xmath20 be zero , i.e. _ = 0,r_=0 . it then follows that all the dimensional commutation functions and matter variables are constant on the hypersurface @xmath55 , which are thus the orbits of a three - parameter group @xmath68 of isometries . the evolution equations ( [ evor ] ) and ( [ evoy ] ) imply that the sh restrictions ( [ shres ] ) define an invariant set of the full evolution equations , which we shall call the _ sh invariant set_. indeed , equation ( [ evor ] ) is trivially satisfied , and equation ( [ evoy ] ) reduces to a system of ordinary differential equations , namely _ t y_a = f_a ( ) . the nontrivial constraints become purely algebraic restrictions on @xmath67 , which we write symbolically as ( ) = 0 . an important consequence of this specialization is that the evolution equation ( [ bsys ] ) for @xmath66 decouples from the evolution equation for @xmath67 , which means that _ the dynamics of sh cosmologies can be analyzed using only equations ( [ sh1 ] ) and ( [ sh2])_. in this context , one can think of the variable @xmath67 in equation ( [ y ] ) as defining a _ reduced hubble - normalized state space _ , of finite dimension , for the sh cosmologies . the bianchi classification of the isometry group then leads to a hierarchy of invariant subsets . in the sh context the restriction @xmath78 defines an invariant subset , giving the so - called _ nontilted sh cosmologies_. this class of models have been analyzed in detail in the literature . we give an overview of the known results in section [ sec_sh ] . we note in passing that tilted sh cosmologies ( i.e. @xmath79 ) were first studied by king & ellis 1973 . considerable progress has been made recently , using hubble - normalized variables ( see hewitt 2001 , hervik 2004 , coley & hervik 2004 ) . the fl cosmologies are obtained by imposing the restrictions @xmath80 which is equivalent to requiring that the fluid has zero shear , vorticity and acceleration . it follows that the density parameters are constant on the hypersurface @xmath55 and the second from the definition of @xmath81 and @xmath52 . ] : @xmath82 in addition the weyl curvature is zero , @xmath83 and the 3-ricci curvature satisfies _ = 0 , _ _ k = 0 . the dynamics of the fl models is governed by the matter evolution equations which simplify to a system of two odes , @xmath84\om_m \\ \lb{sh_ol } \ptl_t \om_\lambda & = 2(q+1 ) \om_\lambda,\end{aligned}\ ] ] with q = 12(3 - 2)_m -_. the gauss constraint ( [ cgauss ] ) simplifies to _ m + _ k + _ = 1 , where all three quantities are functions of time only . for future reference we note that the curvature parameter @xmath61 satisfies the evolution equation _ k = 2q _ k , as follows from ( [ sh_om ] ) , ( [ sh_ol ] ) and ( [ sh_cgauss ] ) . the remaining evolution equations simplify to @xmath85 and hence decouple from ( [ sh_om ] ) and ( [ sh_ol ] ) . these variables reflect the choice of spatial frame , and do not affect the essential dynamics . since @xmath86 , these equations can be integrated to yield @xmath87 where hatted variables depend only on the spatial coordinates @xmath12 , and @xmath88 . these quantities are restricted be equation ( [ sab0 ] ) . models with negative , zero and positive spatial curvature are characterized respectively by @xmath89 and in the corresponding canonical spatial frames we have @xmath90 thus , the fl cosmologies with perfect fluid and cosmological constant are described by a reduced hubble - normalized state space , namely the two - dimensional space with the state vector @xmath91 , and evolution equations ( [ sh_om])([sh_q ] ) . it follows from ( [ sh_cgauss ] ) that for models with negative and zero spatial curvature ( @xmath92 ) , the state space is bounded . this state space is shown in figure [ fig3.1 ] in section [ sec_c_i ] , using @xmath61 and @xmath81 as variables . in the analysis of the dynamics of sh models using the reduced hubble - normalized state space @xmath93 , the equilibrium points ( i.e. fixed points ) of the dynamical system , defined by @xmath94 naturally play a significant role . at this stage the kasner vacuum solutions come into play , given in hubble - normalized variables by n_=0,a_=0,^2=1 , _ m=0,_=0,q=2 , with @xmath95 these equilibrium points ) , consists only of equilibrium points is a consequence of the fact that we are using a fermi - propagated spatial frame . see uggla 2003 , page 7 , for a different spatial gauge choice , which leads to a circle of equilibrium points , and frame transition orbits " . ] form a 4-sphere ( recall the definition ( [ sig_sq ] ) , and that @xmath96 is trace - free ) , which we shall call the _ kasner sphere _ @xmath97 . if @xmath96 is diagonal , then @xmath97 reduces to a circle . two other important equilibrium points lie in the fl invariant set and satisfy @xmath98 and one of the following sets of conditions : . see table [ tab1 ] . ] @xmath99 the local stability of the flat fl and de sitter equilibrium points is of importance in connection with the phenomenon of isotropization , as discussed in section [ sec_iso ] . in the following tables we give the dimensions of the stable and unstable manifolds of these equilibrium points , for both nontilted and tilted sh models . the sh equilibrium points , with the exception of de sitter , represent self - similar solutions of the efe , admitting a four - parameter group @xmath100 of similarities . we refer to we , section 9.1 , for a complete list of these self - similar solutions within the class of nontilted sh models . on account of ( [ bsys ] ) and ( [ evor ] ) , the conditions e_^i=0,r_=0 define an invariant set , the so - called _ _ silent boundary _ _ , that forms part of the boundary of the hubble - normalized state space . solutions of the evolution equations in this invariant set do not correspond to solutions of the efe , however , since the frame variables have to satisfy @xmath101 . nevertheless , it appears that this invariant set plays a fundamental role in describing the asymptotic behaviour of cosmological models . indeed current investigations suggest that the past attractor and the future attractor for @xmath70 cosmologies are subsets of the above invariant set ( subject to @xmath27 in the case of the future attractor ) . we will discuss this matter further in sections [ sec_c_i ] and [ sec_vac_num ] . the condition @xmath102 in the definition of the silent boundary implies that @xmath103 . it thus follows that the evolution equations and constraints , when specialized to the silent boundary , are precisely the sh evolution equations and constraints , as given by ( [ sh1 ] ) and ( [ sh2 ] ) . in addition each self - similar @xmath100 solution determines an equilibrium point in the silent boundary . it is customary to regard a cosmological model as being close to fl " if its metric is a linear perturbation of an fl metric ( e.g. bardeen 1980 ) , i.e. the dynamics of such models are governed by the _ linearized _ efe . an alternative approach is to regard a cosmological model as being close to fl " if an appropriate set of hubble - normalized anisotropy parameters are small . for simplicity we will restrict our attention to two parameters . firstly we consider the shear parameter @xmath104 , defined by ( [ sig_sq ] ) , which describes the anisotropy in the rate of expansion of the fundamental congruence . secondly we consider the weyl parameter @xmath105 defined by @xmath106 where @xmath107 and @xmath108 are the electric and magnetic parts of the weyl curvature tensor relative to the fundamental congruence . the weyl parameter can be thought of as quantifying the intrinsic anisotropy of the gravitational field . as described in section [ sec2.3 ] , the fl cosmologies with negative or zero spatial curvature and positive cosmological constant have a two - dimensional hubble - normalized state space . using @xmath61 and @xmath81 as independent variables , the state space is shown in figure [ fig3.1 ] . the equilibrium points represent the flat fl model ( f ) , de sitter model ( ds ) , and the milne model ( m ) . one sees that f is the past attractor and ds is the future attractor . the milne point m is a saddle , which means that a subset of open fl models with @xmath27 will have a close - to - milne " epoch of finite duration . ( 120,0)(0,0 ) ( 41,64)(0,0)[b]@xmath81 ( 85,48)(0,0)[b]@xmath109 ( 16,47)(0,0)[b]flat fl , @xmath27 ( 87,4)(0,0)[b]open fl , @xmath110 ( 39,11)(0,0)[b]f ( 30,27)(0,0)[b]@xmath111 ( 38,53)(0,0)[b]ds ( 80,11)(0,0)[b]m ( 92,14)(0,0)[b]@xmath61 when one regards the fl models as a subset of the sh models , a complication arises : a specific fl model is represented by infinitely many orbits in the reduced hubble - normalized sh state space . this multiple representation occurs because a flat fl model admits a @xmath68 of bianchi type i and one of bianchi type vii@xmath112 , while an open fl model admits a @xmath68 of bianchi type v and one of bianchi type vii@xmath113 , for any @xmath114 ( see ellis & maccallum 1969 ) . the state space can be described by the variables @xmath115 where @xmath116 here @xmath117 is the group parameter and @xmath118 and @xmath119 are spatial curvature variables defined by @xmath120 the state space , which we will denote by @xmath121 , is shown in figure [ fig3.2 ] . ( 170,0)(0,0 ) ( 42,70)(0,0)[b]@xmath81 ( 115,63)(0,0)[b]@xmath122 , @xmath123 ( 102,72)(0,0)[b]@xmath109 , @xmath124 ( 80,64)(0,0)[b]@xmath109 ( 120,18)(0,0)[b]@xmath109 , @xmath123 ( 79,3)(0,0)[b]@xmath61 ( 39,20)(0,0)[b]f ( 38,58)(0,0)[b]ds ( 113,54)(0,0)[b]@xmath119 ( 82,15)(0,0)[b]m@xmath113 ( 61,6)(0,0)[b]m ( 101,25)(0,0)[b]m@xmath113 ( 120,0)(0,0 ) ( 41,62)(0,0)[b]@xmath81 ( 38,10)(0,0)[b]f ( 79,3)(0,0)[b]flat fl , @xmath110 ( 70,55)(0,0)[b]de sitter ( 33,23)(0,0)[b]@xmath111 ( 17,36)(0,0)[b]flat fl , @xmath27 ( 37,50)(0,0)[b]ds ( 93,11)(0,0)[b]@xmath119 we see that the state space is the union of a one - parameter family of leaves " , and that the milne model is represented by a line of equilibrium points @xmath125 , one in each leaf . the family of orbits in one leaf with @xmath114 are physically equivalent to those in another . the orbits in the leaf with @xmath126 , and hence @xmath127 , are multiple representations of flat fl models with @xmath123 or @xmath128 . in particular , the orbit with @xmath123 , @xmath127 is an unbounded orbit that describes the flat fl model , and the orbit with @xmath127 , @xmath109 is an unbounded orbit that describes the de sitter model . the state space @xmath127 is shown in figure [ fig3.3 ] . the fl state space @xmath121 is an invariant subset of the sh state space @xmath93 , defined by the requirement that the shear of the congruence normal to the @xmath68 orbits is zero . it is natural to regard an sh model as being close to fl during some epoch if its orbit lies in a sufficiently small neighbourhood of the fl state space @xmath121 . the hubble - normalized weyl tensor is unbounded in such a neighbourhood , however , with the result that close - to - fl models can have large hubble - normalized weyl curvature . in particular there are models with @xmath129 , @xmath130 and @xmath105 large ( close to flat fl ) and models with @xmath131 , @xmath130 and @xmath105 large ( close to de sitter ) , during some epoch . it is also of interest to consider the connection with perturbed fl models . are orbits in a neighbourhood of @xmath121 described accurately by solutions of the linearized efe , or do nonlinear effects come into play ? consideration of sh cosmologies of bianchi type vii@xmath112 shows that the latter occurs . linearizing the hubble - normalized evolution equations about the fl invariant set @xmath123 and @xmath127 gives @xmath132 while an analysis of the exact equations for large @xmath133 gives @xmath134 ( wainwright 1999 ) . thus for this class of models , linearization does not give the correct dynamics . it would be of interest to extend this analysis to @xmath70 cosmologies , i.e. consider @xmath121 as an invariant set of the state space of @xmath70 cosmologies and analyze the dynamics of close - to - fl cosmologies in this context . it would be necessary to determine all representations of the fl models within the @xmath70 state space , which would give an invariant set larger than @xmath121 . the term isotropization " is used in three different contexts in cosmology : * asymptotic isotropization into the future , * isotropic initial singularity , * intermediate isotropization . we discuss these dynamic phenomena in turn . a well known result of wald 1983 states that any sh cosmology with zero or negative spatial curvature ( i.e. @xmath92 ) approaches the de sitter solution in the sense that @xmath135 wald proved this result without making specific assumptions about the matter / energy distribution . more detailed analyses of sh cosmologies with a tilted perfect fluid ( goliath & ellis 1999 ) revealed that a bifurcation occurs at @xmath136 : if @xmath137 the peculiar velocity @xmath48 of the fluid does not tend to zero . specifically , @xmath138 , where @xmath139 if @xmath140 and @xmath141 if @xmath142 . this behaviour occurs for the following reason . for the de sitter solution the reduced state vector @xmath67 in ( [ y ] ) is given by n_=0,a_=0,_=0 , _ m=0,_=1 , with @xmath143 subject to the evolution equation @xmath144 in the reduced hubble - normalized state space the de sitter solution is thus represented by a set of equilibrium points , given by ( [ ds_full ] ) and with @xmath143 subject to @xmath145 , or @xmath146 ( a point and a sphere ) if @xmath147 , and subject to @xmath148 ( a solid sphere ) if @xmath140 . the limiting behaviour of @xmath48 determines the future attractor in the reduced hubble - normalized state space , as shown in table [ tab1 ] . the future attractor for sh cosmologies with @xmath27 forms the basis for the description of @xmath70 cosmologies that are future asymptotic to the de sitter solution , as we now explain . the frame variables @xmath66 satisfy equation ( [ bsys ] ) , which , when specialized to the de sitter equilibrium solution ( [ ds_full ] ) can be solved to give @xmath149 where the @xmath150 are constants . it follows that _ t+ e_^i = 0 . in addition the constraints imply that the spatial hubble gradient is zero ( @xmath151 ) . thus in the hubble - normalized @xmath70 state space , the de sitter solution is described by orbits that are future asymptotic to a de sitter equilibrium point on the silent boundary . @xmath70 cosmologies that are future asymptotic to the de sitter solution are described within the hubble - normalized state space as follows . along a fixed timeline , the evolution is described by an orbit that is future asymptotic to a de sitter equilibrium point on the silent boundary . the specific equilibrium point will depend on the equation of state parameter @xmath152 according to table [ tab1 ] . the idea is that the evolution is described by an orbit that is attracted to the silent boundary , i.e. ( [ ds_silent ] ) holds , and within the silent boundary , the evolution is governed by the sh evolution equations , as described in section [ sec_silent_b ] . the asymptotic dependence of the hubble - normalized variables for this class of models has been determined by systematically integrating the evolution equations ( lim 2004 ) . the fact that the orbit is asymptotic to the silent boundary means that the spatial dependence is asymptotically unrestricted , which manifests itself in the appearance of an arbitrary 3-metric in the asymptotic expansion of the spacetime metric ( lim 2004 ) . the issue of how large is the class of @xmath70 cosmologies that are future asymptotic to de sitter has not been completely resolved . for recent progress we refer to rendall 2003 . the second type of asymptotic isotropization , namely isotropization at the initial singularity , arises in connection with the notion of _ quiescent cosmology _ ( cf . , e.g. , barrow 1978 ) , which provides an alternative to cosmic inflation . the idea is that , due to entropy considerations on a cosmological scale , a suitable initial condition for the universe is that the weyl curvature should be zero ( or at least dynamically unimportant ) at the initial singularity ; this is the _ weyl curvature hypothesis _ , proposed by penrose 1979 ( see page 630 ) . this hypothesis leads to the notion of an _ isotropic initial singularity_. recently , khalatnikov 2003 ( see page 3 ) have suggested that a solution with an isotropic singularity could represent an _ intermediate _ asymptotic regime during expansion of the universe after inflation . cosmological initial singularities of this type first arose as a special case in the general analysis of initial singularities performed by lifshitz & khalatnikov 1963 ( see page 203 ) . this type of initial singularity was also encountered in the work of eardley 1972 ( see page 101 ) . subsequently , goode & wainwright 1985 gave a formal definition of an isotropic initial singularity , using a conformally related metric , and derived various properties . heuristically , the model approaches the flat fl model near the singularity : @xmath153 for sh models , the occurrence of an isotropic singularity is characterized by the orbit in the reduced hubble - normalized state space being past asymptotic to the flat fl equilibrium point f. in other words , the orbit of a sh model with an isotropic singularity lies in the unstable manifold of the flat fl equilibrium point , which confirms that isotropic singularities occur only for a set of initial conditions of measure zero . the above behaviour forms the basis for the description of @xmath70 cosmologies that admit an isotropic singularity , as we now explain . the frame variables @xmath66 satisfy equation ( [ bsys ] ) , which , when specialized to the flat fl equilibrium point can be solved to give @xmath154 where the @xmath150 are constants . it follows that @xmath155 assuming @xmath156 . in addition , the constraints imply that the spatial hubble gradient is zero ( @xmath151 ) . thus in the hubble - normalized @xmath70 state space , the flat fl model is described by an orbit that is past asymptotic to an equilibrium point on the silent boundary . @xmath70 cosmologies that have an isotropic singularity are described as follows . along a fixed timeline , the evolution is described by an orbit that is past asymptotic to the flat fl equilibrium point on the silent boundary . the asymptotic time dependence for this class of models has been determined by systematically integrating the evolution equations ( lim 2004 ) . the fact that the orbit is past asymptotic to the silent boundary means that the spatial dependence is asymptotically unrestricted , which manifests itself in the appearance of an arbitrary 3-metric in the asymptotic expansion of the spacetime metric ( see lim 2004 ) . the flat fl equilibrium point f , given by ( [ fl ] ) , is a saddle point in the reduced hubble - normalized state space for sh cosmologies . thus , for any @xmath157 there is a family of orbits , corresponding to an open set of initial conditions , that pass through an @xmath158-neighbourhood of f but are not asymptotic to f. for the corresponding cosmological models there will be a finite time interval during which the rate of expansion of the model is highly isotropic ( i.e. @xmath159 ) and hence compatible with observational constraints . this behaviour is called _ intermediate isotropization _ , and occurs in models of all bianchi types except bianchi type i , in which case f is a sink .. ] at this stage , intermediate isotropization for @xmath70 cosmologies has yet to be investigated in detail . the largest class of cosmologies for which there is a comprehensive and quite detailed knowledge of the dynamics , is the class of nontilted sh cosmologies with @xmath160 and equation of state @xmath161 , with @xmath156 . the reduced hubble - normalized state space @xmath93 is an unbounded subset of @xmath162 . this class of cosmologies has two subsets referred to as class a and class b in the classification of ellis & maccallum 1969 . we focus on the class a models , since they display a wider range of dynamical phenomena , while at the same time permitting a simpler choice of frame . in describing the dynamics of this class of cosmologies , it is helpful to make use of the hierarchy of invariant subsets of lower dimension that arise from specializing the matter / energy content and the bianchi type , vii@xmath112 , viii and ix , with the last two being of maximum generality . we exclude bianchi ix because models of this type can recollapse . ] . we list these invariant subsets in table [ tab_inv ] . we will discuss various dynamical phenomena , indicating how likely it is that each will occur . we say that a dynamical behaviour is * _ generic _ , if it occurs for all initial conditions except for a set of measure zero , * _ typical _ , if it occurs for a set of initial conditions of positive measure , whose complement is also of positive measure . * _ special _ , it it occurs for a set of initial conditions of measure zero . one can say that generic behaviour has probability one , typical behavior has probability between zero and one , and special behaviour has probability zero . we now list the various dynamical phenomena in table [ tab_phe ] and discuss each in turn . within @xmath93 , the mixmaster dynamics is described by a past attractor , which is the union of the kasner circle of equilibrium points and the vacuum bianchi ii orbits , first described by ma & wainwright 1992 . the attractor contains infinite heteroclinic sequences , i.e. infinite sequences of kasner points joined by vacuum bianchi ii orbits . a generic orbit that is asymptotic to the attractor shadows one of these heteroclinic sequences . the evolution of the corresponding cosmological model towards the initial singularity is thus a non - terminating sequence of quasi - equilibrium kasner states , punctuated by increasingly brief curvature transitions . the early work on the attractor was based solely on heuristic local stability arguments and numerical simulations . recently , however , ringstrm 2001 proved the existence of the past attractor , using the hubble - normalized evolution equations . this result may have wider significance in view of the conjectures about the local past attractor for @xmath70 cosmologies that have recently been made ( uggla 2003 and section [ sec_vac_num ] ) . as indicated in table [ tab_phe ] , mixmaster dynamics are generic . examples of special behaviour are provided by models which are past asymptotic to a specific kasner solution and by models with an isotropic singularity . within @xmath93 , inflationary isotropization is simply described by the fact that the de sitter equilibrium point ( [ ds ] ) , which describes the de sitter solution , is a global sink . ] , and hence forms the future attractor in the state space . this behaviour is generic , and special behaviour is restricted to the invariant set @xmath123 . the dimensionless scalars satisfy @xmath163 the asymptotic time dependence can be obtained from lim 2004 and rendall 2003 . generic orbits in the invariant subset @xmath123 escape to infinity , and satisfy _ this result holds for bianchi viii orbits if the equation of state satisfies @xmath164 ( horwood & wainwright 2004 ) and for bianchi vii@xmath112 orbits of @xmath165 ( wainwright 1999 ) . in other words the weyl curvature is dynamically dominant at late times . from a mathematical perspective this result means that the invariant subset @xmath123 does not admit a future attractor . despite this fact , one can still determine the asymptotic dependence of the state vector as @xmath166 ( horwood & wainwright 2004 ) . the asymptotic behaviour in the invariant set @xmath123 as @xmath166 , as epitomized by ( [ weyl_infty ] ) , has a significant effect on the dynamics in the full state space @xmath128 , in that _ it approximates the intermediate dynamics _ of a typical class of models with @xmath128 . one can think of the asymptotic state of models with @xmath123 acting as a _ generalized saddle _ in the full state space , temporarily attracting orbits but eventually repelling them . in summary , for a typical class of models with @xmath128 there will be a finite epoch during which weyl dominance occurs ( i.e. @xmath167 ) . a cosmological model undergoes intermediate isotropization with @xmath169 if there is a finite epoch during which the model is close to the flat fl model . this behaviour is created by the local stability properties of the flat fl model . we have seen that the flat fl model has a multiple representation , as the flat fl equilibrium point f and as an unbounded orbit ( see section [ sec_close_to__fl ] , in particular , figure [ fig3.3 ] ) . orbits that are future asymptotic to f ( i.e. orbits that form the two dimensional stable manifold of f ) satisfy _ t+=0 , _ t+=0 . on the other hand orbits that are future asymptotic to the unbounded orbit satisfy invariant set with @xmath123 . ] _ t+= l & + + & we say that models that satisfy ( [ lim_fl ] ) are _ asymptotic to flat fl in the strong sense _ , while those satisfy ( [ lim_fl_infty ] ) are _ asymptotic to flat fl in the weak sense_. both of these asymptotic states can act as ( generalized ) saddles in the full state space , leading to intermediate isotropization with @xmath168 . thus intermediate isotropization with @xmath168 in the full state space is a result of the occurrence of asymptotic isotropization with @xmath170 in lower dimensional invariant sets . ( 120,0)(0,0 ) ( 27,75)(0,0)[b]@xmath81 ( 23,25)(0,0)[b]f ( 7,1)(0,0)[b]@xmath97 ( 22,64)(0,0)[b]ds ( 117,24)(0,0)[b]@xmath119 intermediate isotropization in the weak sense with @xmath168 is illustrated schematically in figure [ fig_saddle2 ] . an orbit that is past asymptotic to a kasner equilibrium point @xmath97 spirals around the flat fl orbit , creating an epoch of intermediate isotropization with @xmath168 , before approaching the de sitter equilibrium point . within @xmath93 , a cosmological model with an isotropic singularity is simply described by an orbit that is past asymptotic to the flat fl equilibrium point ( [ fl ] ) , i.e. an orbit that lies in the unstable manifold of this equilibrium point . recall that this manifold is of dimension 4 ( see table [ tab_dim1 ] ) . it follows that @xmath171 for these models . the asymptotic time dependence can be obtained from lim 2004 . we now describe the attractor in the hubble - normalized @xmath70 state space , following the discussion in uggla 2003 , but using a different choice of gauge . one of the goals of that paper was to give a precise statement of the so - called bkl conjecture , by describing the past attractor in the hubble - normalized state space . _ the bkl conjecture . _ for almost all cosmological solutions of einstein s field equations , a spacelike initial singularity is _ vacuum - dominated , local and oscillatory_. for cosmological models with a perfect fluid matter source , the phrase vacuum - dominated , " or , equivalently , matter is not dynamically significant , " is taken to mean that the hubble - normalized matter density @xmath172 tends to zero at the initial singularity . one might then expect that the past attractor for cosmological models with a perfect fluid matter source would be the same as the past attractor for an idealized vacuum cosmological model . it turns out ( uggla 2003 ) that this expectation is unwarranted the peculiar velocity @xmath143 plays a role in determining the past attractor . nevertheless , partly in the interests of simplicity and also because the numerical simulations that have been recently performed ( garfinkle 2003 ) have been for vacuum models , we shall restrict our discussion here to the vacuum case . in terms of hubble - normalized variables and the separable volume gauge , the spacelike initial singularity in a @xmath70 cosmology is approached as @xmath173 . we now define a _ silent initial singularity _ to be one which satisfies @xmath174 more precisely , we require that ( [ silent ] ) and ( [ silent2 ] ) are satisfied along typical timelines of @xmath47 . one might initially think that the condition ( [ silent2 ] ) is a consequence of ( [ silent ] ) , since @xmath175 however the analysis of gowdy solutions with so - called spikes shows that the partial derivatives @xmath176 can diverge as @xmath173 . thus the requirement ( [ silent2 ] ) demands that @xmath66 tend to zero sufficiently fast . the asymptotic analysis of the gowdy solutions shows that ( [ silent2 ] ) is satisfied along typical timelines even when spikes occur ( see uggla 2003 , section iv.a ) . in more general @xmath75 cosmologies , however , the validity of ( [ silent2 ] ) is still open and further work is needed . we now refer to uggla 2003 , pages 910 , for further evidence to justify making the following conjecture . * conjecture 1 * : for almost all cosmological solutions of einstein s field equations , a spacelike initial singularity is silent . proving this conjecture entails establishing the limits ( [ silent])([silent2 ] ) . we think of the evolution of the hubble - normalized state vector @xmath177 _ for fixed @xmath12 _ , as being described by an orbit in a finite - dimensional hubble - normalized state space . as @xmath173 , this orbit will be asymptotic to a local past attractor @xmath178 , which in accordance with the definition of silent initial singularity , will be contained in the silent boundary , defined by ( [ silent_cond ] ) . the past attractor is based on the kasner solutions . we have seen that in the reduced hubble - normalized state space for sh models , the kasner solutions are represented by a sphere of equilibrium points . in the full @xmath70 state space , these solutions are represented by orbits that are past asymptotic to kasner equilibrium points on the silent boundary . thus we focus our attention on the _ kasner sphere @xmath97 on the silent boundary _ , defined by @xmath179 in addition to ( [ silent_cond ] ) . each non - exceptional , where @xmath180 denotes tracefree symmetrization . ] equilibrium point on the kasner sphere has one negative eigenvalue and hence has a one - dimensional unstable manifold into the past , which represents a sh vacuum solution of bianchi type ii on the silent boundary . we refer to these orbits as _ curvature transitions _ , since on each such orbit one degree of freedom of the spatial curvature is activated . that one diagonal component of @xmath181 is unstable into the past . ] these curvature transitions satisfy @xmath182 corresponding to the fact that two eigenvalues of @xmath181 are zero . in addition @xmath183 and the gauss constraint simplifies to @xmath184 we now make our second conjecture , motivated by our knowledge of sh dynamics . * conjecture 2 * : the local past attractor @xmath178 for vacuum @xmath70 cosmologies with a silent initial singularity is @xmath185 where @xmath97 is the kasner sphere and @xmath186 is the set of all curvature transitions in the silent boundary . establishing this conjecture entails proving the following limits : @xmath187=0.\end{gathered}\ ] ] the fact that the sequence of transitions is non - terminating implies that @xmath188 do not exist . however it follows from the gauss constraint ( [ cgauss ] ) and equations ( [ silent ] ) , ( [ silent2 ] ) and ( [ n2 ] ) that @xmath189=1.\ ] ] in addition it follows from one of the other constraints in ( [ esys ] ) , the codacci constraint , and the limits ( [ silent ] ) , ( [ silent2 ] ) and ( [ n2 ] ) that _ t- ^ n _ _ ^=0 . we thus expect that in a shear eigenframe , @xmath181 will be asymptotically diagonal " as @xmath14 . the geometrical interpretation of the asymptotic evolution is as follows : the orbit describing the evolution of the state vector @xmath177 for fixed @xmath12 approaches the kasner sphere @xmath97 in the silent boundary and then shadows increasingly closely an infinite sequence of curvature transitions joining kasner equilibrium points . some numerical simulations of the past asymptotic behaviour of vacuum @xmath70 cosmologies using the hubble - normalized evolution equations ( [ bsys])([esys ] ) have been recently performed by garfinkle 2003 . he found it convenient to use the deceleration parameter as a dynamical variable , with the defining equation ( [ q ] ) forming an additional constraint . the evolution equation for @xmath51 has the form of a quasilinear diffusion equation , and it is the only evolution equation to contain second order spatial derivatives . the hubble - normalized evolution equations were solved numerically on a 3-torus , i.e. with periodic boundary conditions , using a crank - nicholson scheme . the initial conditions were chosen to be of the form @xmath190 where @xmath41 , @xmath191 are arbitrary constants . the gauss constraint gives a nonlinear elliptic pde for @xmath192 : @xmath193\ , \ ] ] where @xmath194 is the euclidean laplacian in @xmath195 , @xmath196 and @xmath197 . this pde has to be solved numerically on a 3-torus to obtain explicit initial conditions . the remaining constraints are satisfied identically . note that the above choice of @xmath66 corresponds to a conformally flat initial 3-metric . the numerical simulations show a brief transient epoch after which a sequence of curvature transitions takes place , following the usual bkl transition law , which we now describe . the kasner exponents @xmath198 are related to the shear variables @xmath96 in a shear eigenframe as follows : @xmath199 the exponents satisfy @xmath200 and thus can be expressed in terms of a single parameter @xmath201 according to @xmath202 the transition law between kasner states can then be written in the form @xmath203 the numerical simulations also suggest that asymptotically @xmath96 and @xmath181 have a common eigenframe in accordance with ( [ eigen ] ) . the figures show the results of a numerical simulation with @xmath204 initially , and then becoming negative , i.e. evolution towards the initial singularity ( @xmath22 ) . figures [ fig_u1]a and [ fig_u2]a show the transitions of the kasner parameter @xmath201 along two different timelines and figures [ fig_u1]b and [ fig_u2]b show the diagonal values of @xmath181 in the common asymptotic eigenframe , along the same timelines . overall , the simulations provide support for the limits ( [ silent ] ) and ( [ silent2 ] ) , and ( [ a])([eigen ] ) along typical timelines . the numerical simulation is incomplete in two ways . first , the duration of the kasner epochs in the simulation is far too short , due to the fact that the decay of the eigenvalues of @xmath181 to increasingly small values , which determines when the curvature transitions occur , is not described sufficiently accurately . the reason for this difficulty is that an arbitrarily chosen fermi - propagated frame is not asymptotic to the eigenframe of @xmath181 as @xmath22 . second , the simulation does not have high enough numerical resolution to correctly simulate spiky structures . these structures are known to develop in @xmath75 cosmologies , on approach to the singularity ( berger & moncrief 1993 ) , and we expect that similar structure will develop in @xmath70 cosmologies . from our experience with numerical simulations of @xmath75 cosmologies , these spiky structures occur within the particle horizon of timelines where one of the eigenvalues of @xmath181 crosses zero ( lim 2004 ) . as the particle horizon shrinks into the past , increasingly high numerical resolution is needed to simulate the spiky structures . it may be prohibitively expensive to perform such simulations in @xmath70 cosmologies . the principal deficiency of hubble - normalized variables is that they break down in an expanding cosmological model that reaches a maximum state of expansion ( @xmath205 ) , as is possible in an fl model with positive spatial curvature . this difficulty can be addressed , at least for fl models and sh models of bianchi type ix , by using a modified normalization factor first proposed by uggla ( we section 8.5.2 ) . in the interests of simplicity we introduce the modified normalization procedure for the class of fl models with positive spatial curvature . for these models the normalization factor is defined by @xmath206 where @xmath63 is the curvature scalar of the metric induced on the hypersurfaces @xmath55 . since @xmath207 , this factor is positive throughout the evolution , including at the time of maximum expansion @xmath205 . dimensionless variables are defined in the usual way , using @xmath208 instead of @xmath20 : @xmath209 it follows that @xmath210 introducing a time variable @xmath211 according to @xmath212 one obtains the following evolution equations for @xmath213 and @xmath214 : @xmath215 where @xmath216 we note that @xmath217 is related to the usual deceleration parameter @xmath51 according to @xmath218 whenever @xmath219 . the state space is bounded , being defined by the inequalities @xmath220 and is shown in figure [ fig5.1 ] . the sign of @xmath213 determines whether the model is expanding ( @xmath221 ) or collapsing ( @xmath222 ) . the sign of @xmath217 determines whether the model is decelerating ( @xmath223 ) or accelerating . the fixed points of the evolution equations are as follows : @xmath224 the points @xmath225 and @xmath226 represent expanding models ( @xmath227 ) while @xmath228 and @xmath229 represent the time reversed models ( @xmath230 ) . the orbits @xmath231 represent models that expand from a big - bang singularity and then recollapse to a future singularity . the orbits @xmath232 represent models that expand indefinitely from a big - bang singularity , enter an accelerating epoch , and approach de sitter at late times . ( 120,0)(0,0 ) ( 95,70)(0,0)[b]@xmath226 ( 17,70)(0,0)[b]@xmath229 ( 35,70)(0,0)[b]contracting ( 75,70)(0,0)[b]expanding ( 53,33)(0,0)[b]e ( 105,28)(0,0)[b]@xmath233 ( 110,45)(0,0)[b]accelerating ( 110,20)(0,0)[b]decelerating ( 55,85)(0,0)[b]@xmath214 ( 55,0)(0,0)[b]@xmath234 ( 95,5)(0,0)[b]@xmath225 ( 95,0)(0,0)[b](flat fl , expanding ) ( 17,5)(0,0)[b]@xmath228 ( 15,0)(0,0)[b](flat fl , contracting ) ( 110,9)(0,0)[b]@xmath213 figure [ fig5.1 ] illustrates the central role played by the einstein static model within the class of fl cosmologies with positive and zero spatial curvature . the fixed point e is a saddle point whose stable and unstable manifolds are one - dimensional . the unstable manifold describes a one - parameter family of cosmologies that are past asymptotic to the einstein static model , and hence do not have an initial singularity . these cosmological models have recently been proposed by ellis & maartens 2004 as a possible description of the early universe , a so - called emergent universe . we note that this normalization procedure can be generalized to the class of nontilted sh cosmologies of bianchi type ix . the normalization factor is ( we , equation ( 8.6 ) ) @xmath235 where @xmath236 . the resulting state space , which gives a unified description of the dynamics of nontilted sh cosmologies of bianchi type ix , vii@xmath112 , ii and i , has not been explored in detail . in addition , a variation of this normalization has been given by heinzle 2004 , and has been used to give a detailed analysis of the nontilted sh cosmologies of bianchi type ix that are locally rotationally symmetric . the dynamical systems approach has two key attributes . firstly , it provides a geometric framework for analyzing the space of cosmological models . secondly , it provides a system of first order evolution equations , whose specific form depends on the choice of gauge . this system of equations is potentially useful for giving a rigorous analysis of the asymptotic dynamics of cosmological models , and also for performing numerical simulations . the utility of this approach has been amply demonstrated as regards the sh cosmologies , and there are indications that it may be equally useful as regards the @xmath70 cosmologies . this work has been influenced significantly by our collaboration with claes uggla and henk van elst . it is a pleasure to acknowledge the many discussions that we have had over the past few years . we also thank george ellis for his continued interest and encouragement , and david garfinkle for providing figures [ fig_u1 ] and [ fig_u2 ] , and for helpful discussions on the numerical simulations . j.w . expresses his appreciation to lars andersson and gregory galloway and the department of mathematics at the university of miami for their hospitality during the miami waves conference in january 2004 . this work was partially supported by a grant to j.w . from the natural sciences and engineering research council of canada . 99 l. andersson , h. van elst & c. uggla 2004 , gowdy phenomenology in scale - invariant variables , cqg * 21 * , s29s57 j.m . bardeen 1980 , gauge invariant cosmological perturbations , prd * 22 * , 1882905 j.d . barrow 1978 , quiescent cosmology , _ nature _ * 272 * , 211 b.k . berger 2004 , hunting local mixmaster dynamics in spatially inhomogneous cosmologies , cqg * 21 * , s81s96 b.k . berger and v. moncrief 1993 , numerical investigation of cosmological singularities , prd * 48 * , 467687 a.a . coley 2003 , _ dynamical systems and cosmology _ , kluwer academic publishers a.a . coley , he & lim 2004 , isotropic singularity in inhomogeneous brane cosmological models , cqg * 21 * , 131142 a.a . coley & s. hervik 2004 , bianchi cosmologies : a tale of two tilted fluids , cqg * 21 * , 4193208 a.a . coley & r.j . van den hoogen 2000 , the dynamics of multi - scalar field cosmological models and assisted inflation , prd * 62 * , 023517 c.b . collins 1971 , more qualitative cosmology , _ commun . math . * 23 * , 13756 d. eardley , e.p . liang & r.k . sachs 1972 , velocity - dominated singularities in irrotational dust cosmologies , _ j. math _ * 13 * , 99107 g.f.r . ellis 1967 , dynamics of pressure - free matter in general relativity , _ j. math . * 8 * , 117194 g.f.r . ellis 2004 , the space of cosmological space - times , _ preprint _ g.f.r . ellis & m. bruni 1989 , covariant and gauge - invariant approach to cosmological density fluctuations , prd * 40 * , 180418 g.f.r . ellis , j. hwang & m. bruni 1989 , covariant and gauge - independent perfect - fluid robertson - walker perturbations , prd * 40 * , 181926 g.f.r . ellis & r. maartens 2004 , the emergent universe : inflationary cosmology with no singularity , cqg * 21 * 22332 g.f.r . ellis & m.a.h . maccallum 1969 , a class of homogeneous cosmological models , _ commun . * 12 * , 10841 h. van elst & c. uggla 1997 , general relativistic 1 + 3 orthonormal frame approach , cqg * 14 * , 2673951 h. van elst , c. uggla & j. wainwright 2002 , dynamical systems approach to @xmath75 cosmology , cqg * 19 * , 5182 d. garfinkle 2003 , numerical simulations of generic singularities , gr - qc/0312117 d. garfinkle & m. weaver 2003 , high velocity spikes in gowdy spacetimes , prd * 67 * 124009 m. goliath & g.f.r . ellis 1999 , homogeneous cosmologies with a cosmological constant , prd * 60 * , 023502 s.w . goode & j. wainwright 1985 , isotropic singularities in cosmological models , cqg * 2 * , 99115 j.m . heinzle , n. rohr & c. uggla 2004 , matter and dynamics in closed cosmologies , gr - qc/0406072 s. hervik 2004 , the asymptotic behaviour of tilted bianchi type vi@xmath112 universes , cqg * 21 * , 230117 c.g . hewitt , r. bridson & j. wainwright 2001 , the asymptotic regimes of tilted bianchi ii cosmologies , grg * 33 * , 6594 d. hobill , a. burd & a.a . coley ( ed ) 1994 , _ deterministic chaos in general relativity _ , plenum press j.t . horwood & j. wainwright 2004 , asymptotic regimes of magnetic bianchi cosmologies , grg * 36 * , 799823 i.m . khalatnikov , a.yu . kamenshchik , m. martellini , a.a . starobinsky 2003 , quasi - isotropic solution of the einstein equations near a cosmological singularity for a two - fluid cosmological model , _ journal of cosmology and astroparticle physics _ * 03 * , 001 a.r . king & g.f.r . ellis 1973 , tilted homogeneous cosmological models , _ commun . math . phys . _ * 31 * , 20942 j. kristian & r.k . sachs 1966 , observations in cosmology , apj * 143 * , 37999 v.g . leblanc 1998 , bianchi ii magnetic cosmologies , cqg * 15 * , 160726 e.m . lifshitz & i.m . khalatnikov 1963 , investigations in relativistic cosmology , _ adv . * 12 * , 185249 w.c . lim 2004 , _ the dynamics of inhomogeneous cosmologies _ , phd thesis , university of waterloo w.c . lim , h. van elst , c. uggla & j. wainwright 2004 , asymptotic isotropization in inhomogeneous cosmology , prd * 69 * , 103507 r. maartens , g.f.r . ellis & w.r . stoeger 1995 , limits on anisotropy and inhomogeneity from the cosmic background radiation , prd * 51 * , 152535 r. maartens , g.f.r . ellis & w.r . stoeger 1996 , anisotropy and inhomogeneity of the universe from @xmath237 , _ astron . astrophys . _ * 309 * , l7l10 r. penrose 1979 , singularities and time - asymmetry . in _ general relativity : an einstein centenary survey _ , eds s.w . hawking & w. israel . cambridge university press . rendall 2003 , asymptotics of solutions of the einstein equations with positive cosmological constant , gr - qc/0312020 a.d . rendall & m. weaver 2001 , manufacture of gowdy spacetimes with spikes , cqg * 18 * , 295976 h. ringstrm 2001 , the bianchi ix attractor , _ annales henri poincar _ * 2 * , 405500 c. uggla , h. van elst , j. wainwright & g.f.r . ellis 2003 , past attractor in inhomogeneous cosmology , prd * 68 * , 103502 j. wainwright , m.j . hancock & c. uggla 1999 , asymptotic self - similarity breaking at late times in cosmology , cqg * 16 * , 257798 j. wainwright & g.f.r . ellis ( ed ) 1997 , _ dynamical systems in cosmology _ , cambridge university press r.m . wald 1983 , asymptotic behaviour of homogeneous cosmological models in the presence of a positive cosmological constant , prd * 28 * , 211820

in this paper , we construct a combinatorial algebra of partial isomorphisms that gives rise to a `` projective limit '' of the centers of the group algebras . it allows us to prove a-analogue of a theorem of farahat and higman regarding products of conjugacy classes of permutations . = 1 ' '' '' ' '' ''

1303.4313

in this paper , @xmath2 is a positive integer ; @xmath3 is a prime power ; @xmath4 is a finite field with @xmath3 elements ; and @xmath1 is the group of invertible @xmath5 matrices with coefficients in @xmath4 , or , equivalently , the group of linear isomorphisms of the @xmath4-vector space @xmath6 . it will be convenient to write @xmath7 for the composition of isomorphisms @xmath8 ( beware of the order of composition ) . the matrix of an isomorphism @xmath9 between two @xmath10-vector spaces and with respect to two bases @xmath11 of @xmath12 and @xmath13 of @xmath14 is @xmath15 the vectors being written in columns ; then @xmath16 . if @xmath17 , we shall be interested in the group algebra @xmath18 , and more precisely in its center @xmath19 . a linear basis of @xmath19 is the set of conjugacy classes of @xmath20 , viewed as the formal sums of their elements . we want to address the following kind of problem : let @xmath21 be two elements of @xmath22 , and @xmath23 and @xmath24 the conjugacy classes of the diagonal matrices of size @xmath2 @xmath25 what is the expansion @xmath26 in conjugacy classes of the product @xmath27 computed in the center of the group algebra @xmath28 ? at first sight , this problem might seem to be not so difficult : since @xmath29 and @xmath30 each leave invariant a subspace of dimension @xmath31 , the product @xmath32 of two matrices conjugated to @xmath29 and @xmath30 has to leave invariant a subspace of dimension at least @xmath33 , and as we shall see in a moment , this only leaves a few possibilities for the conjugacy class @xmath34 of @xmath35 . moreover , by looking at how many ways the eigenspaces of the matrices can intersect , one can guess that the coefficients @xmath36 count configurations of subspaces in @xmath6 , and are therefore rational functions in @xmath3 and in its powers @xmath37 , up to @xmath38 . however , it is in practice quite hard to determine the set of possible conjugacy classes @xmath39 ; and also quite hard to compute the structure coefficients . the reader can already have a look at theorem [ productdegree1 ] , to realize that the result depends in particular on : * whether @xmath3 is even or odd ; * in the odd case , whether @xmath40 is a square in @xmath22 or not . also , it is _ a priori _ unclear that the coefficients @xmath36 are in fact all polynomials in @xmath41 , with rational coefficients independent of @xmath2 . one might understand intuitively why this is true from the previous informal discussion : every number of subspaces of fixed dimension @xmath42 in @xmath6 is a polynomial in @xmath41 ( of degree @xmath42 ) , and the same holds when counting `` finite - dimensional configurations '' . however , a formal and rigorous proof demands a lot of combinatorial preparations , especially if one wants to _ compute _ the actual polynomials . the author conjectured this polynomiality a few years ago , and this will be one of the major result of the paper . we shall prove it in the general setting where @xmath29 and @xmath30 are replaced by arbitrary matrices @xmath43 and @xmath44 that are completed by @xmath45 s on the diagonal to obtain matrices of size @xmath2 . we shall also provide a general framework which reduces all the computations to the case when @xmath46 , @xmath42 and @xmath47 being the sizes of @xmath43 and @xmath44 ( @xmath48 for @xmath49 and @xmath50 ) . in other words , knowing the structure coefficients @xmath51 , we shall have at the end of the paper an easy rule to compute all the coefficients @xmath52 . a similar and simpler problem has been solved for permutations by farahat and higman in @xcite , and forty years later , it was given a beautiful explanation by ivanov and kerov , see @xcite . fix two permutations @xmath53 and @xmath54 of size @xmath42 and @xmath47 ; we also assume @xmath53 and @xmath54 without fixed points in @xmath55\!\right]}$ ] and in @xmath56\!\right]}$ ] . for any @xmath57 , one can view @xmath53 and @xmath54 as elements of the symmetric group of order @xmath2 , denoted @xmath58 ; @xmath53 fixes the integers after @xmath42 and @xmath54 fixes the integer after @xmath47 . denote @xmath59 and @xmath60 the conjugacy classes of @xmath53 and @xmath54 in @xmath61 . then , there exists a unique finite set of permutations @xmath62 , and polynomials @xmath63 with rational coefficients and integer values at integers , such that @xmath64 using the well - known labeling of conjugacy classes of @xmath58 by integer partitions of size @xmath2 , one can of course restate this result with two partitions @xmath34 and @xmath65 of size @xmath42 and @xmath47 and without parts of size @xmath45 , and a finite set of partitions of size smaller than @xmath66 and again without parts of size @xmath45:[fhig ] @xmath67 the proof of ivanov and kerov uses the following idea : by manipulating _ partial permutations _ that are permutations with a distinguished support @xmath68\!\right]}$ ] , one can construct an inverse system of graded algebras @xmath69 such that : 1 . each algebra @xmath70 projects onto @xmath71 via a morphism @xmath72 ; 2 . each algebra has a basis @xmath73 labelled by partitions @xmath34 of size smaller than @xmath2 , with @xmath74 ; @xmath75 is an explicit polynomial in @xmath2 ; 3 . the bases @xmath76 are compatible , _ i.e. _ , if @xmath77 , then the map @xmath78 sends @xmath79 to @xmath76 . from this construction , an identity @xmath80 in the projective limit @xmath81 gives immediately the identity @xmath82 in all the centers of the group algebras @xmath71 ; this provides an easy proof of the theorem of farahat and higman . the same techniques and kind of results have been studied more recently by the author for product of geck - rouquier elements in the centers of the hecke algebras @xmath83 of the symmetric groups ( _ cf . _ @xcite ) ; and by o. tout in @xcite for product of classes in the algebras @xmath84 $ ] associated to the gelfand pairs @xmath85 . the objective of this paper is to follow the same program for general linear groups over finite fields ; unfortunately , in this setting , there are a lot of complications in comparison to the framework previously described . the naive idea is of course to replace partial permutations by _ partial isomorphisms _ ; but there are two major obstacles to this idea . 1 . for permutations , if @xmath86 with @xmath68\!\right]}$ ] of size @xmath42 , then there is a unique canonical way to see @xmath53 as an element of @xmath58 ; as explained before , one leaves the integers outside @xmath43 fixed by @xmath53 . the same can not be done in a canonical way for isomorphisms : if @xmath87 is an @xmath4-isomorphism of a subspace @xmath88 of dimension @xmath42 , then @xmath87 admits a lot of extensions to isomorphisms of @xmath6 , related to the choice of complement subspaces of @xmath12 inside @xmath6 . and there is no reason or possibility to distinguish a particular extension among those . 2 . knowing that , a natural thing to do is to take the _ mean of all possible extensions _ ; it seems then possible to define correctly the product of partial isomorphisms . but this still does not work : as we shall explain later ( see remark [ naive ] ) , the algebra that one obtains by this construction is not associative . the solution to these problems , and the main idea of this paper , is to define partial isomorphisms as _ pairs _ of isomorphisms between two subspaces . an alternative construction of partial bijections and partial isomorphisms has been proposed by g. olshanski in @xcite . let @xmath89 an inductive limit of finite groups , and @xmath90 a chain of subgroups of @xmath20 , such that @xmath91 commutes with @xmath92 for all @xmath2 . for instance one can take @xmath93\!\right]}}={\mathrm{id}}_{{\left[\!\left[}1 , n { \right]\!\right]}}\ } ; \\ \text{or}\,\,g(n)&={\mathrm{gl}}(n,{\mathbb{f}}_q)\qquad;\qquad k(n)=\{g \in { \mathrm{gl}}(\infty,{\mathbb{f}}_q)\,\,|\,\,g_{|({\mathbb{f}}_q)^n}={\mathrm{id}}_{({\mathbb{f}}_q)^n}\}.\end{aligned}\ ] ] in these situations , there is a structure of semigroup on the set of double cosets @xmath94 . for instance , when @xmath95 , @xmath96 can be identified with the set of bijections @xmath97 between two parts @xmath43 and @xmath44 of @xmath98\!\right]}$ ] , and the product is @xmath99 notice that the product @xmath100 written above is quite different from the product of partial permutations defined in @xcite , or from the product of partial bijections defined in @xcite . a similar construction of semigroup can be performed when @xmath101 . however , these semigroups ( or the sets of double cosets @xmath102 , studied in okounkov s thesis @xcite ) , do not really help for the specific problem that we are looking at , namely , the understanding of the structure coefficients of the group algebra centers @xmath103 : indeed , there is no natural morphism of semigroups @xmath104 . unfortunately , we do not see how to use _ semigroup algebras _ @xmath105 in order to construct a projective limit of the centers @xmath106 ; we shall however be satisfied with _ combinatorial algebras _ labelled by partial isomorphisms , without an underlying structure of semigroup on the partial isomorphisms . for finite general linear groups , the conjugacy classes are given by jordan s reduction of matrices . if @xmath107 is a monic polynomial , its jordan matrix is the @xmath108 matrix @xmath109 call partition of size @xmath42 a non - increasing sequence of positive integers @xmath110 with @xmath111 . a polypartition of size @xmath2 over the finite field @xmath4 is a family of partitions @xmath112 labelled by monic irreducible polynomials over @xmath4 , all different from @xmath113 , and such that @xmath114 to such a polypartition , we associate the block - diagonal matrix @xmath115 whose blocks are the jordan matrices @xmath116 with @xmath117\!\right]}$ ] and @xmath118\!\right]}$ ] . for instance , if @xmath119 , then @xmath120 jordan s reduction ensures that each conjugacy class of @xmath1 contains a unique matrix @xmath115 with @xmath121 polypartition of size @xmath2 over @xmath4 ; this result is a classical consequence of the classification of finitely generated modules over principal rings ( here , @xmath122 $ ] ) . thus , a linear basis of @xmath28 consists in the classes @xmath123 labelled by the set @xmath124 of these polypartitions . all the elements in @xmath123 have for characteristic and minimal polynomials @xmath125 on the other hand , it can be shown that [ cardinalconjugacy ] @xmath126 where @xmath127 ; @xmath128 is the pochhammer symbol @xmath129 ; and @xmath130 is the number of parts @xmath42 in a partition @xmath65 . this formula is proven by using hall s theory of modules over discrete valuation rings , see ( * ? ? ? * chapters 2 , 4 ) . consider the @xmath131-polypartition @xmath132 a representative of this conjugacy class in @xmath133 is the jordan matrix @xmath134 and the cardinality of the class is @xmath135 for a polypartition @xmath121 of size @xmath136 , denote @xmath137 the polypartition of size @xmath2 obtained by adding parts @xmath45 to the partition @xmath138 . this amounts to complete matrices with @xmath45 s on the diagonal in the bottom right corner . therefore , our initial problem can be reformulated and generalized as follows : fix two polypartitions @xmath139 and @xmath121 of size @xmath42 and @xmath47 . what is the expansion of @xmath140 in completed conjugacy classes @xmath141 in particular , what is the dependence in @xmath2 of the structure coefficients @xmath142 ? denote @xmath143 the set @xmath20 endowed with the opposite of the product of the group @xmath20 ; @xmath144 is then an isomorphism of groups between @xmath20 and @xmath143 . it will be extremely important in our discussion to see @xmath145 as the hecke algebra @xmath146 $ ] this fact is true for every finite group @xmath20 . consider indeed the double action of @xmath20 and @xmath143 on @xmath147 given by @xmath148 these actions correspond to the injective maps @xmath149 let us detemine the orbit of an element @xmath150 under the action on the left by @xmath20 and on the right by @xmath143 . since one can multiply on the right by @xmath151 to obtain @xmath152 , the orbit depends only on @xmath153 . then , if @xmath152 and @xmath154 are in the same orbit , there exists @xmath155 such that @xmath156 thus , @xmath157 , and we have proved that the orbit of @xmath150 consists in pairs @xmath158 such that @xmath153 and @xmath159 are conjugated . for a conjugacy class @xmath160 of @xmath20 , denote @xmath161.\ ] ] suppose that @xmath162 . then , a simple computation allows one to check that @xmath163 . therefore , the map @xmath164 & \to { \mathbb{z}}({\mathbb{c}}g ) \\ c_{\lambda } ' & \mapsto c_{\lambda}\end{aligned}\ ] ] realizes an isomorphism of commutative complex algebras . we are looking for a family of algebras @xmath165 with the following properties . we want them to form an inverse system of graded algebras @xmath166 so that in the end we will be able to look at the projective limit @xmath167 , which will allow us to make generic computations . then , we want these algebras to project onto the centers @xmath28 of the group algebras of the general linear groups over @xmath4 , so that these generic computations will turn into generic identities between conjugacy classes of isomorphisms . finally , we wish to construct the @xmath165 s in a fashion similar to @xcite , which is very combinatorial and natural . this means that we want to define the elements of @xmath165 as linear combinations of partial isomorphisms of @xmath6 , this notion staying for the moment vague and undefined . since @xmath168 can be seen as an hecke subalgebra of the algebra @xmath169 $ ] , the way to do it will be to do the same construction with algebras @xmath170 that project onto the group algebras @xmath169 $ ] . then , to obtain the @xmath165 s , we shall just look at invariant subalgebras of these @xmath170 s . our paper is therefore organized as follows : * in section [ algpartiso ] , we define partial isomorphisms of @xmath6 and their product , and we prove that we obtain indeed an algebra @xmath170 ( theorem [ mainan ] ) . the main difficulty is to see that the product is associative ; this is related to the properties of certain linear operators on the algebra , and the associativity in @xmath170 is essentially equivalent to the commutativity of the algebra formed by these operators . * in section [ constructions ] , we look at some invariant subalgebras @xmath171 , and we prove that they form an inverse system of graded commutative algebras , that project onto the centers @xmath172 . we deduce from it the result of polynomiality of the structure coefficients that we have evoked in this introduction ; see our main theorem [ generalfh ] . * finally , in section [ secdegree1 ] , we give the table of multiplication of all elements of degree @xmath45 in the projective limit @xmath167 . this gives in particular a concrete answer to the first problem stated in this paper . we conclude this introduction by two important remarks . in essence , this paper relies only on linear algebra and combinatorics over finite fields . however , some of the results and proofs will be written using the language of probability : so , we shall speak freely of probability laws , conditional laws , markov chains , _ etc_. , and we refer to @xcite for any detail on these notions . the reason is that the product of the algebra @xmath170 of partial isomorphisms that we shall study will be defined by taking averages of certain extensions of the partial isomorphisms . thus , discrete probability will provide in many situations a convenient way to describe and prove identities in @xmath170 . since we are only dealing with finite sets , for pure algebraists , every `` probabilistic '' statement can be replaced by a statement with a function on a finite set taking non - negative values that sum to @xmath45 . so probability is rather a commodity of language and a way of thinking than an absolute necessity in our reasonings . the reader may naturally ask what could be the uses of the theory of partial isomorphisms developed in this paper . we think of it as a first step in a much larger program , namely , the asymptotic representation theory of the finite general linear groups @xmath1 other important approaches appear in the papers @xcite . for symmetric groups @xmath58 , an important problem is to understand the asymptotic behavior of the random character value @xmath173 , where @xmath53 is a fixed permutation and @xmath174 is a random irreducible character of @xmath58 taken under some spectral measure , for instance the plancherel measure of the group . these random character values are important examples of random variables `` with a non - commutative flavour '' , and their asymptotic behavior is related to random matrix theory , exclusion processes of particules , free probability , and many other topics ; see for instance @xcite . the ivanov - kerov algebra of partial permutations has proven extremely useful in order to compute the moments of these random character values @xmath175 ; thus , they allowed to determine their asymptotic distribution . indeed , the computation of the moments @xmath176 $ ] is directly related to the computation of products and powers of conjugacy classes in symmetric group algebras . we refer in particular to @xcite , where such techniques and results are detailed . we hope to be able to follow the same program for finite linear groups @xmath177 . in particular , among the problems that we wish to solve after this paper , let us ask the following question . take the normalized trace @xmath178 of @xmath0 , which expands as linear combination of all irreducible normalized characters of the group : @xmath179 for @xmath87 fixed ( say , @xmath87 is a diagonal matrix @xmath29 ) and completed by @xmath45 s on the diagonal , we consider @xmath180 as a random variable under the probability measure @xmath181=\frac{(\dim v^{\bblambda})^{2}}{{\mathrm{card}}\ , { \mathrm{gl}}(n,{\mathbb{f}}_{q})},\ ] ] which is the plancherel measure of the group . what are the asymptotics of the law of @xmath182 ? to compute the moments of @xmath183 amounts to compute the powers of @xmath184 in the group algebras @xmath106 , where @xmath121 is the conjugacy class of @xmath87 ; whence the interest of our results for the asymptotic analysis of the random character values @xmath183 . the idea would then be to use these observables @xmath183 to get a better understanding of the properties of asymptotic concentration of the measures @xmath185 ; in particular , one should be able to recover this way the results of @xcite . to help the reader keep track of the notations , and of the various characters appearing as exponents or indices throughout the paper , we have listed hereafter the conventions that we shall use , see the following table . on the other hand , we recall the pochhammer symbol @xmath186 . in the following we shall try to write every enumeration as a polynomial in @xmath3 and in these pochhammer symbols in @xmath187 . @xmath188 [ cols="^ , < " , ] @xmath188 where by number of elements we mean the image by the linear map @xmath189 , or by the linear map @xmath190 . most manipulations are eased by dealing with the normalized versions of classes ( those with a @xmath191 ) . [ ziscommutative ] the algebra @xmath165 is commutative . ultimately the result comes from the commutativity of all the algebras @xmath192 with @xmath12 subspace of @xmath6 , but the verifications are not obvious at all . on the other hand , we really need to do them in order to obtain a formula for the product of two classes @xmath193 and @xmath194 ; so we can not skip this proof of commutativity . take two random independent subspaces @xmath12 and @xmath14 of dimension @xmath42 and @xmath47 inside @xmath6 . the law of @xmath195 is the case @xmath196 of lemma [ complicateproba ] : @xmath197 = q^{(k+l - m)(m - n)}\,\frac{(q^{-1})_{n - k}\,(q^{-1})_{n - l}\,(q^{-1})_{k}\,(q^{-1})_{l}}{(q^{-1})_{k+l - m}\,(q^{-1})_{n - m}\,(q^{-1})_{n}\,(q^{-1})_{m - k}\,(q^{-1})_{m - l}}.\ ] ] then , since the law of the pair @xmath198 is invariant by action by @xmath1 , @xmath199 is a random subspace with a law invariant by @xmath1 , so its law is a mixture of the uniform laws on random subspaces of fixed dimension @xmath200 . hence , @xmath201={\mathbb{p}}[m]\,{\mathbb{u}}_{m}[z]\quad\text{with } m=\dim z.\ ] ] corollary [ uniformproba ] then tells us that knowing @xmath202 , @xmath12 is uniformly distributed among spaces of dimension @xmath42 included into @xmath202 ; and knowing @xmath202 , @xmath14 is uniformly distributed among spaces of dimension @xmath47 included into @xmath202 . notice however that the joint law of @xmath198 is not the product of these uniform laws ; in other words , @xmath12 and @xmath14 are not independent conditionally to @xmath202 . one can rewrite @xmath193 as a multiple of the normalized sum @xmath203\,{\mathbb{u\!f}}_{k}[{\mathcal{f } } ] \,{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{j(\bbmu ) } { \mathrm{span}}({\mathcal{f } } ) ) } , \ ] ] where @xmath204 $ ] and @xmath205 $ ] are the uniform probabilities on free families of size @xmath206 . notice then that @xmath207\,{\mathbb{u\!f}}_{v}[{\mathcal{f } } ] \,{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{j(\bbmu ) } { \mathrm{span}}({\mathcal{f } } ) ) } \right)\\ & = \sum_{\substack{{\mathcal{e}}^{+},{\mathcal{f}}^{+}\text { free families}\\ { \mathrm{span}}({\mathcal{f}}^{+})=v^{+},\,\,{\mathrm{span}}({\mathcal{f}})=v } } { \mathbb{u\!f}}_{k^{+}}[{\mathcal{e}}^{+}]\,{\mathbb{u\!f}}_{v^{+},v}[{\mathcal{f}}^{+ } ] \,{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{k^{+}}}\!\!\rightleftarrows_{j(\bbmu\uparrow k^{+ } ) } { \mathrm{span}}({\mathcal{f}}^{+}))}\end{aligned}\ ] ] if @xmath208 have dimension @xmath42 and @xmath209 ; here @xmath210 $ ] ( resp . , @xmath211 $ ] ) is the uniform law on free families of @xmath6 such that @xmath212 ( resp . , such that @xmath213 and @xmath212 , with @xmath214 the family of the @xmath42 first vectors of @xmath215 ) . a product of two normalized sums @xmath216 and @xmath217 writes now as @xmath218\ , { ( { \mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{j(\bbmu ) } { \mathrm{span}}({\mathcal{f } } ) ) } * { ( { \mathrm{span}}({\mathcal{g } } ) \,\,^{i_{l}}\!\!\rightleftarrows_{j(\bbnu ) } { \mathrm{span}}({\mathcal{h}}))}\\ & = \sum_{\substack{\dim v = k \\ \dim w = l \\ { \mathcal{e}},{\mathcal{f}},{\mathcal{g}},{\mathcal{h}}\\ { \mathrm{span}}({\mathcal{f}})=v\\{\mathrm{span}}({\mathcal{g}})=w}}\binom { { \mathbb{u}}_{k}[v]\,{\mathbb{u}}_{l}[w]\,{\mathbb{u\!f}}[{\mathcal{e}},{\mathcal{h}}]\,{\mathbb{u\!f}}_{v}[{\mathcal{f}}]\,{\mathbb{u\!f}}_{w}[{\mathcal{g}}]\,\ , \times \qquad \qquad \qquad\qquad}{{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{j(\bbmu ) } { \mathrm{span}}({\mathcal{f } } ) ) } * { ( { \mathrm{span}}({\mathcal{g } } ) \,\,^{i_{l}}\!\!\rightleftarrows_{j(\bbnu ) } { \mathrm{span}}({\mathcal{h } } ) ) } } \\ & = \sum_{\substack{z , v , w \\ { \mathcal{e}},{\mathcal{f}},{\mathcal{g}},{\mathcal{h}}\\ { \mathrm{span}}({\mathcal{f}})=v\\{\mathrm{span}}({\mathcal{g}})=w}}\binom { { \mathbb{p}}[m]\ , { \mathbb{u}}_{m}[z]\,{\mathbb{p}}[(v , w)|z]\,{\mathbb{u\!f}}[{\mathcal{e}},{\mathcal{h}}]\,{\mathbb{u\!f}}_{v}[{\mathcal{f}}]\,{\mathbb{u\!f}}_{w}[{\mathcal{g}}]\,\ , \times \qquad\qquad \qquad\qquad}{\mathrm{r}_{v}^{v+w}{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{j(\bbmu ) } { \mathrm{span}}({\mathcal{f } } ) ) } \cdot \mathrm{l}_{w}^{v+w}{({\mathrm{span}}({\mathcal{g } } ) \,\,^{i_{l}}\!\!\rightleftarrows_{j(\bbnu ) } { \mathrm{span}}({\mathcal{h } } ) ) } } \\ & = \!\!\!\!\!\!\!\sum_{\substack{z , v , w \\ { \mathcal{e}}^{+},{\mathcal{f}}^{+},{\mathcal{g}}^{+},{\mathcal{h}}^{+}\\ { \mathrm{span}}({\mathcal{f}}^{+})=z,\,\,{\mathrm{span}}({\mathcal{f}})=v \\ { \mathrm{span}}({\mathcal{g}}^{+})=z,\,\,{\mathrm{span}}({\mathcal{g}})=w}}\!\!\!\!\!\!\!\!\binom { { \mathbb{p}}[m]\ , { \mathbb{u}}_{m}[z]\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]\,{\mathbb{p}}[v|z]\,{\mathbb{u\!f}}_{z , v}[{\mathcal{f}}^{+}]\,{\mathbb{p}}[w|(z , v)]\,{\mathbb{u\!f}}_{z , w}[{\mathcal{g}}^{+}]\,\ , \times}{{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{j(\bbmu \uparrow^{m } ) } { \mathrm{span}}({\mathcal{f}}^{+ } ) ) } \cdot { ( { \mathrm{span}}({\mathcal{g}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{j(\bbnu \uparrow^{m } ) } { \mathrm{span}}({\mathcal{h}}^{+ } ) ) } } \end{aligned}\ ] ] @xmath219\ , { \mathbb{u}}_{m}[z]\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]\,{\mathbb{u\!f}}_{z}[{\mathcal{f}}^{+}]\,{\mathbb{p}}[w|(z , v)]\,{\mathbb{u\!f}}_{z , w}[{\mathcal{g}}^{+}]\,\ , \times}{{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{p^{-1}}\!\!\rightleftarrows_{j(\bbnu\uparrow^{m})pj(\bbmu \uparrow^{m } ) } { \mathrm{span}}({\mathcal{h}}^{+ } ) ) } } \\ & = \!\!\!\!\!\!\!\sum_{\substack{z , v , w \\ { \mathcal{e}}^{+},{\mathcal{f}}^{+},{\mathcal{g}}^{+},{\mathcal{h}}^{+}\\ { \mathrm{span}}({\mathcal{f}}^{+})=z,\,\,{\mathrm{span}}({\mathcal{f}})=v \\ { \mathrm{span}}({\mathcal{g}}^{+})=z,\,\,{\mathrm{span}}({\mathcal{g}})=w}}\!\!\!\!\!\!\!\!\binom { { \mathbb{p}}[m]\ , { \mathbb{u\!f}}_{m}[z]\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]\,{\mathbb{u\!f}}_{z}[{\mathcal{f}}^{+}]\,{\mathbb{p}}[w|(z , v)]\,{\mathbb{u\!f}}_{z , w}[{\mathcal{g}}^{+}]\,\ , \times}{{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{p^{-1}j(\bbnu\uparrow^{m})pj(\bbmu \uparrow^{m } ) } { \mathrm{span}}({\mathcal{h}}^{+ } ) ) } } \end{aligned}\ ] ] where @xmath35 stands for the matrix @xmath220 . notice that this matrix @xmath35 is the only part depending on @xmath221 and @xmath222 in the last average of partial isomorphisms . therefore , the sum will be simplified if we are able to identify the law of @xmath35 under the probability @xmath223\,{\mathbb{p}}[w|(z , v)]\,{\mathbb{u\!f}}_{z , w}[{\mathcal{g}}^{+}]$ ] , which is a conditional probability depending on the random variable @xmath202 . we claim that this law is just the uniform law on matrices in @xmath224 . remember that @xmath215 is chosen uniformly , that the law of @xmath14 knowing @xmath202 and @xmath12 is the uniform law on spaces with dimension @xmath47 and with @xmath225 , and that @xmath222 is then the uniform law on bases of @xmath202 with @xmath226 . this implies that : 1 . the probability of a given matrix @xmath35 is invariant by conjugation by elements of @xmath224 : @xmath227={\mathbb{p}}[{\mathrm{mat}}_{g({\mathcal{f}}^{+})}(g({\mathcal{g}}^{+}))=p]={\mathbb{p}}[{\mathrm{mat}}_{{\mathcal{f}}^{+}}({\mathcal{g}}^{+})=gpg^{-1}].\ ] ] 2 . for @xmath12 fixed , the probability conditionally to @xmath202 and @xmath12 of @xmath222 is invariant by action of the parabolic subgroup @xmath228 of block - triangular matrices @xmath229 since this is the stabilizer of @xmath12 inside @xmath224 . if @xmath230 or @xmath200 the second point proves the claim . otherwise , the group that leaves invariant the probability @xmath231 $ ] is a subgroup of @xmath224 that contains the maximal parabolic subgroup @xmath228 , but also all its conjugates . therefore it contains strictly @xmath228 , so it can only be @xmath224 itself , and the claim is shown in full generality . we can finally write : @xmath232\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]}{{\mathrm{card}}\,{\mathrm{gl}}(m,{\mathbb{f}}_{q})}\,{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{p^{-1}j(\bbnu\uparrow^{m})pj(\bbmu \uparrow^{m } ) } { \mathrm{span}}({\mathcal{h}}^{+ } ) ) } \\ & = \sum_{\substack{{\mathcal{e}}^{+},{\mathcal{h}}^{+}\\ p \in { \mathrm{gl}}(m,{\mathbb{f}}_{q } ) } } \frac{{\mathbb{p}}[m]\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]}{{\mathrm{card}}\,{\mathrm{gl}}(m,{\mathbb{f}}_{q})}\,{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{j(\bbmu \uparrow^{m})p^{-1}j(\bbnu\uparrow^{m})p } { \mathrm{span}}({\mathcal{h}}^{+}))}\\ & = \sum_{\substack{{\mathcal{f}}^{+},{\mathcal{h}}^{+}\\ p \in { \mathrm{gl}}(m,{\mathbb{f}}_{q } ) } } \frac{{\mathbb{p}}[m]\,{\mathbb{u\!f}}_{m}[{\mathcal{f}}^{+},{\mathcal{h}}^{+}]}{{\mathrm{card}}\,{\mathrm{gl}}(m,{\mathbb{f}}_{q})}\,{({\mathrm{span}}({\mathcal{f}}^{+ } ) \,\,^{p}\!\!\rightleftarrows_{j(\bbmu \uparrow^{m})p^{-1}j(\bbnu\uparrow^{m } ) } { \mathrm{span}}({\mathcal{h}}^{+}))}\\ & = \sum_{\substack{{\mathcal{f}}^{+},{\mathcal{g}}^{+}\\ p \in { \mathrm{gl}}(m,{\mathbb{f}}_{q } ) } } \frac{{\mathbb{p}}[m]\,{\mathbb{u\!f}}_{m}[{\mathcal{f}}^{+},{\mathcal{g}}^{+}]}{{\mathrm{card}}\,{\mathrm{gl}}(m,{\mathbb{f}}_{q})}\,{({\mathrm{span}}({\mathcal{f}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{pj(\bbmu \uparrow^{m})p^{-1}j(\bbnu\uparrow^{m } ) } { \mathrm{span}}({\mathcal{g}}^{+}))}\\ & = \sum_{\substack{{\mathcal{f}}^{+},{\mathcal{g}}^{+}\\ p \in { \mathrm{gl}}(m,{\mathbb{f}}_{q } ) } } \frac{{\mathbb{p}}[m]\,{\mathbb{u\!f}}_{m}[{\mathcal{f}}^{+},{\mathcal{g}}^{+}]}{{\mathrm{card}}\,{\mathrm{gl}}(m,{\mathbb{f}}_{q})}\,{({\mathrm{span}}({\mathcal{f}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{p^{-1}j(\bbmu \uparrow^{m})pj(\bbnu\uparrow^{m } ) } { \mathrm{span}}({\mathcal{g}}^{+ } ) ) } \\ & = \widetilde{a}_{\bbnu , n}*\widetilde{a}_{\bbmu , n}\end{aligned}\ ] ] by using on the second line the fact that @xmath233 commutes with everyone in @xmath234 , because it is a multiple of a conjugacy class . starting from the last computation in the proof of commutativity , one can give a useful expression for the product of two normalized classes @xmath216 and @xmath217 : this is @xmath235\,{\mathbb{u\!f}}_{m}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]}{({\mathrm{card}}\,c_{\bbmu\uparrow^{m}})\,({\mathrm{card}}\,c_{\bbnu\uparrow^{m}})}\,{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{hg } { \mathrm{span}}({\mathcal{h}}^{+}))}.\ ] ] with this expansion , it becomes obvious that @xmath236 , because in any group algebra @xmath234 , @xmath237 . we focus now on the construction of a projective limit of the subalgebras of invariants ( in the category of graded algebras ) . the main tool is the following : consider the linear maps @xmath238 @xmath42 denoting the degree of the partial isomorphism . their restrictions @xmath239 are morphisms of commutative algebras and they satisfy the relations @xmath240 where @xmath241 to avoid any ambiguity , we shall precise by indices the notations of the proof of proposition [ ziscommutative ] . thus , we shall denote @xmath242 $ ] the probability of @xmath200 in the previous probabilistic scheme and in @xmath6 , and @xmath243 $ ] the uniform probability over free family of size @xmath42 in @xmath6 . fix two polypartitions @xmath121 and @xmath244 of sizes @xmath42 and @xmath47 . if @xmath42 or @xmath47 is strictly bigger than @xmath2 , then @xmath245 or @xmath246 is equal to zero , and on the other hand , every partial isomorphism appearing in @xmath247 will have degree bigger than @xmath248 , so one has @xmath249 suppose now @xmath250 and @xmath251 . we compute @xmath252 as follows : @xmath253 and @xmath254\,{\mathbb{u\!f}}_{m , n+p}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]}{{\mathrm{card}}(c_{\bbmu\uparrow^{m}}\times c_{\bbnu\uparrow^{m}})}\,\phi_{n}^{n+p}{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{hg } { \mathrm{span}}({\mathcal{h}}^{+}))},\end{aligned}\ ] ] where the families @xmath255 and @xmath256 run _ a priori _ over free families of size @xmath200 in @xmath257 , but in fact over free families of size @xmath200 in @xmath6 , since otherwise @xmath258 vanishes . the quantity @xmath259\,{\mathbb{u\!f}}_{m , n+p}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]$ ] is equal to @xmath260 which multiplied by @xmath261 gives @xmath262 multiplying again by the coefficient in the statement of the proposition , we get : @xmath263 on the other hand , @xmath264}{{\mathrm{card}}\ , c_{\bbmu } } \,\,\phi_{n}^{n+p}{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{g } { \mathrm{span}}({\mathcal{f}}))}\\ & = \sum_{\substack{{\mathcal{e}},{\mathcal{f}}\\t(g)=\bbmu}}\frac{q^{-nk}\,(q^{-1})_{n - k}}{(q^{-1})_{n } } \,\frac{{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{g } { \mathrm{span}}({\mathcal{f}}))}}{{\mathrm{card}}\ , c_{\bbmu}},\end{aligned}\ ] ] where again the families @xmath265 and @xmath214 run _ a priori _ over free families of size @xmath42 in @xmath257 , but in fact over free families of size @xmath42 in @xmath6 , since otherwise @xmath266 vanishes . therefore , one can rewrite this as : @xmath267}{{\mathrm{card}}\ , c_{\bbmu}}\,\,{({\mathrm{span}}({\mathcal{e } } ) \,\,^{i_{k}}\!\!\rightleftarrows_{g } { \mathrm{span}}({\mathcal{f}}))}\\ & = \frac{q^{nk}\,(q^{-1})_{n}}{(q^{-1})_{n - k}}\,\widetilde{a}_{\bbmu , n}=\widehat{a}_{\bbmu , n}.\end{aligned}\ ] ] this proves in particular that @xmath268 . finally , @xmath269\,{\mathbb{u\!f}}_{m , n}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]}{({\mathrm{card}}\,c_{\bbmu\uparrow^{m}})\,({\mathrm{card}}\,c_{\bbnu\uparrow^{m}})}\,{({\mathrm{span}}({\mathcal{e}}^{+ } ) \,\,^{i_{m}}\!\!\rightleftarrows_{hg } { \mathrm{span}}({\mathcal{h}}^{+}))};\ ] ] the quantity @xmath242\,{\mathbb{u\!f}}_{m , n}[{\mathcal{e}}^{+},{\mathcal{h}}^{+}]$ ] is equal to @xmath270 which multiplied by @xmath271 gives @xmath272 this is exactly the coefficient of equation , so the morphism property is shown notice that if @xmath273 is the unity of @xmath170 , then one has indeed also @xmath274 . the compatibility between the maps @xmath275 is obvious from the relations , so one gets as announced an inverse system of graded commutative algebras @xmath276 from there everything gets easy . with @xmath277 and @xmath278 , denote @xmath279 the structure coefficients given by the product @xmath280 in @xmath281 . if @xmath282 , we convene that @xmath283 . since the maps @xmath284 are morphisms of algebras , for every @xmath285 , one also has @xmath286 we claim that this identity still holds for @xmath287 . indeed , denote @xmath288 the structure coefficients in @xmath165 , such that @xmath289 the grading ensures that the sum is indeed over polypartitions of size less than @xmath66 . by applying the map @xmath290 to the previous identity , we get @xmath291 since @xmath292 is a basis of @xmath281 , this shows that @xmath293 . as a consequence : the graded commutative algebra @xmath167 with linear basis @xmath294 , @xmath295 and with product @xmath296 is well - defined and forms a projective limit of the @xmath165 s in the category of graded algebras . the structure coefficients @xmath279 are rational numbers : they are obviously in @xmath297 since every product in @xmath298 involves averages with coefficients in @xmath297 , and assuming @xmath3 fixed , @xmath297 is identified with @xmath299 . denote @xmath300 , where @xmath301 is the canonical projection , and @xmath72 is the map from proposition [ projection ] . this is a morphism of algebras from @xmath167 to @xmath28 , with @xmath302 fix two polypartitions @xmath139 and @xmath121 of sizes @xmath42 and @xmath47 , and with @xmath303 and @xmath304 ; hence , they correspond to classes of partial isomorphisms that are not trivial extensions of smaller partial isomorphisms . this is the analogue of the restriction `` without fixed points '' for permutations in our statement of farahat - higman s theorem on page . notice however that one does not require @xmath305 or @xmath306 to vanish ; so , the isomorphisms considered may have non - zero fixed vectors , but no non - zero fixed vector with a stable complement subspace . under the previous assumption , most of the previous formula simplifies and we get @xmath307 by putting everything independent from @xmath2 in modified structure coefficients @xmath308 we obtain finally : @xmath309 this leads to a @xmath1 version of a theorem of farahat and higman , which to our knowledge was not known before . [ generalfh ] fix two polypartitions @xmath139 and @xmath121 of sizes @xmath42 and @xmath47 , with @xmath303 and @xmath304 . there exists polynomials @xmath310 with rational coefficients such that for every @xmath311 , in the center of the group algebra @xmath0 , @xmath312 where the sum runs over polypartitions @xmath244 of size @xmath313 , and again with @xmath314 . the @xmath2-dependent part of the coefficient of @xmath315 in the previous expansion writes as @xmath316 this is _ a priori _ a laurent polynomial in @xmath41 with rational coefficients . gathering together the polypartitions @xmath244 which give the same completed polypartition @xmath317 ( they differ by the number of parts @xmath45 in @xmath318 ) , we conclude from the previous discussion that @xmath312 where the sum is over the finite set of polypartitions @xmath244 with @xmath319 and @xmath318 without parts of size @xmath45 ; and the coefficients @xmath320 are rational laurent polynomials in @xmath321 $ ] . however , the group algebra of a finite group is defined over @xmath322 , so for every integer @xmath2 , @xmath320 is an integer . with @xmath323 , by looking at the @xmath324-valuation in @xmath299 , one sees that @xmath325 can not have negative powers , so it is in fact a polynomial in @xmath326 $ ] . in this section we make the previous discussion concrete by computing the polynomials @xmath320 when @xmath139 and @xmath121 have degree @xmath45 , hence correspond to irreducible polynomials @xmath327 and @xmath328 with @xmath329 . the generic invariant @xmath330 , that we shall abbreviate as @xmath331 , has for projections in the algebras @xmath165 : @xmath332 here @xmath333 means that one sends the vector @xmath334 to @xmath335 by the first arrow , and @xmath335 to @xmath336 by the second arrow . to compute the generic product @xmath337 in @xmath167 , it suffices to do so in @xmath338 by the discussion of the previous section . take two partial isomorphisms @xmath333 and @xmath339 in @xmath340 . among the @xmath341 possibilities for @xmath342 , the vectors @xmath335 and @xmath343 are colinear in @xmath344 cases , the factor @xmath345 corresponding to the possibilities for the factor of proportionnality @xmath346 such that @xmath347 . in all these cases , @xmath348 has type @xmath349 . so , this situation contributes to a term @xmath350 in @xmath337 . in every other situation , @xmath351 and we have to compute trivial extensions . it should be noticed here that the form of @xmath352 is different when @xmath353 and when @xmath354 . therefore , we have several cases to treat separately , the most interesting being when @xmath354 and @xmath355 . suppose then @xmath335 and @xmath343 not colinear . one has @xmath356 so , outside the terms of , the remaining part of the product @xmath337 is equal to @xmath357 we then need to recognize irreducible polynomials of degree @xmath358 over @xmath4 . the theory is a bit different when @xmath3 is even and when @xmath3 is odd . suppose to begin with that @xmath3 is odd . then , @xmath359 is irreducible if and only if @xmath360 is not a square in @xmath4 . in @xmath361 , there are @xmath362 squares identified by the jacobi symbol @xmath363 , and @xmath362 non - squares identified by the jacobi symbol @xmath364 , @xmath365 being a morphism from @xmath22 to @xmath366 . so : * @xmath367 is irreducible if @xmath368 ; * otherwise , it is equal to @xmath369 , where @xmath370 . this criterion has the following consequence . suppose @xmath371 fixed and consider the family of polynomials @xmath372 . notice that @xmath373 therefore , @xmath374 is of cardinality @xmath375 for @xmath376 . suppose now @xmath3 even ; the solution to the quadratic equation is then a particular case of the artin - schreier theory . recall that on a field of characteristic @xmath358 , @xmath377 is a linear map . the trace of an element of @xmath378 is the linear map @xmath379 the image of the artin - schreier map @xmath380 is the @xmath381-subspace of @xmath378 of elements with trace zero . then , given an arbitrary monic polynomial @xmath367 of degree @xmath358 : * if @xmath382 , then @xmath383 is not irreducible . * if @xmath384 and @xmath385 , then @xmath386 , where @xmath334 and @xmath387 are the two elements of @xmath378 such that @xmath388 . * otherwise , if @xmath389 and @xmath390 , then @xmath367 is irreducible . this implies that for every @xmath376 , @xmath374 is of cardinality @xmath391 . suppose @xmath3 odd and @xmath393 , @xmath355 . going back to equation , the characteristic polynomial of the matrix @xmath394 is @xmath395 . notice that the multiplication map @xmath396 satisfies @xmath397 and @xmath398 for @xmath399 . therefore , with @xmath400 fixed : 1 . one obtains @xmath3 times the polynomial @xmath401 , corresponding for instance to the case @xmath402 and @xmath403 arbitrary . if @xmath404 this gives a contribution @xmath405 otherwise if @xmath406 one gets @xmath407 one can also replace by thanks to the symbol @xmath408 . 2 . then , for every @xmath409 , including @xmath410 , one obtains @xmath345 times the polynomial @xmath411 . by the discussion of [ irrpoly ] , @xmath412 values of @xmath413 give an irreducible polynomial , whence a contribution @xmath414 the other values of @xmath413 correspond to decompositions @xmath415 . if @xmath40 is not a square , then @xmath416 and one obtains a contribution @xmath417 if @xmath40 is a square then one can have @xmath418 ; in this case the type of the matrix @xmath419 can be either @xmath420 or @xmath421 . the first case is excluded since @xmath422 , and therefore @xmath423 so , if @xmath424 is a square , then one obtains a contribution @xmath425 again , one can replace by thanks to the symbol @xmath426 . so , the remaining part of the product @xmath337 is equal to @xmath427 times the sum of , and , that is to say @xmath428 hence , the product @xmath337 is the sum of the quantities and when @xmath429 and @xmath3 is odd . the exact same discussion applies to the case @xmath354 , @xmath355 and @xmath3 even . formula can be kept , and it remains to add for every @xmath409 the @xmath345 terms with characteristic polynomial equal to @xmath411 . half of the cases yield the contribution , the only difference with the odd case being that the sum is over a fixed number of values of @xmath413 ( independent of @xmath40 ) , namely , @xmath391 . for the other values of @xmath413 , remark that @xmath40 is always a square in @xmath4 , namely , the square of @xmath430 . therefore , these over values of @xmath431 give a contribution @xmath432 which replaces . thus , in the even case , the remaining part of @xmath337 is equal to @xmath433 the sum of and gives @xmath337 when @xmath429 and @xmath3 is even . finally , when one of the coefficient @xmath434 or @xmath435 is equal to @xmath45 , the unique corresponding trivial extension on @xmath436 is the identity ( partial ) isomorphism @xmath437 . the computation is then trivial and one obtains for instance for @xmath438 with @xmath393 the result @xmath439 the first term corresponds to colinear vectors @xmath334 and @xmath335 , and the second term to non - colinear vectors . when @xmath434 is also equal to @xmath45 , the same reasoning gives @xmath440 thus , we have proven the following : [ productdegree1 ] the product of two generic degree @xmath45 classes @xmath331 and @xmath441 is given by the following formulas . each time , the writing is decreasing in degree and without terms appearing more than once , but in the sums preceded by a @xmath442 , where each term appears twice . * @xmath21 , @xmath3 odd , @xmath443 , @xmath424 : @xmath444 * @xmath21 , @xmath3 odd , @xmath445 : @xmath446 * @xmath21 , @xmath3 even , @xmath430 : @xmath447 * @xmath448 , @xmath3 odd : @xmath449 * @xmath448 , @xmath3 even : @xmath450 * @xmath354 , @xmath451 : @xmath452 * @xmath453 : @xmath454 the theorem yields readily the product of two completed conjugacy classes of degree @xmath45 for any @xmath455 , since @xmath456 for any polypartition @xmath121 . more specifically , @xmath457 @xmath458 applying these formulas to the seven cases of theorem [ productdegree1 ] , one obtains the expansion in completed conjugacy classes of @xmath459 so for instance , if @xmath3 is odd and @xmath21 , @xmath443 , @xmath424 , then by projection by @xmath460 of the first case of theorem [ productdegree1 ] , the product of completed conjugacy classes @xmath461 in @xmath106 is given by : @xmath462 each structure coefficient is indeed a polynomial in @xmath463 , here of degree @xmath410 or @xmath45 . notice that this formula _ can not _ be specialized to the case @xmath353 or @xmath451 ( in this case , the sixth and seventh cases of theorem [ productdegree1 ] give indeed the much simpler expansion @xmath464 , as can be expected ) . the other cases are similar and easy computations after theorem [ productdegree1 ] and formulas and . from these computations , it becomes clear that the determination of the whole multiplication table of @xmath167 is not possible . however , it might exist simple rules to determine certain structure polynomials @xmath465 , especially those with @xmath466 . if it exists , a general rule for these coefficients `` of maximal degree '' probably involves deeply the arithmetics and galois theory of polynomials over @xmath122 $ ] . the author would like to thank valentin fray for several discussions around the construction of `` generic '' center group algebras ; and the anonymous referee for his valuable comments , that allowed to improve the quality of the paper . v. ivanov and g. olshanski . kerov s central limit theorem for the plancherel measure on young diagrams . in _ symmetric functions 2001 _ , volume 74 of _ nato science series ii . mathematics , physics and chemistry _ , pages 93151 , 2002 . products of geck - rouquier conjugacy classes and the hecke algebra of composed permutations . in _ proceedings of the 22nd international conference on formal power series and algebraic combinatorics ( san francisco , usa ) _ , pages 790801 , 2010 . kerov s central limit theorem for schur - weyl and gelfand measures . in _ proceedings of the 23th international conference on formal power series and algebraic combinatorics ( reykjavk , iceland ) _ , pages 669680 , 2011 . . fluctuations of central measures on partitions . in _ proceedings of the 24th international conference on formal power series and algebraic combinatorics ( nagoya , japan ) _ , pages 387398 , 2012 .

we present long - slit optical spectra along the major axis of the planetary nebula ngc 7009 . these data allow us to discuss the physical , excitation and chemical properties of all the morphological components of the nebula , including its remarkable systems of knots and jets . the main results of this analysis are the following : _ i ) _ the electron temperature throughout the nebula is remarkably constant , = k ; _ ii ) _ the bright inner rim and inner pair of knots have similar densities of , whereas a much lower density of 1500 is derived for the outer knots as well as for the jets ; _ iii ) _ all the regions ( rim , inner knots , jets and outer knots ) are mainly radiatively excited ; and _ iv ) _ there are no clear abundance changes across the nebula for he , o , ne , or s. there is a marginal evidence for an overabundance of nitrogen in the outer knots ( ansae ) , but the inner ones ( caps ) and the rim have similar n / h values that are at variance with previous results . our data are compared to the predictions of theoretical models , from which we conclude that the knots at the head of the jets are not matter accumulated during the jet expansion through the circumstellar medium , neither can their origin be explained by the proposed hd or mhd interacting - wind models for the formation of jets / ansae , since the densities as well as the main excitation mechanisms of the knots , disagree with model predictions .

astro-ph0307265

in addition to their large - scale structures , mainly identified in the forbidden [ oiii ] emission line , many planetary nebulae ( pne ) have a number of smaller - scale structures which are instead more prominent in low - ionization lines , such as [ n ii ] , and . in a previous paper ( gonalves , corradi , & mampaso 2001 ) , we have used the acronymous liss to identify these low - ionization structures . liss appear with different morphological and kinematical properties in the form of pairs of knots , filaments , jets , or isolated features moving with supersonic velocities through the large - scale components in which they are located , or instead as structures with the above morphologies but with low velocities that do not differ substantially from that of the main nebula . liss can be easily seen in the imaging catalogues of balick ( 1987 ) , schwarz , corradi , & melnick ( 1992 ) , manchado et al . ( 1996 ) , corradi et al . ( 1996 ) and grny et al . ( 1999 ) . for a detailed study concerning the morphological and kinematical classification of liss in pne , see gonalves et al . ( 2001 ) . a very interesting class of liss these are highly collimated outflows ( with aspect ratios varing between 3 and 20 ; lpez 1997 ) , which are directed radially outward from the central star , appear in opposite symmetrical pairs , and move with velocities substantially larger than those of the ambient gas that forms the main bodies of the nebulae . it should be noted in addition that jets are much more collimated than other bipolar outflows of pne , i.e , more collimated than the lobes of a bipolar pn . ngc 7009 , the `` saturn nebula '' , is a well studied elliptical pn , which possesses a jet - like system as well as two pairs of low - ionization knots along its major axis . a number of previous studies showed that high - excitation lines dominate the inner regions along the minor axis , while the low - excitation ones are enhanced at the ends of the major axis . the ionization structure is further enriched by the fact that the low - excitation region present strong variations in excitation level and clumpiness . ngc 7009 was classified as an oxygen - rich pn ( hyung & aller 1995 ) , with an o / c ratio exceeding 1 , and anomalous n , o , and c abundances ( baker 1983 ; balick et al . 1994 , hyung & aller 1995 ) . its central star is an h - rich o - type star , with effective temperature of 82000 k ( mndez , kudritzki , & herrero 1992 ; kingsburgh & barlow 1992 ) . the kinematics of ngc 7009 was studied by reay & atherton ( 1985 ) and balick , preston & icke ( 1987 ) . the distance determinations of ngc 7009 vary from 0.5 to 2.3 kpc from various indirect methods ( acker et al . 1992 ; cahn , kaler & stanghellini 1992 ; maciel 1995 ) . according to the theoretical models for the formation of jets and of the pairs of knots that are often found at their tips ( soker 1990 ; garca - segura 1997 ; garca - segura et al . 1999 ; garca - segura & lpez 2000 ; steffen , lpez , & lim 2001 ; gardiner & frank 2001 ; blackman , frank , & welch 2001 ) , a significant density contrast between the jet itself and the knots at its tip is expected . this is a quantity that can in principle be determined by optical spectroscopy . in addition , previous observational work suggests that the liss of ngc 7009 and other pne have higher abundances of nitrogen than their main bodies ( see balick et al . 1994 and references therein ) . these authors proposed that such an overabundance might indicate that the knots of ngc 7009 were ejected at a different epoch than the main body of the nebula , when the stellar surface was enriched of these elements due to the drastic changes of nitrogen in agb stars and some dredge - up episode , or that the knots come from chemically peculiar regions in the stellar surface or its interior . in this paper , we test the above two issues by obtaining spectroscopic data for ngc 7009 , deriving its physical and chemical properties , and finally showing that neither a density contrast between the jets and knots is found , nor are their abundances strongly enhanced compared to the rest of the nebula . in the following sections , we first describe the data , their reduction and the various morphological / kinematical structures of the nebulae ( sections 2 and 3 ) . then we proceed with the analysis of the pn spectra in terms of physical , excitation and chemical parameters ( section 4 ) and , finally , we discuss the implication of the present results for the formation of the liss of ngc 7009 and give our conclusions , in sections 5 and 6 . spectra of ngc 7009 ( pn g037.7 - 34.5 ) were obtained on 2001 august 29 at the 2.5 m isaac newton telescope ( int ) at the observatorio del roque de los muchachos ( european northern observatory , la palma , spain ) , using the intermediate dispersion spectrograph ( ids ) . the ids was used with the 235 mm camera and the r300v grating , providing a spectral coverage from 3650 to 7000 with a spectral reciprocal dispersion of 3.3 pix@xmath5 . the spatial scale of the instrument was [email protected] pix@xmath5 with the tek5 ccd . seeing varied from [email protected] to [email protected] . the slit width and length were 1.5 and 4 , respectively . the slit was positioned through the center of the nebula at p.a . = 79@xmath7 , and the exposure times were 3@xmath830 s , 3 @xmath8 120 s , and 3 @xmath9 600 s. during the night , bias frames , twilight and tungsten flat - field exposures , wavelength calibrations and exposures of standard stars ( bd+332642 , cyg ob2#9 , hd19445 , and bd+254655 ) were obtained . spectra were reduced following the iraf instructions for long - slit spectra , being bias - subtracted , flat - fielded , combined in order to improve the signal - to - noise ratio and eliminate cosmic rays , wavelength calibrated , and sky - subtracted . finally , they were flux - calibrated using the above - mentioned standard stars and the mean atmospheric extinction curve for la palma . we also retrieved _ hst _ and archive images of ngc 7009 obtained on 1996 april 28 , and on 2000 april 7 , respectively , with the wfpc2 camera in the f658n and f502n filters , with total exposure times of 1200 s and 320 s. the _ hst _ and images of ngc 7009 are shown in figure [ i - n7009 ] . the nebula consists of several components . the brightest one is the elliptical inner shell ( _ rim _ ) roughly elongated in the ew direction , where it extends out to @xmath10 from the center . the inner portions of the rim cut by our spectroscopic slit are indicated in figure [ i - n7009 ] ( bottom ) by the labels r@xmath11 and r@xmath12 . surrounding the rim , a fainter elliptical outer shell ( _ shell _ ) that shares its main axis with the rim is also visible in the _ hst _ images . in , at fainter intensity levels the shell shows two `` polar '' bubbles . inside these bubbles , collimated liss are clearly seen in the images . these are composed of a `` pencil '' of emission ( the _ jetlike _ structures , labeled j@xmath11 and j@xmath12 ) that connects the inner rim with two _ outer knots _ ( k@xmath11 and k@xmath13 ) . the image also shows another pair of microstructures , which appear as `` compact groups of bright knots '' that we call the _ inner knots _ ( k@xmath12 and k@xmath14 ) . the extension of all features is given in the caption of figure [ i - n7009 ] . the rim of ngc 7009 is likely to be formed by the interaction between the fast post - agb and the slow agb winds ( kwok , purton , & fitzgerald 1978 ; frank & mellema 1994 ; mellema & frank 1995 ; zhang & kwok 1998 , schnberner 2002 ) . the cavity delimited by the rim is filled with a hot ( @xmath15 k ) gas emitting thermal x - rays and its size is in agreement with that of the optical rim ( guerrero , gruendl , & chu , 2002 ) providing a direct proof for the shocked fast stellar wind in the interacting winds scenario . the formation of the shell has been ascribed to the action of the photoionization front on the agb matter not yet reached by the shock produced by the fast wind ( e.g. , schnberner 2002 ) . the origin and nature of the liss in ngc 7009 are rather unclear at present . the inner and outer pairs of knots have been studied by many authors ( reay & atherton 1985 ; balick et al . 1994 ; hyung & aller 1995 ; lame & pogge 1996 ; balick et al . 1998 ) , who often called them _ caps _ and _ ansae _ , respectively . reay & atherton ( 1985 ) have studied the kinematics of the inner and outer knots in the 6300 line using fabry perot data . they derived an expansion velocity for the inner knots of @xmath16 with an inclination with respect to the line of sight @xmath17 , and a velocity of @xmath18 with @xmath19 for the outer pair . note that the determination of the deprojected velocities and inclination angles by reay & atherton ( 1985 ) is mainly based on the proper motion measurements of liller ( 1965 ) , as the velocity in the plane of the sky appears one order of magnitude larger than the doppler shift . we have estimated the separation of the outer knots in the _ hst _ image taken in 1996 , finding @xmath2050.8 @xmath21 0.5 arcsec , the uncertainty depending on the highly irregular morphology of these knots that makes it difficult to define a centroid . this is clearly not consistent with the extrapolation of liller s measurements ( separation of 48 arcsec in 1965 and apparent expansion of 1.6 arcsec per century for each knot ) , making the velocity and inclination figures of reay & atherton ( 1985 ) quite uncertain . their analysis also depends on the distance adopted , which is poorly known , as mentioned in the introduction . no velocity information is available from the literature for the jetlike structures connecting the rim to the outer knots . if we assume a continuity between the velocity of the outer knots and these jetlike structures , then the latter might be real supersonic jets . a more detailed study of the kinematics of ngc 7009 , and in particular of the faint `` pencil '' and large - scale structures surrounding it ( shell and polar bubbles ) , is clearly needed to understand its nature better . line intensities were mainly measured from the deep spectrum with 3@xmath9600 s exposure time . the shorter - exposure spectra were used to measure the bright , , , and lines that are saturated in the deep spectra . measurements are given in table 1 , which lists the wavelength of the line identification and the corresponding ion ( column 1 ) , the line fluxes measured for the different spatial regions selected ( columns 2 to 9 , as defined in fig . 1 and section 3 ) , and the flux integrated along the slit for the whole nebula , labeled as neb ( column 10 ) . all fluxes are normalized to @xmath22 ( ) = 100 in each of the regions considered . errors in the fluxes were calculated taking into account the statistical error in the measurement of the fluxes , as well as systematic errors of the flux calibrations , background determination , and sky subtraction . table 2 gives the estimated accuracy of the measured fluxes for a range of line fluxes ( relative to ) in each of the selected regions . absolute fluxes integrated along the slit in each region are also given in table 2 . fluxes were extinction - corrected by using = 0.16@xmath21 0.01 ( the logarithmic ratio between observed and dereddened fluxes ) , determined from the / ratio in the brightest regions of the nebula ( and ) . for the derivation of , we assumed @xmath23 = 10@xmath24 k and @xmath25 = 10@xmath24 @xmath3 , appropriate for the values that we have computed in the rim ( section 4.2 ) . theoretical balmer line ratios from osterbrock ( 1989 ) and the reddening law of cardelli , clayton , & mathis ( 1989 ) were used . our derived value is in fair agreement with the average value ( 0.15 ) reported by other authors : = 0.08 @xmath21 0.03 ( hyung & aller 1995 ) ; = 0.24 @xmath21 0.04 ( lame & pogge 1996 ) ; = 0.20 ( liu et al . 1995 ) ; = 0.10 ( rubin et al . 2002 ) . note that we applied = 0.16 for the fluxes of all the different zones under analysis , because of the highest s / n ratio of the rim , thus avoiding the larger errors in the measured balmer line ratios of the fainter zones . this assumption of a constant extinction throughout the nebula is supported by the work of bohigas , lpez , & aguilar ( 1994 ) , who reported that the for the different structures of ngc 7009 are always lower than 0.28 , implying small corrections to the observed fluxes . as described above , eight different regions were selected in the spectrum of ngc 7009 ( see also figure [ i - n7009 ] ) . the extinction - corrected fluxes were then used in order to obtain the densities and temperatures for each of the regions that are plotted in figure [ p - temden ] . these physical parameters were obtained using the _ nebular _ package of iraf ( shaw & dufour 1995 ) , based on the five - level atom program ( de robertis , dufour , & hunt 1987 ) to derive the physical conditions in a low - density nebular gas , including diagnostics for a large number of ions and emission lines . to estimate densities , we assumed an electron temperature of 10@xmath24 k ( see below ) . in most positions , we were able to measure the densities from both the 6716 /6731 and 5517 /5537 line ratios . these provide a simultaneous estimate of densities in low - ionization ( ) and higher ionization ( ) regions . in general , we found a good agreement between the results from these two density estimators for most of the structures ( see table 1 and figure [ p - temden ] ) . using the total flux integrated along the slit , we obtain @xmath26/@xmath26 @xmath27 1.13 ; the horizontal dashed - dashed line in figure [ p - temden ] indicates the @xmath26overall value , while the continuous line shows the corresponding @xmath26 overall density . figure [ p - temden ] clearly shows that the densities of the outermost knots are very similar to those of the jets connecting them to the edge of the inner rim , with @xmath26 ( ) = 1900 @xmath21 500 @xmath3 and @xmath26()= 1300@xmath21200 @xmath3 and , on the opposite side , @xmath26 ( ) = 1300 @xmath21 200 @xmath3 and @xmath26 ( ) = 1300 @xmath21 500 @xmath3 . on the other hand , the inner pair of knots have higher electron densities , namely @xmath26 ( ) = 4500 @xmath21 400 @xmath3 and @xmath26 ( ) = 5000 @xmath21 500 @xmath3 , and similar to those of the rim ( @xmath26()= 5500 @xmath21 600 @xmath3 and @xmath26 ( ) = 5900 @xmath21 800 @xmath3 ) . as for the densities , the temperature estimators used are appropriate for zones of low- and high - excitation . we used the following line ratios : @xmath22(4959 + 5007 )/@xmath22(4363 ) for @xmath0 ; @xmath22(6548 + 6583 )/@xmath22(5755 ) for @xmath0 ; and @xmath22(6716 /6731 )/@xmath22(4069 + 4076 ) for @xmath0 . the general trend of temperatures ( figure [ p - temden ] , bottom ) is that , within the errors , they are constant throughout the nebula , having an average value of @xmath0 = 10200 k and @xmath0 = 11100 k. in spite of a small discrepancy of @xmath0 at the position of , the temperature is remarkably constant across this nebula . there are a number of papers that discuss the physical properties of ngc 7009 ( rubin et al . 2002 ; hyung & aller 1995 ; bohigas et al . 1994 ; balick et al . 1994 ) . in general , there is good overall agreement between our temperature determinations and previous ones . a more detailed zone - by - zone comparison can be made with some of these papers . balick et al . ( 1994 ) studied the west rim , cap , and ansae . they found @xmath0 and @xmath0 of 10000 k and 9400 k , respectively , for the rim , 9600 k and 9600 k for the cap , and 8100 k and 11500 k for the ansa . with the exception of @xmath0 of that ansa , these figures are very similar to ours ( see table 1 ) . from the six regions of the pn studied by bohigas et al . ( 1994 ) , two coincide with ours ( those of the w and e ansae ) , and they found , for both structures , an temperature of 9800 k , again consistent , within the errors , with our measurements . the @xmath0 determined from the _ hst_/wfpc2 data by rubin et al . ( 2002 ) is also consistent with the values we derived , since they found 9000 k @xmath28 @xmath0 @xmath28 11000 k. their stis long - slit results for temperature are limited to the 13 central arcseconds along the major axis . the average @xmath0 for , , , and are @xmath2010300 k @xmath2010700 k , @xmath2010600 k , and @xmath2010100 k , respectively , which are very similar to our figures in table 1 . the above papers also report density estimates , whose agreement with our measurements is not , however , as good as for the electron temperatures . for instance , balick et al . ( 1994 ) found ne = 4900 @xmath3 , 4100 @xmath3 , and 1000 @xmath3 for the western rim , cap , and ansae , respectively . they also found a large difference between the @xmath26/@xmath26 ratio at the position of the w rim and cap , which was measured to be @xmath203 and @xmath200.4 , respectively , while ours is @xmath201.0 and @xmath200.8 ( in line with @xmath26 @xmath29 @xmath26found by copetti & writzl 2002 for a large sample of pne ) . lame & pogge ( 1996 ) found that the rim plus inner knots of ngc 7009 have densities as high as 8000 @xmath3 , and that the rim without including and has a significantly lower density , @xmath204000 @xmath3 . there is also only moderate agreement between our densities and those of bohigas et al . ( 1994 ) , who obtained 2300 @xmath21 600 @xmath3 and 4300 @xmath21 2700 @xmath3 for the w and e ansae , respectively . the reason of the discrepancies between different authors might be the s / n ratio of the relevant lines , and the seeing . note that our observations are , in general , much deeper and were obtained during better seeing conditions than the above studies , allowing a more precise definition of the spatial regions . concerning the general excitation structure of ngc 7009 , the low - ionization ( , , , ) line emission has local maxima at the position of the pairs of knots ( k@xmath30 ) . the higher ( , ) ionization emission lines , however , are much more prominent within the rim of the nebula . as the line ratio profiles might provide additional interesting information about the excitation conditions of a nebula , we show in figure [ p - ratio ] the / , / , / , and , spatial profiles along the major axis of the nebula . the top panel clearly shows the inner and outer pairs of knots as prominent zones of low ionization . moreover , the fact that / , / and / ratios are very similar in shape suggests that their local maxima might not be due to the overabundance of a given element . these line ratios are rather low at the position of the jets and the rim implying that the excitation of the jets is quite different from that of the knots . the average values of / in the outer knots are 1.00 ( ) and 0.96 ( ) , much higher than those usually found for spherical and elliptical pne ( e.g. , aller & liller 1968 ) . the inner pair of knots , on the other hand , have less extreme / line ratios , namely 0.13 and 0.14 for and , respectively . the lower two panels of figure [ p - ratio ] show the and spatial profiles at the position of the inner ( top ) and outer ( bottom ) pairs of knots . the and profiles have been normalized in the nebular regions just ahead of the knots in order to see whether the variation of the and emission through the knots shows any effects related to collisional quenching of the emission lines . the critical density for collisional de - excitation is @xmath31 @xmath3 and @xmath32 @xmath3 for the 3726 3729 doublet , respectively , while it is @xmath33 @xmath3 for the 6583 line . figure 3 shows that the and profiles are very similar in the outer , lower density ( n@xmath34 @xmath35 2000 @xmath3 ) knots , whereas in both inner knots ( n@xmath34 @xmath29 5000 @xmath3 ) on the contrary , the profile is much flatter than the one . this represents a direct proof that collisional quenching is indeed affecting the doublet in regions of relatively high densities , and caution should be exercised when , for example , computing densities using the doublet . peculiar line ratios from structures like the knots of ngc 7009 have been attributed to anomalous abundances and/or shock excitation ( balick et al . 1987 ; miranda & solf 1992 ) . in order to test whether shock excitation plays a role in ngc 7009 , we show in figure [ p - shockphot ] two relevant diagnostic diagrams ( c.f . cant 1981 ; phillips & cuesta 1999 ) . the behavior of the different components of ngc 7009 in the / ( 6548 + 6583 ) versus / ( 6716 + 6731 ) diagram is very interesting , revealing the eight regions to be neatly separated and ordered in pairs along the pn band . this is in line with the trends found in other pne ( see the case of m 2 - 9 , phillips & cuesta 1999 ) : the rim appears on the right of the diagram , whereas the outer zones of the nebula lie in the bottom left of the diagram . an interesting result from this diagram is that the jets and rim , which are high - excitation zones are on the right side , while both pairs of knots appear to be separated from the former in terms of degree of excitation . the bottom panel of figure [ p - shockphot ] , showing ( 4959 + 5007)/ versus ( 6716 + 6731)/ , separates the zone of radiatively excited emission lines ( the pn zone , adapted from phillips & cuesta 1999 ) from the zone mainly excited by shocks ( from the predictions of the plane parallel and bow - shock models of hartigan , raymond , & hartmann 1987 ) . it is clear that all eight structures in ngc 7009 are mainly radiatively excited by the pn central star . this is remarkable , since other pne show , at least in some of their morphological components , indications of shock excitation , whereas in ngc 7009 not even the jet and outer knots approach the shock area in figure [ p - shockphot ] . on the other hand , the fact that forbidden lines are radiatively excited in this nebula validates our calculations of physico - chemical parameters for the different regions . the chemical composition of a pn results from the mixing of the elements produced by the central star and dredged up to the surface with those present in the original gas from which the star was made . the abundances of o , ne , ar , and s are thought to represent the chemical composition of the interstellar medium when the progenitor star was born , since these elements are neither produced by the pn progenitor nor very much affected by nucleosynthesis ( see iben & renzini 1983 and stasiska 2002 for reviews ) . on the contrary , he , c , and n changes dramatically in the progenitor star enriching the stellar surface and , owing to the dredge - up episodes , enrich the pn envelopes . abundances of ionized nebulae are calculated based either on empirical methods or on model fitting ( stasiska 2002 ) . in the former case , complete multiwavelength data would give the emission for all ions of a given element , and the total abundance is the sum of the ionic ones . usually , however , only a few lines of some ions can be measured and one must correct for unseen ions using ionization correction factors . here , we have computed the ionic and total abundances ( relative to hydrogen ) for the different regions of ngc 7009 , and also for the whole nebula via the ionization correction factors ( _ icf _ ) following the scheme in kingsburgh & barlow ( 1994 ) , as done , for example , by corradi et al . the uncertainties in this method have been discussed by alexander & balick ( 1997 ) and appear to be significant for the analysis of spatially resolved long - slit observations . table 3 lists the ionic and total abundances and the _ icf _ used . total abundance profiles along the major axis of the nebula are shown in figure [ p - anew ] . errors in the ionic abundances are derived by taking into account both the errors in the line ratios and those in the adopted temperatures . errors in the total abundances are obtained by propagating the errors in the ionic abundances as well as on the _ icf_. typical resulting errors ( 1@xmath36 ) in the total abundances are less than 10@xmath37 for he ( for which no _ icf _ is applied ) , 1020@xmath37 for o , less than 35@xmath37 for ne , and 2540@xmath37 for n and s. first of all note in figure [ p - anew ] that the abundances obtained for the regions and should be interpreted with great caution because these are indeed very faint regions in the nebula . in particular , the @xmath385755 line could not be measured at the jet positions , and we could not determine @xmath0 . instead , we have assumed @xmath0 = @xmath0 for and , based on the fact that the @xmath0 to @xmath0 ratio of all the other regions ranges between 1.01 and 1.25 . therefore , although the adopted temperature for and seems reasonable , the ionic abundances of the low - ionization ions for these regions are probably more uncertain than the quoted errors . figure [ p - anew ] shows that the abundance of neon remains constant throughout the nebula . oxygen and sulfur abundances appear larger in and than in the other regions , but the variations are under the 2@xmath36 level . the constancy of these elements is expected from stellar evolution theory and is consistent with previous results for most spatially resolved pne studied so far ( cf . perinotto & corradi 1998 ) . helium also remains practically constant , implying no or little enrichment in the nebular ejecta . figure [ p - anew ] also shows evidence for an increase of the n abundance in the outer knots and , as previously suggested by balick et al . ( 1994 ) . they reported a much higher overabundance , up to a factor of 5 for both and ( their western ansa and cap ) , although they also warned that their uncertainties in the abundances could be as high as a factor of 2 . our results can also be compared with those of other authors , who found `` peculiar '' abundances of n and o for certain zones of ngc 7009 . an overabundance of n / h in the outer knots of a factor of 2 or more with respect to the other regions of the nebula was first derived by czyzak & aller ( 1979 ) . baker ( 1983 ) , on the other hand , found that the n abundance does not vary substantially through the pn , but that o / h is marginally ( by a factor of @xmath39 ) lower at the position of the outer knots . note , however , that based on much better data and an analysis of four pne containing ansae and including ngc 7009 , balick et al . ( 1994 ) concluded that _ only _ n / h increases at the ansae , by a factor between 2 and 5 as compared with the rim . such behavior was interpreted as the result of the recent ejection of n - enriched high - velocity material by the pn central star . a good test to investigate further this point is the inspection of the profile of the n@xmath40/o@xmath40 ionic abundance ratio ( commonly taken as a measure of the total n / o abundance that does not depend on _ icf _ corrections ) . this is shown in the bottom right panel of figure [ p - anew ] , and shows the same trend as the n / h ratio . n / o ratios deduced by czyzak & aller ( 1979 ) also show an apparent gradient from the rim to the ansae . note , however , that perinotto & corradi ( 1998 ) and hajian et al . ( 1997 ) did not find any evidence of peculiar n or o abundances in the 17 pne studied , some of which possess ansae . finally , our total abundances of he , o , n , ne , and s for the whole nebula ( integrated over the entire slit , i.e. those labeled neb in table 3 ) are in a general agreement with those obtained by baker ( 1983 ) , liu et al . ( 1995 ) , and hyung & aller ( 1995 ) . more accurate abundances were derived by pottasch ( 2000 ) combining iue and iso / iras line fluxes with optical data from the literature , and are also listed in table 3 . these abundances have _ icf _ of order of unity and it is therefore relevant to compare them with our values . percentual differences ( from 4@xmath37 for he to 30 - 40@xmath37 for n and ne ) are similar to our estimated errors and indicate a fair level of agreement between the two sets . we find that the sulfur abundance for the whole nebula is lower than the corresponding solar value , while the he , o and ne ones are approximately solar ( grevesse & anders 1989 ; allende - prieto et al . 2001 ) ; the n abundance appears similar or marginally higher than the solar value . in conclusion , we find a homogeneous elemental abundance across the nebula , to within 9% , 17% , and 35% for he , o , ne and s , respectively . nitrogen seems to be enhanced in the outermost knots of ngc 7009 by a factor @xmath41 , but this evidence is only marginal considering the errors and the enormous range in the derived _ icf_. the uncertainties intrinsic to the method are also rather large ( alexander & balick 1997 ) , but the present data seem to discard variations at the level found by balick et al . how the observed physico - chemical characteristics of the microstructures in ngc 7009 fit into current models for lis formation and evolution ? let us first note that discrete knots such as to could form during the evolution of continuous jets : dense knots develop at the tips of jets as matter accumulates at the head of the jet , while knots in the main body of the jet can originate via several forms of instabilities ( kink , vishniac , rayleigh taylor , and kelvin helmholtz ) . sometimes , both jets and knots are observed , whereas in other instances we observe only the latter . from a theoretical point of view , the cooling conditions on the post - shock gas and the stage of the jet s evolution mainly determine whether or not both jets and knots are seen ( see , for instance , soker 1990 ; frank , balick & livio 1996 and garca - segura 1997 ) . according to hydrodynamical ( hd ) and magnetohydrodynamical ( mhd ) studies ( garca - segura et al . 1999 ; garca - segura & lpez 2000 ; steffen , lpez , & lim 2001 ; gardiner & frank 2001 ; blackman et al . 2001 ) , the interplay between the stellar agb and post - agb winds ( for single - star progenitors ) or between stellar and disk winds ( if the progenitor is a binary star ) can form highly collimated jets in pne . jets that originate in this way are supersonic , highly collimated , two - sided , and roughly coeval with the main shell ( single star ) or younger ( binaries ) . another characteristic of some of the models is the linear increase of the jets expansion velocity with distance from the center ( garca - segura et al . 1999 ; steffen et al . the orientation of jets that result from these models depends on the alignment between the shell axis and the disk / rotation ( or magnetic ) axis ( garca - segura & lpez 2000 ; blackman et al . 2001 ) , as well as on the presence of precession or wobbling ( cliffe et al . 1995 ; garca - segura 1997 ; livio & pringle 1997 ) . if the kinematical results of reay & atherton ( 1985 ) are correct ( but see the discussion in section 3 ) , then , as we have shown in a previous paper ( gonalves et al . 2001 ) , the morphology and kinematics of the jet and knots of ncg 7009 can be qualitatively explained in the light of hd or mhd single - star interacting stellar wind model . however , as far as the comparison of our results with models in terms of physical parameters is concerned , the situation becomes worse . on the one hand , most of the above models do not show clear predictions for the densities and temperatures of jets and knots . on the other hand , those which do show such predictions are hardly in agreement with the physical parameters derived from the present observations , as we will show in the following . the simple process of mass accumulation implies that knots at the jets tips are necessarily denser than the jets themselves . as we showed in section 4 this is not the case in ngc 7009 . an additional , very intriguing , property of the jets / knots of this pn is that we find no evidence for shock excitation from the optical lines ( the line ratios of the jets and outer knots appear to be essentially radiatively excited , see section 4.3 ) . further insights can be gained by comparing our results with the following detailed models . steffen et al . ( 2001 ) developed hd stagnation knot / jet 2-d models which are based on the slow agb and fast post - agb winds interaction with lack of momentum in the polar direction of the main shell . they show that this interaction gives rise to `` microstructures '' which first produce a slightly elongated knot attached to the rim , then a knot further elongated and detached to the rim , and finally a polar jet with a characteristic outward linear increase in velocity . the stagnation knot / jet model has transient lower density jets with higher density knots at the tips , but the opposite occurs with the more permanent jet structures . therefore , the stagnation knot / jet models do not explain the outer knots and jets of ngc 7009 , whose outer pair of knots and jets have very similar densities , contrary to the model predictions . recent 3d mhd models by garca - segura & lpez ( 2000 ) can also be qualitatively compared with our results . these models seem to be able to reproduce successfully the formation of large- and small - scale structures in pne . these authors state that the mass - loss rate during the post - agb is a key parameter in distinguishing between the formation of pne with highly collimated jets and pne with knots ( ansae - like structures ) : @xmath42 @xmath43 yr@xmath5 leads to the formation of jets , while lower mass - loss rates ( @xmath44 @xmath43 yr@xmath5 ) are only able to produce knots , instead of jets . although the low resolution of these models precludes making quantitative predictions for the physical parameters of the knots and jets , their results suggest ( their figure 7 ) that jets and knots are indeed shock - dominated while our analysis indicates that the jets and the outer knots of ngc 7009 are mainly radiatively excited . an alternative possibility for the formation of polar knots / jets in elliptical pne is the model proposed by frank et al . they suggest that jets and ansae can be formed during the `` momentum - conserving '' proto - pn stage , i.e. , before the beginning of the fast wind . the key ingredients of this model are the gradual change from a slow to a fast wind , as well as an appreciable equator - to - pole density contrast in the slow wind . the accelerated wind would flow preferentially parallel to the shock , and the gas would follow the shocks prolate boundary toward the polar axis . the final result is that the gas flows converge at the polar axis where they collide , form a new shock that redirects the flow into a jet , and moves outward along the main axis of the bubble . jets / ansae formed this way can be accelerated up to 70 and have densities @xmath45 @xmath3 , before being ionized by the uv photons of the post - agb star . in fact the jet / ansae of ngc 7009 look like a continuity of its elliptical rim ( remembering that rim and jet are very similar in excitation degree ) , which is very much in the line of the `` converging polar flows '' model for the formation of jets and ansae . however , such flows do not tend to converge into stable `` microstructures '' , since the fast / slow wind interaction could destroy these structures ( frank et al . 1996 ; dwarkadas & balick 1998 ; lpez 2002 ) . finally , if the nitrogen overabundance of the outer pair of knots is taken as real ( but see the discussion in section 4.4 ) , the formation of and might be related to recent high - velocity ejections of enriched material from the central star . such a conclusion requires that the high - velocity nature of the structures is effectively confirmed by more robust kinematical data ( especially concerning the proper motions ) . another issue to be further investigated is the relation between the thin jet and the `` fat '' polar bubbles of the nebular shell . the jet seems to be located along the polar axis of the bubbles , and the outer knots lie exactly at the edge of the bubbles . it is then conceivable that the outer knots are the results of the interaction of the jet with the walls of the bubbles . however , would such interaction enhance the low - ionization emission without developing a significant density contrast , as observed ? the formation of the inner system of knots is also uncertain . morphologically , the knots seem to be located in the shell , and might well share its expansion velocity , as found in other pne , such as ic 2553 and ngc 5882 ( corradi et al if so , they could well be the results of in situ instabilities , or survivors of pre - existing condensations in the agb wind . our suggestions for the formation of the outer and inner pair of knots can readily be tested with adequate kinematical modeling of the different structures in ngc 7009 . a physico - chemical investigation of the symmetrical pairs of structures rim , inner and outer knots , and jetlike systems connecting the latter in ngc 7009 is presented . this study , the most extensive ( eight individual regions along the major axis ) , better resolved ( 1 arcsec spatial resolution within the central 1 arcmin ) and complete to date ( including all diagnosis from available optical lines ) , has revealed a clear and coherent pattern of physical , excitation , and chemical characteristics through the nebula : 1 ) the electron temperature , both for low- and high- excitation species , is remarkably constant ; 2 ) electron densities are similar for both the jetlike structures and the outer knots , but they are roughly one - third of those found in the pair of inner knots and the rim ; 3 ) no sign of shock excitation is found for any of the microstructures or the rim ; and 4 ) a notable chemical uniformity is found for the eight regions studied , with the exception of a marginal , moderate ( @xmath9 2 ) enhancement of nitrogen at the outernmost knots . the origin of the system of microstructures in ngc 7009 is discussed in the light of its physico - chemical parameters and the available theoretical models , but none of these appears able to account for both the lack of a strong density contrast between the outer knots and jetlike structures , and the absence of shocks in them . more than 200 years after its discovery by herschel , the saturn nebula is still challenging our comprehension . we would like to thank the referee , letizia stanghellini , for several useful comments and suggestions that had help us to improve the paper . we are grateful to f. sabbadin for detecting a misidentification ( now corrected ) of features in fig . 1 of the preprint . the work of drg , rlmc and am is partially supported by a grant from the spanish ministry of science and technology ( aya 2001 - 1646 ) . acker , a. , ochsenbein , f. , stenholm , b. , tylenda , r. , marcout , j. , & schohn , c. , 1992 , strasbourg - eso catalogue of galactic planetary nebulae , eso alexander , j. , & balick , b. 1997 , aj , 114 , 713 allende - prieto , c. , lambert , d. l. , & asplund , m. 2001 , apj , 556 , 63 aller , l. h. , & liller , w. 1968 , in nebulae and interstellar matter , ed . b. m. middlehurst & aller ( chicago : univ . press ) , 498 baker , t. 1983 , apj , 267 , 630 balick , b. 1987 , aj , 94 , 671 balick , b. , alexander , j. , hajian , a. r. , terzian , y. , perinotto , m. , & patriarchi , p. 1998 , aj , 116 , 360 balick , b. , owen , r. , bignell , c. r. , & hjellming , r. m. 1987 , aj , 94 , 948 balick , b. , perinotto , m. , maccioni , a. , alexander , terzian , y. , & hajian , a. r. 1994 , apj , 424 , 800 balick b. , preston , h. l , & icke , v. 1987 , aj , 94 , 164 balick b. , rugers , m. , terzian , y. , & chengalur , j. n. 1993 , apj , 411 , 778 blackman , e. g. , frank , a. , & welch , c. 2001 , apj , 546 , 288 bohigas , j. , lpez , j. a. , & aguilar , l. 1994 , a&a , 291 , 595 cahn , j. h. , kaler , j. b. , & stanghellini , l. 1992 , a&as , 94 , 399 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , apj , 345 , 245 cant , j. 1981 , in : investigating the universe , by d. reidel publishing co. , p. 95 cliffe , j. a. , frank , a. , & livio , m. , jones , t. w. 1995 , apj , 447 , l49 copetti , m. v. f. , & writzl , b. c. 2002 , a&a , 382 , 282 corradi , r. l. m. , manso , r. , mampaso , a. , & schwarz , h. e. 1996 , a&a , 313 , 913 corradi , r. l. m. , gonalves , d. r. , villaver , e. , mampaso , a. , & perinotto , m. 2000 , apj , 542 , 861 corradi , r. l. m , perinotto , m. , schwarz , h. e. , claeskens , j - f 1997 , a&a , 322 , 975 czyzak , s. j. , & aller , l. h. 1979 , mnras , 188 , 229 dwarkadas , v. v. , & balick , b. 1998 , apj , 497 , 267 de robertis , m. , dufour , r. & hunt , r. 1987 , jrasc , 81 , 195 frank , a. , balick , b. , & livio , m. 1996 , apj , 471 , l53 frank , a. , & mellema , g. 1994 , apj , 430 , 800 garca - segura , g. 1997 , apj , 489 , 189 garca - segura , g. , langer , n. , rzyczka , m. , & franco , j. 1999 , apj , 517 , 767 garca - segura , g. , & lpez , j. a. 2000 , apj , 544 , 336 gardiner , t. a. , & frank , a. 2001 , apj , 557 , 250 gonalves , d. r. , corradi , r. l. m. , & mampaso , a. 2001 , apj , 547 , 302 grny , s. k. , schwarz , h. e. , corradi , r. l. m. , & van winckel , h. 1999 , a&as , 136 , 145 grevesse , n. , & anders , e. 1989 , in waddington j. , ed . , cosmic abundances in matter , aip , new york , p. 1 guerrero , m. a. , gruendl , r. a. , & chu , y .- h . 200 2 , a&a , 387 , l1 hajian , a. r. , balick , b. , terzian , y. , & perinotto , m. 1997 , apj , 487 , 313 hartigan , p. , raymond , j. , & hartmann , l. 1987 , apj , 316 , 323 hyung , s. , & aller , l. h. 1995 , mnras , 273 , 958 iben , i , & renzini , a. 1983 , ara&a , 21 , 271 kingsburgh , r. l. , & barlow , m. j. 1992 , mnras , 257 , 317 kingsburgh , r. l. , & barlow , m. j. 1994 , mnras , 271 , 257 kwok , s. , purton , c. r. , fitzgerald , p.m. 1978 , apj , 219 , l125 lame , n. j. , & pogge , r. w. 1996 , aj , 111 , 2320 liller , w. 1965 , pasp , 77 , 25 liu , x .- w . , storey , p. j. , barlow , m. j. , clegg , r. e. s. 1995 , mnras , 272 , 369 livio , m. & pringle , j. e. 1997 , apj , 486 , 835 lpez , j. a. , vzquez , r. , & rodrguez , l. f. 1995 , apj , 455 , l63 lpez , j. a. 1997 , in planetary nebulae , proccedings of the 180th symposium of the iau , ed . h. j. habing . and h. j. g. l. lamers . publisher : dordrecht kluwer academic publ . , p. 197 lpez , j. a. 2002 , in `` emission lines from jet flows '' , eds . henney , w. steffen , a.c . raga & l. binette , revmexaa ( conference series ) , 13 , p. 139 maciel , w. j. 1995 , ap&ss , 229 , 203 manchado , a. , guerrero , m. a. , stanghellini , l. , & serra - ricart , m. 1996 , the iac morphological catalog of northern galactic pne , iac publ . mellema , g. , & frank , a. 1995 , mnras , 273 , 401 mndez , r. h. , kudritzki , r. p. , & herrero , a. 1992 , a&a , 260 , 329 miranda , l. f. , & solf , j. 1992 , a&a , 260 , 397 osterbrock , d. e. , 1989 , astrophysics of gaseous nebulae and active galctive nuclei . university science books , ed . mill valey , california . perinotto , m. & corradi , r. m. l. 1998 , a&a , 332 , 721 phillips , j. p. , & cueta , l. 1999 , aj , 118 , 2919 pottasch , s. r. 2000 , in asymmetrical planetary nebulae ii : from origins to microstructures , asp conference series , vol . , edited by j. h. kastner , n. soker , and s. rappaport , p. 289 reay , n. k. , & atherton , p.d . 1985 , mnras , 215 , 233 rubin , r. h. , et al . 2002 , mnras , 334 , 777 shaw , r. a. , & dufour , r. j. 1995 , pasp , 107 , 896 schwarz , h. e. , corradi , r. l. m. , & melnick , j. 1992 , a&as , 96 , 23 schnberner , d. 2002 , in `` planetary nebulae and their role in the universe '' , proceedings of iau symposium no . 209 , asp conference series soker , n. 1990 , aj , 99 , 1869 stasiska , g. 2002 , to be published in the proceedings of the xiii canary islands winter school of astrophysics ( astro - ph/0207500 ) steffen , w. , lpez , j. a. , & lim , a. 2001 , apj , 556 , 823 zhang , c. y. , & kwok , s. 1998 , apjs , 117 , 341 llllllllll [ oii ] 3726.0 + 3728.8 & 182 & 63.5 & 27.3 & 13.8 & 6.52 & 24.8 & 27.0 & 138 & 21.5 + h12 3750.15 & - & 1.87 & 1.28 & 2.12 & 1.64 & 2.05 & 3.53 & 1.72 & 2.05 + h11 3770.6 & 2.14 & 3.25 & 1.65 & 2.31 & 2.10 & 2.53 & 3.95 & 2.35 & 2.48 + h10 3797.9 & 2.07 & 3.37 & 2.77 & 3.67 & 3.05 & 3.68 & 5.52 & 3.49 & 3.69 + h9 3835.4 & 3.08 & 7.90 & 4.49 & 6.01 & 5.40 & 6.21 & 10.2 & 7.35 & 6.10 + [ neiii ] 3868.7 & 74.0 & 114 & 74.2 & 95.5 & 94.4 & 99.4 & 150 & 127 & 97.1 + hei 3888.7 + h8 3889.1 & 12.9 & 25.9 & 15.4 & 20.0 & 14.6 & 19.3 & 32.2 & 22.6 & 18.8 + [ neiii ] 3967.5 + h@xmath46 3970.1 & 32.0 & 53.5 & 36.2 & 46.9 & 40.5 & 46.1 & 70.5 & 54.2 & 44.4 + hei 4026.0 & 1.70 & 2.94 & 1.90 & 2.32 & 2.00 & 2.35 & 3.78 & 2.66 & 2.23 + [ sii ] 4068.6 & 6.81 & 3.20 & 1.71 & 1.38 & 2.17 & 2.15 & 2.63 & 4.62 & 1.82 + [ sii ] 4076.4 & 2.17 & 1.53 & 0.78 & 0.80 & 1.07 & 0.98 & 0.84 & 1.68 & 1.02 + h@xmath47 4101.8 & 14.9 & 30.0 & 21.0 & 28.0 & 22.3 & 28.0 & 40.5 & 27.8 & 27.0 + hei 4120 . & - & - & - & 0.49 & - & - & - & - & 0.52 + oii + hei 4143.7 & - & - & 0.28 & 0.41 & - & - & - & - & 0.41 + [ fev ] 4228 . & - & - & - & 0.26 & - & 0.31 & - & - & - + cii 4267.2 & - & - & 0.30 & 0.64 & 0.76 & 0.68 & 0.47 & - & 0.59 + h@xmath48 4340.5 & 29.2 & 56.1 & 37.9 & 48.1 & 40.5 & 48.5 & 67.2 & 48.5 & 46.8 + [ oiii ] 4363.2 & 6.57 & 9.40 & 5.98 & 7.35 & 8.19 & 7.78 & 13.0 & 9.13 & 7.72 + hei 4387.9 & - & - & 0.65 & 0.64 & - & 0.55 & - & - & 0.65 + hei 4471.5 & 3.71 & 6.07 & 4.20 & 5.01 & 4.27 & 5.11 & 7.57 & 5.84 & 4.85 + heii 4541.7 & - & - & 0.19 & 0.45 & - & - & - & - & 0.35 + niii + oii 4641.0 & - & 1.50 & - & 3.07 & 4.61 & 2.68 & 1.99 & - & 3.09 + ciii 4647.1 & - & 1.28 & 0.59 & 1.13 & - & 0.92 & 1.33 & - & 1.04 + heii 4685.7 & 0.99 & 2.62 & 10.2 & 25.6 & 23.0 & 8.42 & 2.16 & 1.21 & 15.6 + [ ariv ] + hei 4711 + 13 & 2.16 & 4.62 & 3.23 & 5.40 & 4.89 & 2.78 & 5.08 & 1.97 & 4.04 + [ ariv ] 4740.2 & 1.13 & 3.63 & 2.78 & 5.50 & 4.49 & 2.54 & 3.95 & 1.48 & 3.88 + h@xmath49 4861.3 & 100 & 100 & 100 & 100 & 100 & 100 & 100 & 100 & 100 + [ oiii ] 4958.9 & 428 & 473 & 435 & 405 & 428 & 424 & 458 & 454 & 421 + [ oiii ] 5006.86 & 1256 & 1339 & 1268 & 1179 & 1243 & 1227 & 1381 & 1329 & 1223 + [ ariii ] 5191.8 & - & - & 0.09 & - & - & - & - & - & 0.07 + [ ni ] 5200.2 & 6.06 & - & 0.06 & - & - & - & - & 2.82 & 0.11 + [ feiii ] 5270 . & - & - & - & - & - & - & - & - & 0.08 + [ feiii ] + [ cav 5292 . & - & - & - & - & - & - & - & - & 0.09 + heii 5411.6 & - & - & 0.72 & 1.48 & 1.67 & 0.95 & - & - & 1.12 + [ cliii ] 5517.7 & - & - & 0.71 & 0.46 & 0.46 & 0.55 & 0.91 & 1.03 & 0.58 + [ cliii ] 5537.9 & - & 0.40 & 0.85 & 0.57 & 0.59 & 0.71 & 0.79 & 0.96 & 0.68 + [ oi ] 5577.4 & 0.51 & - & 0.14 & - & - & - & - & - & 0.01 + [ oiii ] 5592.4 & - & - & - & 0.03 & 0.40 & - & - & - & 0.05 + [ nii ] 5754.6 & 6.95 & - & 0.72 & 0.15 & 0.20 & 0.69 & - & 4.19 & 0.50 + hei 5875.7 & 20.3 & 17.0 & 17.9 & 15.0 & 15.3 & 15.8 & 15.8 & 16.2 & 15.7 + [ kiv ] 6102 . & - & - & - & 0.16 & 0.18 & 0.11 & - & - & 0.15 + heii 6166 + 6170 & - & - & - & - & - & - & - & - & 0.06 + heii 6234 . & - & - & - & - & - & - & - & - & 0.06 + [ oi ] 6300.3 & 31.0 & - & 1.02 & - & - & 1.69 & - & 14.8 & 1.06 + [ siii ] 6312.1 & 4.29 & 1.80 & 2.84 & 1.41 & 1.26 & 1.97 & 1.50 & 3.59 & 1.86 + siii 6347.1 & - & - & - & 0.07 & - & - & - & - & - + [ oi ] 6363.8 & 11.3 & - & 0.38 & 0.12 & 0.17 & 0.56 & - & 5.33 & 0.36 + siii 6371 . & - & - & - & 0.10 & 0.12 & - & - & - & - + heii 6406.5 & - & - & - & 0.19 & 0.19 & - & - & - & - + [ nii ] 6548.0 & 132 & 10.8 & 17.9 & 3.57 & 3.57 & 13.9 & 7.05 & 72.5 & 11.2 + h@xmath50 6562.8 & 394 & 337 & 408 & 312 & 341 & 280 & 217 & 224 & 313 + [ nii ] 6583.4 & 397 & 19.5 & 53.8 & 7.93 & 6.26 & 39.2 & 14.7 & 217 & 30.2 + hei 6678.1 & 8.41 & 5.52 & 6.16 & 4.35 & 4.44 & 4.00 & 3.74 & 3.57 & 4.46 + [ sii ] 6716.5 & 41.4 & 2.52 & 4.82 & 0.54 & 0.43 & 3.13 & 2.59 & 26.1 & 2.63 + [ sii ] 6730.8 & 57.4 & 3.09 & 8.19 & 0.95 & 0.78 & 5.42 & 3.05 & 32.0 & 4.34 + + @xmath26 & 2000 & 1300 & 4500 & 5500 & 5900 & 5000 & 1100 & 1300 & 4000 + @xmath26 & - & - & 4700 & 5200 & 5900 & 6000 & 1300 & 1900 & 4500 + @xmath0 & 9600 & 10400 & 9300 & 10000 & 10200 & 10100 & 11600 & 10400 & 10100 + @xmath0 & 11000 & - & 9400 & 10400 & 12800 & 10400 & - & 11700 & 10300 + @xmath0 & 7100 & - & 8300 & - & - & - & - & 9400 & - + lccccccccc ( 0.010.05)i@xmath51 & 26 & 31 & 5 & 6 & 10 & 4.5 & 11 & 13.5 & 4 + ( 0.050.15)i@xmath51 & 14 & 10 & 3.5 & 4 & 6.5 & 3.5 & 7.5 & 10 & 3.5 + ( 0.150.30)i@xmath51 & 9 & 8 & 3.5 & 3.5 & 4.5 & 3.5 & 6 & 7.5 & 3 + ( 0.302.0)i@xmath51 & 6 & 6 & 3.5 & 3.5 & 3.5 & 3 & 4.5 & 5.5 & 3 + ( 2.05.0)i@xmath51 & 5 & 4.5&3 & 3.5 & 3.5 & 3 & 4 & 5 & 3 + ( 5.010.0)i@xmath51 & 4.5&4 & 3 & 3 & 3 & 3 & 4 & 5 & 3 + @xmath52 10 i@xmath51 & 4 & 4 & 3 & 3 & 3 & 3 & 4 & 5 & 3 + + i@xmath51 & 1.05(6 ) & 1.24(6 ) & 30.2(4 ) & 30.7(4 ) & 15.5(4 ) & 70.2(3 ) & 1.60(5 ) & 1.28(6 ) & 188(3 ) + lllllllllll he@xmath40/h & 1.02e-1(06 ) & 1.23e-1(05 ) & 1.04e-1(04 ) & 9.54e-2(04 ) & 9.18e-2(04 ) & 9.46e-2(04 ) & 1.07e-1(05 ) & 9.55e-2(06 ) & 9.77e-2(04 ) & + he@xmath53/h & - & 1.2e-3(10 ) & 5.02e-3(06 ) & 1.24e-2(07 ) & 2.43e-2(06 ) & 1.07e-2(07 ) & 1.87e-3(08 ) & - & 1.30e-2(06 ) & + * he / h * & * 1.02e-1(09 ) * & * 1.24e-1(08 ) * & * 1.09e-1(06 ) * & * 1.08e-1(08 ) * & * 1.16e-1(06 ) * & * 1.05e-1(06 ) * & * 1.08e-1(08 ) * & * 9.55e-2(09 ) * & * 1.11e-1(06 ) * & * 1.107e-1 * + o@xmath54/h & 4.5e-5(11 ) & - & 2.40e-6(09 ) & 5.0e-7(13 ) & 3.1e-7(16 ) & 2.61e-6(08 ) & - & 1.8e-5(13 ) & 1.71e-6(08 ) & + o@xmath40/h & 7.5e-5(19 ) & 2.2e-5(13 ) & 2.4e-5(14 ) & 7.0e-6(15 ) & 1.3e-6(16 ) & 1.4e-5(14 ) & 6.06e-6(17 ) & 4.2e-5(21 ) & 1.2e-5(13 ) & + o@xmath53/h & 5.12e-4(09 ) & 4.32e-4(08 ) & 5.84e-4(06 ) & 4.12e-4(07 ) & 4.11e-4(06 ) & 4.25e-4(06 ) & 2.9e-4(10 ) & 4.1e-4(10 ) & 4.21e-4(06 ) & + _ icf_(o ) & 1.00 & 1.00 & 1.03 & 1.08 & 1.16 & 1.07 & 1.01 & 1.00 & 1.08 & + * o / h * & * 5.8e-4(15 ) * & * 6.4e-4(12 ) * & * 6.28e-4(09 ) * & * 4.5e-4(11 ) * & * 4.82e-4(09 ) * & * 4.7e-4(11 ) * & * 3.0e-4(15 ) * & * 4.5e-4(17 ) * & * 4.71e-4(09 ) * & * 5.2e-4 * + n@xmath54/h & 8.7e-6(16 ) & - & 2.2e-7(12 ) & - & - & - & - & 2.7e-6(19 ) & 2.7e-7(11 ) & + n@xmath40/h & 5.00e-5(08 ) & 3.09e-6(07 ) & 9.66e-6(06 ) & 1.10e-6(07 ) & 5.19e-7(08 ) & 5.59e-6(06 ) & 1.7e-6(10 ) & 2.38e-5(09 ) & 4.45e-6(06 ) & + _ icf_(n ) & 7.8 & 20.1 & 25.7 & 64.7 & 355 & 32.1 & 50.5 & 10.6 & 38.2 & + * n / h * & * 3.8e-4(35 ) * & * 6.2e-5(26 ) * & * 2.4e-4(24 ) * & * 7.0e-5(26 ) * & * 1.8e-4(29 ) * & * 1.8e-4(24 ) * & * 8.6e-5(34 ) * & * 2.5e-4(38 ) * & * 1.7e-4(23 ) * & * 1.3e-4 * + ne@xmath53/h & 9.8e-5(14 ) & 1.1e-4(12 ) & 1.1e-4(10 ) & 1.0e-4(10 ) & 9.6e-5(10 ) & 1.0e-4(10 ) & 9.3e-5(17 ) & 1.2e-4(15 ) & 1.05e-4(09 ) & + _ ic_f(ne ) & 1.14 & 1.06 & 1.07 & 1.10 & 1.17 & 1.11 & 1.03 & 1.10 & 1.11 & + * ne / h * & * 1.1e-4(29 ) * & * 1.1e-4(26 ) * & * 1.2e-4(20 ) * & * 1.1e-4(21 ) * & * 1.1e-4(20 ) * & * 1.1e-4(20 ) * & * 9.6e-5(34 ) * & * 1.3e-4(30 ) * & * 1.1e-4(18 ) * & * 1.9e-4 * + s@xmath40/h & 2.2e-6(13 ) & 8.99e-8(08 ) & 5.11e-7(08 ) & 4.4e-8(11 ) & 2.2e-8(13 ) & 2.7e-7(10 ) & 7.0e-8(10 ) & 1.0e-6(14 ) & 2.09e-7(06 ) & + s@xmath53/h & 7.45e-6(21 ) & 3.2e-6(15 ) & 7.4e-6(14 ) & 2.1e-6(16 ) & 9.7e-7(19 ) & 3.4e-6(14 ) & 1.6e-6(19 ) & 4.8e-6(23 ) & 3.3e-6(13 ) & + _ icf_(s ) & 1.43 & 1.91 & 2.07 & 2.79 & 4.91 & 2.22 & 2.58 & 1.57 & 2.35 & + * s / h * & * 1.39e-5(30 ) * & * 6.3e-6(23 ) * & * 1.6e-5(23 ) * & * 6.1e-6(26 ) * & * 4.9e-6(30 ) * & * 8.1e-6(23 ) * & * 4.4 e-6(29 ) * & * 9.3e-6(35 ) * & * 8.3e-6(26 ) * & * 11.1e-6(26 ) * +

we define a family of maps on lattice paths , called _ sweep maps _ , that assign levels to each step in the path and sort steps according to their level . surprisingly , although sweep maps act by sorting , they appear to be bijective in general . the sweep maps give concise combinatorial formulas for the-catalan numbers , the higher-catalan numbers , the-square numbers , and many more general polynomials connected to the nabla operator and rational catalan combinatorics . we prove that many algorithms that have appeared in the literature ( including maps studied by andrews , egge , gorsky , haglund , hanusa , jones , killpatrick , krattenthaler , kremer , orsina , mazin , papi , vaill , and the present authors ) are all special cases of the sweep maps or their inverses . the sweep maps provide a very simple unifying framework for understanding all of these algorithms . we explain how inversion of the sweep map ( which is an open problem in general ) can be solved in known special cases by finding a `` bounce path '' for the lattice paths under consideration . we also define a generalized sweep map acting on words over arbitrary alphabets with arbitrary weights , which is also conjectured to be bijective .

1406.1196

this paper introduces a family of sorting maps on words that we call _ sweep maps_. in its simplest form , a sweep map @xmath1 uses coprime parameters @xmath2 and @xmath3 to associate a _ level _ @xmath4 to each letter @xmath5 in a word @xmath6 consisting of @xmath7 copies of the letter @xmath8 and @xmath9 copies of the letter @xmath10 . ( note that @xmath2 or @xmath3 may be negative . ) this assignment is done recursively : use the convention that @xmath11 ; for @xmath12 we set @xmath13 if @xmath14 and @xmath15 if @xmath16 . the word @xmath17 is then obtained by sorting the letters in @xmath18 according to level , starting with @xmath19 , then continuing with @xmath20 . figure [ fig : sweepmap1 ] provides an example of @xmath21 acting on the word @xmath22 . ( here we have identified @xmath18 with a lattice path in the plane : each @xmath8 corresponds to a unit - length north step , while each @xmath10 corresponds to a unit - length east step . ) surprisingly , even though sweep maps act by sorting , they are ( apparently ) bijective . the reader may find it useful to check this bijectivity by hand for @xmath23 acting on the set of all lattice paths from @xmath24 to @xmath25 . as detailed in conjecture [ conj : gen - sweep ] , bijectivity seems to hold even for the general sweep maps over arbitrary alphabets with arbitrary weights , described in section [ subsec : gen ] . the bijectivity of the general sweep maps appears to be a very subtle and difficult fact . the order in which the levels are traversed is a key ingredient to bijectivity . for example , in the case of @xmath26 , @xmath27 , if we scan levels in the order @xmath28 , both of the paths @xmath29 and @xmath30 map to @xmath31 . the sweep maps encode complex combinatorial information related to @xmath0-catalan numbers , the bergeron - garsia nabla operator , and other constructions arising in representation theory , algebraic geometry , and symmetric functions . researchers have discovered special cases of the sweep map in many different guises over the last fifteen years or so . one of the goals of this paper is to present a unifying framework for all of this work . in @xcite , loehr introduced bijections on @xmath32-dyck paths , as well as generalizations to lattice paths contained in certain trapezoids , that turn out to be special cases of the sweep map . the bijection in the case @xmath33 also appears in a paper of haglund and loehr @xcite and is foreshadowed by a counting argument in haglund s seminal paper on @xmath0-catalan numbers ( * ? ? ? * proof of thm . the inverse bijection in the case @xmath33 appears even earlier , where it was used by andrews et al . @xcite in their study of @xmath34-nilpotent @xmath35-ideals in the lie algebra @xmath36 . ( see also @xcite . ) more recently , special cases of the sweep map have arisen while studying lattice paths in squares @xcite ; partition statistics @xcite ; simultaneous core partitions @xcite ; and compactified jacobians @xcite . we discuss a number of these connections in more detail in section [ sec : alg - sweep ] . we suspect that to the typical mathematician , the most interesting question regarding the sweep maps is whether they are bijective ( as conjectured in conjecture [ conj : gen - sweep ] ) . for a researcher interested in the @xmath0-catalan numbers , however , of comparable interest is the connection between the sweep maps and statistics on lattice paths such as @xmath37 , @xmath38 and @xmath39 . since shortly after haiman s introduction of @xmath39 , it has been known that a `` zeta map '' takes @xmath39 to @xmath37 to @xmath38 . one point of view , then , is that rather than having three statistics on dyck paths , we have one statistic @xmath37 and a sweep map . many polynomials related to the @xmath0-catalan numbers can be defined using only an `` area '' and an appropriate sweep map . that these polynomials are jointly symmetric ( conjecturally ) supports the utility of this view ( see section [ sec : area - qtcat ] ) . the structure of this paper is as follows . section [ sec : basic ] introduces the necessary background on lattice paths . we then define sweep maps and present conjecture [ conj : gen - sweep ] on their bijectivity in section [ sec : intro - sweep ] . section [ sec : alg - sweep ] reviews various algorithms that have appeared in the literature that are equivalent to special cases of the sweep map , while section [ sec : invert - sweep ] describes how to invert these maps ( when known ) . finally , section [ sec : area - qtcat ] shows how the sweep maps may be used to give concise combinatorial formulas for the higher @xmath0-catalan numbers and related polynomials formed by applying the nabla operator to appropriate symmetric functions and then extracting the coefficient of @xmath40 . an extended abstract of this paper appears as @xcite . this section introduces our basic conventions regarding partitions , words , and lattice paths . integer parameters @xmath41 and @xmath42 will serve to restrict our attention to such objects fitting within a rectangle of height @xmath41 and width @xmath42 . integer parameters @xmath2 and @xmath3 will be used to assign a `` level '' to the components of the various objects considered . of particular interest is the case of @xmath43 and @xmath44 . the constraint of @xmath45 arises naturally in some particular sweep maps such as the map due to armstrong , hanusa and jones @xcite and the map due to gorksy and mazin @xcite ( see sections [ subsec : zetamap ] and [ subsec : gorsky - mazin ] , respectively ) . let @xmath46 . integer partition diagrams with at most @xmath41 parts and largest part at most @xmath42 ( drawn according to the english convention ) fit in the rectangle with vertices @xmath24 , @xmath47 , @xmath48 and @xmath49 . we denote the set of such partitions ( identified with their diagrams , which are collections of unit squares in the first quadrant ) by @xmath50 . let @xmath51 denote the set of all words @xmath52 , @xmath53 , for which each @xmath54 , and let @xmath55 denote the subset of words consisting of @xmath41 copies of @xmath56 and @xmath42 copies of @xmath57 . finally , let @xmath58 denote the set of lattice paths from @xmath24 to @xmath49 consisting of @xmath41 unit - length north steps and @xmath42 unit - length east steps . there are natural bijections among the three sets @xmath59 , @xmath60 and @xmath61 . each word @xmath62 encodes a lattice path in @xmath61 by letting each @xmath10 correspond to an east step and each @xmath8 correspond to a north step . the frontier of a partition @xmath63 also naturally encodes a path in @xmath61 . we write @xmath64 or @xmath65 for the word associated with a path @xmath66 or a partition @xmath67 , respectively . operators @xmath68 and @xmath69 are defined analogously . for example , taking @xmath70 and @xmath71 , the path @xmath66 shown on the left in figure [ fig : sweepmap1 ] has @xmath72 and @xmath73 . for the word @xmath74 , @xmath75 is the path shown on the right in figure [ fig : sweepmap1 ] , whereas @xmath76 . let @xmath77 . we assign a * level * to each square of a partition and each step of a path in the following manner . first , assign to each lattice point @xmath78 the * @xmath79-level * @xmath80 . assign to each lattice square the level of its southeast corner . we will have occasion to consider two different conventions for associating levels to north and east steps of paths . for the * east - north ( e - n ) * convention , each east ( resp . north ) step inherits the level of its eastern ( resp . northern ) endpoint . for the * west - south ( w - s ) * convention , each east ( resp . north ) step inherits the level of its western ( resp . southern ) endpoint . figure [ fig : labeling ] illustrates the various levels relevant to the word @xmath81 for @xmath82 and @xmath83 . let @xmath84 denote the set of lattice paths in @xmath85 whose steps all have nonnegative @xmath86-levels under the e - n convention . we call these paths * @xmath79-dyck paths*. a word @xmath87 is an * @xmath79-dyck word * iff @xmath75 is an @xmath86-dyck path ; let @xmath88 denote the set of such words . a partition @xmath89 is an * @xmath79-dyck partition * iff @xmath90 is an @xmath86-dyck path ; let @xmath91 denote the set of such partitions . we begin in section [ subsec : def - sweep ] by giving an algorithmic description of the basic sweep maps for words over the alphabet @xmath92 . some minor variations are presented in sections [ subsec : irrat - sweep ] and [ subsec : minor - var ] . we then present a general sweep map in section [ subsec : gen ] that acts on words over any alphabet with arbitrary weights assigned to each letter . let @xmath93 . we first describe the * @xmath94-sweep map * , @xmath95 . given @xmath96 , assign levels using the east - north convention applied to @xmath75 . define a word @xmath97 by the following algorithm . initially , @xmath98 is the empty word . for @xmath99 and then for @xmath100 , scan @xmath18 from right to left . whenever a letter @xmath5 is encountered with level equal to @xmath101 , append @xmath5 to @xmath98 . the * @xmath102-sweep map * @xmath103 is defined the same way as @xmath104 , except that : the value @xmath105 is the first value of @xmath101 used rather than the last ; and for each value of @xmath101 , @xmath18 is scanned from left to right . figure [ fig : negsweep - ex ] illustrates the action of both @xmath106 and @xmath107 on a path in @xmath108 . we define the action of @xmath104 on a partition @xmath67 as @xmath109 and the action of @xmath110 on a path as @xmath111 ; similarly for @xmath103 . geometrically , we think of each step in @xmath75 as a _ wand _ whose _ tip _ is located at the lattice point at the end of each step . this visual description reminds us that the maps @xmath112 are assigning levels to steps in the path using the east - north convention . for @xmath113 and @xmath114 , @xmath115 acts by scanning southwest along each diagonal line @xmath116 ( taking @xmath101 s in the appropriate order ) and `` sweeping up '' the wands whose tips lie on each of these diagonals . the wands are laid out in the order in which they were swept up to produce the output lattice path . the labels on the wand tips are _ not _ part of the final output . it is clear from this description , or from the original definition , that the sweep map depends only on the `` slope '' @xmath117 ; i.e. , @xmath118 for every positive integer @xmath32 . the @xmath119 statistic introduced in section [ subsec : area - stat ] is computed using the @xmath86-levels of the steps in @xmath18 , which depend not just on the ratio @xmath117 but on the specific values of @xmath2 and @xmath3 . so it is not always safe to assume that @xmath2 and @xmath3 are relatively prime . since the sweep maps rearrange the letters in the input word @xmath18 , it is immediate that both @xmath104 and @xmath103 map each set @xmath120 into itself . we will see later that @xmath104 maps each set @xmath88 into itself . so far , for each rational @xmath117 , we have defined the basic sweep map @xmath115 ( which we also call the * negative - type * sweep map ) and the * positive - type * sweep map @xmath121 . we can extend this setup to define sweep maps @xmath122 indexed by _ irrational _ numbers @xmath123 . we regard inputs to @xmath122 as lattice paths @xmath124 consisting of `` wands '' with tips at their north ends and east ends . there is a `` sweep line '' @xmath125 that sweeps through the plane as @xmath101 decreases from just below zero to negative infinity , then from positive infinity to zero . because @xmath123 is irrational , every sweep line intersects at most one wand tip in @xmath75 . we obtain @xmath126 by writing the wands in the order in which the sweep line hits the wand tips . for fixed @xmath127 and fixed @xmath128 , one readily checks that @xmath129 for all irrationals @xmath123 with @xmath130 and @xmath123 sufficiently close to @xmath117 . on the other hand , @xmath131 for all irrationals @xmath123 with @xmath132 and @xmath123 sufficiently close to @xmath117 . this explains the terminology `` positive - type '' and `` negative - type '' sweep map . one approach to studying the sweep map is to understand the `` jump discontinuities '' between @xmath103 and @xmath104 that occur at certain critical rationals @xmath117 . for irrational @xmath123 , we say @xmath18 is a * @xmath123-dyck word * iff for all lattice points @xmath78 visited by @xmath75 , @xmath133 . \(a ) for all irrational @xmath123 , if @xmath134 is a @xmath123-dyck word then @xmath135 is a @xmath123-dyck word . ( b ) for all @xmath136 , if @xmath134 is an @xmath86-dyck word , then @xmath137 is an @xmath86-dyck word . \(a ) if @xmath138 then all words are @xmath123-dyck words , so the proposition certainly holds . therefore , in the following we assume @xmath139 . fix @xmath140 ; it suffices to show that the point @xmath78 reached by the first @xmath141 steps of @xmath142 satisfies @xmath143 . since @xmath123 is irrational and @xmath18 is a @xmath123-dyck word , there exists a real @xmath144 such that @xmath145 consists of all the symbols in @xmath18 with wand tips at levels higher than @xmath101 . in other words , @xmath145 is a rearrangement of all the steps of @xmath75 that end at points @xmath78 with @xmath146 . now , @xmath75 begins at the origin , which has level zero . in general , this path enters and leaves the region @xmath147 several times . let @xmath148 be the maximal substrings of consecutive letters in @xmath18 such that every step of @xmath149 ends in @xmath150 . every @xmath149 begins with a north step that enters @xmath150 from below , and every @xmath149 except possibly @xmath151 is followed by an east step that exits @xmath150 to the right . suppose @xmath149 consists of @xmath152 north steps and @xmath153 east steps ; since the boundary of @xmath150 is a line of slope @xmath123 , the geometric fact in the previous sentence implies that @xmath154 for @xmath155 . by definition , @xmath145 is some rearrangement of @xmath156 north steps and @xmath157 east steps . now @xmath158 for all @xmath159 implies @xmath160 . thus , @xmath145 is a path from @xmath24 to @xmath49 , where @xmath160 . since @xmath141 was arbitrary , @xmath142 is a @xmath123-dyck path . \(b ) it suffices to treat the case @xmath113 , @xmath114 . fix @xmath127 and @xmath161 . we can choose an irrational @xmath162 so close to @xmath117 that the region @xmath163 contains no lattice points , and @xmath164 . since @xmath18 is an @xmath86-dyck word and @xmath162 , @xmath18 is also a @xmath123-dyck word . by part ( a ) , @xmath135 is a @xmath123-dyck word . since @xmath165 contains no lattice points , @xmath166 is an @xmath86-dyck word , as needed . for a fixed choice of @xmath2 and @xmath3 , there are four parameters that can be used to define a potential sweep map : * the level to start sweeping at , * the direction of sweep for a given level ( i.e. , right - to - left or left - to - right ) , * the relative order in which to visit levels ( i.e. , @xmath167 after level @xmath101 versus @xmath168 after level @xmath101 ) , * the convention for levels assigned to steps ( i.e. , using the west - south convention or the east - north convention ) . empirical evidence suggests that for each of the @xmath169 possible choices for the second through fourth parameters , there is a unique choice of starting level that will lead to a bijective sweep map for general @xmath85 . in fact , each of these maps is closely related to the others through the following two natural involutions on words . let @xmath170 be the * reversal map * given by @xmath171 . let @xmath172 be the * transposition map * that acts by interchanging @xmath8 s and @xmath10 s . evidently , both @xmath173 and @xmath174 are involutions ; @xmath173 maps @xmath85 bijectively onto itself , whereas @xmath174 maps @xmath85 bijectively onto @xmath175 . we can modify the sweep maps by composing on the left or right with @xmath173 and/or @xmath174 . the new maps are bijections ( between appropriate domains and codomains ) iff the original sweep maps are bijections . table [ tab : sweeps ] displays the eight maps along with their relationships to @xmath104 and @xmath103 . one can also check that @xmath176 the following degenerate cases of the sweep map are readily verified . * if @xmath177 and @xmath114 , then @xmath178 , the identity map on @xmath51 . * if @xmath113 and @xmath179 , then @xmath180 , the reversal map . * for @xmath181 , @xmath182 and @xmath183 . * if @xmath184 and @xmath114 , then @xmath185 , whereas @xmath186 maps @xmath187 ( where @xmath188 ) to @xmath189 . similar statements hold in the cases : @xmath184 and @xmath179 ; @xmath190 and @xmath177 ; @xmath190 and @xmath113 . .symmetries of sweep maps . [ cols="^,^,^,^,^,^ " , ] where the zero in parentheses is the level following the final east step . the subwords @xmath191 are @xmath192 by definition of the levels , one sees that the maximum level of any letter in @xmath18 is the same value @xmath193 appearing in the definition of @xmath194 . since @xmath195 , it will suffice to prove that @xmath196 , @xmath197 for @xmath198 , and @xmath199 for all @xmath159 with @xmath200 ( as illustrated by example [ ex : phi ] ) . define @xmath201 for all @xmath159 ; this is the word obtained by scanning @xmath18 from left to right and taking all letters at level @xmath159 . by reversing everything , it is enough to prove that @xmath202 , @xmath203 for @xmath198 , and @xmath204 for all @xmath159 with @xmath200 . * step 2 . * fix a level @xmath205 . we define an * event sequence * in @xmath206 associated with a left - to - right scan of level @xmath159 . it follows from step 1 that the levels of the north steps of @xmath66 will be @xmath207 in this order . as we scan @xmath18 during this iteration , the following * events * may occur : * we scan an @xmath8 of @xmath18 at level @xmath159 , which appends an @xmath8 onto both @xmath208 and @xmath209 . * we scan an @xmath10 of @xmath18 at level @xmath159 , which appends an @xmath10 onto @xmath209 . * we scan an @xmath8 of @xmath18 with level in @xmath210 , which appends an @xmath10 onto @xmath208 . consider the sequence of events a , b , c that occur during the @xmath205 scan . in our example , the @xmath211 scan has event sequence bcbcabcab , whereas the @xmath212 scan has event sequence caabcbcabcb . * * we prove that @xmath202 . for the @xmath213 scan , events b and c are impossible , since the path stays within the trapezoid . so the event sequence consists of @xmath141 a s for some @xmath141 , and @xmath214 and @xmath215 both consist of @xmath141 @xmath56 s . * step 4 . * for @xmath216 , we analyze the possible transitions between events a , b , and c that may occur during the @xmath205 scan . note that events a and b can only occur when the level of the current character in @xmath18 is @xmath217 , whereas event c only occurs when this level is @xmath218 . moreover , the only way to transition from a level @xmath217 to a level @xmath218 is via event b , and the only way to transition from a level @xmath218 to a level @xmath217 is via event c. consequently , in the event sequence for @xmath205 , every a ( not at the end ) can only be followed by a or b ; every b ( not at the end ) can only be followed by c ; and every c ( not at the end ) can only be followed by a or b. the path @xmath66 ends at level @xmath219 , so the event sequence must end in a b. * step 5 . * we prove that @xmath220 for @xmath198 . since @xmath221 and the origin has level @xmath101 , the first letter in the event sequence must be a or b. by step 4 , the event sequence is some rearrangement of a s and bc s , except there is an unmatched b at the end . by definition of the events in step 2 , this means that @xmath209 and @xmath208 agree , except for an extra @xmath10 at the end of @xmath209 . * * we prove that @xmath222 for @xmath200 . since @xmath223 and the origin has level @xmath101 , the first letter in the event sequence must be an unmatched c. thereafter , the event sequence consists of a s and matched bc pairs , with one unmatched b at the end . the initial c gives the initial @xmath10 in @xmath208 that is deleted to form @xmath224 . as in step 5 , we see that @xmath209 and @xmath224 agree , except for an extra @xmath10 at the end of @xmath209 caused by the unmatched b. the proof structure above can be readily adapted to show that other algorithms based on area vectors are equivalent to suitable sweep maps . for this reason , we will omit the details of these proofs in the remainder of this subsection . for instance , one can modify the preceding proof to show that the map @xmath225 ( which acts on dyck paths of order @xmath226 ) is also a sweep map . for all @xmath227 and @xmath228 , @xmath229 . similarly , let @xmath230 denote the map described in ( * ? ? ? * , bijections ) that sends unlabeled dyck paths to unlabeled dyck paths . for all @xmath227 and @xmath228 , @xmath231 . we note that the partition @xmath232 is the transpose of the partition @xmath233 . since @xmath234 for all partitions @xmath67 ( where @xmath235 denotes the transpose of @xmath67 ) , the theorem for @xmath230 follows from the theorem for @xmath225 and vice versa . in @xcite , loehr and warrington modified the map @xmath230 to obtain a bijection @xmath236 on @xmath237 , the set of lattice paths in an @xmath238 square . given @xmath239 , we define its * area vector * @xmath240 by letting @xmath241 be the number of complete squares in the strip @xmath242 that lie to the right of @xmath66 and to the left of @xmath243 . ( this reduces to the previous area vector if @xmath66 is a dyck path . ) the area vectors of paths in @xmath237 are characterized by the following properties : @xmath244 ; @xmath245 for @xmath246 ; and @xmath247 for @xmath248 . given @xmath239 , we define a new path @xmath249 as follows . for all @xmath250 , let @xmath251 be the subword of @xmath252 consisting of all occurrences of @xmath159 and @xmath253 . create words @xmath208 from @xmath251 by replacing each @xmath159 by @xmath10 and each @xmath253 by @xmath8 . for all @xmath254 , let @xmath255 be the reversal of @xmath208 . for all @xmath256 , @xmath208 must end in @xmath10 , so we can write @xmath257 ; let @xmath258 . finally , define @xmath259 and set @xmath260 . let @xmath261 be such that @xmath262 then @xmath263 , so ( for instance ) @xmath264 , @xmath265 , @xmath266 , @xmath267 , @xmath268 , @xmath269 , and so on . next , @xmath270 is the concatenation of the words @xmath271 , @xmath269 , @xmath272 , @xmath273 , @xmath266 , and @xmath274 , and @xmath260 . on the other hand , the reader can check that @xmath275 . in fact , for all @xmath250 , the subword @xmath255 is precisely the subword @xmath191 of letters in @xmath276 coming from letters in @xmath18 with @xmath277-level ( using the e - n convention ) equal to @xmath159 . one can prove this always happens , by adapting the ideas in the proof of theorem [ thm : trapz - vs - sweep ] , to obtain the following theorem . for all @xmath227 and all @xmath239 , @xmath278 . a * schrder path * of order @xmath226 is a path from the origin @xmath24 to @xmath279 , never going below @xmath280 , with the allowed steps being north steps of length @xmath281 , east steps of length @xmath281 , and a northeast step of length @xmath282 . in ( * theorem 6 ) , egge , haglund , killpatrick , and kremer extend @xmath230 to a bijection @xmath283 acting on schrder paths . after converting from paths to words , @xmath283 is the sweep map @xmath284 associated to the alphabet @xmath285 with weight function @xmath286 , @xmath287 , and @xmath288 . throughout this subsection , fix positive integers @xmath41 and @xmath42 with @xmath45 . in @xcite , d. armstrong , c. hanusa and b. jones investigate the combinatorics of @xmath289-cores . in the process , they define a map @xmath290 . for @xmath291 , the partition @xmath292 is defined in two stages . first , create a partition @xmath293 as follows . consider the levels of all lattice squares lying above @xmath294 and below the path @xmath90 . since @xmath45 , these levels must all be distinct . sort these levels into increasing order , and write them in a column from bottom to top . let @xmath293 be the unique partition such that these levels are the hook - lengths of the cells in the first column . ( recall that the * hook * of a cell @xmath295 in a partition diagram consists of @xmath295 itself , all cells below @xmath295 in its column , and all cells right of @xmath295 in its row . the * hook - length * of @xmath295 , denoted @xmath296 , is the number of cells in the hook of @xmath295 . ) the second stage maps @xmath297 to a new partition @xmath298 as follows . there will be one nonzero row of @xmath299 for each row of @xmath297 whose first - column hook - length is the level of a square directly east of a north step of @xmath90 . to determine the length of each row in @xmath299 , count the number of cells of @xmath297 in the corresponding row whose hook - length is less than or equal to @xmath42 . the @xmath300 map is then defined by @xmath301 . see figure [ fig : drewmap - ex ] for an example . [ thm : drew - sweep ] for all @xmath302 , @xmath303 . to prove this theorem , we will introduce alternate formulations of the maps @xmath304 and @xmath305 , denoted @xmath306 and @xmath307 , that focus attention on the lattice paths making up the frontiers of @xmath67 , @xmath297 , and @xmath299 . this will enable us to compare the action of the @xmath300 map on partitions to the action of the sweep map on lattice paths . to start , define @xmath308 by setting @xmath309 , where @xmath310 , and for all @xmath311 , @xmath312 since @xmath90 begins at @xmath24 and ends at @xmath49 , we must have @xmath313 for all @xmath314 . [ lem : nu ] for all @xmath315 , @xmath316 . we first observe that the partitions @xmath317 and @xmath318 will have the same number of ( positive - length ) rows . for , on one hand , the first - column hook - lengths in @xmath318 will be the levels of the squares between @xmath90 and @xmath294 . on the other hand , these same levels will be the indices of the north steps of @xmath319 . now we show that the row lengths will be the same . consider the @xmath159-th row from the bottom ( starting with @xmath320 for the bottom row ) in each partition . suppose the first - column hook - length in this row in @xmath318 is @xmath101 . by definition , the length of this row in @xmath318 will be @xmath321 . additionally , since we are in the @xmath159-th row from the bottom , it follows that of the values @xmath322 , exactly @xmath253 are levels of lattice squares below @xmath90 , while the remaining @xmath321 values are levels of lattice squares above @xmath90 . but the levels corresponding to squares above @xmath90 map to @xmath10 s under @xmath306 . so @xmath323 , which implies that the number of cells in the @xmath159-th row from the bottom of @xmath317 will also be @xmath321 . we now define an analog of the map @xmath305 that maps @xmath324 to a new partition @xmath299 . with the word @xmath325 defined as above , let @xmath326 be the subword of @xmath327 formed by retaining only those @xmath328 for which @xmath329 is the level ( using the w - s convention ) of a step of @xmath90 ; then set @xmath330 . ( technically , @xmath307 depends not only on the partition @xmath297 , but on @xmath41 , @xmath42 and @xmath331 as well . however , @xmath41 and @xmath42 are fixed and we consider @xmath307 only as part of the composition @xmath332 . ) [ lem : drew - sweep ] for all @xmath291 , @xmath333 . recall from table [ tab : sweeps ] that the sweep map variation on the right side of the lemma acts on @xmath65 by scanning the levels @xmath334 in this order , sweeping up path steps with levels assigned according to the w - s convention . ( since @xmath45 , each level appears at most once in @xmath65 . also , for @xmath43 and @xmath44 , @xmath335 . ) to compare this sweep map to the action of @xmath336 , first note that a north step on the frontier of @xmath67 has level @xmath329 iff the lattice square directly east of that north step has level @xmath159 . such lattice squares are encoded as @xmath337 by @xmath306 and are then retained by @xmath307 . similarly , an east step on the frontier of @xmath67 has level @xmath329 iff the lattice square directly north of that east step has level @xmath159 . such lattice squares are encoded as @xmath338 by @xmath306 and are then retained by @xmath307 . all other lattice squares not of these two types are discarded by @xmath307 . thus @xmath339 , which is precisely the subword of @xmath340 consisting of letters retained by @xmath307 , will be the same word produced by the sweep map . intuitively , one can think of @xmath306 as sweeping up _ all _ lattice - square levels , and then @xmath307 keeps only those levels of squares that are `` adjacent '' to the frontier of @xmath67 in the sense described above . below , we will call these squares _ frontier squares _ of @xmath67 . before proving our final lemma , we need to introduce some temporary notation for describing the cells and rows of @xmath341 . first , let @xmath342 be the set of hook - lengths of cells in the first ( leftmost ) column of @xmath297 . in our running example , @xmath343 . each square @xmath295 in the diagram of @xmath297 lies due north of an east step on the frontier of @xmath297 , say @xmath338 ; and @xmath295 lies due west of a north step on the frontier of @xmath297 , say @xmath344 . identify the square @xmath295 with the ordered pair of labels @xmath345 $ ] . observe that @xmath345 $ ] is the label of some cell @xmath295 in the diagram of @xmath297 iff @xmath346 and @xmath338 and @xmath344 ; in this case , we must have @xmath347 . it is routine to check that the hook - length @xmath296 is @xmath348 . for all @xmath347 , the _ row of @xmath297 indexed by @xmath32 _ is the row with leftmost cell @xmath349 $ ] , whose hook - length is @xmath32 . [ lem : rho ] for all @xmath350 , @xmath351 . let @xmath352 . we must show @xmath353 . * step 1 . * we show that @xmath305 and @xmath307 keep the same rows of @xmath297 . on one hand , the definition of @xmath305 tells us to keep the rows of @xmath297 indexed by those @xmath347 appearing as the level of a square immediately east of a north step in @xmath90 . these squares are the frontier squares of @xmath67 below @xmath90 . on the other hand , the definition of @xmath307 tells us to retain the frontier steps @xmath344 of @xmath297 for those @xmath347 such that @xmath354 is the level of a north step of @xmath90 . as observed in the earlier lemma , these @xmath32 s correspond to @xmath32 s that are the levels of frontier squares of @xmath67 below @xmath90 . so @xmath305 and @xmath307 do retain the same rows of @xmath297 . * for each fixed @xmath347 , we compare the cells in the row of @xmath297 indexed by @xmath32 that are discarded by @xmath305 and @xmath307 . on one hand , let @xmath355 the cells in @xmath356 are erased by @xmath305 . so , for those row indices @xmath32 retained by @xmath305 , @xmath357 is the difference between the length of this row in @xmath297 and the length of the corresponding row in @xmath358 . on the other hand , let @xmath359 the values @xmath360 index the east steps @xmath361 prior to the north step @xmath362 that are discarded by @xmath307 . so , for those row indices @xmath32 retained by @xmath307 , @xmath363 is the difference between the length of this row in @xmath297 and the length of the corresponding row in @xmath364 . since @xmath305 and @xmath307 retain the same row indices @xmath32 ( by step 1 ) , it will now suffice to show that @xmath365 for all @xmath347 . * step 3 . * fix @xmath347 ; we define bijections @xmath366 and @xmath367 . given @xmath360 , let @xmath368 $ ] . given a cell @xmath345\in c_m$ ] , let @xmath369)=i+b$ ] . it is clear that @xmath370 and @xmath371 are identity maps , so the proof will be complete once we check that @xmath372 does map @xmath373 into @xmath356 , and @xmath374 does map @xmath356 into @xmath373 . consider a fixed @xmath360 . since @xmath361 , @xmath141 is the level of a square above @xmath90 , but this square is not a frontier square of @xmath67 . hence , the square directly below this square ( whose level is @xmath375 ) is also above @xmath90 . this implies @xmath376 and @xmath377 . moreover , since @xmath378 , the hook - length of the cell @xmath379 $ ] is @xmath380 , proving that @xmath368\in c_m$ ] . now consider a fixed cell @xmath345\in c_m$ ] . by definition of @xmath356 , we must have @xmath338 and @xmath381 . so the square with level @xmath159 is above @xmath90 , and hence the square with level @xmath382 is also above @xmath90 and is not a frontier square of @xmath67 . in particular , finally , @xmath384 and @xmath385 , so @xmath386 . we conclude that @xmath369)\in b_m$ ] . in our running example , the row indices @xmath32 retained by both @xmath305 and @xmath307 are @xmath387 . for @xmath388 , we have @xmath389 @xmath390,[3,25],[6,25],[7,25],[10,25],[12,25],[13,25],[14,25]\}.\ ] ] ( @xmath391 is the set of the leftmost eight cells in the top row of @xmath297 . ) the map @xmath392 $ ] defines a bijection from @xmath393 to @xmath391 . in @xcite , e. gorsky and m. mazin relate the @xmath0-catalan numbers and their generalizations to the homology of compactified jacobians for singular plane curves with puiseux pair @xmath289 . in the course of their investigations , they introduce the following map @xmath394 on partitions in @xmath395 ( we follow the notation of @xcite ) . let @xmath396 with @xmath45 and @xmath291 . for @xmath397 , define the * @xmath398-generators of @xmath399 * , denoted @xmath400 , to be the levels of the squares immediately above @xmath90 . define @xmath401 to be the set of levels of _ all _ lattice squares lying north or west of @xmath90 ( i.e. , including squares not adjacent to @xmath90 ) . equivalently , @xmath402 where @xmath403 is the finite set of levels of lattice squares between @xmath90 and @xmath294 . we then define a new partition @xmath404 by setting the @xmath159-th column of @xmath299 to have length @xmath405 for our running example where @xmath406 , @xmath407 , and @xmath408 , the @xmath409-generators are @xmath410 , @xmath411 and @xmath412 . it follows that @xmath413 the vector @xmath414 gives the column lengths of the partition @xmath415 . see figure [ fig : gmmap ] . the preceding example suggests that the gorsky - mazin map coincides with the map discussed in [ subsec : zetamap ] . we now prove this fact . [ thm : gm ] for @xmath396 with @xmath45 and all @xmath302 , @xmath416 let @xmath291 have @xmath42-generators @xmath417 . recall from table [ tab : sweeps ] that @xmath418 uses the west - south convention to assign levels to steps of a lattice path . it follows that the levels of east steps in @xmath90 are precisely the numbers @xmath419 . the @xmath159-th east step in the output of @xmath418 will be preceded by all the north steps whose levels are less than @xmath420 and followed by all the north steps whose levels are greater than @xmath420 . since the output has exactly @xmath41 north steps total , it will suffice to prove ( for each fixed @xmath159 ) that @xmath421 for each north step of level @xmath422 , the lattice square with level @xmath101 lies below @xmath90 , but all of the squares to the left lie west of @xmath90 and have levels of the form @xmath423 for some @xmath424 . on the other hand , for each @xmath42-generator @xmath425 , @xmath426 is the number of levels in the set @xmath427 that are the levels of squares below @xmath90 . for each north step of level @xmath422 with @xmath428 , @xmath429 but for each north step of level @xmath422 with @xmath430 , then the cardinality will be exactly @xmath281 . thus the number of levels removed from the @xmath41-element set @xmath431 to obtain @xmath426 is the same as the number of north steps of level @xmath432 , as needed . the main open problem in this paper is to prove that _ all sweep maps are bijections_. even in the two - letter case , this problem appears to be very difficult in general . nevertheless , many special cases of the sweep map are known to be invertible . after discussing the basic strategy for inversion ( which involves recreating the labels on the output steps by drawing a suitable `` bounce path '' ) , we describe the inverse sweep maps that have appeared in the literature in various guises . we omit detailed proofs that the inverse maps work , since these appear in the references . in figure [ fig : negsweep - ex ] , we showed the computation of @xmath433 where @xmath434 are paths in @xmath108 . the output @xmath435 is the path shown on the far right of the figure , _ not including labels_. suppose we were given @xmath435 and needed to compute @xmath436 . if we could somehow recreate the labels on the steps of @xmath435 ( as shown in the figure ) , then the sweep map could be easily inverted , as follows . by counting the total number of north and east steps , we deduce that @xmath66 must end at level @xmath437 . we now reconstruct the steps of @xmath66 in reverse order . the last step of @xmath66 must be the first step of @xmath435 in the collection of steps labeled 4 ( since , when sweeping @xmath66 to produce @xmath435 , level 4 is swept from right to left ) . we mark that step of @xmath435 as being used . since it is a north step , the preceding step of @xmath66 must end at level 1 . we now take the first unused step of @xmath435 labeled 1 ( which is a north step ) , mark it as used , and note that the preceding step of @xmath66 must end at level @xmath438 . we continue similarly , producing @xmath66 in reverse until reaching the origin , and marking steps in @xmath435 as they are used . because @xmath435 is in the image of the sweep map , this process must succeed ( in the sense that all steps of @xmath435 are used at the end , and we never get stuck at some level where all steps in @xmath435 with that label have already been used ) . evidently , the strategy outlined here works for any choice of weights , including the general case of alphabets with more than one letter . variations of the sweep map ( such as @xmath439 ) can be handled analogously . the crucial question is _ how to recreate the labels on the steps of @xmath435_. this question has been answered in the literature for dyck paths , @xmath32-dyck paths , trapezoidal lattice paths , square paths , schrder paths , @xmath440-dyck paths , and @xmath441-dyck paths . in every known case , the key to recreating the labels is to define a _ bounce path _ for a lattice path @xmath435 . the steps of @xmath435 associated with the `` @xmath159-th bounce '' in the bounce path receive label @xmath159 . once labels have been assigned , one can reverse the sweep map as described in the previous paragraph . we begin by discussing the simplest instance of the bounce path , which is used to invert the map @xmath442 ( see [ subsubsec : trapz ] ) acting on dyck paths . ironically , several different authors independently introduced inverse sweep maps even before the map @xmath230 was proposed in the context of @xmath0-catalan numbers . we describe these inverses in [ subsec : vaille ] and [ subsec : andrews ] below . figure [ fig : invert - hl ] shows the computation of @xmath443 for a dyck path @xmath444 . to understand how to find @xmath445 given @xmath435 , it suffices ( by the above remarks ) to see how to pass from the unlabeled path @xmath435 on the right side of the figure to the labeled path in the middle of the figure . to recreate the labels , we draw the _ bounce path _ for the dyck path @xmath435 , using the following definition due to haglund @xcite . the bounce path starts at @xmath279 and makes a sequence of horizontal moves @xmath446 and vertical moves @xmath447 , for @xmath448 , until reaching @xmath24 . each horizontal move is determined by moving west from the current position as far as possible without going strictly left of the path @xmath435 . then the next vertical move goes south back to the diagonal @xmath280 . figure [ fig : sweep - bounce ] shows the bounce path for our example path @xmath435 . in this example , the labels we are trying to recreate are related to the bounce path as follows : every step of @xmath435 located above the bounce move @xmath446 and to the left of bounce move @xmath449 has label @xmath159 . as special cases , the steps of @xmath435 above @xmath450 have label zero , and the steps of @xmath435 to the left of the last bounce move @xmath451 have label @xmath452 . we claim that this relation between the labels and the bounce path holds in general , for any dyck path @xmath435 of the form @xmath453 . this claim implies that @xmath66 can be uniquely recovered from @xmath435 , so that @xmath230 is injective and hence bijective . to prove the claim , let @xmath454 be the length of the horizontal move @xmath446 of the bounce path for @xmath435 , and let @xmath455 be the length of the vertical move @xmath447 of the bounce path for @xmath435 . also define @xmath456 for @xmath256 and @xmath457 , where @xmath452 is the total number of horizontal bounces . finally , let @xmath458 ( resp . @xmath459 ) denote the number of north ( resp . east ) steps of @xmath66 with @xmath277-level equal to @xmath159 . we first prove this lemma : @xmath460 for all @xmath250 . for any level @xmath159 , the number of times the path @xmath66 arrives at this level ( via a north step of level @xmath159 or an east step of level @xmath159 ) equals the number of times @xmath66 leaves this level ( via a north step of level @xmath461 or an east step of level @xmath253 ) . this holds even when @xmath462 , since @xmath66 begins and ends at level zero . it follows that @xmath463 for all @xmath250 . also @xmath464 for all @xmath465 , and @xmath466 for all @xmath256 . thus @xmath467 for all integers @xmath465 . fix an integer @xmath254 , and assume that @xmath468 . then @xmath469 , so the lemma follows by induction . now we show that for all @xmath250 , @xmath470 and @xmath471 . since @xmath66 is a dyck path , the assertion holds for all @xmath256 . we prove the two equalities for @xmath472 by induction on @xmath159 , starting with the base case of @xmath462 . recall that the steps of @xmath66 are swept in decreasing order by level . we know that @xmath473 . since @xmath474 , the path @xmath435 ends in @xmath475 east steps . hence @xmath476 . since @xmath66 starts at level @xmath105 , the first step in @xmath66 at _ any _ level @xmath311 must be a north step . it follows that the last step in @xmath435 labeled with a 1 is a north step and , consequently , that @xmath477 . assume now that @xmath478 and @xmath479 for some fixed @xmath480 . it follows from the bounce mechanism that @xmath481 . we know that @xmath478 by the induction hypothesis . finally , @xmath482 by the above discussion . combining these equalities , we find that @xmath483 . we know that @xmath484 using @xmath483 and the fact that the east steps of @xmath66 at level @xmath167 must be swept after any steps at level @xmath485 . as observed above , the first step in @xmath66 at level @xmath485 ( if it exists ) is a north step . hence @xmath486 . in 1997 , vaill @xcite defined a bijection @xmath487 mapping dyck paths to dyck paths , which is the inverse of the map @xmath225 defined in [ subsubsec : trapz ] . ( recall that @xmath225 differs from @xmath230 by reversing and flipping the output lattice path . ) vaill gives this example of his bijection @xmath487 in ( * ? ? ? 3 , p. 121 ) : @xmath488 the bounce path of @xmath66 is clearly visible in the left panel of that figure , although here the bounce path moves from @xmath24 north and east to @xmath279 as a result of the reversal and flipping . andrews , krattenthaler , orsina , and papi @xcite described a bijection mapping dyck partitions to dyck paths that is essentially the inverse of @xmath489 . they give an example starting with an input partition @xmath490 in ( * ? ? ? 2 , p. 3841 ) . the word of this partition ( after adding one more zero part at the end ) is @xmath491 this partition maps to the output dyck path shown in ( * ? ? ? 3 , p. 3846 ) , which has word @xmath492 one may check that @xmath493 , and similarly for other objects , so these authors have inverted the sweep map on dyck paths . here too , haglund s bounce path construction ( this time proceeding from @xmath279 to @xmath24 ) is visible in figure 2 of @xcite . loehr describes @xmath496 and its inverse in @xcite . the maps @xmath494 , @xmath495 , and their inverses are treated in @xcite . the key to inversion is defining the bounce path for a trapezoidal lattice path @xmath497 . this bounce path starts at @xmath24 and moves north and east to @xmath498 . for @xmath254 , the @xmath159-th bounce moves north @xmath499 steps from the current location as far as possible without going strictly north of the path @xmath435 . the @xmath159-th bounce continues by moving east @xmath500 steps , where @xmath501 for @xmath502 , @xmath503 for @xmath504 , and @xmath190 for @xmath505 . one can show that if @xmath435 is produced from @xmath66 via sweeping ( as described in step 1 of the proof of theorem [ thm : trapz - vs - sweep ] ) , then the steps of @xmath435 located north of the @xmath506-th horizontal bounce move and west of the @xmath159-th vertical bounce move receive label @xmath159 . thus , we can invert the sweep map in this case . loehr and warrington describe the inverse of @xmath236 in @xcite using the language of area vectors . their result amounts to inverting the sweep map @xmath507 on the domain @xmath237 of lattice paths in an @xmath238 square . as usual , it suffices to discuss the construction of the `` square bounce path . '' given @xmath508 with @xmath509 , first choose the maximum integer @xmath101 such that the path @xmath435 touches the line @xmath510 . call this line the _ break diagonal_. the _ break point _ of @xmath435 is the lowest point @xmath78 of @xmath435 on the line @xmath510 . the _ positive bounce path _ of @xmath435 starts at @xmath279 and moves to the break point as follows . first go south @xmath101 steps from @xmath279 to @xmath511 ; call this move @xmath512 . repeat until reaching the break point , taking @xmath448 : in move @xmath446 , go west until blocked by the north end of a north step of @xmath435 ; in move @xmath447 , go south to the break diagonal . next , the _ negative bounce path _ of @xmath435 starts at @xmath24 and moves to the break point as follows . first go east @xmath101 steps from @xmath24 to @xmath513 ; call this move @xmath514 . repeat until reaching the break point , taking @xmath515 : in move @xmath447 , go north to the first lattice point on @xmath435 ; in move @xmath446 , go east to the break diagonal . ( since the negative bounce path is blocked by lattice points , not edges , on @xmath435 , this rule is not the reflection of the rule for the positive bounce path . ) one can check that when sweeping @xmath66 to produce @xmath435 , the steps of @xmath435 located north of move @xmath446 and west of move @xmath449 receive label @xmath159 . it is now straightforward to invert @xmath236 . egge , haglund , killpatrick , and kremer describe the inverse of their map @xmath283 in @xcite . given a schrder path @xmath435 , their algorithm to compute @xmath516 begins by dividing the steps of @xmath435 into regions based on a version of the bounce path defined for schrder paths . this amounts to reconstructing the labels of the steps of @xmath435 created when applying the sweep map to @xmath66 . they then reconstruct the area vector of @xmath66 ( modified to allow diagonal steps ) by an insertion process that reverses the action of the sweep map . this special case of sweep inversion is notable because it inverts a sweep map on a three - letter alphabet ( although one of the letters has weight zero ) . in @xcite , gorsky and mazin describe how to invert special cases of their map @xmath394 using the language of semi - module generators and bounce paths . their results amount to inverting @xmath517 on the domain @xmath518 in the case where @xmath519 for some @xmath32 . the case @xmath520 essentially duplicates the @xmath32-bounce path construction of @xcite ( although there is a lot of new material relating this construction to semi - modules ) . the case @xmath521 is a new inversion result obtained via a modification of the @xmath32-bounce paths . specifically , one constructs the bounce path s vertical moves @xmath499 ( for @xmath254 ) as described in [ subsec : invert - trapz ] above , but now @xmath522 , where @xmath523 for @xmath524 and @xmath525 for all other @xmath159 . this section applies the sweep map to provide new combinatorial generalizations of the @xmath0-catalan numbers @xcite and the @xmath0-square numbers @xcite . we make several conjectures regarding the joint symmetry of these polynomials and their connections to the nabla operator @xmath526 on symmetric functions introduced by a. garsia and f. bergeron @xcite . for any word @xmath134 , let @xmath527 be the number of pairs @xmath528 with @xmath16 and @xmath529 . this is the area of the partition diagram @xmath530 consisting of the squares above and to the left of steps in @xmath75 . for @xmath136 , let @xmath18 have @xmath86-levels @xmath531 ( e - n convention ) . let @xmath532 , and set @xmath533 . note that @xmath534 in general , so we can not necessarily assume that @xmath535 when using @xmath536 . the function @xmath537 appears in @xcite as @xmath538 . the correspondence between partitions and paths lying in a fixed triangle has led to inconsistent terminology : `` area '' can refer to either the number of squares lying in the partition determined by a path ( as above ) or the number of squares between the path and a diagonal ( such as in @xcite ) . for @xmath136 and @xmath127 , define the * @xmath0-catalan numbers for slope @xmath539 ending at @xmath540 * by @xmath541 [ conj : qt - joint ] for all @xmath136 and all @xmath127 , @xmath542 . note that the conjectured bijectivity of @xmath186 on the domain @xmath88 would imply the weaker univariate symmetry property @xmath543 . the rational @xmath0-catalan polynomials defined in @xcite arise from a sweep map that , in the case @xmath45 considered in that paper , reduces to @xmath544 . as such , the joint symmetry conjecture ( * ? ? ? 19 ) is not quite a special case of conjecture [ conj : qt - joint ] . let @xmath545 be a `` classical '' dyck path with area vector @xmath546 ( see [ subsubsec : trapz ] ) . let @xmath547 , which is the number of area squares between the path @xmath18 and the line @xmath280 , and let @xmath548 be the number of @xmath528 with @xmath549 . the garsia - haiman @xmath0-catalan numbers @xcite can be defined by the combinatorial formula @xmath550 to relate this polynomial to the one defined above , note that @xmath551 . similarly , it follows from theorem [ thm : trapz - vs - sweep ] and @xcite that @xmath552 . therefore , @xmath553 combining this with a theorem of garsia and haglund @xcite , we get @xmath554 more generally , for any positive integers @xmath555 , the higher - order @xmath0-catalan numbers @xcite satisfy @xmath556 the main conjecture for these polynomials can be stated as follows : @xmath557 an interesting open problem is to find formulas relating the general polynomials @xmath558 to nabla or related operators . next we generalize the @xmath0-square numbers studied in @xcite . for @xmath127 , define the * @xmath559-rectangle numbers for the @xmath560 rectangle * by @xmath561 for all @xmath562 , @xmath563 . the joint symmetry conjecture is known to hold when @xmath564 . this follows from the stronger statement @xmath565 which was conjectured in @xcite and proved in @xcite . we conjecture the following more general relationship between certain @xmath0-rectangle numbers and higher powers of @xmath526 . for all @xmath566 and @xmath227 , @xmath567 recall the definitions of @xmath569-integers , @xmath569-factorials , and @xmath569-binomial coefficients : @xmath570_q=1+q+q^2+\cdots+q^{n-1}$ ] , @xmath570!_q=[n]_q[n-1]_q\cdots [ 2]_q[1]_q$ ] , and @xmath571}{0pt}{}{a+b}{a , b}_{q}}=[a+b]!_q/([a]!_q[b]!_q)}$ ] . in ( * ? 21 ) , the authors make the following conjecture for coprime @xmath41 and @xmath42 : @xmath572_q}{\displaystyle\genfrac{[}{]}{0pt}{}{a+b}{a , b}_{q}}.\ ] ] we conjecture here that @xmath544 can be replaced by @xmath573 : for all coprime @xmath574 , @xmath575_q}{\displaystyle\genfrac{[}{]}{0pt}{}{a+b}{a , b}_{q}}.\ ] ] we also introduce two conjectures regarding the @xmath568 specialization of paths in a rectangle . for all @xmath576 , @xmath577_{q^n}}{\displaystyle\genfrac{[}{]}{0pt}{}{mn+n}{mn , n}_{q}}.\ ] ] this conjecture generalizes to arbitrary rectangles as follows : for all @xmath127 , write @xmath578 and @xmath579 for integers @xmath580 with @xmath581 . then @xmath582_{q^k } } { \displaystyle\genfrac{[}{]}{0pt}{}{a+b}{a , b}_{q}}.\ ] ] a sage worksheet containing code to compute the images of paths under various versions of the sweep map as well as to check the conjectures of section [ sec : area - qtcat ] can be found at the third author s web page @xcite . the authors gratefully acknowledge discussions with jim haglund , mark haiman and michelle wachs . 99 george andrews , christian krattenthaler , luigi orsina , and paolo papi , `` @xmath34-nilpotent @xmath35-ideals in @xmath583 having a fixed class of nilpotence : combinatorics and enumeration , '' _ trans . soc . _ * 354 * no . 10 ( 2002 ) , 38353853 . f. bergeron , a. garsia , m. haiman , and g. tesler , `` identities and positivity conjectures for some remarkable operators in the theory of symmetric functions , '' _ methods and applications of analysis _ vii * 3 * ( 1999 ) , 363420 . eric egge , jim haglund , kendra killpatrick , and darla kremer , `` a schrder generalization of haglund s statistic on catalan paths , '' _ electron . j. of combin . _ * 10 * ( 2003 ) , research paper 16 , 21 pages ( electronic ) . a. garsia and j. haglund , `` a proof of the @xmath0-catalan positivity conjecture , '' _ lacim 2000 conference on combinatorics , computer science , and applications _ ( montreal ) , _ discrete math . _ * 256 * ( 2002 ) , 677717 . evgeny gorsky and mikhail mazin , `` compactified jacobians and @xmath0-catalan numbers , ii , '' _ j. algebraic combin . _ * 39 * ( 2014 ) , no . 1 , 153186 . james haglund , `` conjectured statistics for the @xmath0-catalan numbers , '' _ adv . in math . _ * 175 * ( 2003 ) , 319334 .

we present a new numerical code designed to solve the einstein field equations for axisymmetric spacetimes . the long term goal of this project is to construct a code that will be capable of studying many problems of interest in axisymmetry , including gravitational collapse , critical phenomena , investigations of cosmic censorship , and head - on black hole collisions . our objective here is to detail the ( 2 + 1)+1 formalism we use to arrive at the corresponding system of equations and the numerical methods we use to solve them . we are able to obtain stable evolution , despite the singular nature of the coordinate system on the axis , by enforcing appropriate regularity conditions on all variables and by adding numerical dissipation to hyperbolic equations .

gr-qc0301006

in this paper we introduce a numerical code designed to solve the einstein field equations for axisymmetric spacetimes . even though the predominant focus in numerical relativity in recent years has been to study situations of relevance to gravitational wave detection , and hence lacking symmetries , there are still numerous interesting problems , both physical and computational , that can be tackled with an axisymmetric code . the advantages of restricting the full 3d problem to axisymmetry ( 2d ) are that the complexity and number of equations are reduced , as are the computational requirements compared to solving a similar problem in 3d . prior numerical studies of axisymmetric spacetimes include head - on black hole collisions @xcite , collapse of rotating fluid stars @xcite , the evolution of collisionless particles applied to study the stability of star clusters @xcite and the validity of cosmic censorship @xcite , evolution of gravitational waves @xcite , black hole - matter - gravitational wave interactions @xcite , and the formation of black holes through gravitational wave collapse @xcite and corresponding critical behavior at the threshold of formation @xcite . our goals for creating a new axisymmetric code are not only to explore a wider range of phenomena than those studied before , but also to provide a framework to add adaptive mesh refinement ( amr ) and black hole excision to allow more thorough and detailed investigations than prior works . the outline of the rest of the paper is as follows . in section [ sec : formalism ] we describe the @xmath0 decomposition @xcite of spacetime that we adopt to arrive at our system of equations . the @xmath0 formalism is the familiar adm space + time decomposition ( in this case @xmath1 ) applied to a dimensionally reduced spacetime obtained by dividing out the axial killing vector , following a method devised by geroch @xcite . in section [ sec : coords_and_vars ] we specialize the equations to our chosen coordinate system , namely cylindrical coordinates with a conformally flat @xmath2metric . at this stage we do not model spacetimes with angular momentum , and we include a massless scalar field for the matter source . in section [ sec : apph ] we discuss how we search for apparent horizons during evolution . in section [ sec : implementation ] we describe our numerical implementation of the set of equations derived in section [ sec : coords_and_vars ] . a variety of tests of our code are presented in section [ sec : test ] , which is followed by conclusions in section [ sec : conclusion ] . some details concerning our finite difference approximations , solution of elliptic equations via the multi - grid technique , and a spherically symmetric code used for testing purposes are given in appendices a and b. unless otherwise specfied , we use the units and conventions adopted by misner , thorne and wheeler @xcite . the most common approach in numerical relativity is to perform the so - called @xmath3 , or adm , split of the spacetime that one would like to evolve . in this procedure , a timelike vector field is chosen together with spatial hypersurfaces that foliate the spacetime . if axisymmetry is assumed , it is usually incorporated once the adm decomposition has been done , and is reflected in the independence of the various metric and matter quantities on the ignorable , angular coordinate , @xmath4 . our approach , which follows geroch @xcite and nakamura _ et al _ @xcite , reverses this procedure . we assume axisymmetry from the outset and perform a reduction of the spacetime based on this assumption . once we have projected out the symmetry , we perform an adm - like split ( now a @xmath1 split ) of the remaining @xmath5manifold . more specifically , we begin with a @xmath6dimensional spacetime metric manifold and latin letters to denote coordinates in the reduced @xmath5manifold . ] @xmath7 on our manifold @xmath8 . the axisymmetry is realized in the existence of a spacelike killing vector x^= ( ) ^ , with closed orbits . we define the projection operator @xmath9 , allowing us to project tensors from the 4dimensional manifold @xmath10 with metric @xmath11 to a @xmath5dimensional manifold @xmath12 with metric @xmath13 , as g _ = _ - x_x _ , where @xmath14 is the norm of the killing vector s^2=x^x_. with the definition of the vector y_= x _ , the metric on the full , @xmath6dimensional spacetime can be written as _ = ( cc g_ab + s^2 y_a y_b & s^2 y_a + s^2 y_b & s^2 ) . projecting , or dividing out the symmetry amounts to expressing @xmath6dimensional quantities in terms of quantities on the @xmath5manifold . for instance , the connection coefficients @xmath15 associated with the @xmath6dimensional metric are ^(4 ) ^ _ & = & ^(3 ) ^ _ + ^ _ + & = & ^(3 ) ^ _ + s^2 g^ + y^ , [ christof ] where @xmath16 are the connection coefficients constructed from the @xmath5metric @xmath9 , and we have defined the antisymmetric tensor z _ = _ y _ - _ y_. notice that @xmath17 is an intrinsically @xmath5dimensional object , in that @xmath18 . with some algebra , the ricci tensor on the @xmath6manifold , @xmath19 , can now be written as the ricci tensor on the reduced space , @xmath20 , together with additional terms involving fields coming from the dimensional reduction ^(4)r _ = ^(3)r _ + d _ ^ _ - d_^ _ + ^ _ _ ^- ^ _ ^ _ , [ ricci_4 ] where @xmath21 is the covariant derivative on the @xmath5 dimensional manifold . expressed in terms of @xmath14 , @xmath22 , @xmath23 and @xmath24 , the components of ( [ ricci_4 ] ) are ^(4)r _ & = & 14 s^4 z_bcz^bc - s d^a d_a s , [ eq : r_phiphi ] + ^(4)r_a & = & 12s d^c(s^3 z_ac ) + y_a , [ eq : r_rhia ] + ^(4)r_ab & = & ^(3)r_ab - 1s d_a d_bs - 12 s^2 z_acz_b^c - 1s d^cy_b ) + y_a y_b . [ eq : r_ab ] taking the trace of ( [ ricci_4 ] ) , and using the definitions described above , gives the decomposition of the ricci scalar as ^(4)r = ^(3)r - 2s d^a d_as - 14 s^2 z_bcz^bc . [ eq : ricci ] the @xmath6dimensional einstein equations , with a stress - energy tensor @xmath25 , are ^(4)r _ - ^(4)r _ = 8t_. [ eq : einstein ] using equations ( [ eq : r_phiphi]-[eq : ricci ] ) , we can write the einstein equations as d^ad_a s & = & - w_aw^a 2s^3 - 8s ( t _ - 12 s^2 t_^ ) , [ eq : s_eom ] + d_[a w_b ] & = & 8 s _ ab^c t_c , [ eq : w_eom ] + ^(3)r_ab & = & 1s d_a d_bs + 12 s^2 z_acz_b^c + 8 ( t _ g^_a g^_b - 12 g_ab t_^ ) , [ eq : r_eom ] where we have introduced the twist @xmath26 of the killing vector w_= s^42 _ y^z^ , and the four and three dimensional levi - civita symbols @xmath27 and @xmath28 , respectively . the twist vector @xmath26 is intrinsically @xmath5dimensional , _ i.e. _ @xmath29 . furthermore , @xmath30 is divergence free d^a=0 . [ eq : div_w_eom ] at this point , the first reduction is essentially done . equation ( [ eq : r_eom ] ) can be viewed as the @xmath5dimensional einstein equations , coupled to the projection of the stress - energy tensor @xmath25 and to induced `` matter fields '' @xmath14 and @xmath31 ( or @xmath23 ) . the equations of motion for @xmath14 and @xmath31 are given by ( [ eq : s_eom ] ) and ( [ eq : w_eom ] ) respectively ; additional equations of motion will need to be specified for whatever true matter fields one incorporates into the system . this procedure so far is completely analogous to the kaluza - klein reduction from five to four dimensions , in which the @xmath32dimensional geometry becomes gravity in @xmath6 dimensions coupled to electromagnetism and a scalar field @xcite . here , however , the reduced @xmath5manifold has no dynamics and we have the rather appealing picture of having divided out " the dynamics of the four dimensional gravitational system and reinterpreted the two degrees of freedom in the gravitational field as scalar ( @xmath14 ) and `` electromagnetic '' ( @xmath23 , or @xmath33 ) degrees of freedom . dimensions has only a single degree of freedom as compared to the two degrees of freedom in @xmath6 dimensions . ] we now perform the adm split of the remaining spacetime . this is done by first foliating the @xmath5dimensional spacetime into a series of spacelike hypersurfaces with unit , timelike normal vector @xmath34 . then , similar to the dimensional reduction above , we decompose quantities into components orthogonal and tangent to @xmath34 , using the projection tensor @xmath35 h_ab = g_ab + n_a n_b . we now define the components of @xmath34 , and the induced @xmath2dimensional metric @xmath36,-dimensional tensor components . ] using the following decomposition of the @xmath5metric g_ab dx^a dx^b = -^2 dt^2 + h_ab ( dx^a + ^adt ) ( dx^b + ^bdt ) , where @xmath37 is the lapse function and @xmath38 the shift vector . the gravitational equations now become the evolution equations for the components of the @xmath2metric @xmath36 and the @xmath2extrinsic curvature @xmath39 _ t h_ab & = & -2k_ab + _ a _ b + _ b _ a [ eq : hdot ] + _ t k_ab & = & ^c _ c k_ab + k_ac _ b ^c + k_bc _ a ^c + + & & - 2k_ack_b^c - _ a_b - ^(3)r_ab , [ eq : kdot ] [ eq : k_ab_dot ] the hamiltonian constraint equation ^(2)r - k^a_bk^b_a + k^2 = ^(3)r + 2 ^(3)r_abn^a n^b [ eq : hc ] and the @xmath40 and @xmath41 momentum constraint equations _ a k_b^a - _ ak = - ^(3)r_cd n^d h^c_b . [ eq : mc ] in the above , @xmath42 is the covariant derivative compatible with the @xmath2metric @xmath36 , @xmath43 , and @xmath44 and @xmath45 are the @xmath2dimensional ricci tensor and ricci scalar , respectively . note that because gravity in @xmath5 dimensions has no propagating degrees of freedom , the constraint equations fix the @xmath5dimensional geometry completely . thus , if desired , one can use the constraint equations ( [ eq : hc]-[eq : mc ] ) instead of the evolution equations ( [ eq : hdot]-[eq : kdot ] ) to solve for @xmath36 . the freely specifiable degrees of freedom of the @xmath6manifold are encoded in @xmath14 and @xmath31 , which are evolved using ( [ eq : s_eom ] ) , ( [ eq : w_eom ] ) and ( [ eq : div_w_eom ] ) . note that ( [ eq : w_eom ] ) and ( [ eq : div_w_eom ] ) constitute four equations for the three components of @xmath33the purely spatial have four equations for the three components of @xmath33 . the purely spatial part of ( [ eq : w_eom ] ) is , in @xmath3 language , the angular momentum constraint equation and only needs to be solved at the initial time in a free evolution of @xmath33 . we note that the restricted class of axisymmetric spacetimes having no angular momentum ( rotation ) is characterized by the existence of a _ scalar twist _ , @xmath46 , such that @xmath47 . in the vacuum case , @xmath46 generally represents odd parity gravitational waves , while @xmath14 encodes even parity , or brill waves@xcite . we further note that this class includes the special case @xmath48 , which will be the focus of our discussion below . in this section we describe a particular coordinate system and set of variables which , in the context of the formalism described in the previous section , provides us with the concrete system of partial differential equations that we solve numerically . we also detail the outer boundary conditions we use , and the on - axis regularity conditions necessary to obtain smooth solutions to these equations . we only consider spacetimes with zero angular momentum , and no odd - parity gravitational waves ; therefore @xmath49 . we choose a conformally flat , cylindrical coordinate system for the @xmath2metric h_ab dx^a dx^b = ( , z , t)^4 ( d^2 + dz^2 ) . [ hab_def ] this choice for @xmath36 exhausts the coordinate freedom we have to arbitrarily specify the two components of the shift vector@xmath50 and @xmath51 . in order to maintain the form ( [ hab_def ] ) during an evolution , we use the momentum constraints , which are elliptic equations , to solve for @xmath52 and @xmath53 at each time step . the hamiltonian constraint provides a third elliptic equation that we can use to solve for the conformal factor @xmath54 . for a slicing condition , we use maximal slicing of @xmath55 hypersurfaces _ in the @xmath6dimensional manifold__i.e . _ we impose @xmath56 , where @xmath57 is the trace of the extrinsic curvature tensor of @xmath55 slices of @xmath11 . this condition ( specifically @xmath58 ) gives us an elliptic equation for the lapse . instead of directly evolving the norm of the killing vector , @xmath14 , we evolve the quantity @xmath59 , defined by s = ^2 e^| , and furthermore , we convert the resultant evolution equation for @xmath59 ( [ eq : s_eom ] ) to one that is first order in time by defining the quantity @xmath60 , which is `` conjugate '' variable to @xmath59 , via |&= & - 2k_^- k_z^z + & = & - n^a ( s)_,a + . [ eq : bo_def ] part of the motivation behind using @xmath59 and @xmath60 as fundamental variables is to simplify the enforcement of on - axis regularity . in particular , regularity as @xmath61 implies that @xmath59 and @xmath60 must exhibit leading order behavior of the form @xmath62 and @xmath63 respectively , and experience has shown it to be easier to enforce such conditions , than to enforce the leading order behavior of @xmath14 ( or its time derivative ) near the axis , which is @xmath64 . as mentioned previously , the only matter source we currently incorporate is a massless scalar field @xmath65 , which satisfies the usual @xmath6dimensional wave equation @xmath66 . we convert this equation to first - order - in - time form by defining a conjugate variable @xmath67 : ^2 n^a _ , a . the stress - energy tensor for the scalar field is t _ = 2 _ , _ , - _ ^ , _ , . using all of the above definitions and restrictions within the formalism detailed in the previous section , we end up with the following system of equations that we solve with our numerical code , described in the next section . the maximal slicing condition results in the following elliptic equation for @xmath37 @xmath68 - \frac{\psi^4}{6\alpha } \bigl [ 2 \alpha \rho \bo + \beta^\rho{}_{,\rho } - \beta^z{}_{,z } \bigr]^2 = 16 \pi \alpha \pi^2 . \label{eq : cons_alpha}\end{aligned}\ ] ] the hamiltonian constraint gives an elliptic equation for @xmath54 @xmath69 + { \psi^4 \over 6\alpha^2 } \bigl [ 2\alpha\rho\bar{\omega } + \beta^\rho{}_{,\rho } - \beta^z{}_{,z } \bigr]^2 \quad\qquad\qquad \nonumber\\ \qquad \ , = \ , - 16 \pi \left ( \pi^2 + \phi_{,\rho}{}^2 + \phi_{,z}{}^2 \right ) % - 6 \left ( \rho^2 \left(\rho \bs\right)_{,\rho } \right)_{,\rho^3 } - 6 \bigl ( \rho^2 \left(\rho \bs\right)_{,\rho } \bigr)_{,\rho^3 } \nonumber\\ % - 2 \left ( \left ( \rho \bs \right)_{,\rho } \right)^2 % - 2 \left ( \rho \bs \right)_{,zz } % - 2 \left ( \left ( \rho \bs \right)_{,z } \right)^2 . - 2 \bigl ( \left ( \rho \bs \right)_{,\rho } \bigr)^2 - 2 \bigl ( \rho \bs \bigr)_{,zz } - 2 \bigl ( \left ( \rho \bs \right)_{,z } \bigr)^2 . \label{eq : cons_psi}\end{aligned}\ ] ] the @xmath40 and @xmath41 momentum constraints , which provide elliptic equations that we use to solve for @xmath52 and @xmath53 , are @xmath70 % - \frac{8}{3 } \alpha \bar{\omega } % -\frac{2 \alpha \rho } { 3 } \left [ % 6 \bo \frac{\psi_{,\rho } } { \psi } % + \bo_{,\rho } % + 3 \bo \left(\rho \bs \right)_{,\rho } % \right]=0 , \frac{2}{3}\beta^{\rho}{}_{,\rho\rho } + \beta^{\rho}{}_{,zz } + \frac{1}{3 } \beta^z{}_{,z\rho } -\frac{2 \alpha \rho } { 3 } \bigl [ 6 \bo \frac{\psi_{,\rho } } { \psi } + \bo_{,\rho } + 3 \bo \left(\rho \bs \right)_{,\rho } \bigr ] - \frac{8}{3 } \ , \alpha \bar{\omega } \qquad\qquad\qquad\qquad\qquad\qquad \nonumber\\ - \frac{2}{3 } \bigl [ \frac { \alpha_{,\rho } } { \alpha } - 6 \frac { \psi_{,\rho } } { \psi } \bigr ] \bigl ( \beta^\rho{}_{,\rho } - \beta^z{}_{,z } \bigr ) - \bigl [ \frac { \alpha_{,z } } { \alpha } - 6 \frac { \psi_{,z } } { \psi } - \left ( \rho\bs \right)_{,z } \bigr ] \bigl ( \beta^\rho{}_{,z } + \beta^z{}_{,\rho } \bigr ) \ ; = \ ; - 32\pi { \alpha \over \psi^2 } \pi_{,\rho } , \label{eq : cons_betarho}\end{aligned}\ ] ] and @xmath71 % \left ( \beta^\rho{}_{,z } + \beta^z{}_{,\rho } \right ) % \nonumber\\ % + \left [ 2 \left ( \rho \bs \right)_{,z } % - \frac{4}{3 } \left ( \frac { \alpha_{,z } } { \alpha } % - 6 \frac { \psi_{,z } } { \psi } \right ) % \right ] % \left ( \beta^z{}_{,z } - \beta^\rho{}_{,\rho } \right ) % - \frac{2 \alpha\rho}{3 } \left ( % 6 \bo \frac{\psi_{,z}}{\psi } % + \bo_{,z } % \right ) % \nonumber\\ % + 32\pi { \alpha \over \psi^2 } \pi_{,z } % - 2\alpha(\bs_{,z})\rho^2\bo=0 . \beta^z{}_{,\rho\rho } + \frac{4}{3}\beta^z{}_{,zz } - \frac{1}{3 } \beta^{\rho}{}_{,z\rho } - \frac{2 \alpha\rho}{3 } \bigl [ 6 \bo \frac{\psi_{,z}}{\psi } + \bo_{,z } + 3 \bo ( \rho \bs)_{,z } \bigr ] \qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad\quad \nonumber\\ + \frac{4}{3 } \ , \bigl [ \frac { \alpha_{,z } } { \alpha } - 6 \frac { \psi_{,z } } { \psi } - \frac{3}{2 } \left ( \rho \bs \right)_{,z } \bigr ] \bigl ( \beta^\rho{}_{,\rho } - \beta^z{}_{,z } \bigr ) + \bigl [ \frac { 2 \alpha } { \psi^6 } \bigl ( \frac{\rho \psi^6}{\alpha } \bigr)_{,\rho^2 } \ , + \left ( \rho \bs \right)_{,\rho } \bigr ] \bigl ( \beta^\rho{}_{,z } + \beta^z{}_{,\rho } \bigr ) \ ; = \ ; - 32\pi { \alpha \over \psi^2 } \pi_{,z } . \label{eq : cons_betaz}\end{aligned}\ ] ] the definition of @xmath60 in eq . ( [ eq : bo_def ] ) gives an evolution equation for @xmath59 ( where the overdot denotes partial differentiation with respect to @xmath72 ) @xmath73_{,\rho}. - \bigl ( { \beta^{\rho } \over \rho } \bigr)_{,\rho}.\ ] ] the evolution equation for @xmath60 is @xmath74 + 64\pi{\alpha\over \psi^4 } \rho ( \phi_{,\rho^2})^2 . \label{eq : evol_bo}\end{aligned}\ ] ] we also have an evolution equation for @xmath54 , which we optionally use instead of the hamiltonian constraint ( [ eq : cons_psi ] ) to update @xmath54 @xmath75 the definition of @xmath67 and the wave equation for @xmath76 give @xmath77 and @xmath78 + \frac{\alpha}{\psi^2 } \left [ \left ( \rho \bs \right)_{,\rho } \phi_\rho + \left ( \rho \bs \right)_{,z } \phi_z \right ] . \label{eq : evol_pi}\end{aligned}\ ] ] to complete the specification of our system of equations , we need to supply boundary conditions . in our cylindrical coordinate system , where @xmath40 ranges from @xmath79 to @xmath80 and @xmath41 ranges from @xmath81 to @xmath82 , we have two distinct boundaries : the physical outer boundary at @xmath80 , @xmath83 , and @xmath84 ; and the axis , at @xmath79 . historically , the axis presented a stability problem in axisymmetric codes . we solve this problem by enforcing regularity on the axis , and , as described in section [ sec : implementation ] , adding numerical dissipation to evolved fields . the regularity conditions can be obtained by inspection of the equations in the limit @xmath85 , or more formally , by transforming to cartesian coordinates and demanding that components of the metric and matter fields be regular and single valued throughout @xcite . garfinkle and duncan @xcite have further proposed that in order to ensure smoothness on the axis , one should use quantities that have either even or odd power series expansions in @xmath40 as @xmath61 , but which do not vanish faster than @xmath86 . it is interesting that the quantities which we found to work best also obey this requirement . as discussed earlier , the particular choice of @xmath59 and @xmath60 as fundamental variables was partly motivated by regularity concerns . the results are _ , ( 0,z , t ) & = & 0 [ eq : bc_alpha ] + _ , ( 0,z , t ) & = & 0 + ^z_,(0,z , t ) & = & 0 + ^(0,z , t ) & = & 0 [ eq : bc_betarho ] + ( 0,z , t ) & = & 0 + ( 0,z , t ) & = & 0 + _ , ( 0,z , t ) & = & 0 + _ , ( 0,z , t ) & = & 0 at the outer boundary , we enforce asymptotic flatness by requiring _ r ( r , t ) & = & 1 + + o(r^-2 ) [ eq : obc_alpha ] + _ r ( r , t ) & = & 1 + + o(r^-2 ) + _ r ^z(r , t ) & = & + o(r^-2 ) + _ r ^(r , t ) & = & + o(r^-2 ) , [ eq : obc_betarho ] for undetermined functions @xmath87 , @xmath88 , @xmath89 , @xmath90 , and @xmath91 . these latter relations are converted to mixed ( robin ) boundary conditions ( see appendix a for details ) and then are imposed at the outer boundaries of the computational domain : @xmath92 , @xmath93 , and @xmath94 . we have also experimented with the use of dirichlet conditions on @xmath95 and @xmath53 at the outer boundaries ( specifically @xmath96 and @xmath97 there ) , and have found that these work about as well as the robin conditions . for the scalar field , we assume that near the outer boundary we can approximate the field as purely radially outgoing , and require ( r ) _ , t + ( r ) _ , r = 0 . for scalar field configurations far from spherical symmetry , this approximation suffers and reflections are relatively large . however , in general , the reflections do not grow and are somewhat damped . for the other two evolved quantities , @xmath59 and @xmath60 , we use this same naive condition for lack of any better , more physically motivated conditions . while this condition proves to be stable with damped reflections , a better condition is sought and this issue remains under investigation . for initial conditions , we are free to set @xmath98 . once the free data is chosen , we then use the constraint and slicing equations to determine @xmath99 . specifically , we define a general pulse shape g_x(,z ) = a_x [ g_x ] characterized by six parameters @xmath100 and then choose initial data of the form ( 0,,z ) & = & g_(,z ) ( 0,,z ) & = & g_(,z ) ( 0,,z ) & = & g_(,z ) ( 0,,z ) & = & 0 . for @xmath101 , these pulses are gaussian , spherical shells centered at @xmath102 with radius @xmath103 and pulse width @xmath104 . for @xmath101 and @xmath105 , the pulses are spherical . the factor of @xmath40 in the initial data for @xmath59 and @xmath60 ensures the correct behavior on axis for regularity . for the evolutions presented here , we let @xmath106 so that the initial configuration represents a moment of time symmetry . we note , however , that we are also able to generate and evolve approximately ingoing initial data . in this section we describe the equation and technique we use to search for apparent horizons ( ahs ) within @xmath107 spatial slices of the spacetime ( see @xcite for descriptions of some of the methods available to find ahs in axisymmetry ) . we restrict our search to isolated , simply connected ahs . in axisymmetry , such an ah can be described by a curve in the @xmath108 plane , starting and ending on the axis at @xmath79 . we define the location of the ah as the level surface @xmath109 , where @xmath110 and @xmath111 the ah is the outermost , marginally trapped surface ; hence , we want to find an equation for @xmath112 such that the outward null expansion normal to the surface @xmath109 , is zero . to this end , we first construct the unit spatial vector @xmath113 , normal to @xmath114 @xmath115 then , using @xmath113 and the @xmath107 hypersurface normal vector @xmath34 , we construct future - pointing outgoing ( @xmath116 ) and ingoing ( @xmath117 ) null vectors as @xmath118 the normalization of the null vectors is ( arbitrarily ) @xmath119 . the outward null expansion @xmath120 is then the divergence of @xmath121 projected onto @xmath114 @xmath122 using the definition of the extrinsic curvature @xmath39 , and substituting ( [ null ] ) into ( [ exp1 ] ) , we arrive at the familiar form for the null expansion when written in terms of adm variables @xmath123 note that because the normalization of @xmath121 is arbitrary , so ( to some extent ) is that of @xmath120 . the above normalization is chosen so that @xmath120 measures the fractional rate of change of area with time measured by an observer moving along @xmath34 . substituting ( [ fs2 ] ) and ( [ sadef ] ) into ( [ exp2 ] ) , and setting @xmath124 , we are left with an ordinary differential equation for @xmath112 . this equation takes the following form , where a prime @xmath125 denotes differentiation with respect to @xmath126 : @xmath127 @xmath128 is a rather lengthy function of its arguments , non - linear in @xmath129 and @xmath130 ; for brevity we do not display it explicitly . all of the metric functions and their gradients appearing in ( [ rpp_eqn ] ) are evaluated along a given curve of integration , and hence are implicitly functions of @xmath126 . @xmath126 ranges from @xmath131 to @xmath132 , and regularity of the surface @xmath109 about the axis requires @xmath133 . integration of ( [ rpp_eqn ] ) therefore proceeds by specifying @xmath129 at @xmath134 ( for instance ) , and then `` evolving '' @xmath129 until either @xmath135 , or @xmath129 diverges at some value of @xmath136 . if an ah exists , and assuming @xmath137 is inside the ah , then the ah can be found by searching for the ( locally ) unique ) augmented with the conditions @xmath133 . we want the outermost of these surfaces . ] initial value @xmath138 such that integration of ( [ rpp_eqn ] ) ends at @xmath139 , with @xmath140 and @xmath141 finite . for @xmath142 slightly larger than @xmath143 ( outside the ah ) , the integration will end at @xmath135 with @xmath144 , indicating an irregular point on the surface ; similarly , for @xmath142 slightly smaller than @xmath143 ( inside the ah ) the integration will end with @xmath145 . therefore , if we can find a reasonable bracket about the unknown @xmath143 , we can use a bisection search to find @xmath143 . currently , we find a bracket to search by testing a set of initial points , equally spaced in @xmath41 at intervals of @xmath146 . this seems to work well in most situations , and the search is reasonably fast . we use a second - order runge - kutta method to integrate equation ( [ rpp_eqn ] ) . the metric functions appearing in @xmath128 are evaluated using bilinear interpolation along the curve . in this section we describe the numerical code that we have written to solve the equations listed in section [ sec : coords_and_vars ] . some details are deferred to appendix [ app : solver ] . we use a uniform grid of size @xmath147 points in @xmath40 by @xmath148 points in @xmath41 , with equal spacing @xmath149 in the @xmath40 and @xmath41 directions . the value of a function @xmath150 at time level @xmath151 and location @xmath152 within the grid , corresponding to coordinate @xmath153 , is denoted by @xmath154 . for the temporal discretization scale we use @xmath155 , where @xmath156 is the courant factor , which for the type of differencing we employ , should be less than one for stability ; typically we use @xmath157 . the hyperbolic equations ( [ eq : evol_bs]-[eq : evol_pi ] ) are discretized using a second - order accurate crank - nicholson type scheme , whereby we define two time levels , @xmath72 and @xmath158 , and obtain our finite difference stencils by expanding in taylor series about @xmath159 . this gives the following second - order accurate approximation to the first derivative of @xmath150 with respect to time @xmath160 second - order accurate approximations to functions and spatial derivative operators at @xmath159 are obtained by averaging the corresponding quantity , @xmath161 , in time : @xmath162 thus , after discretization of the evolution equations using ( [ ref_cn_time ] ) and ( [ ref_cn_spat ] ) , function values are only referenced at times @xmath72 and @xmath158 , even though the stencils are centered at time @xmath163 . specific forms for all the finite difference stencils that we use can be found in appendix [ app : solver ] . we add kreiss - oliger dissipation @xcite to the evolution of equations of @xmath164 and @xmath60 ( in addition to @xmath54 during partially constrained evolution ) , as described in appendix [ app : solver ] . to demonstrate that this is _ essential _ for the stability of our numerical scheme , we compare in fig . [ fig : diss ] the evolution of @xmath59 from simulations without and with dissipation , but otherwise identical . the elliptic equations ( [ eq : cons_alpha]-[eq : cons_betaz ] ) are solved using brandt s fas multigrid ( mg ) algorithm @xcite , described in some detail in appendix [ app : solver ] . there are no explicit time derivatives of functions in these equations , and we discretize them at a single time level @xmath151 ( _ i.e. _ we do not apply the crank - nicholson averaging scheme for the elliptics ) . we use either a fully constrained evolution , solving for @xmath165 and @xmath54 using the constraint equations and slicing condition , or a partially constrained evolution where instead of using the hamiltonian constraint to update @xmath54 , we use the evolution equation ( @xmath166 ) . partially constrained evolution has proven to be useful due to the occasional failure of the mg solver in the strong - field regime ( _ i.e. _ close to black hole formation ) . use of the evolution equation for @xmath54 ( rather than the hamiltonian constraint ) circumvents this problem in many instances ; however , in certain brill - wave dominated spacetimes , free evolution of @xmath54 is not sufficient to ensure convergence of the mg process . we are currently working to make the mg solver more robust in these situations . the code is written in a combination of rnpl ( rapid numerical prototyping language @xcite ) and fortran 77 . the hyperbolic equations are implemented in rnpl , which employs a point - wise newton - gauss - seidel iterative relaxation scheme to solve these equations , while the mg solver is implemented in fortran ( see appendix [ app : solver ] for more mg details ) . a pseudo - code description of the time - stepping algorithm used is as follows : .... a ) as an initial guess to the solution at time t+dt , copy variables from t to t+dt b ) repeat until ( residual norm < tolerance ) : 1 : perform 1 newton - gauss - seidel relaxation sweep of the evolution equations , solving for the unknowns at time t+dt 2 : perform 1 mg vcycle on the set of elliptic equations at time t+dt end repeat .... for the residual norm used to terminate the iteration we use the the infinity norm of the residuals of all updated unknowns . in this section , we describe some of the tests we have performed to check that we are solving the correct set of equations . the first test consists of checking the equations against those derived with a computer algebra system @xcite . by inputting the metric and coordinate conditions , the computer derived equations can then be subtracted from our equations and simplified . by finding that the differences simplify to 0 , we can conclude that two sets of equations agree . for diagnostic purposes and as tests of the equations and of their discretization , we compute several quantities during the numerical evolution . the first is the adm mass @xcite @xmath167 where the integral is evaluated on a flat 3-space , _ i.e. _ with metric @xmath168 . the spatial 3-metric @xmath169 is that from our curved space solution , but has its indices raised and lowered with the flat metric . integrating around the boundaries of our numerical grid , the normal vectors @xmath170 are @xmath171 and @xmath172 . after some algebra , the adm mass becomes m_adm & = & _ z_max ^4 d & & - _ z_min ^4 d & & + _ _ max ^4 dz . a second set of quantities we calculate are the @xmath173-norms of the residuals of the evolution equations for the extrinsic curvature ( [ eq : k_ab_dot ] ) , which we denote @xmath174 . because we do not directly evolve individual components of the extrinsic curvature , these residuals will not be zero ; however , they _ should _ converge to zero in the limits as the discretization scale @xmath175 , and the outer boundary positions @xmath176 . note that we include these last conditions because it is only in the limit @xmath177 that our outer boundary conditions are fully consistent with asymptotic flatness . the convergence properties of our code are measured by computing the convergence factor , @xmath178 , associated with a given variable , @xmath179 , obtained on grids with resolution @xmath180 , @xmath181 and @xmath182 via q_u = . in particular , for the case of @xmath183 ( second order ) convergence , we expect @xmath184 as @xmath185 . the first set of tests we present here are comparisons of @xmath76 and @xmath54 from the evolution of spherically symmetric initial data to the corresponding functions computed by a 1d spherically symmetric code , the details of which are presented in appendix [ app : spherical ] . in general , the results from the two codes are in good agreement . a sample comparison is illustrated in fig . [ fig : spherical ] which shows the scalar field obtained with the 1d code as well as two radial slices of the corresponding solution calculated using the axisymmetric code . note , however , that we do not expect _ exact _ agreement in the limit @xmath175 for a fixed outer boundary location , as the `` rectangular '' boundaries of the axisymmetric code are , in general , incompatible with precise spherical symmetry . in the second series of tests , we examine evolutions of brill waves and non - spherical scalar pulses . figs . [ fig : brill_10][fig : oblate_20 ] show results from two typical initial data sets , each computed using two distinct outer boundary positions . each figure plots ( a ) the adm mass @xmath186 , ( b ) the @xmath173-norm of the residual of the @xmath187 component of the evolution equation for the extrinsic curvature @xmath188 , and , ( c ) the convergence factor @xmath189 of @xmath54 , as functions of time ( the convergence factor @xmath161 for other functions exhibit similar behavior as @xmath189 , and so for brevity we do not show them ) . here , one expects to see an improvement of the results namely trends toward mass conservation early on , a zero residual , and a convergence factor of 4in the limits @xmath190 and @xmath191 . after energy has reached the outer boundary , and to a lesser extent before ( as is evident in the scalar field example in figs . [ fig : oblate_10 ] and [ fig : oblate_20 ] ) , we fail to get consistency with the evolution equation ( [ eq : k_ab_dot ] ) as @xmath190 , for _ fixed _ @xmath192 . this is a measure of the inaccuracy of our outer boundary conditions ( [ eq : obc_alpha]-[eq : obc_betarho ] ) ; though the trends suggest that we _ do _ achieve consistency in the limit @xmath191 . = 8.5 cm = 8.5 cm = 8.5 cm = 8.5 cm finally , we show some results of a simulation of black hole formation from the collapse of a spherically symmetric distribution of scalar field energy . again , by looking at spherically symmetric collapse we can compare with the 1d code ( obtaining the same level of agreement as seen in the example in fig . [ fig : spherical ] ) . however , here we want to show the behavior of our coordinate system ( in particular the maximal slicing ) in the strong - field regime , which demonstrates the need to incorporate black hole excision techniques and/or adaptive mesh refinement ( amr ) before attempting any serious investigation of physics with this code . [ fig : bh_plots ] shows plots of the adm mass estimate ( [ madm ] ) , an estimate of the black hole mass @xmath193 , where @xmath194 is the area of the apparent horizon , and the minimum value of the lapse as a function of time from the simulation . [ fig : bh_psi ] shows the conformal factor @xmath54 at several times , in the central region of the grid . maximal slicing is considered _ singularity avoiding _ , because as the singularity is approached in a collapse scenario , the lapse @xmath37 tends to 0 , as demonstrated in fig . [ fig : bh_plots ] . this effectively freezes the evolution inside the black hole , though it causes a severe distortion in the @xmath107 slices as one moves away from the black hole . this particular coordinate pathology is evident in fig . [ fig : bh_psi ] . recall from the @xmath2metric ( [ hab_def ] ) that @xmath195 determines proper length scales in the @xmath40 and @xmath41 directions ; thus the rapid growth with time of @xmath54 shown in fig . [ fig : bh_psi ] means that a given coordinate area represents increasing proper area . furthermore , the increase in magnitude of @xmath54 in the strong - field regime ( which happens even when black holes do not form , and in non - spherical scalar field and brill wave evolution , though not to the same extent as shown in fig . [ fig : bh_psi ] ) implies that our effective numerical resolution decreases in those regions , as some feature of the solution with a given characteristic size will span less of the coordinate grid . thus , in the end , even though maximal slicing may prevent us from reaching a physical singularity , the `` grid - stretching '' effect is just as disastrous for the numerical code , preventing any long - term simulation of black hole spacetimes . for these reasons we will add black hole excision techniques and amr before exploring physics with this code ; our efforts in this regard are well underway , and will be described elsewhere . we have described a @xmath0 gravitational evolution model which evolves axisymmetric configurations of gravitational radiation and/or a scalar field . a thorough battery of tests confirms that the correct equations are being solved . in particular , we have provided evidence that the code is second - order convergent , consistent and conserves mass in the limit where the outer boundary position goes to infinity . the unigrid code described here is the first step towards our long - term goal of studying a range of interesting theoretical and astrophysical phenomena in axisymmetry . these include gravitational collapse of various matter sources and gravitational waves , the corresponding critical phenomena at the threshold of black hole formation , head - on black hole collisions and accretion disks . to this end , we need to include support for angular momentum and additional matter fields in the code , as well as to add additional computational and mathematical infrastructure adaptive mesh refinement , black hole excision and the capability of running in parallel on a network of machines . all of these projects are under development , and results will be published as they become available . another goal of this project is to provide a platform from which to develop computational technology for 3d work . in particular , we see development of amr in axisymmetry as a precursor to its incorporation in 3d calculations . likewise , accurate and stable treatment of boundary conditions presents a continual challenge in numerical relativity , and it is possible that we can develop an effective treatment of boundaries in axisymmetry that will generalize to the 3d case . the authors would like to thank david garfinkle for helpful discussions . mwc would like to acknowledge financial support from nserc and nsf phy9722068 . fp would like to acknowledge nserc , the izaak walton killam fund and caltech s richard chase tolman fund for their financial support . ewh and sll would like to acknowledge the support of nsf grant phy-9900644 . sll acknowledges support from grant nsf phy-0139980 as well as the financial support of southampton college . ewh also acknowledges the support of nsf grant phy-0139782 . the majority of the simulations described here were performed on ubc s * vn * cluster ( supported by cfi and bckdf ) , and the * maci * cluster at the university of calgary ( supported by cfi and asra ) in this appendix we briefly mention some aspects of our multigrid ( mg ) routine , and list the set of finite difference approximations that we use . the constraint equations ( [ eq : cons_alpha]-[eq : cons_betaz ] ) are four elliptic equations which , for a fully constrained system , must be solved on every time slice ( _ i.e. _ spatial hypersurface ) . as such , it can be expected that the time taken by a given evolution will be dominated by the elliptic solver and hence we look for the fastest possible solver . currently , multigrid methods are among the most efficient elliptic solvers available , and here we have implemented a standard full approximation storage ( fas ) multigrid method with v cycling ( see @xcite ) to solve the four nonlinear equations simultaneously . ( when using the evolution equation for @xmath54 in a partially constrained evolution , we use the same multigrid routine described here , except we only solve for the three quantities @xmath95 and @xmath53 during the v - cycle ; @xmath54 is then simply considered another `` source function '' . ) a key component of the mg solver is the relaxation routine that is designed to _ smooth _ the residuals associated with the discretized elliptic equations . we use point - wise newton - gauss - seidel relaxation with red - black ordering ( see fig . [ fig : redblack ] ) , simultaneously updating all four quantities @xmath196 at each grid - point during a relaxation sweep . in addition to its use for the standard pre - coarse - grid - correction ( pre - cgc ) and post - cgc smoothing sweeps , the relaxation routine is also used to _ solve _ problems on the coarsest grid . we use half - weighted restriction to transfer fields from fine to coarse grids and ( bi)linear interpolation for coarse to fine transfers . we generally use 3 pre - cgc and 3 post - cgc sweeps per @xmath197-cycle , and likewise normally use a single @xmath197-cycle per crank - nicholson iteration . one complicating factor here is the treatment of the boundary conditions eqs . ( [ eq : bc_alpha]-[eq : bc_betarho ] ) and eqs . ( [ eq : obc_alpha]-[eq : obc_betarho ] ) . in accordance with general multi - grid practice , we view the boundary conditions as logically and operationally distinct from the interior equation equation . the outer boundary conditions can generally be expressed as rx [ ap : bc ] where @xmath198 . taking the derivative of ( [ ap : bc ] ) with respect to @xmath199 , we arrive at the differential form that is applied on the outer boundaries of the computational domain ( @xmath200 ) : x - x _ , - z x_,z = 0 . on the @xmath41-axis ( @xmath79 ) , the conditions eqs . ( [ eq : bc_alpha]-[eq : bc_betarho ] ) take one of the following two forms : a_i , = 0 [ eq : qfit ] or a_i = 0 [ eq : zfit ] equations of the former form are discretized using an @xmath183 backwards difference approximation to the @xmath40 derivative . the interior and boundary differential equations are solved in tandem via the multigrid approach : 1 . the residual is smoothed using some number of relaxation sweeps . for the interior , eqs . ( [ eq : cons_alpha]-[eq : cons_betarho ] ) are relaxed using red - black ordering as discussed above . after each call of this relaxation routine , a second routine that `` relaxes '' the boundary points is called @xcite . 2 . for quantities restricted from a fine to a coarse grid , discrete forms of ( [ eq : qfit ] ) and ( [ eq : zfit ] ) are applied during the @xmath197-cycle . at the other boundaries , straightforward injection is used . the key idea here is to ensure that the boundary relaxation process does not substantially impact the smoothness of the interior residuals , because it is only for smooth residuals that a coarsened version of a fine - grid problem can sensibly be posed . finally , in table [ table : diff ] , we show all of the difference operators we use to convert the differential equations listed in [ sec : coords_and_vars ] to finite difference form , using the crank - nicholson scheme described in section [ sec : implementation ] . in addition , as discussed in section [ sec : implementation ] , we use kreiss - oliger dissipation @xcite to maintain smoothness in the evolved fields . specifically , we add the kreiss - oliger filter to discretized evolution equations _ t a^n = _ t f^n ( ) by replacing the crank - nicholson time difference operator @xmath201 with @xmath202 : ^_t a^n = _ t f^n ( ) empirically , we find that a value of @xmath203 generally keeps our fields acceptably smooth . .finite difference operators and their correspondence to differential operators . here , @xmath204 is an arbitrary grid function defined via @xmath205 , where @xmath180 and @xmath206 are the spatial and temporal grid spacings , respectively . @xmath207 denotes either of the two spatial coordinates @xmath40 or @xmath41 , with the dependence of @xmath204 on the other suppressed . the parameter @xmath208 represents a user - specifiable `` amount '' of kreiss - oliger dissipation . [ cols="<,<,<",options="header " , ] one simple test of the code compares the results for spherically symmetric initial data with the output of a code which explicitly assumes spherical symmetry . here we present the equations for this 1d code . the spacetime metric is : ds^2 = -(^2 + ^4 ^ 2)dt^2 + 2 ^ 4dt dr + ^4 ( dr^2 + r^2 d^2 ) , where @xmath37 , @xmath209 and @xmath54 , are functions of @xmath199 and @xmath72 , @xmath210 is the line element on the unit 2-sphere , and @xmath209 is the radial component of the shift vector ( i.e. @xmath211 ) . adopting maximal slicing to facilitate direct comparison to the axisymmetric code , we have k^i_j = diag(k^r_r(r , t),0,0 ) . then a sufficient set of equations for the coupled einstein - massless - scalar system is @xcite + + 2 + ( k^r_r)^2 ^5 & = & 0 [ eq : ss_psi ] + ( k^r_r ) + 3 k^r_r + & = & 0 [ eq : ss_krr ] + - & = & 0 [ eq : ss_alpha ] ( ) & = & [ eq : ss_beta ] + & = & + [ eq : ss_ph ] + & = & ( + ) [ eq : ss_phi ] + & = & + & & - . [ eq : ss_pi ] here dot and prime denote derivatives with respect to @xmath72 and @xmath199 , respectively . the evolution equations ( [ eq : ss_ph]-[eq : ss_pi ] ) are discretized using an @xmath183 crank - nicholson scheme . ( [ eq : ss_psi ] ) and ( [ eq : ss_krr ] ) are similarly discretized using @xmath183 finite difference approximations , then solved iteratively for @xmath54 and @xmath212 at each time step . once @xmath76 , @xmath67 , @xmath54 and @xmath212 have been determined , @xmath37 and @xmath209 are found from @xmath183 finite - difference versions of eqs . ( [ eq : ss_alpha ] ) and ( [ eq : ss_beta ] ) . the code is stable and second - order convergent . 1 l. l. smarr , ph.d . dissertation , university of texas at austin , unpublished ( 1975 ) k. r. eppley , ph.d . dissertation , princeton university , unpublished ( 1977 ) s. l. shapiro and s. a. teukolsky , `` collisions of relativistic clusters and the formation of black holes '' , phys . rev . * d45 * , 2739 ( 1992 ) p. anninos , d. hobill , e. seidel , l. smarr and w. suen , `` the collision of two black holes , '' phys . lett . * 71 * , 2851 ( 1993 ) [ gr - qc/9309016 ] . j. baker , a. abrahams , p. anninos , s. brandt , r. price , j. pullin and e. seidel , `` the collision of boosted black holes '' , phys . * d55 * , 829 ( 1997 ) p. anninos and s. brandt `` head on collision of two unequal mass black holes , '' phys . lett . * 81 * , 508 ( 1998 ) [ gr - qc/9806031 ] . t. nakamura , `` general relativistic collapse of axially symmetric stars leading to the formation of rotating black holes , '' _ prog . of theor . physics _ * 65 * , 1876 - 1890 ( 1981 ) . r. f. stark and t. piran , `` gravitational wave emission from rotating gravitational collapse , '' phys . . lett . * 55 * , 891 ( 1985 ) . t. nakamura , k. oohara and y. kojima , `` general relativistic collapse of axially symmetric stars '' , prog . . suppl . * 90 * , 13 ( 1987 ) m. shibata , `` axisymmetric simulations of rotating stellar collapse in full general relativity criteria for prompt collapse to black holes '' prog . phys . * 104 * , 325 ( 2000 ) . a. m. abrahams , g. b. cook , s. l. shapiro and s. a. teukolsky , `` solving einstein s equations for rotating space - times : evolution of relativistic star clusters , '' phys . d * 49 * , 5153 ( 1994 ) . shapiro and s.a . teukolsky , `` formation of naked singularities : the violation of cosmic censorship , '' phys . * 66 * , 994 ( 1991 ) . d. garfinkle and g. c. duncan , `` numerical evolution of brill waves , '' phys . rev . * d 63 * , 044011 ( 2001 ) [ gr - qc/0006073 ] . m. alcubierre , s . brandt , b . brugmann , d . holz , e . seidel , r . takahashi , j . thornburg `` symmetry without symmetry : numerical simulation of axisymmetric systems using cartesian grids '' int . j. mod . d * 10 * , 273 ( 2001 ) . s. brandt , j.a . font , j.m . ibanez , j. masso and e. seidel , `` numerical evolution of matter in dynamical axisymmetric black hole spacetimes '' , comput . . commun . * 124 * 169 ( 2000 ) s. brandt and e. seidel , `` the evolution of distorted rotating black holes i : methods and tests '' , phys . rev . * d52 * , 856 ( 1995 ) s. brandt and e. seidel , `` the evolution of distorted rotating black holes ii : dynamics and analysis '' , phys . rev . * d52 * , 870 ( 1995 ) s. brandt and e. seidel , `` the evolution of distorted rotating black holes iii : initial data '' , phys . rev . * d54 * , 1403 ( 1996 ) a. m. abrahams and c. r. evans , `` trapping a geon : black hole formation by an imploding gravitational wave . , '' phys . d. * 46 * , 4117 ( 1992 ) . a. m. abrahams and c. r. evans , `` critical behavior and scaling in vacuum axisymmetric gravitational collapse , '' phys . * 70 * , 2980 ( 1993 ) . k. maeda , m . sasaki , t . nakamura , s . miyama `` a new formalism of the einstein equations for relativistic rotating systems . '' phys . * 63 * , 719 ( 1980 ) . k. maeda , `` [ ( 2 + 1)+1]-dimensional representation of the einstein equations , '' in _ proceedings of third marcel grossmann meeting on general relativity _ , ed . hu ning , ( science press , 1983 ) . r. geroch , `` a method for generating solutions of einstein s equations , '' j. math . * 12 * , 918 ( 1971 ) . misner , k.s . thorne and j.a . wheeler , _ gravitation _ , new york , w.h . freeman and company ( 1973 ) j.m . bardeen and t. piran , `` general relativistic axisymmetric rotating systems : coordinates and equations '' , phys . rep . * 96 * 205 ( 1983 ) . t. appelquist , a. chodos , and p.g.o . freund , `` modern kaluza - klein theories , '' addison - wesley , menlo park , 1987 ) . j. thornburg , `` finding apparent horizons in numerical relativity '' , phys . rev . * d54 * , 4899 ( 1996 ) m. alcubierre , s. brandt , b. brugmann , c. gundlach , j. masso , e. seidel and p. walker , `` test beds and applications for apparent horizon finders in numerical relativity '' , class . * 17 * 2159 ( 2000 ) kreiss , h .- o . , and oliger , j. , `` methods for the approximate solution of time dependent problems '' , global atmospheric research program publication no . 10 , world meteorological organization , case postale no . 1 , ch-1211 geneva 20 , switzerland ( 1973 ) . software available from http://laplace.physics.ubc.ca/members/matt/rnpl/index.html we use waterloo maple along with a tensor package written by one of the authors ( mwc ) . a. brandt , `` multilevel adaptive solutions to boundary value problems , '' _ math . of computation _ * 31 * , 333 - 390 ( 1977 ) . a. brandt , `` guide to multigrid development , '' in _ lecture notes in mathematics _ * 960 * , 220 - 312 , ( springer - verlag , new york , 1982 ) . choptuik , `` a study of numerical techniques for radiative problems in general relativity , '' ph.d . thesis , the university of british columbia , unpublished ( 1982 ) . a subtle point here concerns the `` relaxation '' occurring on the boundaries . the finite - difference equations on the boundary yield algebraic equations which determine the given fields there `` exactly . '' the subtlety arises because these algebraic conditions couple neighboring boundary points and thus despite the fact that we solve these equations exactly , the residual ( as computed after the entire boundary is `` relaxed '' ) will not be identically zero . we find this procedure suffices to keep residuals sufficiently smooth over both interior and boundary domains .

we present single pointing observations of sio , hco and hco from the james clerk maxwell telescope towards 23 massive star forming regions previously known to contain molecular outflows and ultracompact hii regions . we detected sio towards 14 sources and suggest that the non - detections in the other nine sources could be due to those outflows being older and without ongoing shocks to replenish the sio . we serendipitously detected so towards 17 sources in the same tuning as hco . we detected hco towards all sources , and suggest that it is tracing infall in nine cases . for seven infall candidates , we estimate mass infall rates between 1 and 2 m yr . seven sources show both sio detections ( young outflows ) and hco infall signatures . we also find that the abundance of hco tends to increase along with the abundance of sio in sources for which we could determine abundances . we discuss these results with respect to current theories of massive star formation via accretion . from this survey , we suggest that perhaps both models of ionized accretion and halted accretion may be important in describing the evolution of a massive protostar ( or protostars ) beyond the formation of an hii region .

0704.1245

the dynamics in massive star forming regions are , in general , much more complex than in regions which form only low mass stars . for instance , in the early stages within a low mass star forming region , the dynamics can be understood in terms of a few broad categories : large scale infall , which causes a disk to form , accretion through the disk , and outflow to release angular momentum ( see for example , di francesco et al . 2001 , andr et al . 1993 , muzerolle et al . 2003 ) . in intermediate and high mass star forming regions , turbulence , stellar winds , multiple sites of star formation , and , for regions with massive star formation , the presence of hii regions , all contribute to the dynamics in these regions as well ( i.e. beuther et al . 2006 , shepherd & churchwell 1996 , mckee & tan 2003 , krumholz et al . the more complicated source dynamics make the processes involved in the formation of the most massive stars much more difficult to understand than those involved in the formation of lower mass stars . adding to the complexity , massive stars do not form as often as their lower mass counterparts and so we must look to larger distances before finding examples of high mass star formation . for instance , the average distance to the 63 sources in shirley et al . ( 2003 , hereafter s03 ) is 5.3 kpc . if we assume massive stars form through accretion , that this accretion occurs in the inner regions of disks ( i.e. pudritz & norman 1986 , or shu et al . 1994 ) , and that these disks have radii of a few thousand au ( chini et al . 2004 , beltran et al . 2004 , cesaroni et al . 2005 ) , we do not yet quite have the resolving power to detect accretion directly at the distances to massive star forming regions ( 1000 au at 5.3 kpc is @xmath7 ) . here , we define accretion as the infall motions from the disk onto the forming star , in contrast to the larger scale motions of envelope material falling onto the disk . however , while we can not observe accretion directly , its presence can be inferred from the presence of accretion tracers such as larger scale infall and outflow . infall can act to replenish disk material as mass accretes onto a protostar ( nakamura 2000 ) , while molecular outflows serve as a release mechanism for the angular momentum which builds up during the accretion process ( e.g arce et al . these large scale motions are seen in star forming regions of all mass scales ( see for instance beuther & shepherd 2005 ) . in massive star forming regions , the accretion rates are orders of magnitude greater than in low mass star forming regions ( i.e. beuther et al . 2002 ) , while the accreted masses are only approximately one order of magnitude greater . these accretion rates and masses result in accretion timescales that are much shorter than in low mass star forming regions , which allows the kelvin - helmholtz timescale to become important in the evolution of the protostar ( e.g. , there is no pre - main sequence stage for massive star formation ) . the outward radiation and thermal pressure from the forming star becomes strong enough that it can ionize the surrounding medium and a small , highly ionized hii region ( either hypercompact ( hchii ) or ultracompact ( uchii ) region , keto 2003 ) can form . it is still unclear whether the outward pressure needed to create the hii region is strong enough to halt accretion , or whether accretion can continue in some form ( either through a molecular or ionized disk , or through an ionized accretion flow ) after the formation of an hii region . some models suggest that accretion must halt before the onset of a visible uchii region ( i.e. garay & lizano 1999 , yorke 2002 ) , while other models suggest that an ionized accretion flow can continue through an hii region ( i.e. keto 2003 , 2006 ) . there is now also observational evidence which suggests that , once the protostar becomes hot enough to ionize its surroundings , both modes of massive star formation ( halted and ionized accretion ) are possible . g10.6 - 0.4 has been shown to have an ionized accretion flow by sollins et al . ( 2005 ) and keto & wood ( 2006 ) , while accretion in g5.89 - 0.39 seems to have halted at the onset of the uchii region ( klaassen et al . although sample statistics at this point are still quite small , these two examples pose interesting questions . we do not yet have enough data to determine whether the apparently conflicting models of halted and ionized accretion can both be correct . however , we can begin with a uniform survey of infall and outflow tracers in massive star forming regions in order to constrain massive star formation scenarios . in this paper , we present a survey of 23 massive star forming regions . because we are interested in the relationship between accretion and outflow after the formation of an hii region , our source selection criteria include ( 1 ) the presence of an uchii region , which indicates that there is a massive protostar forming , and ( 2 ) previous evidence of outflows , which suggests ongoing accretion in most formation scenarios . sources were selected based on inclusion in the wood & churchwell ( 1989 ) and kurtz et al . ( 1994 ) catalogs of uchii regions as well as having molecular outflow signatures in the plume et al . ( 1992 ) survey of massive star forming regions . additional sources were taken from hunter ( 1997 ) which were shown to have both uchii regions and molecular outflows . we describe the observations collected for this survey in section [ sec : observations ] , we discuss the results of these observations in section [ sec : results ] , and present our conclusions in section [ sec : conclusions ] . observations of sio ( j=8 - 7 ) , hco@xmath0 , and h@xmath1co@xmath0 ( j=4 - 3 ) were taken at the james clerk maxwell telescope ( jcmt ) in 2005 ( as parts of projects m05ac11 and m05bc04 ) . sio ( 347.330 ghz ) and h@xmath1co@xmath0 ( 346.999 ghz ) were observed simultaneously in the same sideband by tuning the receiver to 347.165 ghz . thirteen or twenty minute observations , depending on the source elevation , were taken towards each source with a velocity resolution of 1.08 km s@xmath6 , which resulted in rms noise levels of [email protected] k. separately , we observed hco@xmath0 ( 356.370 ghz ) with a velocity resolution of 0.53 km s@xmath6 to an rms noise limit of [email protected] k in twenty minute integrations . both sets of observations were taken in position switching mode with dual mixers and a sideband rejection filter in place . table [ tab : sources ] shows the positions , rms noise limits of both tunings , the local standard of rest velocity , and distances to all sources in this survey . the half power beam width for these observations is 15@xmath9 , and the main beam efficiency is @xmath10=0.62 . data were obtained using the das autocorrelator system and reduced using the specx software package . linear baselines were removed from all spectra except for those towards g10.47 . for this source , there were so many different chemical species in the observed spectrum that we were unable to fit a linear baseline over the entire @xmath11 700 km s@xmath6 bandwidth of the 347 ghz observations or the @xmath11 450 km s@xmath6 bandwidth of the 356 ghz observations . in this case , no baseline was removed . we also present @xmath12 maps of one source ( g45.07 ) in the same emission lines . these raster maps are sampled every 5 and have rms noise limits of 0.10 and 0.14 k ( t@xmath13 ) for the 347 and 356 ghz tunings , respectively . note that these values are different than the ones reported in table [ tab : sources ] for the single pointing observations . the das autocorrelator was configured with the same tunings as were described above for the single pointing observations , and the maps were centered at the same position . this survey of single pointing observations towards 23 massive star forming regions is meant as an initial , uniform survey from which to base future observations . with these observations , we can only comment on the molecular gas component within our beams ; we can not discuss the larger scale molecular dynamics , or the ionized gas components of these regions . for our sources , the selection criteria of having an hii region confirms that these regions are forming massive stars . figures [ fig : sno_sio ] through [ fig : hdet_sio ] show the single pointing sio and hco@xmath0/h@xmath1co@xmath0 spectra towards all sources ; the spectra are ordered according to sio integrated intensity . for each panel in these figures , line brightnesses have been corrected for the jcmt main beam efficiency and centered on the @xmath14 of the source ( table [ tab : sources ] ) . figures [ fig : sno_sio ] and [ fig : hno_sio ] show sio and hco@xmath0/h@xmath1co@xmath0 spectra , respectively , towards the sources with no sio detections . figures [ fig : sdet_sio ] and [ fig : hdet_sio ] show the sio and hco@xmath0/h@xmath1co@xmath0 spectra , respectively , towards sources with sio detections . peak line strengths and integrated intensities are given for all three lines in table [ tab : fits ] . sio was only detected in 14 out of our 23 sources , where we define a detection as a minimum of 4 @xmath15 in integrated intensity . the rms noise limits in integrated intensity were calculated using @xmath16 where @xmath17 is the rms noise level in k , @xmath18 is the velocity resolution of the observations , and @xmath19 is the number of channels over which the integrated intensity is calculated . hco@xmath0 was detected in all sources and h@xmath1co@xmath0 was detected in all but two sources . along with hco@xmath0 , we serendipitously observed so@xmath2 ( j=10@xmath20 - 10@xmath21 at 356.755 ghz ) in 17 of our sources . the hco@xmath0 and so@xmath2 lines are only separated by 17 km s@xmath6 and thus the lines were blended in eight sources . for the sources with so@xmath2 detections , we have also plotted ( in gray ) the lower intensity hco@xmath0 observations in order to highlight the so@xmath2 emission ( figures [ fig : hno_sio ] and [ fig : hdet_sio ] ) . double peaked hco@xmath0 line profiles were observed towards 10 sources , with nine of them having stronger blue peaks than red . this blue line asymmetry in an optically thick tracer such as hco@xmath0 is often suggestive of infall ( i.e myers et al . 1996 ) . we discuss the possibility of our observations tracing large scale infall further in section [ sec : hcop ] . the distances to our sources , as taken from the literature , are shown in table [ tab : sources ] . the average distance is 5.7 @xmath22 3.8 kpc , where the error quoted reflects the 1 @xmath15 dispersion in the distances . since our observations were taken with a 15@xmath9 beam , this resolution corresponds an average linear size of 0.4 pc for our observations . sio is a well known outflow tracer , since in the general interstellar medium , si is frozen out onto dust grains . when the gas in a region is shocked ( i.e. the gas through which a protostellar outflow is passing ) the dust grains can sublimate and si is released into the gas phase . after the passage of a shock , the sio abundance ( [ sio]/[h@xmath2 ] ) can jump to almost 10@xmath24 , whereas the dark cloud abundance of sio is often closer to 10@xmath25 ( see for example schilke et al . 1997 , caselli et al . 1997 , or van dishoeck & blake , 1998 ) . while sio is easily identified as an outflow tracer , the emitting region for hco@xmath0 is much less certain . many authors suggest that hco@xmath0 can be used to trace the envelope material surrounding a protostellar region ( i.e. hogerheijde et al . 1997 , rawlings et al . 2004 ) , while others suggest that it traces disk material ( i.e. dutrey et al . one thing that is apparent , however , is that it becomes optically thick very quickly and readily self absorbs . we detected so@xmath2 in 17 of our sources , suggesting that our beam contains at least some molecular gas at temperatures greater than 100 k ( see for instance doty et al . 2002 , charnley 1997 ) . however , fontani et al . ( 2002 ) have determined the average temperature in twelve massive star forming regions to be 44 k , using observations with beam sizes comparable to those presented here . for seven of our sources , which were observed in the fontani et al ( 2002 ) sample , the average temperature is also 44 k. thus , in the following analysis , we adopt an ambient temperature of 44 k for all sources . table [ tab : results ] shows the column densities for each region derived for both sio and h@xmath1co@xmath0 . the column density was calculated assuming that each tracer is optically thin , in local thermodynamic equilibrium , and at an ambient temperature of 44 k. for optically thin lines , the column density of the observed transition scales directly with the integrated intensity of the line ( see for instance , tielens 2005 ) : @xmath26 where @xmath27 is the column density in the upper state of the transition , @xmath28 is the einstein a coefficient , @xmath29 is the frequency of the j=@xmath30-@xmath31 transition , and @xmath32 is the integrated intensity of the line . the column density of this one state can then be related to the total column density of that molecule through the partition function . it is the total column density for the molecule ( not the observed state ) that is presented in table [ tab : results ] . for each source we determined the column density , or upper limit to the column density , in sio ( table [ tab : results ] ) using the methods described above . these column densities can be compared to the column densities of other molecules ( i.e. cs ) for the same regions in order to determine the fractional abundance of sio , if the abundance of the other molecule is known . we were able to obtain the cs or c@xmath33s column densities for fifteen of our sources from the literature . column densities for fourteen sources were taken from plume et al . ( 1997 , hereafter p97 ) , with the column density for one additional source taken from wang et al . ( 1993 ) . for those sources with c@xmath33s column densities instead of cs column densities , we assumed an abundance ratio of [ cs]/[c@xmath33s ] = 22 ( wilson & rood , 1994 ) to determine a cs column density . the abundance of cs , relative to h@xmath2 , was calculated by s03 for 13 of these sources , and we assume a cs abundance of @xmath34 for the other two source for which the cs column density is known , since this was the average cs abundance as calculated by s03 . we then compare the column density and abundance of cs to our observed sio column density , or column density upper limit , to determine the abundance of sio relative to h@xmath2 in our sources ( table [ tab : results ] ) . despite our source selection criteria requiring previous evidence of outflows , we detected sio towards only 14 of our 23 sources . this raises a number of questions , such as : is the observed sio in fact tracing outflow if we do not detect it in all sources ? why do we not detect sio in all sources ? is the signal being beam diluted at large distances ? has the si evolved into other species ? below we first address whether the detected sio can be used as an outflow tracer , and then discuss reasons for our non - detections of sio in nine of our sources . si is liberated in shocks , and if these shocks are not due to the outflow , they must be due to the photo dissociation region ( pdr ) surrounding the uchii region . evidence for diffuse ( not collimated ) sio can be seen in w75n ( shepherd , kurtz & testi , 2004 ) suggesting that the sio may be due to the pdr and not an outflow . models and observations of sio in the pdrs around high mass star forming regions suggest moderate sio enhancement , and that the sio abundance is independent of the ambient radiation field ( i.e. schilke et al . 2001 ) . schilke et al . ( 2001 ) find sio column densities of @xmath35 @xmath36 in their observed pdrs . this is , admittedly , below our detection threshold ; however , we detect average sio column densities of @xmath37 @xmath36 . this suggests a possibly higher sio abundance than found in pdrs . the enhanced sio column density alone is not enough to discount the origin of the sio in our sources as the pdr and so we can consider how our sio abundance varies with the ambient radiation field . for the gas near an hii region , we can approximate the strength of the ambient radiation field using the far infrared ( fir ) luminosity of the region . for twelve of our sources with sio detections and abundance calculations , we obtained the fir luminosity from either wood & churchwell ( 1989 ) , kurtz et al . ( 1994 ) or evans et al . we then scaled their values for the different source distances used in this study ( see table [ tab : results ] ) . comparing the sio abundance to the source luminosity ( see figure [ fig : iras ] ) , we find that the sio abundance increases with source luminosity . there is a 10% chance that this relationship could arise from uncorrelated data . thus , it is possible that our result is contrary to the findings of schilke et al ( 2001 ) , and we suggest that the sio we observe does come primarily from outflow shocks . we can also compare our detection rate of sio to that of the sio survey towards maser sources of harju et al ( 1998 ) who observed sio ( j=2 - 1 ) and sio ( j=3 - 2 ) . for comparison to our results , we only consider the sources in harju et al . which are listed as uchii regions . our detection rate is 61% , compared to their rate of 29% . while our detection threshold is slightly lower ( our observations have rms noise levels generally below 0.05 k at 347 ghz , while their rms noise levels are generally below 0.08 k ) , we suggest that the different detection rates are due to differences in the source selection criteria . although both samples contain uchii regions , our sample contains sources with previous observations of outflows , while harju et al . have selected sources based on previous observations of masers . for the 12 sources which overlap between the two studies , we detect sio towards 9 sources , while they detect sio towards 10 . they detected sio ( j=2 - 1 ) in g31.41 while we did not detect it in sio ( j=8 - 7 ) . based on this comparison to sio observations of uchii regions _ not _ selected by outflows , which have a lower sio detection rate , and that our sio abundance increases with source luminosity , we suggest that the sio , in the 14 sources in which it is detected , is being generated in the outflow . previous , high resolution observations of sio also suggest that sio can be enhanced in the outflows from high mass stars ( i.e. beuther et al . 2004 , beuther , schilke & gueth 2004 ) just as it is in the outflows from low mass stars . the nine non - detections in our sample could be caused by beam dilution if these sources are on average further away . however , if we compare the distances for sources with and without sio detections , we find average distances of 6.3 @xmath22 4.4 and 4.4 @xmath22 2.5 kpc , respectively . thus , the non - detections can not be attributed to larger average source distances and so beam dilution can play only a minimal role in the non - detections of sio . if these sio non - detections are not due to distance effects , there must be some local phenomenon which can explain why sio is not being detected in regions known to contain protostellar outflows . it is possible that the si is evolving into different species and the sio abundance is dropping back down to dark cloud values . pineau des forts et al . ( 1997 ) suggest that a few@xmath38 yr after the si is liberated from dust grains and forms sio , it can either freeze out back onto dust grains or oxidize and form sio@xmath2 . thus , the lack of sio may be due to silicon moving into other species if it was liberated more than 10@xmath39 yr ago . this interpretation implies that we did not detect sio in some of our sources because the outflow generating mechanism shut off more than 10@xmath39 years ago , and the outflow observed in hco@xmath0 ( or in other molecules by other authors ) is a remnant of previous accretion . the kinematic ages of nine of our sources are listed in the wu et al ( 2004 ) catalog of high velocity outflows . of these nine sources , five were also included in the p97 and s03 studies . this results in five sources for which we have both the kinematic age of the outflow , and the abundance of sio . the relationship between outflow age and sio abundance is shown in figure [ fig : age_abund ] along with the model predictions of pineau des forts et al . ( 1997 ) . with only five points , it is difficult to draw conclusions about the relationship between outflow age and sio abundance , especially given the uncertain beam filling factor . at higher resolution , these points would likely move upwards to higher abundances . a larger beam filling factor would move the points for the two young outflows ( g5.89 and cep a ) towards the model predictions . as for the other three outflows , this would move them further from the model predictions . we suggest that the outflow generating mechanism is continuing to shock these regions , replenishing the sio . the oldest of these sources ( g192.58 ) shows an infall signature in hco@xmath0 , which also suggests that the outflow is still being powered . given the large average distance to our sources ( 5.7 kpc ) , our 15@xmath9 beam subtends an average linear distance of 0.4 pc . thus , it is quite likely that the hco@xmath0 emitting region does not entirely fill our single jcmt beam . in general we can determine the beam filling factor , @xmath40 , for each source using @xmath41 where @xmath42 is the line brightness temperature measured at the telescope ( corrected for telescope efficiencies ) and @xmath43 is an approximation to the ambient temperature ( 44 k ) which is valid at densities greater than @xmath44 @xmath45 ( rholfs & wilson , 1994 ) . in a number of our sources , the hco@xmath0 is asymmetric , and thus can not be consistently used to determine a beam filling factor . instead , we can use the optically thin h@xmath1co@xmath0 line , and we can simplify the above equation to @xmath46 . these values are shown in table [ tab : results ] . hco@xmath0 becomes optically thick quite quickly due to its relatively high abundance with respect to h@xmath2 ( [ hco@xmath0]/[h@xmath2 ] @xmath48 ) , and as such , can be used to roughly trace outflow and to trace infall ( i.e. myers et al . 1996 ) if the line profile shows a double peak . we determined the optical depth of hco@xmath0 towards each source using equation 1 of choi et al . ( 1993 ) , and found that in all but one case ( g139.9 ) , it is optically thick . in all cases , the optical depth of hco@xmath0 is less than 77 ( the abundance ratio between hco@xmath0 and h@xmath1co@xmath0 , wilson & rood , 1994 ) , resulting in optically thin h@xmath1co@xmath0 towards all sources . because hco@xmath0 is optically thick in its line center , the line wings can be used to detect outflows , and so , if there is an outflow , it should be detectable in hco@xmath0 even if it goes undetected in sio . gaussian profiles were fit to our hco@xmath0 spectra ( either single or double gaussians , depending on the observed line shape ) , and the fits were subsequently subtracted from the spectra to leave only the residual outflowing gas . when using two gaussians to fit the self absorbed spectra , we employed a method similar to the single gaussian fitting of purcell et al . ( 2006 ) because we used the sides of the detected lines to fit our profiles ( see their figure 3 ) . comparing the two gaussian fits to single gaussian fits showed no significant differences in distinguishing line wing intensity . because of the possibility of contamination from so@xmath2 emission in the blue shifted outflow wing ( at -17 km s@xmath6 ) , the peak brightness of the residual emission was determined using only the red shifted wing emission . in all cases ( except for g10.47 for which we could not find a linear baseline ) , we found a minimum of a 5@xmath15 peak brightness temperature in the residual line wing emission , with 19 of our sources having a minimum of a 10@xmath15 peak . this result suggests that we can detect outflow motions in all sources using our detections of hco@xmath0 , despite not detecting sio towards every source . our observations show that for ten of our sources , the spectral line profile of the hco@xmath0 emission has a double peak . this profile could either be due to self absorption of the optically thick hco@xmath0 line or from multiple velocity components within our 15@xmath9 beam . to break this degeneracy , we observed the optically thin h@xmath1co@xmath0 . if the h@xmath1co@xmath0 line has a single peak at the same velocity as the hco@xmath0 absorption feature , then it is likely that the hco@xmath0 line is self absorbed . if , however , the h@xmath1co@xmath0 also has two peaks , and they are at approximately the same velocities as the two hco@xmath0 peaks , it would suggest that there are multiple components within the beam . of our ten sources with double hco@xmath0 peaks , only one shows a double peak profile in h@xmath1co@xmath0 ( g20.08 ) . this results in nine sources with double peaked optically thick hco@xmath0 . in addition to the nine optically thick sources , similar line asymmetries appear in a number of other sources . however , in these sources , there is no clear emission gap producing a double peak profile , only an emission shoulder ( i.e. de vries & myers 2005 ) . if we take g75.78 as an example , the hco@xmath0 line peak is red shifted from the rest velocity of the source , with a blue shifted emission shoulder . there are a number of different kinds of source dynamics that can lead to the double peaked line profiles seen in our spectra , such as infall , outflow and even rotation . however , infall is the only one of these processes which would produce line asymmetries which are consistently blue ( i.e. the blue peak is higher than the red peak or shoulder ) . if these profiles were due to outflow or rotation , there would be no statistical reason to have more sources with higher blue peaks than red peaks . many previous studies have investigated the statistical significance of using this type of optically thick blue line asymmetry to trace infall as opposed to other dynamical motions ( i.e. mardones et al . 1997 and gregersen et al . 1997 for low mass star forming regions , and fuller et al . 2005 for high mass star forming regions ) . of the 10 sources in our survey which have double peaked hco@xmath0 profiles , we suggest eight may be indicative of infall . the other two sources are g20.08 and g45.47 . g20.08 has already been shown to have multiple components in the beam from the double peaked h@xmath1co@xmath0 profile , and g45.47 has a brighter red peak than blue . there are two additional sources ( g19.61 , and g240.3 ) in which hco@xmath0 has a strong red shifted shoulder , which we suggest may also be tracing infall . this analysis gives a total of ten infall candidates in our sample of 23 sources . a recent survey of hco@xmath0 ( j=1 - 0 ) towards sources with methanol masers shows an even distribution of sources with blue and red line asymmetries , and a higher percentage of self absorbed lines than in our study ( purcell et al . 2006 ) . of the six sources which overlap between our survey and that of purcell et al , all six are self absorbed in hco@xmath0 ( j=1 - 0 ) . five of them have blue line asymmetries consistent with infall , while only one ( g31.41 ) has its red peak brighter than its blue peak . we only find self - absorption in hco@xmath0 ( j=4 - 3 ) for three of these six sources . in two of the sources for which we do not see a clear self - absorption feature , we do see evidence for a red shifted shoulder which may be showing unresolved infall . the sources in purcell et al . ( 2006 ) have an even distribution of red and blue line asymmetries , while we have a clear bias towards detecting blue line asymmetries . this comparison could suggest that the higher energy j=4 - 3 transition of hco@xmath0 is a better tracer of infalling gas because it does not self absorb as readily as the j=1 - 0 transition . for each of the 8 sources with blue , double peaked hco@xmath0 profiles , we can determine an infall velocity ( @xmath49 ) using the two layer radiative transfer model of myers et al . ( 1996 ) . using their equation 9 , we find infall velocities for all eight double peaked infall sources ( table [ tab : infall_rates ] ) . the mass infall rate can then be determined using : @xmath50 where @xmath51 is the mean molecular weight ( @xmath52 ) , the geometric mean radius ( @xmath53 ) is the unresolved circular radius of the hco@xmath0 emitting region derived from the beam radius and the beam filling factor ( @xmath54 ) , and @xmath55 is the ambient source density . for seven of the sources in table [ tab : infall_rates ] , the ambient density was determined by either p97 , hofner et al . ( 2000 ) , or wang et al . we could not find the ambient density for the eighth source ( g192.58 ) . from this analysis , we determine mass infall rates ranging from 1@xmath3 to 2@xmath4 m@xmath5 yr@xmath6 . these values are slightly higher than those generally observed for low mass star forming regions , but are consistent with the accretion rates derived for high mass star forming regions by mckee & tan ( 2003 ) . since outflow rates are orders of magnitude higher in high mass star forming regions ( i.e. beuther et al . 2002 ) it is not unreasonable to suggest that infall rates are also much higher in these regions . the mass outflow rate for only one of these sources ( cep a ) can be determined from the wu et al . ( 2004 ) survey of high velocity outflows by dividing the mass in the outflow by the kinematic age of the outflow . we find the ratio of the mass outflow rate to the mass infall rate to be @xmath56 . this value is only slightly higher than values seen in other high mass star forming regions ( i.e. behrend & maeder 2001 ) . also , models suggest that a mass equivalent to 20 - 30% of the mass accreted onto a protostar is ejected as a wind ( i.e. pelletier & pudritz 1992 , shu et al . 1994 ) , and that this wind entrains 5 - 20 times its mass in the outflow ( matzner & mckee 1999 ) . at a distance of 9.7 kpc , g45.07 is one of our furthest sources . this source was known from previous observations to have multiple continuum sources ( de buizer et al . 2003 , 2005 ) . the three continuum sources were observed in the mid - infrared ( mir ) and all three fall within 6@xmath9 of our map center . a fairly young outflow has also been mapped at high resolution ( @xmath57 synthesized beam ) in co and cs towards this region ( hunter et al . they observed a bipolar outflow with a position angle of -30@xmath58 ( east of north ) , as well as a red shifted absorption feature in their cs observations which they take to be indicative of infall . due to the large distance to this source , we should be able to detect all of the emission associated with this source in a fairly small map . in the left panel of figure [ fig : g45_maps ] we present a map of the sio ( contours ) and h@xmath1co@xmath0 ( halftone ) emission in this region . the right panel of figure [ fig : g45_maps ] shows the hco@xmath0 emission . in both figures , the 5@xmath15 ( 2.1 k km s@xmath6 ) h@xmath1co@xmath0 emission contour is plotted as a dashed line to help guide the eye . the first contour for sio in the left panel and only contour of hco@xmath0 in the right panel are also 5@xmath15 ( 2.2 and 3.5 k km s@xmath6 respectively ) . the differences in the 5@xmath15 contour levels for each tracer come from the different single channel rms levels between the two tunings , and the width of each line as given in table [ tab : fits ] . also plotted in both figures are the three mir continuum sources observed by de buizer et al ( 2003 , 2005 ) . if we did not have the added information provided by this map , there would be two main conclusions we could draw from our single pointing observations towards this source . the first is that the enhanced blue emission in all three tracers suggests that our pointing is observing more of the blue shifted outflow lobe than the red . second , since we only have one peak in the spectrum of each tracer , there is only one source and we can not classify it as infalling . the left panel of figure [ fig : g45_maps ] shows contours of sio emission superimposed on the h@xmath1co@xmath0 halftone . with beam spacings of 5@xmath9 , these maps are oversampled ; however we note that much of the structure in the sio emission is on scales comparable to the size of the jcmt beam . for instance , the structure at @xmath59 , @xmath60 is offset from the map center by more than the radius of our beam and could be independent from the emission at the map center . there is also sio emission at @xmath61 , @xmath62 , which is more than a full beam away from the map center , and suggests that the sio emission is more extended that the primary beam of our observations . in fact , it appears as though there is a second sio emission peak towards the upper left of the left panel of figure [ fig : g45_maps ] . interestingly , there does not appear to be as much h@xmath1co@xmath0 emission at this northern position . this comparison shows that the sio and h@xmath1co@xmath0 lines are tracing different gas populations in this region . the excess sio emission is offset from the map center in the same direction as the co emission shown in hunter et al . ( 1997 ) at much higher resolution . the right panel of figure [ fig : g45_maps ] shows the hco@xmath0 emission for this region . it appears that the hco@xmath0 emission extends much further than the sio emission , suggesting it is tracing the larger scale envelope material . the line through the middle of this plot indicates the cut taken for the position - velocity ( pv ) diagram along the outflow axis as described by hunter et al . ( 1997 ) the two panels of figure [ fig : pv ] show the pv diagrams for sio and hco@xmath0 in our maps both perpendicular and parallel to the outflow axis defined by hunter et al . our single pointing hco@xmath0 spectrum ( figure [ fig : hdet_sio ] ) suggests we are observing more blue shifted outflow emission than red shifted emission ; however , from our pv diagrams , we see that there is excess blue emission at all positions in our map . this excess blue emission can not be due to outflow alone ; instead , it could be due to an inherent velocity shift between the three continuum sources in our beam . we can , in fact , fit three gaussian components to most of our hco@xmath0 spectra . these gaussians peak at velocities of 60 , 52 and 44 km s@xmath6 , with the peak temperature for each component decreasing with velocity . the third component ( at 44 km s@xmath6 ) could not be fit at all positions because it was intrinsically weaker than the other two peaks , and was lost in the noise towards the edges of the map . it appears as though this third component might be contamination from so@xmath2 , which should occur at an apparent velocity of 41 km s@xmath6 ( or -17 km s@xmath6 in figure [ fig : hdet_sio ] ) . perpendicular to the outflow axis , the mean velocity of the hco@xmath0 line peak appears to shift from @xmath11 58 km s@xmath6 at an offset of @xmath63 from the source center to @xmath64 km s@xmath6 at an offset of @xmath65 from the source ( figure [ fig : pv ] ) . given our velocity resolution ( 1.08 km s@xmath6 ) and spatial resolution ( @xmath66 ) , is unclear whether this velocity shift is real . if it is , it could indicate large scale ( @xmath11 1.4 pc ) rotation within the core , on a much larger scale than would be expected for a rotating accretion disk . previously , we discussed the reasons why we do not detect sio towards a number of our sources , and have calculated the mass infall rates for the sources with double peaks in their hco@xmath0 emission , but we have not yet discussed the correlations between the two species . in figure [ fig : abund ] we plot the logarithm of the abundance of h@xmath1co@xmath0 against the same quantity for sio ( open circles ) , as well as the column densities of both species ( filled circles ) . the abundance of h@xmath1co@xmath0 was calculated in the same manner as the abundance of sio described above ( using the cs column density from p97 and the cs abundance from s03 ) . the probability of obtaining these correlations if the data are , in fact , uncorrelated is 6@xmath67 for the abundances , and 2@xmath68 for the column densities . as stated earlier , sio is a well known shock tracer , and as such , an increased abundance of sio would suggest more shocked material within our 15@xmath9 beam . the ( generally ) infall tracing hco@xmath0 has been shown by some authors to be destroyed in strong shocks ( i.e. bergin et al . 1998 , jrgensen et al . 2004 ) . however , wolfire & knigl ( 1993 ) suggest that hco@xmath0 can be enhanced in regions with high energy shocks , where electron abundances are much greater . this enhancement in the electron abundance increases the formation rate of ions , and we suggest that this is responsible for the h@xmath1co@xmath0 abundance enhancement in our sources . this correlation between the abundances of h@xmath1co@xmath0 and sio suggests that hco@xmath0 and h@xmath1co@xmath0 are not only tracing infalling gas , but also the outflowing gas as well . this conclusion is supported by the strong , and broad , line wing emission detected in hco@xmath0 ( see section [ sec : hcop ] ) . hco@xmath0 over abundances have been seen in high mass star forming regions not included in this study like ngc 2071 ( girart et al . 1999 ) and orion irc2 ( vogel et al . 1984 ) . in these two papers , the over abundances of h@xmath1co@xmath0 are with respect to ambient cloud tracers such as co and h@xmath2 , rather than the high density or shock tracers like the cs and sio with which we are comparing our h@xmath1co@xmath0 abundances . however , viti & williams ( 1999a , b ) show that hco@xmath0 is indeed over abundant with respect to cs in the gas surrounding hh objects , and jimnez - serra et al . ( 2006 ) also show that the abundance of h@xmath1co@xmath0 can be enhanced with respect to sio by up to a factor of ten in the same regions ahead of hh objects . without maps of each region , it is impossible to tell how much of the hco@xmath0 emission in the line center and in the line wings is due directly to infall and outflow motions ; however , based on the arguments we have presented above , we suggest that ten sources show infall motions , and all 23 source show outflow motions based on the hco@xmath0 line profiles . we have found evidence for recent outflow activity ( sio emission ) in 14 out of our 23 sources . seven of these outflow and infall sources overlap . m17s , g192.6 and g240 appear to show only infall signatures and no sio outflow signatures . they do , however , appear to have hco@xmath0 outflow signatures of a minimum of 8@xmath15 . detection of line wing emission in hco@xmath0 and the relationship between h@xmath1co@xmath0 and sio abundances described in the previous section suggest that while sio is tracing outflow in most sources and hco@xmath0 is tracing infall in some sources , hco@xmath0 is also observable in the outflowing gas for all regions . we find that the non - detection of sio in nine of our sources is not due to beam dilution or larger average distances to the source , but possibly to older outflows for which the si has likely either frozen back onto dust grains or evolved into sio@xmath2 . for these sources , it appears that the accretion may have ceased , and the observed outflowing gas is a remnant of previous accretion . we have found seven sources with sio outflow signatures but no infall signatures in hco@xmath0 . this result could be due to a number of factors such as beam dilution of the infalling gas which masks the spectral line profile we would expect for large scale infall . it is possible that , as the outflow ages and widens , it may impinge on the region in which we could detect infalling gas . for outflow cones oriented along the line of sight , the younger , narrower outflows would have infalling gas with large line of sight velocities and be likely to produce an observable infall signature . however , for the older outflows which have widened , the largest infall velocities will be in the plane of the sky , and unobservable at the resolution of the jcmt . however , this effect would not be as pronounced for outflows in the plane of the sky ( such as g5.89 , for which we do not see an infall signature ) . it is difficult to assess the importance of this effect without detailed information on the outflow orientation in each source . thus , we suggest that some of these sources may have finished accreting , and what we observe are remnant outflows from a previous phase of accretion . this scenario was suggested by klaassen et al . ( 2006 ) to explain the large scale outflow in g5.89 , and this source is one of these seven sources with an sio outflow and no apparent infall signature . the seven sources which show recent outflow activity ( those with sio emission ) and which appear to be undergoing infall are suggestive of ongoing accretion beyond the onset of the hii region . if accretion is ongoing in the presence of an hii region , then it seems likely that this accretion flow may be ionized . this ionized accretion scenario could be similar to low angular momentum accretion with high ionization as suggested by keto ( 2006 ) . thus , from this survey , we suggest that both models of ionized accretion and halted accretion may be important in describing the evolution of a massive protostar ( or protostars ) beyond the formation of an hii region . we would like to acknowledge the support of the national science and engineering research council of canada ( nserc ) . we thanks the referee for helpful comments which improved the paper . p.d.k would also like to thank e. keto and d. johnstone for helpful discussions during the preparation of this manuscript . 99 andre , p. , ward - thompson , d. , & barsony , m. 1993 , , 406 , 122 arce , h. g. , shepherd , d. , gueth , f. , lee , c .- f . , bachiller , r. , rosen , a. , & beuther , h. 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0603071 behrend , r. , & maeder , a. 2001 , , 373 , 190 beltrn , m. t. , cesaroni , r. , neri , r. , codella , c. , furuya , r. s. , testi , l. , & olmi , l. 2004 , , 601 , l187 bergin , e. a. , neufeld , d. a. , & melnick , g. j. 1998 , , 499 , 777 beuther , h. , schilke , p. , sridharan , t. k. , menten , k. m. , walmsley , c. m. , & wyrowski , f. 2002 , , 383 , 892 beuther , h. , et al . 2004 , , 616 , l23 beuther , h. , schilke , p. , & gueth , f. 2004 , , 608 , 330 beuther , h. , churchwell , e. b. , mckee , c. f. , & tan , j. c. 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0602012 beuther , h. , & shepherd , d. 2005 , cores to clusters : star formation with next generation telescopes , 105 caselli , p. , hartquist , t. w. , & havnes , o. 1997 , , 322 , 296 cesaroni , r. , neri , r. , olmi , l. , testi , l. , walmsley , c. m. , & hofner , p. 2005 , , 434 , 1039 charnley , s. b. 1997 , , 481 , 396 chini , r. , hoffmeister , v. , kimeswenger , s. , nielbock , m. , nrnberger , d. , schmidtobreick , l. , & sterzik , m. 2004 , , 429 , 155 choi , m. , evans , n. j. , ii , & jaffe , d. t. 1993 , , 417 , 624 churchwell , e. , walmsley , c. m. , & cesaroni , r. 1990 , , 83 , 119 de buizer , j. m. , osorio , m. , & calvet , n. 2005 , , 635 , 452 de buizer , j. m. , radomski , j. t. , telesco , c. m. , & pia , r. k. 2003 , , 598 , 1127 de vries , c. h. , & myers , p. c. 2005 , , 620 , 800 di francesco , j. , myers , p. c. , wilner , d. j. , ohashi , n. , & mardones , d. 2001 , , 562 , 770 doty , s. d. , van dishoeck , e. f. , van der tak , f. f. s. , & boonman , a. m. s. 2002 , , 389 , 446 dutrey , a. , guilloteau , s. , & guelin , m. 1997 , , 317 , l55 evans , n. j. , ii , et al . 1981 , , 244 , 115 fontani , f. , cesaroni , r. , caselli , p. , & olmi , l. 2002 , , 389 , 603 fuller , g. a. , williams , s. j. , & sridharan , t. k. 2005 , , 442 , 949 garay , g. , & lizano , s. 1999 , , 111 , 1049 girart , j. m. , ho , p. t. p. , rudolph , a. l. , estalella , r. , wilner , d. j. , & chernin , l. m. 1999 , , 522 , 921 gregersen , e. m. , evans , n. j. , ii , zhou , s. , & choi , m. 1997 , , 484 , 256 hanson , m. m. , luhman , k. l. , & rieke , g. h. 2002 , , 138 , 35 harju , j. , lehtinen , k. , booth , r. s. , & zinchenko , i. 1998 , , 132 , 211 hofner , p. , wyrowski , f. , walmsley , c. m. , & churchwell , e. 2000 , , 536 , 393 hogerheijde , m. r. , van dishoeck , e. f. , blake , g. a. , & van langevelde , h. j. 1997 , , 489 , 293 hunter , t. r. 1997 , ph.d . thesis , hunter , t. r. , phillips , t. g. , & menten , k. m. 1997 , , 478 , 283 jimnez - serra , i. , martn - pintado , j. , viti , s. , martn , s. , rodrguez - franco , a. , faure , a. , & tennyson , j. 2006 , , 650 , l135 jrgensen , j. k. , schier , f. l. , & van dishoeck , e. f. 2004 , , 416 , 603 keto , e. , & wood , k. 2006 , , 637 , 850 keto , e. 2003 , , 599 , 1196 keto , e. 2006 , arxiv astrophysics e - prints , arxiv : astro - ph/0603856 klaassen , p. d. , plume , r. , ouyed , r. , von benda - beckmann , a. m. , & di francesco , j. 2006 , , 648 , 1079 krumholz , m. r. , mckee , c. f. , & klein , r. i. 2005 , , 438 , 332 kumar , m. s. n. , fernandes , a. j. l. , hunter , t. r. , davis , c. j. , & kurtz , s. 2003 , , 412 , 175 kurtz , s. , churchwell , e. , & wood , d. o. s. 1994 , , 91 , 659 mardones , d. , myers , p. c. , tafalla , m. , wilner , d. j. , bachiller , r. , & garay , g. 1997 , , 489 , 719 matzner , c. d. , & mckee , c. f. 1999 , star formation 1999 , proceedings of star formation 1999 , held in nagoya , japan , june 21 - 25 , 1999 , editor : t. nakamoto , nobeyama radio observatory , p. 353 - 357 , 353 mckee , c. f. , & tan , j. c. 2003 , , 585 , 850 muzerolle , j. , calvet , n. , hartmann , l. , & dalessio , p. 2003 , , 597 , l149 myers , p. c. , mardones , d. , tafalla , m. , williams , j. p. , & wilner , d. j. 1996 , , 465 , l133 nakamura , f. 2000 , , 543 , 291 pelletier , g. , & pudritz , r. e. 1992 , , 394 , 117 pineau des forets , g. , flower , d. r. , & chieze , j .- 1997 , iau symp . 182 : herbig - haro flows and the birth of stars , 182 , 199 plume , r. , jaffe , d. t. , evans , n. j. , ii , martin - pintado , j. , & gomez - gonzalez , j. 1997 , , 476 , 730 plume , r. , jaffe , d. t. , & evans , n. j. , ii 1992 , , 78 , 505 pudritz , r. e. , & norman , c. a. 1986 , , 301 , 571 purcell , c. r. , et al . 2006 , , 367 , 553 rawlings , j. m. c. , redman , m. p. , keto , e. , & williams , d. a. 2004 , , 351 , 1054 rholfs , k. , wilson , t. l. 2004 , tools of radio astronomy ( 4th ed ; berlin ; springer ) schilke , p. , walmsley , c. m. , pineau des forets , g. , & flower , d. r. 1997 , , 321 , 293 schilke , p. , pineau des forts , g. , walmsley , c. m. , & martn - pintado , j. 2001 , , 372 , 291 shepherd , d. s. , & churchwell , e. 1996 , , 472 , 225 shepherd , d. s. , kurtz , s. e. , & testi , l. 2004 , , 601 , 952 shirley , y. l. , evans , n. j. , ii , young , k. e. , knez , c. , & jaffe , d. t. 2003 , , 149 , 375 shu , f. , najita , j. , ostriker , e. , wilkin , f. , ruden , s. , & lizano , s. 1994 , , 429 , 781 sollins , p. k. , zhang , q. , keto , e. , & ho , p. t. p. 2005 , , 624 , l49 tielens , a. g. g. m. 2005 , the physics and chemistry of the interstellar medium ( cambridge university press ) , 537 , 283 van dishoeck , e. f. , & blake , g. a. 1998 , , 36 , 317 viti , s. , & williams , d. a. 1999 , , 310 , 517 viti , s. , & williams , d. a. 1999 , , 305 , 755 vogel , s. n. , wright , m. c. h. , plambeck , r. l. , & welch , w. j. 1984 , , 283 , 655 yorke , h. w. 2002 , asp conf . ser . 267 : hot star workshop iii : the earliest phases of massive star birth , 267 , 165 wang , y. , jaffe , d. t. , evans , n. j. , ii , hayashi , m. , tatematsu , k. , & zhou , s. 1993 , , 419 , 707 wilson , t. l. , & rood , r. 1994 , , 32 , 191 wolfire , m. g. , & koenigl , a. 1993 , , 415 , 204 wood , d. o. s. , & churchwell , e. 1989 , , 69 , 831 wu , y. , wei , y. , zhao , m. , shi , y. , yu , w. , qin , s. , & huang , m. 2004 , , 426 , 503 lllcccccc g5.89 & 18 00 30.3 & -24 03 58 & 0.060 & 0.111 & 9 & 2&1 & w28a2 ( 1 ) + g5.97 & 18 03 40.4 & -24 22 44 & 0.044 & 0.069 & 10 & 2.7&5 & + g8.67 & 18 06 19.0 & -21 37 32 & 0.068 & 0.074&36 & 8.5&2 & 8.67 - 0.36 + g10.47 & 18 08 38.4 & -19 51 52 & & 0.100&67 & 12&2 & w31 ( 1 ) + g12.21 & 18 12 39.7 & -18 24 21 & 0.042 & 0.076&24 & 16.3&2 & 12.21 - 0.1 + m17s & 18 20 24.8 & -16 11 35 & 0.044 & 0.077&20 & 2.3&5 & m17 ( 2 ) + g19.61 & 18 27 38.1 & -11 56 40 & 0.048 & 0.065&43 & 4.5&1 & 19.61 - 0.23 + g20.08 & 18 28 10.4 & -11 28 49 & 0.044 & 0.068 & 42 & 4.1&1 & 20.08 - 0.13 + g29.96 & 18 46 03.9 & -02 39 22 & 0.044 & 0.079 & 98 & 9&1 & w43s + g31.41 & 18 47 33.0 & -01 12 36 & 0.044 & 0.061 & 97 & 8.5&1 & 31.41 + 0.31 + g34.26 & 18 53 18.5 & 01 14 58 & 0.047 & 0.131 & 58 & 3.7&1 & w44 + g45.07 & 19 13 22.1 & 10 50 53 & 0.044 & 0.073 & 59 & 9.7&1 & 45.07 + 0.13 + g45.47 & 19 14 25.6 & 11 09 26 & 0.037 & 0.063 & 58 & 8.3&6 & + g61.48 & 19 46 49.2 & 25 12 48 & 0.037 & 0.048 & 12 & 2&1 & s88 b + k3 - 50a & 20 01 45.6 & 33 32 42 & 0.035 & 0.077 & -24 & 8.6&3 & k3 - 50 + g75.78 & 20 21 44.1 & 37 26 40 & 0.038 & 0.068 & 0 & 5.6&1 & on 2n + cep a & 22 56 17.9 & 62 01 49 & 0.031 & 0.076&-10 & 0.7&1 & cep a + w3(oh ) & 02 27 03.8 & 61 52 25 & 0.026 & 0.079 & -48 & 2.4&2 & w3 ( oh ) + g138.3 & 03 01 29.2 & 60 29 12 & 0.066 & 0.073&-38 & 3.8&1 & s201 + g139.9 & 03 07 23.9 & 58 30 53 & 0.058 & 0.071 & -39 & 4.2&1 & + g192.58 & 06 12 53.6 & 17 59 27 & 0.037 & 0.053&9 & 2.5&3 & s255/7 + g192.6 & 06 12 53.6 & 18 00 26 & 0.074 & 0.098&9 & 2.5&3 & s255/7 + g240.3 & 07 44 51.9 & -24 07 40 & 0.035 & 0.048&68 & 6.4&4 & + + [ tab : sources ] lcc@@xmath22cccc@@xmath22cccc@@xmath22cc g5.89 & 2.3 & 75.3 & 0.7 & 120 & 39.2 & 690.0 & 0.8 & 105 & 8.9 & 65.2 & 0.4 & 35 + g5.97 & @xmath690.1 & @xmath690.3&0.1 & 8 & 12.4 & 50.5 & 0.2 & 15 & 0.6 & 1.6 & 0.1 & 8 + g8.67 & 0.2 & 1.6 & 0.3 & 13 & 13.8 & 80.0 & 0.3 & 38 & 5.6 & 20.9 & 0.2 & 10 + g10.47 & 1.0 & 9.9&&22 & 7.7 & 79.3 & 0.4 & 35 & 1.2 & 9.2&&15 + g12.21 & 0.2 & 2.5 & 0.2 & 25 & 10.0 & 93.8 & 0.3 & 30 & 0.8 & 6.5 & 0.2 & 15 + m17s & @xmath690.1 & @xmath690.7&0.2 & 15 & 16.2 & 79.0 & 0.2 & 15 & 2.9 & 9.4 & 0.2 & 11 + g19.61 & 0.8 & 15.9 & 0.3 & 45 & 10.8 & 162.2 & 0.4 & 55 & 1.8 & 16.1 & 0.2 & 20 + g20.08 & 0.4 & 6.9 & 0.3 & 40 & 6.5 & 75.0 & 0.3 & 40 & 1.1 & 9.7 & 0.2 & 25 + g29.96 & 0.5 & 7.9 & 0.3 & 40 & 23.6 & 160.4 & 0.3 & 30 & 3.8 & 13.2 & 0.1 & 10 + g31.41 & @xmath690.1 & @xmath690.7&0.2 & 15 & 1.6 & 6.8 & 0.2 & 13 & @xmath690.1 & @xmath690.3 & 0.1 & 3 + g34.26 & 1.3 & 26.0 & 0.4 & 55 & 27.8 & 228.2 & 0.6 & 35 & 4.3 & 33.2 & 0.2 & 25 + g45.07 & 0.7 & 10.3 & 0.3 & 37 & 15.0 & 169.4 & 0.4 & 45 & 1.5 & 11.2 & 0.3 & 33 + g45.47 & @xmath690.1 & @xmath690.3 & 0.1 & 12 & 10.1 & 59.4 & 0.2 & 20 & 1.3 & 6.1 & 0.1 & 12 + g61.48 & @xmath690.1 & @xmath690.3 & 0.1 & 8 & 10.8 & 58.5 & 0.1 & 18 & 1.5 & 4.7 & 0.1 & 8 + k3 - 50a & 0.3 & 2.1 & 0.1 & 16 & 18.6 & 154.8 & 0.3 & 25 & 1.8 & 10.6 & 0.2 & 18 + g75.78 & 0.3 & 2.2 & 0.1 & 13 & 12.5 & 136.8 & 0.3 & 45 & 2.5 & 10.3 & 0.2 & 17 + cep a & 0.5 & 7.3 & 0.2 & 45 & 23.2 & 234.7 & 0.4 & 60 & 4.5 & 23.4 & 0.2 & 28 + w3(oh ) & 1.4 & 14.5 & 0.2 & 50 & 18.5 & 148.2 & 0.3 & 30 & 2.4 & 11.6 & 0.1 & 15 + g138.3 & @xmath690.2 & @xmath690.3 & 0.1 & 2 & 5.0 & 10.6 & 0.2 & 8 & 0.3 & 0.5 & 0.1 & 4 + g139.9 & @xmath690.2 & @xmath690.3&0.1 & 11 & 8.6 & 18.8 & 0.1 & 6 & @xmath690.2 & @xmath690.3 & 0.1 & 2 + g192.58 & 0.3 & 1.3 & 0.1 & 8 & 14.5 & 84.2 & 0.2 & 25 & 1.1 & 5.5 & 0.1 & 15 + g192.6 & @xmath690.2 & @xmath690.6&0.2 & 30 & 12.0 & 54.1 & 0.3 & 15 & 0.8 & 2.2 & 0.2 & 8 + g240.3 & @xmath690.1 & @xmath690.4&0.2 & 10 & 7.3 & 59.1 & 0.2 & 30 & 0.4 & 2.5 & 0.1 & 14 + [ tab : fits ] lccrccccc g5.89 & y & n & 0.79 & 39.7 & 421.0 & -8.90 & 5.25&0.2 + g5.97 & n & n & 0.28 & 1.0 & @xmath691.7 & & 5.23 & + g8.67 & y & y & 0.24 & 12.7 & 9.0 & & 5.70 & + g10.47 & y & y & 0.16 & 5.6 & 55.4 & -9.53 & 6.26 & + g12.21 & y & n & 0.22 & 4.0 & 14.0 & -10.15 & 6.17 & + m17s & n & y & 0.33 & 5.7 & @xmath693.9 & @xmath69 - 10.73 & 5.72 & + g19.61 & y & y & 0.22 & 9.8 & 89.0 & -9.72 & 5.42 & + g20.08 & y & n & 0.13 & 5.9 & 38.6 & -10.48 & 4.86 & + g29.96 & y & n & 0.49 & 8.0 & 44.2 & -9.12 & 6.30 & + g31.41 & n & n & & @xmath690.2&@xmath693.9 & @xmath69 - 11.89&5.45 & + g34.26 & y & y & 0.58 & 20.2 & 145.0 & -10.43 & 5.77 & + g45.07 & y & n & 0.32 & 6.8 & 57.6 & -9.54 & 6.15&4 + g45.47 & y & n & 0.21 & 3.7 & @xmath691.7 & & 6.04 & + g61.48 & n & n & 0.23 & 2.9 & @xmath691.7 & @xmath69 - 10.80 & 5.01&7 + k3 - 50a & y & n & 0.40 & 6.5 & 11.7 & -11.60 & 6.35 & + g75.78 & y & n & 0.25 & 6.3 & 12.3 & & 5.65&3.7 + cep a & y & y & 0.47 & 14.2 & 40.8 & -9.78 & 4.40&0.2 + w3(oh ) & y & y & 0.39 & 7.1 & 81.1 & -11.16 & 5.12 & + g138.3 & n & n & 0.11 & 0.3 & @xmath691.7 & & 4.57&17 + g139.9 & n & n & & @xmath690.2 & @xmath691.7 & & 4.82 & 6 + g192.58 & y & y & 0.32 & 3.4 & 7.3 & -10.94 & 4.79&50 + g192.6 & y & y & 0.26 & 1.3 & @xmath693.4 & & & + g240.3 & y & y & 0.16 & 1.5 & @xmath692.2 & & & 2.3 + [ tab : results ] lc@@xmath22ccc@@xmath22c g8.67 & 0.4 & 0.1 & 1.8 & 4&2 + g10.47 & 1.8 & 0.3 & 7.2 & 100&80 + m17s & 1.4 & 0.5 & 5.0 & 4 & 2 + cep a & 0.23&0.07 & 10.0 & 0.17&0.07 + w3(oh ) & 0.06 & 0.02 & 60.0 & 3&1 + g192.58 & 0.8 & 0.3 & & + g192.6 & 0.9 & 0.4 & 4.0 & 2&1 + g34.26 & 1.5 & 0.3 & 3.6 & 14&4 + [ tab : infall_rates ]

in this work , we examine a two - higgs - doublet extension of the standard model in which one higgs doublet is responsible for giving mass to both up- and down - type quarks , while a separate doublet is responsible for giving mass to leptons . we examine both the theoretical and experimental constraints on the model and show that large regions of parameter space are allowed by these constraints in which the effective couplings between the lightest neutral higgs scalar and the standard - model leptons are substantially enhanced . we investigate the collider phenomenology of such a `` leptophilic '' two - higgs - doublet model and show that in cases where the low - energy spectrum contains only one light ,-even scalar , a variety of collider processes essentially irrelevant for the discovery of a standard model higgs boson ( specifically those in which the higgs boson decays directly into a charged - lepton pair ) can contribute significantly to the discovery potential of a light - to - intermediate - mass ( gev ) higgs boson at the lhc .

0903.0667

one of the primary goals of the large hadron collider ( lhc ) , a proton - proton collider with a center of mass energy @xmath2 tev , will be to investigate the sector responsible for the breaking of the electroweak symmetry . in the standard model ( sm ) , a single higgs doublet is responsible for the spontaneous breakdown of the @xmath3 gauge group to @xmath4 . the coupling constants of the sole physical higgs scalar to the rest of the sm particles are completely determined by their masses , and consequently there is little guesswork involved in determining the most promising channels @xcite in which one might hope to discover such a scalar . for a relatively light ( @xmath5 ) sm higgs boson , those channels are @xmath6 and @xmath7 , while for an intermediate - mass ( @xmath8 ) higgs , the single most promising channel is the weak - boson fusion ( wbf ) @xcite process @xmath9 @xcite . for a heavier higgs , with @xmath10 , the most relevant channels are @xmath11 and @xmath12 , with the higgs produced via either gluon fusion or wbf @xcite . in models where the higgs sector differs significantly from that of the standard model , however , the situation can change dramatically . this is true even in cases where the low - energy effective theory describing a given model at the weak scale contains only a single , light , @xmath0-even higgs scalar . indeed , at low energies , many models with extended higgs sectors have effective descriptions that are `` standard - model - like '' in the sense that they contain a single light higgs boson , but one whose couplings to the standard model fermions and gauge bosons differ potentially significantly from those of a sm higgs . such discrepancies , in turn , can translate into vast differences in lhc phenomenology : some ( or , in severe cases , even all ) of the standard detection channels for a sm higgs may disappear as a result of such modifications , while others , related to processes buried beneath background in the sm , may become crucial for discovery . one set of channels which are not terribly significant for the discovery of a sm higgs , but could become so in models with modified higgs sectors , consists of those involving direct decays of the higgs boson to a pair of high-@xmath13 leptons . in the sm , a light higgs boson ( with mass @xmath14 gev ) decays predominantly into @xmath15 , and the ratio @xmath16 ( where @xmath17 ) is roughly proportional to @xmath18 , due to the fact that in the sm , the same higgs doublet is responsible for giving mass to both quarks and leptons . consequently , attention has been focussed predominately on processes in which the higgs boson decays to a tau pair ( with a branching ratio of about 10% ) , and in particular on the weak - boson fusion process @xmath9 . this is the only process particularly relevant for sm higgs discovery in which the higgs decays directly to leptons , though it is now regarded as one of the most promising discovery channels for a sm higgs in the intermediate mass region @xcite . searching for the higgs in the @xmath19 and @xmath20 channels is more difficult , due to a combination of factors , including enhanced sm backgrounds and suppressed signal cross - sections . by contrast , processes in which a sm higgs boson decays into first- or second - generation leptons are generally assumed to be irrelevant for discovery . this is because under the assumption of yukawa - coupling universality among the lepton generations ( an assumption we will be making throughout the present work ) , the small size of @xmath21 compared to @xmath22 results in @xmath23 being roughly two orders of magnitude smaller than @xmath24 , with @xmath25 nearly three orders of magnitude smaller still . consequently , the rates for processes involving @xmath26 and @xmath27 are extremely suppressed relative to those involving tau pairs , both in the sm and in most simple extensions of the higgs sector . on the other hand , there are strong motivations for considering processes of this sort at the lhc . experimentally , a signal involving a pair of high-@xmath28 muons or electrons will be easy to identify , as the muon- and electron - identification efficiencies at each of the lhc detectors are each greater than 90%@xcite . furthermore , once a higgs boson is discovered in these channels , its mass could be readily reconstructed with high precision . such channels could also be of use in determining the higgs yukawa couplings to leptons . two - higgs - doublet models ( 2hdm ) , which stand as perhaps the simplest , most tractable example of a non - minimal electroweak - symmetry - breaking sector , provide a useful context in which to study the role of leptonic higgs - decay processes . these models arise in a number of beyond - the - standard - model contexts from supersymmetry to little higgs scenarios @xcite and have a rich phenomenology , many of whose consequences for lhc physics are still being uncovered . in general , 2hdm can be categorized according to how the higgs doublets couple to the sm quarks and leptons . in what has become known as a type i 2hdm , one doublet is responsible for the masses of both quarks and leptons , while the other decouples from the fermions entirely . in a type ii 2hdm , one higgs doublet couples to the up - type quark sector , while the other higgs doublet couples to both the down - type quark sector and the charged leptons as is the case , for example , in the minimal supersymmetric standard model ( mssm ) . in both of these standard scenarios , the leptonic branching ratios for a light higgs do not differ much from their sm values throughout most of parameter space coupling is suppressed due to radiative corrections @xcite , and consequently @xmath29 becomes large . ] , since the same doublet gives masses to both the bottom quark and the charged leptons . one interesting alternative possibility , which will be the primary focus of the present work , is a 2hdm scenario in which one higgs doublet couples exclusively to ( both up- and down - type ) quarks , while the other couples exclusively to leptons a scenario which we will henceforth dub the leptophilic two - higgs - doublet model ( l2hdm ) . this model has been discussed previously in the literature in relation to its effect on higgs branching fractions and decay widths @xcite , flavor physics @xcite , and potential implications for neutrino phenomenology @xcite and dark matter studies @xcite . some analyses of the lhc phenomenology of the model were presented in ref . @xcite , which focused on the non - decoupling region of the parameter space where additional physical higgs scalars are light . in this work , we discuss the leptonic decays of the lightest @xmath0-even higgs scalar in the l2hdm at the lhc . in particular , we examine the discovery potential in a decoupling regime in which only one light scalar , which resembles the sm higgs , appears in the low - energy effective description of the model . we begin in section [ sec : param ] by presenting the model and reviewing how the coupling structure of the lightest neutral higgs particle is modified from that of a sm higgs . in section [ sec : bounds ] , we discuss the applicable experimental constraints from flavor physics , direct searches , etc . and show that they still permit substantial deviations in the couplings between the higgs boson and the other sm fields away from their standard - model values . in section [ sec : brxsec ] , we discuss the implications of such modifications on the higgs branching ratios and production rates . in section [ sec : lhcsignatures ] , we discuss potential higgs discovery channels in which the higgs boson decays directly into a pair of charged leptons , and in section [ sec : combined ] , we calculate the discovery potential for a light , leptophilic higgs using the combined results from all of these leptonic channels . in section [ sec : conclusion ] we conclude . the l2hdm , as defined here , is a modification of the sm in which the higgs sector consists of two @xmath3 scalar doublets , both of which receive nonzero vacuum expectation values . the first of these doublets , which we call @xmath30 , couples only to ( both up- and down - type ) quarks , while the other , which we call @xmath31 , couples only to leptons . in other words , the yukawa interaction lagrangian is specified to be @xmath32 where @xmath33 , @xmath34 , and @xmath35 are @xmath36 yukawa matrixes , @xmath37 and @xmath38 respectively denote the left - handed quark and lepton fields , @xmath39 and @xmath40 respectively denote the right - handed up- and down - type quark fields , and @xmath41 denotes the right - handed lepton fields . this coupling structure can be achieved by imposing a @xmath42 symmetry under which @xmath43 and @xmath41 are odd , while all the other fields in the model are even . we will assume that this symmetry is broken only softly , by a term of the form @xmath44 in the scalar potential . in the l2hdm , that scalar potential takes the usual form common to all two - higgs doublet models . assuming that there is no @xmath0-violation in the higgs sector , this potential can be parameterized as follows @xcite : @xmath45\label{eq : vstandard}\end{aligned}\ ] ] it is assumed that the parameters of the theory are assigned such that both @xmath30 and @xmath43 acquire nonzero vevs ( which we respectively denote @xmath46 and @xmath47 ) , and that @xmath48 . we define @xmath49 as @xmath50 so that large @xmath49 corresponds to small @xmath47 , and therefore to large intrinsic lepton yukawa couplings . in the broken phase of the theory , the spectrum of the model includes the three massless goldstone modes which become the longitudinal modes of the @xmath51 and @xmath52 bosons , as well as five massive scalar degrees of freedom : two @xmath0-even fields @xmath53 and @xmath54 , a pseudoscalar @xmath55 , and a pair of charged fields @xmath56 . the relationship between the physical @xmath0-even higgs scalars @xmath53 , @xmath54 and the real , neutral degrees of freedom in @xmath30 and @xmath43 is parameterized by the mixing angle @xmath57 : @xmath58\\ \mathrm{re}[\phi_q^0-v_q]\end{array}\right ) . \label{eq : defalpha}\ ] ] in what follows , we will focus primarily on the physics of @xmath53 , the lightest of these two scalars . since the potential given in eqn . ( [ eq : vstandard ] ) includes eight model parameters @xmath59 @xmath60 , @xmath61 , @xmath62 , and @xmath63 which are subject to the constraint @xmath64 , seven of these eight parameters may be considered free . in what follows , it will be useful to work in a different , more physically meaningful basis for these parameters : @xmath65 where @xmath66 , @xmath67 , @xmath68 , and @xmath69 are the masses of the corresponding physical higgs scalars . in order to study the collider phenomenology of the l2hdm , it will be necessary to characterize how the effective couplings between @xmath53 and the sm fields differ from their sm values . ( [ eq : yuks ] ) indicates that the effective couplings between the fermions and @xmath53 are given in terms of these mixing angles ) and ( [ eq : defalpha ] ) used in defining @xmath57 and @xmath70 , and hence frequently differ from source to source within the literature . ] by @xmath71 following @xcite , we can define a set of parameters @xmath72 which represent the ratios of these effective couplings to their sm values . at tree level , @xmath73 similarly , one can also define @xmath74-parameters for the trilinear couplings of @xmath53 with the electroweak gauge bosons , with the result that @xmath75 since a certain set of effective couplings whose leading contributions occur at one loop namely @xmath76 and @xmath77 are also relevant to the collider phenomenology of higgs bosons , it is worth deriving @xmath74-factors for them as well . the effective operators that give rise to @xmath76 and @xmath77 are @xcite @xmath78 @xmath79 where @xmath80 , @xmath81 is the electric charge of quark @xmath82 , and @xmath83\label{eq : f12def } \\ f_1(\tau ) & = & 2 + 3\tau+3\tau(2-\tau)f(\tau)\end{aligned}\ ] ] and @xmath84 ^ 2 & \tau<1 \end{array}\right.\ ] ] with @xmath85 . when @xmath86 and @xmath87 are complex ( which occurs when @xmath88 ) , it corresponds to internal lines going on shell . this allows us to define a scaling factor for each of these effective vertices : @xmath89 since @xmath90 has an overall @xmath91 prefactor ( from the @xmath92 ) , the contribution from top quarks running in the loops will still dominate over the contribution from leptons unless @xmath93 ; thus the lepton loops generally can be neglected . it is worth noting that since the effective higgs - gluon - gluon coupling receives contributions solely from quark loops , @xmath94 to leading order in @xmath95 , whereas @xmath96 depends on @xmath97 , @xmath98 , and @xmath99 in a nontrivial way . the mixing angles @xmath57 and @xmath70 are constrained by several theoretical consistency conditions , as well as a number of experimental constraints . we will put off discussion of the latter until section [ sec : bounds ] and focus on the former . first of all , we require that the higgs sector not be strongly coupled , in the sense that all @xmath59 may be considered perturbatively small ( i.e. @xmath100 for all @xmath101 ) and that the @xmath102-matrix satisfies all relevant tree - unitarity constraints . this implies that the quartic couplings @xmath59 appearing in eqn . ( [ eq : vstandard ] ) must satisfy @xcite @xmath103 perturbativity constraints also apply to the yukawa couplings @xmath104 , @xmath105 , and @xmath106 appearing in eqn . ( [ eq : yuks ] ) , which are modified from their sm values according to eqn . ( [ eq : hffcouplings ] ) . however , since @xmath104 and @xmath105 are suppressed relative to their sm values rather than enhanced when @xmath107 ( the case of interest here ) , no stringent constraints arise on account of such modifications . in addition to these perturbativity constraints , we must also require that the scalar potential given in eqn . ( [ eq : vstandard ] ) is finite at large field values and contains no flat directions . these considerations translate into the bounds @xcite @xmath108 in this work , we will be primarily interested in examining situations in which the additional physical scalars @xmath56 , @xmath54 , and @xmath55 are heavy enough to `` decouple '' from the collider phenomenology of the theory in the sense that the only observable signals of beyond - the - standard - model physics at the lhc at low luminosity involve the light @xmath0-even scalar @xmath53 . for our present purposes , it will be sufficient to define our `` decoupling regime '' by the condition that @xmath109 , where @xmath110 is some high scale . of course this regime includes the strict decoupling limit in which @xmath111 and the mixing angles satisfy the condition @xmath112 . however , it also includes substantial regions of parameter space within which the values of @xmath57 and @xmath70 deviate significantly from this relationship . 3.5 truein the extent of parameter space allowed according to our definition of the decoupling regime is illustrated in fig . [ fig : sinatanbexcl ] . this figure shows the decoupling regions of @xmath113 - @xmath49 parameter space in which all of the aforementioned constraints are satisfied for a variety of different values of @xmath110 . contours corresponding to @xmath114 gev , @xmath115 gev , and @xmath116 tev are displayed , along with a dash - dotted line representing the pure decoupling limit , where @xmath117 and @xmath112 . the contours in fig . [ fig : sinatanbexcl ] were obtained by fixing @xmath66 to a particular value ( @xmath118 gev ) and surveying over the remaining parameters . a given combination of @xmath113 and @xmath49 is considered to be `` allowed '' in this sense as long as there exists some combination of model parameters for which @xmath109 , and for which all of the constraints in eqs . ( [ eq : pertbounds ] ) and ( [ eq : vacstabbounds ] ) are simultaneously satisfied . it is readily apparent from the figure that sizable regions of parameter space exist within which all constraints are satisfied , yet the masses of all scalars other than @xmath53 are large enough to effectively decouple from the low - energy effective description of the model . it is also apparent that for @xmath119 tev , the decoupling region , as we have defined it , approaches the pure decoupling limit . 2.1 truein 2.1 truein 2.1 truein it is interesting to inquire to what extent the effective higgs couplings can be modified in the decoupling regime without running afoul of the aforementioned constraints . in the three panels shown in fig . [ fig : sinatanbetacont ] we plot a number of contours in @xmath113 - @xmath49 parameter space corresponding to different values of @xmath120 ( left ) , @xmath97 ( center ) and @xmath121 ( right ) . on each panel , we have also superimposed the @xmath122 gev contour from fig . [ fig : sinatanbexcl ] . it is evident from these plots that while @xmath123 in the @xmath111 limit , large regions of parameter space are still allowed in the decoupling regime within these @xmath74-factors can deviate substantially from unity . the message , then , is that the effective couplings of a light higgs boson in the decoupling regime do not have to approximate those which correspond to the pure decoupling limit , in which they approach those of a sm higgs . on the contrary , a wide variety of possibilities for the mixing angles @xmath57 and @xmath70 are still open in this regime , and as we shall soon see , many of these possibilities have a dramatic effect on in the collider phenomenology of the scalar @xmath53 . in addition to the perturbativity and vacuum - stability bounds discussed in the previous section , the l2hdm is constrained by additional considerations related to flavor - physics experiment , direct searches , etc . we now proceed to investigate these constraints in an effort to show that they can easily be satisfied in the decoupling regime even in the region of parameter space most interesting for collider phenomenology , where @xmath49 is large and @xmath113 deviates substantially from zero . let us begin with those bounds related to direct searches for beyond - the - standard - model scalars at lep . the current direct detection bounds ( at 95% cl ) for the masses of charged and neutral @xmath0-odd higgs bosons , as reported by the particle data group @xcite , are @xmath124 and @xmath125 . these clearly present no problem for the model in the decoupling limit considered here . far more stringent constraints on models with more than one higgs doublet can be derived , however , from experimental limits on flavor - violating processes that receive contributions at the one - loop level from diagrams involving charged higgs bosons . let us first consider flavor violation in the lepton sector , which is constrained by analyses of @xmath126 , @xmath127 , @xmath128 , and @xmath129 conversion in nuclei . in the absence of neutrino masses , the matrix of effective @xmath130 couplings is proportional to the charged - lepton mass matrix ; hence there is no additional source of lepton - flavor violation ( lfv ) . in the presence of neutrino masses this is no longer true , and nonzero contributions to lfv processes arise at one loop due to diagrams with charged higgs bosons running in the loop . however , it has been shown @xcite that the resultant flavor - violating amplitudes are several orders of magnitude below current experimental bounds . therefore , even in cases in which the effective @xmath131 coupling receives a substantial @xmath49-enhancement factor , such sources of lfv will not present any phenomenological difficulties . now let us turn to consider the situation in the quark sector , where , by contrast , flavor - violation rates can be sizeable . this is because the effective @xmath132 couplings in two - higgs - doublet models include flavor - violating terms proportional to the off - diagonal elements of the cabibbo - kobayashi - maskawa ( ckm ) matrix @xmath133 : @xmath134d_jh^+ -\frac{\tan\beta}{v}m_{e_i}\bar{\nu}_i p_r e_ih^+ + h.c.,\ ] ] as a result , such models are constrained by experimental bounds on @xmath135 , @xmath136 , @xmath137 , @xmath138 , rare kaon decays , etc . , which translate into bounds on the model parameters relevant to the charged - scalar sector : @xmath69 and @xmath49 . since the flavor mixing in the charged higgs couplings to the quark sector is proportional to @xmath139 , it is the region where both @xmath49 and @xmath69 are small which is most tightly constrained by these bounds . the most stringent constraints are those associated with @xmath140 and with mixing in the @xmath141 and @xmath142 systems . in the l2hdm , the same higgs doublet couples to both up- and down - type quarks , just as it does in type i 2hdm @xcite ; hence the bounds on @xmath69 and @xmath49 due to flavor mixing in the quark sector will be essentially identical to those applicable in type i models . we now turn to review the bounds from each of these processes , updating the results obtained in @xcite . 2.3 truein 2.3 truein + 2.3 truein 2.3 truein the first bounds we consider are those associated with the observed branching ratio for the flavor - violating decay @xmath140 . the combined result from the cleo and belle experiments @xcite is @xmath143 this is consistent with the expected standard model result @xmath144 . in models with a non - minimal higgs sector , additional contributions to the amplitude for @xmath140 arise at the loop level from diagrams involving virtual charged higgs bosons , as discussed above . these diagrams are compiled in fig . [ fig : btosgammadiagrams ] ( sm contributions to this amplitude come from diagrams of the same sort , but with @xmath145 in place of @xmath146 . ) the rate for the process can be calculated in the usual manner . after incorporating the effect of qcd corrections ( which can be quite large @xcite ) , one finds that @xcite @xmath147 where @xmath148 is the coefficient of the effective operator @xmath149 in the conventions of ref . @xcite , evaluated at the scale @xmath150 . this coefficient takes the form @xmath151- \frac{3x}{28}\left [ \left(\frac{\alpha_{3}(m_{w})}{\alpha_{3}(m_{b})}\right)^{28/23}-1\right]\right],\end{aligned}\ ] ] where the weak - scale amplitude function @xmath152 in the l2hdm is given by @xmath153 . \label{eq : c7weak}\ ] ] in these formulae , @xmath154 and @xmath155 , @xmath156 , @xmath157 , @xmath158 is the fermi constant , @xmath133 are elements in the ckm matrix , and @xmath159 . the functions @xmath160 , @xmath161 , and @xmath162 , which represent the loop integral contributions to the @xmath140 amplitude , are given in @xcite . the constraints on @xmath163 and @xmath49 from @xmath140 are displayed in fig . [ fig : bsgammagraph ] . each curve therein represents the value of @xmath135 for a given choice of @xmath69 as a function of @xmath49 . note that for the case under consideration here , in which @xmath164 , the experimental constraints are satisfied as long as @xmath165 . 4 truein 2.3 truein 2.3 truein constraints on @xmath163 and @xmath49 can also be obtained from limits on the observed mixing in the mesonic @xmath141 and @xmath142 systems . the diagrammatic contributions to @xmath141 mixing are shown in fig . [ fig : mesonboxdiagram ] , and these contributions translate into shift in the mass - splitting @xmath166 between @xmath167 and @xmath168 . in the l2hdm , this splitting , including sm contributions , is given by @xcite @xmath169,\label{eq : deltambb}\end{aligned}\ ] ] where the @xmath170 are defined as below eqn . ( [ eq : c7weak ] ) , @xmath171 is the @xmath172-meson decay constant , @xmath173 is the `` bag factor '' ( which encompasses all deviations from the vacuum saturation approximation ) . expressions for the factor @xmath174 , which accounts for qcd effects , along with the functions @xmath175 , @xmath176 , and @xmath177 can be found in @xcite . as for @xmath171 and @xmath173 , there is a good deal of uncertainty as to their precise numerical values . since they appear in eqn . ( [ eq : deltambb ] ) in the combination @xmath178 , it is easier simply to deal with the uncertainty in this single quantity . estimates of @xmath178 have been made using a variety of lattice qcd sum rules in conjunction with experimental evidence on heavy meson decays from slac , and the uncertainties in their values depend on the summation methods employed and the assumptions made . following @xcite , we take the range of uncertainty to be @xmath179 instead of dealing with @xmath166 directly , it is easier to deal with the combination @xmath180 , since the time - integrated mixing probability in the @xmath141 system depends on this combination of variables . the accepted experimental value for @xmath181 , as reported by the heavy flavor averaging group , is @xmath182 @xcite . using the observed lifetime of the @xmath167 meson ( @xmath183 ) and the expression in eqn . ( [ eq : deltambb ] ) , one may obtains a theoretical value for @xmath181 , which can be compared to this experimental result . in fig . [ fig : mixinggraphs ] , we show how the @xmath141 mixing bound constrains @xmath163 and @xmath49 . as there is a large uncertainty in @xmath178 [ eqn . ( [ eq : uncertaintyfbrootb0 ] ) ] , and in fact one far larger than that associated with the measured value of @xmath181 , the theoretical prediction for a given choice of @xmath163 translates into a broad band in @xmath184 space , rather than a thin line . in fig . [ fig : mixinggraphs ] , the upper and lower bounds of each such band are demarcated by a pair of thick , dark lines of the same type ( solid , dotted and dot - dashed ) . the thin , shaded , horizontal stripe represents the experimentally - allowed window . if any part of this stripe falls within the band corresponding to a given value of @xmath69 for a given @xmath49 , that parameter combination is permitted by the @xmath138 constraint . we see from the plot that this constraint only becomes relevant for very small values of @xmath185 , and thus is not particularly stringent . 3.75 truein similar calculations to those outlined above for the @xmath141 system can also be performed for mixing in the @xmath142 and @xmath186 systems @xcite . in addition , limits can also be derived on the @xmath0-violating parameters @xmath187 and @xmath188 . however , due to large theoretical uncertainties in the hadronic matrix elements , the resulting bounds on new physics from these considerations are not particularly stringent in the l2hdm , especially when @xmath107 @xcite . experimental bounds on leptonic charged - meson decays @xmath189 , @xmath190 , @xmath191 , @xmath192 etc . can also be used to constrain 2hdm @xcite . in general , the partial width for the leptonic decay of a given meson is modified by a @xmath49-dependent factor @xmath193 , which in many scenarios ( e.g. type ii models ) can be quite sizeable when @xmath49 is large @xcite . in the l2hdm , however , the @xmath193 are independent of @xmath49 due to the cancellation of the @xmath49 factors between the quark and the lepton couplings . as a result the model is essentially unconstrained by these considerations . experimental limits on the rates for leptonic decays such as @xmath194 can also constrain models with enhanced higgs couplings to leptons @xcite . however , such constraints only become relevant when the charged - higgs mass is @xmath195 or lower , or when @xmath49 is extremely large , and thus have little bearing on the decoupling regime studied here . the above analysis shows that in the decoupling region ( as we have defined it ) , where @xmath196 gev , all constraints from direct charged - higgs searches , neutral meson mixing , @xmath0-violation , charged - meson decay , etc . can be satisfied as long as @xmath49 is greater than @xmath197 . this is mainly due to the fact that in the l2hdm , there is no new source of flavor violation except the sm ckm matrix . the effective couplings between @xmath56 and the sm quarks are proportional to @xmath139 , which implies that the most stringent constraints become weaker as @xmath49 increases . thus , we conclude that experimental constraints from flavor violation , direct searches , etc . do not pose any significant issues for the l2hdm as long as the charged higgs scalars are heavy . ( indeed , a relatively low value of @xmath198 and a charged higgs light enough to be discovered at the lhc are by no means incompatible . ) this is true even in the region of parameter space most interesting for collider physics , in which both @xmath113 and @xmath49 are large , and the effective couplings between the lightest @xmath0-even higgs and the sm leptons differ drastically from their sm values . we now turn to examine the effect of these coupling modifications on the production cross - sections and decay widths of a light higgs boson . since the overall amplitudes for higgs decays into any two - particle final state @xmath199 scale as @xmath200 ( i.e. , the appropriate @xmath74-factor for that final state ) , the associated branching ratios scale like @xmath201 in order to provide a concrete example of the effect such a modification can have on higgs phenomenology , let us focus on a particular benchmark point : @xmath202 , @xmath203 , which we have indicated by a dot in fig . [ fig : sinatanbexcl ] . we pick this particular point as a benchmark because it yields only a moderate deviation from the sm couplings and is consistent with the bounds ( [ eq : pertbounds ] ) and ( [ eq : vacstabbounds ] ) when @xmath204 . the @xmath74-factors corresponding to this particular point are @xmath205 fig . [ fig : brplot ] illustrates the effect of this coupling - constant modification on the branching ratios of a light , @xmath0-even higgs scalar . in the left - hand panel , we have plotted the sm branching ratios for a number of higgs decay processes as a function of @xmath66 . all branching ratios used in the construction of the figure were calculated using hdecay @xcite . in the right - hand panel , we have plotted branching ratios for the same set of processes in the l2hdm at our chosen benchmark point . it is evident that even this moderate modification of the couplings has a dramatic effect on the decay behavior of a light higgs : for example , the rates for @xmath206 and @xmath207 are on the same order . since @xmath208 is the dominant decay channel for a higgs boson with a mass in the range @xmath209 , this clearly represents a substantial effect on higgs phenomenology . it is also worth noting that @xmath210 and @xmath211 are also on the same order for this choice of parameters . this suggests that processes involving direct decays of a light higgs boson to a pair of high-@xmath13 muons could play an important role in the collider phenomenology of the light higgs @xmath212 a suggestion we will explore further in section [ sec : lhcsignatures ] . the branching ratios for a number of other decay channels relevant to the study of a light sm higgs boson at the lhc , such as @xmath213 , are clearly suppressed here relative to their sm values . 2.5 truein 2.5 truein the effect of the coupling modifications on the total higgs width is shown in fig . [ fig : gamtotplot ] . here , we have plotted the ratio of the total higgs width @xmath214 to its sm value for three different points in the allowed region of @xmath113 @xmath212 @xmath49 parameter space as a function of @xmath66 . the first of these points is our chosen benchmark ( @xmath202 , @xmath203 ) , for which @xmath214 ( indicated by the solid line ) is slightly lower than its sm value due to the suppression of @xmath215 when @xmath66 is small . when @xmath66 increases and decays to electroweak gauge bosons begin to dominate the higgs width , @xmath214 drops even further , since @xmath216 at this point [ see eqn . ( [ eq : etasforbm ] ) ] . the second of these points , ( @xmath217 , @xmath218 ) , is located very near the pure decoupling " line in fig . [ fig : sinatanbexcl ] ; hence for this point @xmath214 ( indicated by the dotted line ) is essentially equal to @xmath219 . at the third point , ( @xmath220 , @xmath221 ) , a substantial enhancement in @xmath222 overcomes the suppression factor in @xmath215 , and consequently @xmath223 for @xmath224 gev ( as indicated by the dash - dotted line ) . for larger values of @xmath66 , gauge - boson decays once again dominate the higgs width , which becomes suppressed relative to its sm value . even in the most extreme cases permitted by the model consistency constraints outlined in section [ sec : param ] , however , @xmath225 this implies that the narrow - width approximation remains valid over the higgs mass range @xmath209 , which will be the mass region of primary focus of the present work . 4.0 truein since we have shown that the narrow - width approximation to be valid , we can proceed in a straightforward manner from the decay width calculations above to determine how the cross - sections for full collider processes are modified . in this approximation , one assumes that essentially all the higgs bosons produced in any such process are produced on - shell . this allows one to approximate the cross - section for any process of the form @xmath226 by @xmath227 furthermore , if the sm production cross - section @xmath228 for the process is known , one can use the fact that @xmath229 to obtain the relation @xmath230 which allows us to calculate the cross - sections for these overall processes in the modified model . for the benchmark point that we have chosen ( @xmath202 , @xmath49=3 ) , the cross - sections for most of the conventional higgs search modes at the lhc are suppressed relative to their sm values , due to the suppressed higgs couplings to quarks and to gauge bosons . many of the processes in which the higgs decays directly to charged - lepton pairs , on the other hand , are substantially enhanced . we will discuss the implications these modifications can have for higgs searches in detail in section [ sec : combined ] . one of the most interesting aspects of the l2hdm is that in certain regions of parameter space , new channels for the discovery of a light higgs boson can open up . in particular , when the effective coupling between @xmath53 and the sm leptons is substantially increased while its couplings to sm quarks and/or electroweak gauge bosons are not dramatically suppressed , a number of processes in which the higgs boson decays directly to a pair of high-@xmath13 leptons can become far more important for the discovery of a light higgs than they are in the sm . in our analyses , we focus on the discovery of @xmath53 in the light - to - intermediate - mass region @xmath231 . for heavier higgs bosons , @xmath232 dominates and leptonic higgs decays play a less important role . in the decoupling limit case studied here , in which the additional higgs scalars @xmath146 , @xmath54 , and @xmath55 are heavy , such processes might well constitute the only evidence for physics beyond the standard model accessible within the first @xmath233 of integrated luminosity at the lhc , and are therefore of crucial importance . this situation is quite different from the one studied in ref . @xcite , in which some of these additional scalars are light and play a significant role in collider phenomenology . since the largest leptonic contribution to the higgs total width comes from @xmath234 , processes involving higgs decays directly to tau leptons will play a significant role in the collider phenomenology of the l2hdm . however , the analysis of such processes is complicated by subtleties associated with tau decay . each @xmath235 lepton can decay either leptonically or hadronically , with respective branching ratios @xcite , @xmath236 and @xmath237 . we will henceforth denote a hadronically - decaying tau as @xmath238 and a leptonically decaying one as @xmath239 . for processes involving @xmath0-even higgs boson decays into a @xmath240 pair , there are two final states which permit successful identification of both taus : @xmath241 and @xmath242 , where @xmath243 . final states resulting from fully hadronic decays have a large background from dijet processes with narrow jets misidentified as taus . final states involving two leptons of like flavor ( @xmath244 and @xmath245 ) are also less useful due to the overwhelming sm background from @xmath246 processes . a hadronically - decaying tau will decay into either a `` one - prong '' ( approximately 77% of the time ) or `` three - prong '' ( approximately 23% of the time ) final state . these final states involve narrow , well - collimated jets including one or three charged pions , respectively . the identification of a jet as coming from a hadronically - decaying @xmath235 , as opposed to some qcd process , is far from trivial . one of the principal discrimination variables is jet radius @xmath247 ( see @xcite for more details regarding @xmath235 identification ) . at the tevatron ( run ii ) , @xmath248 for a @xmath249 gev cut . at the lhc , a @xmath235 identification efficiency of around @xmath250 is expected @xcite . processes involving direct decays of @xmath53 to muon pairs can also be of interest for higgs discovery in the l2hdm . the disadvantage of such channels for higgs searches , relative to those involving direct decays to taus , is the suppressed branching ratio . since yukawa coupling universality dictates that @xmath251 , both in the sm and in the l2hdm , @xmath252 . however , this is compensated for to a great extent by the fact that the dimuon signal is exceptionally clean . indeed , the muon identification efficiency at the lhc is more that 90% @xcite . in addition , the measurement of muon momenta allows for a precise reconstruction of the higgs mass within @xmath253 gev . this permits the implementation of an extremely efficient cut on @xmath254 , the invariant mass of the muon pair , and a substantial reduction in background levels for all channels involving direct higgs - boson decays to muon pairs . we now turn to address the prospects for detecting a light sm - like @xmath0-even higgs boson at the lhc on a channel - by - channel basis . in the present work , as discussed in section [ sec : param ] , we will assume generation universality among the lepton yukawa couplings . therefore , we will ignore the @xmath255 channel and focus only on @xmath256 and @xmath26 . the channels of primary interest , then , are those in which the higgs is produced by gluon fusion , weak - boson fusion , or @xmath257 associated production and decays to either @xmath258 or @xmath259 . associated @xmath51 and @xmath52 production processes generally have smaller rates , but may also potentially be of interest , and as such we briefly discuss them as well . bottom - quark - fusion processes with a leptonically - decaying higgs boson @xcite , while potentially interesting for type ii 2hdm in which the @xmath260 vertex receives a large @xmath49-enhancement , are less useful in the l2hdm , since the effective down - type quark couplings are suppressed in that scenario rather than enhanced . in this section , we briefly summarize the results of the existing studies of the leptonic - higgs - decay channels at the lhc , with an eye toward their utility for the discovery of a leptophilic higgs . we begin with a discussion of the weak - boson - fusion process @xmath261 , which is the only channel involving direct higgs - boson decay to a pair of charged leptons that contributes significantly to the higgs discovery potential in the sm . indeed , it is a particularly promising channel for sm higgs discovery in the intermediate mass region ( @xmath8 ) @xcite . discriminating between signal and sm background can be facilitated by requiring that events have two leading tagging jets in the forward - backward direction and imposing a minijet veto in the central region of the detector . a great deal of attention has been devoted to this channel , with an emphasis on @xmath262 and @xmath263 final states . combining all channels , a statistical significance of more than @xmath264 can be reached for higgs masses around 120 @xmath212 130 gev with @xmath265 of integrated luminosity at atlas @xcite . the detection prospects are similar at cms @xcite . the prospects for detecting a light , sm higgs boson produced by gluon fusion and decaying to @xmath259 at the tevatron were examined in ref . @xcite . in order to effectively reconstruct the higgs mass from the various final - state particles produced during tau decay , it is necessary to focus on events in which the transverse momentum of the tau pair is nonzero ; hence the authors elected to focus on the process @xmath267 . taking into account both the @xmath268 and @xmath269 ratios , @xmath262 turns out to be the most promising channel for signal identification , but that an integrated luminosity of @xmath270 would be needed at the tevatron in order to exclude a 120 gev sm higgs boson at the 95% c.l . however , preliminary studies at atlas @xcite indicate that this will be a promising channel in which to look for a higgs boson with enhanced coupling to leptons at the lhc . this process was examined in a standard model context in @xcite . in order to be able to reconstruct the two top quarks effectively , the authors restricted their analysis to cases in which one of the @xmath271 bosons produced during top decay decays leptonically , while the other decays hadronically . only events with hadronic tau decays were considered , as reconstructing both tops proves to be slightly easier in this scenario . thus the overall process of interest is @xmath272 . since the production cross section drops quickly with increased higgs mass , this channel is only important when the higgs is light . for @xmath66 around 120 gev , a statistical significance of @xmath273 can be obtained with @xmath274 of integrated luminosity . in @xcite , semileptonic tau decays were considered in particular , decays of the form @xmath275 , and it was found that such a process could provide evidence for a 120 gev higgs boson at the @xmath276 level for an integrated luminosity of @xmath233 . the weak boson fusion process @xmath278 was analyzed in @xcite . after the appropriate cuts on the tagging jets are imposed , the leading sm background comes from irreducible @xmath279 or @xmath280 processes , with the @xmath281 decaying to muon pairs . due to the extremely suppressed sm @xmath282 branching ratio , an integrated luminosity of @xmath283 or more is generally required to claim a @xmath284 discovery for a higgs mass less than 140 gev . in the l2hdm , however the detection prospects can be substantially improved if @xmath285 and @xmath286 . the prospects for the detection of a light higgs boson of @xmath288 produced by gluon fusion and decaying directly into @xmath258 were discussed in @xcite . the irreducible background for @xmath287 is dominated by the drell - yan processes @xmath289 . the sharp invariant mass resolution of the muon pair allows for a substantial reduction in this background via a stringent cut on @xmath254 . consequently , a significance level similar to that in the wbf channel as discussed in @xcite can be attained in this channel as well . the prospects for discovering a standard model higgs boson in the @xmath290 channel were recently studied in @xcite . this channel tends to be more important when the higgs mass is light ( around 120 gev ) since the production cross section drops quickly for a heavier higgs . the primary irreducible backgrounds , which come from @xmath291 and @xmath292 production , with the @xmath281 decaying into a muon pair , can be reduced quite effectively by a cut on that muon pair s invariant mass . additional , reducible backgrounds such as @xmath293 can be effectively eliminated by reconstructing the masses of both top quarks , which is possible in the case where the tops decay either fully hadronically or semileptonically . the statistical significances for the @xmath290 channel are of roughly the same order as those in the @xmath287 and @xmath278 channels , and hence could contribute significantly to the discovery potential for a light higgs scalar with enhanced couplings to leptons . higgs production via the processes @xmath296 and @xmath297 could also potentially play a role in the discovery of a leptophilic higgs , though the prospects in these channels are not as favorable as the other , aforementioned ones . sm cross - sections for these processes , taking into account the leptonic decay of the higgs boson , are given in table [ table : whzh ] for the case in which @xmath298 gev . these were determined from leading - order results obtained using madgraph @xcite and modified by the appropriate @xmath299-factors : @xmath300 for signal @xcite , @xmath301 for background @xcite . for processes in which the higgs decays to @xmath258 , the signal is clearly too small to be of any use . however , for processes involving decays to @xmath259 , the signal is only about a factor of @xmath302 smaller than the background . by optimizing cuts to eliminate the sm background , this channel might potentially be of use particularly if @xmath303 is enhanced , as in the l2hdm . little analysis of these processes exists in the literature , and we leave the detailed study of these channels for future work . .sm production cross sections at the lhc for @xmath304 , @xmath305 associated production , with @xmath53 decays into muon pair or tau pair . the higgs mass is taken to be 120 gev . also shown are the sm background @xmath306 , @xmath307 with one @xmath52 decays into muons or taus . the numbers are obtained using madgraph @xcite . [ cols="^,^,^,^,^,^",options="header " , ] now that we have discussed the channels in which one might look for a leptonically - decaying higgs boson at the lhc , let us investigate the prospects for the discovery of such a higgs boson in the l2hdm , using the combined results from all channels discussed above ( excepting the @xmath304 , @xmath305 channels , which we have shown do not contribute significantly to the discovery potential ) . in particular , we focus on the region of @xmath113 - @xmath49 parameter space in which @xmath120 is large and @xmath308 . in this case , the cross - sections for processes involving a @xmath309 coupling are substantially increased , while those for processes involving @xmath310 , @xmath311 , or @xmath76 are only slightly reduced . as before , for purposes of illustration , we will focus on the benchmark point ( @xmath202 , @xmath203 ) , which exemplifies this situation nicely . in fig . [ fig : sigplot ] , we show the effect of the coupling - constant modifications on the discovery potential of a light higgs boson for this particular benchmark point . in the right - hand panel , the statistical significance associated with each of the relevant leptonic channels discussed in section [ sec : lhcsignatures ] is displayed as a function of higgs mass for our chosen benchmark point in the l2hdm . the sm results for the same processes are shown in the left - panel for comparison . the results in each panel correspond to an integrated luminosity of @xmath312 . it is apparent from fig . [ fig : sigplot ] that @xmath313 is one of the most promising detection channels for the chosen benchmark point in the l2hdm , as in the sm . for this particular choice of parameters , @xmath314 and @xmath214 does not deviate drastically from @xmath219 ( see fig . [ fig : gamtotplot ] ) , and consequently the overall significance level in this channel is essentially unchanged from its sm value . however , in other regions of parameter space , drastic amplifications can occur : for example , the choice ( @xmath315 , @xmath221 ) results in a amplification of the statistical significance for the same process by a factor of @xmath316 . it should also be noted that in the ( @xmath317 , @xmath203 ) case , the significance levels for both @xmath266 and @xmath318 also exceed @xmath264 . the processes in which the higgs decays to muons are statistically less significant , but also provide strong evidence at the @xmath284 level with @xmath319 100 @xmath320 of integrated luminosity . indeed , the evidence for such a higgs boson would be dramatic and unmistakable . furthermore , once the higgs is observed in any of the muonic channels , the excellent invariant - mass resolution of the muon pairs can be used to determine the value of @xmath66 with a very high degree of precision . 2.75 truein 2.75 truein 2.75 truein 2.75 truein while the significances in those channels which involve a leptonically - decaying higgs can potentially be amplified in l2hdm , those in other channels useful for the detection of a sm higgs may be substantially suppressed . this is illustrated in fig . [ fig : sigplotother ] , which shows the significance of discovery in each individual channel which contributes meaningfully to the discovery potential of a sm higgs boson in the low to intermediate - mass region , both in the sm ( left - hand panel ) and in the l2hdm at the benchmark point ( @xmath202 , @xmath203 ) ( right - hand panel ) . in the latter case , there is no single , non - leptonic channel in which evidence for the higgs boson can be obtained at the @xmath264 level . to further illustrate the point , in fig . [ fig : sigscomb ] , we display the combined statistical significances for the leptonic channels discussed in section [ sec : lhcsignatures ] , as well as the combined significances for all other relevant channels for higgs discovery , both in the sm and in the l2hdm at the benchmark point ( @xmath317 , @xmath203 ) . indeed , for this particular parameter choice , all relevant non - leptonic channels are suppressed relative to their standard - model to such an extent that , for most of the @xmath321 mass window displayed in the plot , their combined significance does not even provide @xmath284 evidence for much less a @xmath264 discovery of a light higgs boson . on the other hand , statistical significance for leptonic higgs decay channels are enhanced , therefore becoming the dominant discovery channels for the light @xmath0-even higgs in the l2hdm model . this clearly illustrates the crucial role leptonic channels can play in the lhc phenomenology of models with extended ( and particularly leptophilic ) higgs sectors . 3.5 truein we emphasize that these plots represent the results for a single benchmark point , and one in which the @xmath74-factors are not particularly extreme . there exist other points in the parameter space of the model allowed by all constraints for which the deviations of the effective couplings of @xmath53 to the other fields in the theory are even more severe . as an example , consider the case in which @xmath322 and @xmath323 , for which @xmath324 , @xmath325 , and @xmath326 . for this choice of parameters , most of the standard higgs discovery channels those involving @xmath327 and @xmath328 , as well as all weak - boson - fusion processes not involving direct higgs decays to leptons are strongly suppressed ; furthermore , other contributing channels such as @xmath329 and @xmath7 are also moderately suppressed . in such a case , the leptonic channels discussed in section [ sec : lhcsignatures ] especially ones such as @xmath330 , which do not involve a direct coupling between @xmath53 and the electroweak gauge bosons may well constitute the only observable evidence of the higgs boson , and would thus be crucial for its discovery at the lhc . the phenomenology of a light higgs boson in two - higgs - doublet models can differ drastically from that of a sm higgs . in this work , we have focussed on one particularly interesting example : a leptophilic 2hdm , in which different higgs bosons are responsible for giving masses to the quark and lepton sectors . we have examined the effect of such a modification on the collider phenomenology of a light higgs boson in a decoupling regime in which the only light scalar is a standard - model - like higgs boson , and have shown that a number of collider processes involving the direct decay of the higgs to a pair of charged leptons can play a crucial role in its discovery . in particular , we have shown that there are regions of parameter space in which the higgs - boson couplings to leptons can be greatly enhanced . this can have a potentially dramatic effect on the higgs discovery potential , as signals involving direct , leptonic decays of the higgs can be substantially amplified . at the same time , signals in some ( or in some cases , even all ) of the other conventional channels useful for the detection of a standard model higgs boson can suffer a dramatic suppression . even when coupling modifications are not severe , leptonic decay processes will also play an important rule in differentiating between the higgs sector of the standard model and that of other , more complicated scenarios . we would like to thank keith dienes , tao han , chung kao and lisa randall for fruitful discussions and correspondence . we would also like to thank the kavli institute of theoretical physics at santa barbara for its hospitality during the completion of this work . this work was supported in part by the department of energy under grant de - fg02 - 04er-41298 . note added : after the completion of the work reported in this paper , a number of papers @xcite appeared which discuss the phenomenology of the l2hdm . @xcite gives a brief presentation on the effect of effective - coupling modification in the decoupling regime . their results agree with ours . @xcite and @xcite focussed on the situation in which @xmath54 , @xmath55 and @xmath56 are light . s. asai _ et al . _ , eur . j. c * 32s2 * , 19 ( 2004 ) [ arxiv : hep - ph/0402254 ] . s. abdullin _ et al . _ , eur . j. c * 39s2 * , 41 ( 2005 ) . d. l. rainwater and d. zeppenfeld , jhep * 9712 * , 005 ( 1997 ) [ arxiv : hep - ph/9712271 ] . d. l. rainwater , d. zeppenfeld and k. hagiwara , phys . d * 59 * , 014037 ( 1999 ) [ arxiv : hep - ph/9808468 ] . [ atlas collaboration ] , cern - lhcc-99 - 15 , atlas - tdr-15 . g. l. bayatian _ et al . _ [ cms collaboration ] , j. phys . g * 34 * , 995 ( 2007 ) . n. arkani - hamed , a. g. cohen and h. georgi , phys . b * 513 * , 232 ( 2001 ) [ arxiv : hep - ph/0105239 ] . h. e. haber , m. j. herrero , h. e. logan , s. penaranda , s. rigolin and d. temes , phys . d * 63 * , 055004 ( 2001 ) [ arxiv : hep - ph/0007006 ] ; m. s. carena , h. e. haber , h. e. logan and s. mrenna , phys . d * 65 * , 055005 ( 2002 ) [ erratum - ibid . d * 65 * , 099902 ( 2002 ) ] [ arxiv : hep - ph/0106116 ] . v. barger , h. e. logan and g. shaughnessy , arxiv:0902.0170 [ hep - ph ] . h. s. goh , l. j. hall and p. kumar , arxiv:0902.0814 [ hep - ph ] . r. m. barnett , g. senjanovic , l. wolfenstein and d. wyler , phys . b * 136 * , 191 ( 1984 ) . r. m. barnett , g. senjanovic and d. wyler , phys . rev . d * 30 * , 1529 ( 1984 ) . v. d. barger , j. l. hewett and r. j. n. phillips , phys . d * 41 * , 3421 ( 1990 ) . y. grossman , nucl . b * 426 * , 355 ( 1994 ) [ arxiv : hep - ph/9401311 ] . m. aoki , s. kanemura and o. seto , arxiv:0807.0361 [ hep - ph ] . j. f. gunion and h. e. haber , phys . d * 67 * , 075019 ( 2003 ) [ arxiv : hep - ph/0207010 ] . d. phalen , b. thomas and j. d. wells , phys . rev . d * 75 * , 117702 ( 2007 ) [ arxiv : hep - ph/0612219 ] . j. f. gunion , h. e. haber , g. l. kane and s. dawson , _ the higgs hunter s guide _ , addison - wesley : redwood city , ca , 1990 . a. g. akeroyd , a. arhrib and e. m. naimi , phys . b * 490 * , 119 ( 2000 ) [ arxiv : hep - ph/0006035 ] ; a. arhrib , arxiv : hep - ph/0012353 ; i. f. ginzburg and i. p. ivanov , arxiv : hep - ph/0312374 . w. m. yao _ et al . _ [ particle data group ] , j. phys . g * 33 * , 1 ( 2006 ) . w. grimus and l. lavoura , phys . d * 66 * , 014016 ( 2002 ) [ arxiv : hep - ph/0204070 ] . h. e. haber , g. l. kane and t. sterling , nucl . phys . b * 161 * , 493 ( 1979 ) . n. g. deshpande , p. lo , j. trampetic , g. eilam and p. singer , phys . * 59 * , 183 ( 1987 ) . s. bertolini , f. borzumati and a. masiero , phys . lett . * 59 * , 180 ( 1987 ) ; b. grinstein and m. b. wise , phys . b * 201 * , 274 ( 1988 ) ; b. grinstein , r. p. springer and m. b. wise , phys . b * 202 * , 138 ( 1988 ) . w. s. hou and r. s. willey , phys . b * 202 * , 591 ( 1988 ) ; t. g. rizzo , phys . d * 38 * , 820 ( 1988 ) ; j. l. hewett , s. nandi and t. g. rizzo , phys . rev . d * 39 * , 250 ( 1989 ) . b. grinstein and m. b. wise , phys . b * 201 * , 274 ( 1988 ) . j. franzini , phys . * 173 * , 1 ( 1989 ) . w. s. hou , phys . d * 48 * , 2342 ( 1993 ) . a. g. akeroyd and f. mahmoudi , arxiv:0902.2393 [ hep - ph ] . m. krawczyk and d. temes , eur . j. c * 44 * , 435 ( 2005 ) [ arxiv : hep - ph/0410248 ] . a. djouadi , j. kalinowski and m. spira , comput . commun . * 108 * , 56 ( 1998 ) [ arxiv : hep - ph/9704448 ] . f. tarrade [ atlas collaboration ] , _ prepared for hadron collider physics symposium 2005 , les diablerets , switzerland , 4 - 9 jul 2005 _ ; a. f. saavedra [ atlas collaboration ] , atl - phys - proc-2009 - 007 . v. d. barger and c. kao , phys . b * 424 * , 69 ( 1998 ) [ arxiv : hep - ph/9711328 ] ; s. dawson , d. dicus and c. kao , phys . b * 545 * , 132 ( 2002 ) [ arxiv : hep - ph/0208063 ] ; s. dawson , d. dicus , c. kao and r. malhotra , phys . lett . * 92 * , 241801 ( 2004 ) [ arxiv : hep - ph/0402172 ] . a. belyaev , t. han and r. rosenfeld , jhep * 0307 * , 021 ( 2003 ) [ arxiv : hep - ph/0204210 ] . e. richter - was and t. szymocha , atl - phys-2004 - 012 . t. plehn and d. l. rainwater , phys . b * 520 * , 108 ( 2001 ) [ arxiv : hep - ph/0107180 ] . t. han and b. mcelrath , phys . b * 528 * , 81 ( 2002 ) [ arxiv : hep - ph/0201023 ] . s. su and b. thomas , arxiv:0812.1798 [ hep - ph ] . j. alwall _ et al . _ , jhep * 0709 * , 028 ( 2007 ) [ arxiv:0706.2334 [ hep - ph ] ] . o. brein , a. djouadi and r. harlander , phys . b * 579 * , 149 ( 2004 ) [ arxiv : hep - ph/0307206 ] . s. frixione , p. nason and g. ridolfi , nucl . b * 383 * , 3 ( 1992 ) . m. aoki , s. kanemura , k. tsumura and k. yagyu , arxiv:0902.4665 [ hep - ph ] .

we present new measurements of the interstellar gas - phase oxygen abundance along the sight lines towards 19 early - type galactic stars at an average distance of 2.6 kpc . we derive o i column densities from _ hst_/stis observations of the weak 1355 intersystem transition . we derive total hydrogen column densities [ n(h i)+2n(h ) ] using _ hst_/stis observations of lyman- and _ fuse _ observations of molecular hydrogen . the molecular hydrogen content of these sight lines ranges from f(h ) = 2n(h)/[n(h i)+2n(h ) ] = 0.03 to 0.47 . the average of 6.3 mag with a standard deviation of 15% is consistent with previous surveys . the mean oxygen abundance along these sight lines , which probe a wide range of galactic environments in the distant ism , is 10 = ( 1 in the mean ) . we see no evidence for decreasing gas - phase oxygen abundance with increasing molecular hydrogen fraction and the relative constancy of suggests that the component of dust containing the oxygen is not readily destroyed . we estimate that , if 60% of the dust grains are resilient against destruction by shocks , the distant interstellar total oxygen abundance can be reconciliated with the solar value derived from the most recent measurements of 10 = 517 58 ( 1 ) . we note that the smaller oxygen abundances derived for the interstellar gas within 500 pc or from nearby b star surveys are consistent with a local elemental deficit .

astro-ph0303586

the chemical evolution of galaxies is a long - standing open question in astronomy ( audouze & tinsley 1976 ) . of all atomic species , oxygen , which is mainly produced in type ii supernovae , has proven to be the most useful element for probing the astration of the ism by massive stars , both because of its large abundance and origin ( for a review see henry & worthey 1999 ) . the value of the oxygen abundance within a few hundred parsec from the sun has been thoroughly investigated in the past two decades toward dozens of sight lines with _ copernicus _ ( de boer et al . 1981 , york et al . 1983 , keenan et al . 1985 ) , _ hst _ ( meyer et al . 1998 , cartledge et al . 2001 ) and _ fuse _ ( moos et al . 2002 , and references therein ) . the accurate study by meyer et al . ( 1998 ) revealed a well mixed medium within 500 pc from the sun : @xmath12 ppm ( error in the mean ) . note that this value has been corrected linearly with an updated f - value ( see 3.3 ) . however , investigations of more distant sight lines with the fuv absorption technique were limited by the unknown higher molecular fractions and the fainter background continua . measurements of oxygen abundances beyond 1 kpc have been carried out in two different environments : h ii regions and b star atmospheres . to this day , these methods do not agree on a single picture for the oxygen distribution in the disk . in the most recent survey of the oxygen abundance in galactic h ii regions , deharveng et al . ( 2001 ) confirm the existence of a metalicity gradient from the center out to 18 kpc in the outer disk and find a local abundance consistent with ism investigations . however they point out that their value is highly dependent on the model adopted for the electron temperature in the h ii region . as for the b star surveys , the considerable scatter in the measurements makes it difficult to pin down an accurate value for the gradient ( smartt & rolleston 1997 , primas et al . 2001 ) while , at the same time , the origin of the scatter is still debated . with the high sensitivity of _ fuse _ ( 10,000 times more sensitive than _ copernicus _ ; see moos et al . 2000 ) it is now possible to probe the molecular hydrogen content of the ism beyond 1 kpc and in denser media as well . along with lyman-@xmath1 data from the stis instrument , this allows us to obtain accurate total hydrogen column densities and hence derive a reliable o / h ratio . since oxygen and hydrogen have about the same ionization potential , no ionization correction is needed and the analysis is largely model independent . the principal question we try to answer in this work is the robustness of a constant o / h for large sight line distances . the sight lines studied sample a vast diversity of environments . the enhanced depletion of oxygen in dust grains along the densest sight lines , the imprint of the galactic gradient along the most distant sight lines and inefficient mixing processes are expected to play significant roles in the final scatter . in this paper , we present new o / h measurements toward 19 lines of sight ( table 1 ) obtained with data from _ fuse _ and stis . for each one of these sight lines we performed a complete analysis of the o i , h i , and h@xmath0 content , using different techniques . the datasets and the reduction processing are described in 2 . the analysis of o i , h i , and h@xmath13 are presented in 3 . in 4 the results are discussed in view of the most recent investigations . discussions about the dust content and local oxygen deficit are given in 5 and conclusions are given in 6 . in this work , we use high - resolution stis echelle observations of 19 stars ( table 1 ) to derive the column densities of h i , o i , and other atomic species . the design and construction of stis are described by woodgate et al . ( 1998 ) , while information about the on - orbit performance of stis is summarized by kimble et al . a summary of the stis observations is given in table 2 . all observations in this work used the far - ultraviolet mama detector with the e140h grating . because the datasets are drawn from several programs , three different apertures were used in obtaining the data . the @xmath14 aperture ( the so - called `` jenkins slit '' ) provides a spectral resolution of @xmath15 ( velocity resolution of @xmath16 km s@xmath6 , fwhm ) . these datasets have been discussed by jenkins & tripp ( 2001 ) . the remainder of the data , taken through the @xmath17 or @xmath18 apertures , provide a spectral resolution of @xmath19 ( velocity resolution of @xmath20 km s@xmath6 , fwhm ) . data taken through the latter apertures have a line spread function ( lsf ) with power in a halo extending @xmath21 km s@xmath6 from line center . the lsf for all of the data have weak , broad wings stretching to @xmath22 km s@xmath6 ( see the stis instrument handbook ) . the data were calibrated and extracted and the background estimated as described by howk & sembach ( 2000a ) . for spectral regions covered in multiple orders , or when more than one observation existed , we co - added the flux - calibrated data weighting each spectrum by the inverse square of its error . this work makes use of _ fuse _ observations for deriving molecular hydrogen column densities along the 19 lines of sight studied . these column densities are in turn used to derive the total ( atomic+molecular ) hydrogen column densities along each sight line . the _ fuse _ mission , its planning and on - orbit performance are discussed by moos et al . ( 2000 ) and sahnow et al . briefly , the _ fuse _ observatory consists of four co - aligned prime - focus telescopes and rowland - circle spectrographs that produce spectra over the wavelength range 905 - 1187 with a spectral resolution of @xmath23 15 - 20 km s@xmath6 for point sources ( depending on the wavelength ) . two of the optical channels employ sic coatings providing reflectivity in the wavelength range @xmath23 905 - 1090 while the other two have lif coatings for maximum sensitivity above 1000 . dispersed light is focused onto two photon - counting microchannel plate detectors . table 3 presents a summary of the _ fuse _ observations used in this work . all of the datasets were obtained through the large 30@xmath330 ( lwrs ) aperture . the two - dimensional _ fuse _ spectral images were processed using the standard calfuse pipeline ( version 1.8.7 ) . the reduction and calibration procedure includes thermal drift correction , geometric distortion correction , heliocentric velocity correction , dead time correction , and wavelength and flux calibration . the observations of each star were broken into several sub - exposures , which were aligned by cross - correlating the individual spectra over a short wavelength range that contained prominent spectral features . the aligned sub - exposures were co - added , weighting each by the exposure time . no co - addition of data from different channels was performed ; instead each channel was used independently in the fitting procedure as a different measurement . in this section we discuss the analysis of atomic species o i , h i , kr i , and molecular hydrogen ( h@xmath13 ) . whenever possible we compare our results with those found in the literature . the analysis of h i ( from stis data ) and h@xmath0 ( from _ fuse _ data ) are described respectively in sections 3.1 and 3.2 . 3.3 discusses the analysis of o i with other atomic species ( using stis data ) . one of the most important factors limiting the accurate measurement of the o / h ratio is the uncertainty associated with the available h i measurements , which are derived primarily from analyses of the damping wings of h i lyman-@xmath1 . for ob stars , the uncertainties in the h i measurements are largely systematic , particularly due to the complexity of the stellar continuum against which the interstellar lyman-@xmath1 is viewed . the systematic uncertainties can be minimized , however , in some cases by using high signal - to - noise datasets to probe further into the core of the lyman-@xmath1 profile ( e.g. , howk et al . 1999a ) . we have endeavored to measure the h i column densities towards the stars in our sample , making use of the stis e140h spectra discussed above . because the lyman-@xmath1 transition covers several , we have co - added several echelle orders for our analysis . the calstis flux calibration of the e140h data shows a systematic error in that the short wavelength portion of an order @xmath24 shows a larger flux than the long wavelength portion of the order @xmath25 . we have corrected this by applying a polynomial flux correction to each order . the correction was assumed to be the same for each order within an observation and derived from a comparison of stis and ghrs data of hd218915 . unfortunately no other ghrs target was available for this comparison but we are confident in our correction for it produces a symmetrical profile of the lyman-@xmath1 transition in each spectrum . we used a weighted average scheme to combine adjacent flux - corrected orders . given the confusing presence of stellar photospheric and wind lines in the region near interstellar lyman-@xmath1 , we make use of the continuum reconstruction method ( bohlin 1975 ; diplas and savage 1994 ; howk et al . 1999a ) for deriving n(h i ) . this method multiplies the observed lyman-@xmath1 profile by @xmath26 , where @xmath27 is the optical depth as a function of wavelength . the h i column density is determined by the value of @xmath28 $ ] which best reconstructs the damping wings of lyman-@xmath1 to the estimated position of the stellar continuum . the method relies on human judgment to set the stellar continuum , which is not rigorously defined given the undulating stellar absorption features . for each of the stars , two authors have independently determined the @xmath11 2 @xmath9 limits following howk et al . ( 1999a ) with the final value being negotiated between them . the @xmath11 2 @xmath9 limits are those column densities which showed reconstructed continua that were highly implausible . the best fit values were then calculated to be the midpoint , assuming symmetric errors . finally , we find that our column densities are , in all cases , within @xmath29 of those given by diplas & savage ( 1994 ) . however , the higher signal to noise ratio of our _ hst _ data allows us to use regions closer to the line core , leading to smaller systematic uncertainties . an example of this procedure is shown in figure 1 , where the reconstructed continua for hd210839 are displayed . one potential difficulty in the h i analysis is the contamination of the ism lyman-@xmath1 profile by stellar h i. in particular , for stars earlier than b2 the contamination is important for `` low '' interstellar column densities ( @xmath30 ; see diplas and savage 1994 ) . since our sight lines probe long pathlengths , most of which are in the galactic disk , we are always in the regime where @xmath31 . therefore , the corrections for stellar lyman-@xmath1 are always negligible . we have tested the consistency of our lyman-@xmath1 h i analysis for nine of our targets using the higher - order lyman transitions in the _ fuse _ spectra of these stars . for the other stars , the blending of h@xmath0 and h i is too severe to adequately check the n(h i ) determination . for each of these nine targets , we constructed a complete model of the line of sight ( velocity components , atomic and molecular content ) which was then compared with the _ fuse _ profiles of the weaker h i transitions . these models provided two checks of our lyman-@xmath1 analysis : first , they confirmed , using the weak damping wings of lyman-@xmath32 , the total h i column derived from the stis data ( although with larger uncertainties ) ; second , modeling the high - order lines ruled out the existence of high velocity , low-@xmath33 components which might adversely affect our simple assumptions about the distribution of lyman-@xmath1 optical depth with wavelength ( see appendix a of howk et al . this eliminated a potentially significant bias in our h i column density determinations caused by hot , low column - density clouds similar to those discussed in this context by lemoine et al . ( 2002 ) and vidal - madjar & ferlet ( 2002 ) for sight lines through the local ism . the _ fuse _ bandpass contains some 20 ro - vibrational bands of molecular hydrogen ( see , e.g. , shull et al . 2000 ; tumlinson et al . 2002 ) which can be used for the determination of h@xmath0 column densities . most of the sight lines studied here lie predominantly within the galactic disk and have quite high molecular hydrogen column densities . we adopt wavelengths , oscillator strengths , and damping constants for the molecular hydrogen transitions from abgrall et al . ( 1993a ) for the lyman system and abgrall et al . ( 1993b ) for the werner system . our analysis of h@xmath0 with the _ fuse _ data makes use of the @xmath34 profile fitting software ( lemoine et al . 2002 , hbrard et al . 2002a ) to fit the h@xmath0 line profiles assuming a single maxwellian component . we assume that the _ fuse _ lsf is well described by a single gaussian with a fwhm of 20 km s@xmath6 . while this is an oversimplification of the _ fuse _ lsf ( see wood et al . 2002 , hbrard et al . 2002a ) , it is sufficient for our purposes . in the course of determining the h@xmath0 content , we proceed in two steps . first , we analyse the population of the rotational levels j=2 and 3 that contain information on the excitation temperature within the gas . although most of the corresponding absorption lines are deeply saturated , a few lines are weak enough to be near the linear part of the curve of growth , thereby allowing us to get reliable column density estimates . the profile fitting is done simultaneously over the different absorption lines ( @xmath35 10 spectral windows ) . then , we investigate the j=0 and 1 rotational levels that represent more than 90% of the total h@xmath0 column density . for all of the sight lines presented in this work , the strongest j=0,1 lines show damping wings implying @xmath36 @xmath5 ( tumlinson et al . 2002 ) . the specifics of the unresolved component structure were unimportant for these rotational levels . the main difficulty in analyzing these lines is determining the stellar continuum . catanzaro et al . ( 2001 ) showed that , for early type stars with low rotational velocities , stellar continua are relatively smooth and featureless over the v1 - 0 and v2 - 0 ro - vibrational bands . we follow catanzaro et al . ( 2001 ) in analyzing the j=0 and j=1 levels using the v1 - 0 and v2 - 0 lyman bands . the j=0 transitions in these bands are centered at @xmath37 and 1077 . the profile fitting is simultaneously performed over 5 spectral windows corresponding to the different channels . an example of the fit quality is shown in figure 2 . we find that the rotational levels j=2 and 3 represent only a few percent of the total h@xmath0 column density in all lines of sight studied here . as a result , they make relatively minor contributions to the error budget . finally the errors are dominated by uncertainties in the continuum placement near the strong j=0,1 h@xmath0transitions . we have estimated the contribution of the continuum - placement uncertainties by fitting the data with three different continua : the best continuum , the best continuum plus 10% , and the best continuum minus 10% . we then assume that the h@xmath0 column densities derived with the @xmath11 10% continua represent @xmath29 deviations from the best - fit columns . table 4 presents the derived h@xmath0 column densities for each line of sight , as well as the fractional abundance of molecular hydrogen , @xmath38 $ ] and the t@xmath39 rotational temperature . we have compared our h@xmath13 results with previous analyses of the _ fuse _ data along two high extinction cloud sight lines studied by rachford et al . ( 2002 ) . in table 5 we compare our results with theirs . to determine the column density of oxygen along our 19 sight lines , we use the weak o i intersystem transition at 1355.598 ; the extreme weakness of this transition allows us to avoid most of the saturation effects seen in the other o i lines observable by stis and _ fuse_. in agreement with meyer ( 2001),we adopt the f - value for the 1355 transition suggested by welty et al ( 1999 ) : f=@xmath40 . this value was derived from the mean of the lifetime measurements of johnson ( 1972 ) , wells and zipf ( 1974 ) , nowak , borst and fricke ( 1978 ) and mason ( 1990 ) and the theoretical branching fraction of bimont and zeippen ( 1992 ) . this value is slightly smaller than the one previously adopted ( @xmath41 , morton 1991 ) . when comparing our results with previous works that make use of the older f - value , we linearly correct their column densities to reflect our choice of oscillator strength . our derived errors do not include uncertainties in the f - value ( which are in large part unknown ) . although the o i 1355 line is typically quite weak , the presence of cold components with significant amounts of matter could cause difficulties when deriving the o i column densities if the lines are sufficiently narrow to be unresolved by stis . we used 2 different approaches to derive the o i column densities : profile fitting ( pf ) and the apparent optical depth ( aod ) method . the aod method ( savage & sembach 1991 , lehner et al . 2002 ) yields apparent column densities , n@xmath42 , that are equivalent to the true column density if no unresolved saturated structure is present . if such structure is present , then n@xmath43 n , and the apparent column density is a lower limit to the true column density . our aod analysis , including estimation of the errors , follows the procedures outlined in sembach & savage ( 1992 ) . profile fitting , in which a detailed component model of the interstellar absorption is compared with the data , was performed with the profile fitting code @xmath34 developed by martin lemoine and the _ fuse _ french team . we assume that the stis lsf is well represented by a gaussian with a fwhm of 2.67 km s@xmath6 for the spectra obtained through the @xmath17 and @xmath18 slits . for the @xmath14 slit the adopted lsf is a gaussian with a fwhm of 1.5 km s@xmath6 . the errors were calculated with the @xmath44 technique for each of the fitted components ( hbrard et al . 2002a , lemoine et al . 2002 ) . the profile fitting analysis of the o i column density was done simultaneously with other atomic species with absorption lines visible in the stis data . each fit made use of the following species to constrain the line of sight velocity structure : cl i @xmath451347 , c i @xmath451276 and c i @xmath451270 , s i @xmath451295 and kr i @xmath451235 . results of the fits for these species will be presented in a companion paper . cl i and s i are useful tracers of dense clouds , while c i also traces cool material along diffuse lines of sight . because these atomic species have different masses we were able to constrain the kinetic temperature and the non - thermal broadening for each component along the line of sight . an example of the full fit to all of the atomic species is shown in figure 3 for the sight line towards hd210839 . figures 4 and 5 show the resulting fits for o i along the 19 sight lines . table 6 shows the comparison between the results obtained with the two methods , the equivalent width and the adopted n(o i ) . the adopted values are weighted averages of the two techniques , while the adopted uncertainties are the largest of the two errors . previous o i measurements can be found in the literature for six of our targets : hd75305 , hd104705 , hd177989 , hd185418 , hd218915 and hd303308 . our measurements are compared with the literature values in table 5 . with one exception our results are consistent with previous measurements , within 1 @xmath9 . for the sight line towards hd218915 , we found that there is no need for the saturation correction applied by howk et al ( 2000 ) , so our value is 30% smaller than previously stated . meyer et al . ( 1998 ) and cartledge et al . ( 2001 ) have discussed the use of kr i , a noble gas , as a proxy for neutral hydrogen , demonstrating the relative constancy of o / kr in local gas . for four stars we were able to derive kr i column densities , which , when compared with the o i column densities , offer another opportunity to compare our results with others . table 7 summarizes our o / kr measurements , showing that our measurements are consistent with previous results . the dataset is composed of 19 sight lines toward distant stars ranging from .8 to 5.0 kpc . the average distance from the sun is 2.6 kpc while the average galactocentric radius is 7.7 kpc . we are thus probing material well beyond the local ism . the average distance from the sun in our sample is about 5 times the one found in meyer et al ( 1998 ) : @xmath46d@xmath47 = 530 pc . the molecular fraction ranges from 3% to 47% while the e@xmath48 ranges from 0.16 to 0.64 . although most of our sight lines are `` translucent sight lines '' as defined in rachford et al ( 2002 ) , a handful are purely diffuse sight lines with the gas largely in the local ism . as shown in table 1 , the stars studied in this work are mostly distributed in the disk of the galaxy with a few of them being in the low halo ( @xmath49 @xmath50 250 pc ) : hd88115 , hd99857 , hd157857 , hd218915 , and hd177989 . hd177989 is the most distant star in the sample ( 5 kpc ) and is , at the same time , the highest above the galactic plane with @xmath49=1032 pc . this line of sight goes through the ejecta of the scutum supershell ( gs 018 - 06 + 44 ) , in the scutum spiral arm . previous observations carried out by savage et al . ( 2001 ) reveal strong and broad absorption in the lines of si iv and c iv centered on lsr velocities of + 18 and + 42 km s@xmath6 and weaker absorption from these ions near @xmath5150 and @xmath5113 km s@xmath6 . however , the neutral gas does not show any signature of these weak high velocity components . among the densest sight lines in our dataset are hd210839 and hd192639 with e@xmath48=0.62 and 0.64 , respectively . hd210839 is much closer to the sun than hd192639 and has been long recognized to have a peculiar extinction curve ( diplas & savage 1994 ) . the average density is about 80 hydrogen atoms per cubic centimeter as derived from o i excited lines ( jenkins & tripp 2001 ) . new diffuse interstellar band features ( dibs ) have also been reported recently toward this star ( weselak 2001 ) . with the highest molecular fraction of our sample ( 47% ) this sight line is likely to probe dense material in which the oxygen might be more depleted onto grains . hd192639 is part of the cyg ob1 association located at 1.8 kpc from the sun . however , for this latter sight line a deep investigation of the c i transitions with the stis e140h echelle , by sonnentrucker et al . ( 2002 ) shows that there is a dense , albeit relatively small , knot along this sight line , which has diffuse global properties ( with an expected average of 11 atoms @xmath52 ) . two peculiar sight lines in our sample show intermediate velocity components : hd93205 and hd93222 . both of these stars are part of the carina nebula and are only 400 parsec away from each other . in a recent survey of the carina nebula ( garcia & walborn 2000 ) , it has been demonstrated that the high and intermediate velocity structures are produced by an interaction between the stellar winds of the earlier type stars and immediately surrounding interstellar material . small angular variations in na i and ca ii abundances are typical in this nebula and we shall expect the same for the oxygen abundance if the absorption is mainly due to the surrounding clouds . table 6 summarizes all the measurements performed for . using the 19 stars of the sample , we find a mean abundance of = 408 ppm with an error in the mean of 13 ppm . the error in the mean we report here is the weighted average variance which is computed with the individual error estimates and the scatter around the mean . we also computed the _ a priori _ estimate of the error in the mean using only the individual error estimates . while this usually should give a smaller value , we find 14 ppm and this result strongly suggests that some of our errors have been overestimated . the weighted average variance is certainly a more reliable estimate for the error in the mean in this case . throughout this section , we also use the weighted average variance as the best estimate of the error in the mean when comparing different samples . the straight standard deviation is 59 ppm and represents only a 15% deviation from the mean . the reduced @xmath53 of 0.9 indicates that the scatter is consistent with the errors assuming a gaussian parent distribution around a constant value . if we only consider the low halo targets ( @xmath54 pc ) , we have 5 stars with a mean of = @xmath55 ppm . these sight lines to halo stars appear to be statistically consistent with the rest of the sample ; this is due to the fact that these sight lines are probing mostly disk material . in particular , in the case of hd177989 , we have seen that none of the typical velocity structures associated with high - z features along this sight line are detected in the neutral gas . the fact that we find a homogeneous oxygen abundance throughout the disk contrasts with the large scatter reported in stellar atmospheres studies ( smartt & rolleston 1997 , primas et al . 2001 ) . it could be that this scatter vanishes as we integrate the column density over long pathlengths . on the other hand , our scatter is consistent with the rms scatter found in the h ii survey by deharveng et al . ( 2000 ) ( @xmath35 22 % ) and is also consistent with the picture of an efficient mixing of oxygen in the galactic disk as shown by theoretical calculations ( see roy & kunth 1995 ) . the constancy of the oxygen abundance holds for a large range of molecular fractions as shown in figure 6 . it is remarkable to observe that , although many of the targets in our sample have molecular fractions similar to the ones reported in cartledge et al . ( 2001 ) , we do not report any trend toward an enhanced oxygen depletion . one possible explanation is that the sight lines in our sample are made of a mixture of diffuse and dense clouds and are indeed dominated by the diffuse components in most cases , contrary to the most reddened sight lines in their study which are spatially associated with dark cloud complexes . however , in the case of the sight line toward hd210839 , which presents both high e@xmath48 and large density ( 80 hydrogen atoms per cubic centimer ) , we still do not detect the hint of an enhanced oxygen depletion reported by cartledge et al . ( 2001 ) . in order to test how much the oxygen abundance is varying with distance and density we combine our measurements with measurements from the recent literature . the 13 sight lines from meyer et al . ( 1998 ) are similar to the sight lines in our sample ( referred as sample a ) in the sense that they are mostly diffuse and are not expected to show o i enhanced depletion onto grains . noteworthy , the average distance in sample a is about one order of magnitude larger than in the meyer et al . ( 1998 ) sample . the comparison of the mean also reveals a systematic 16% underabundance of oxygen within 500 pc , as illustrated in figure 7 . by combining meyer et al . ( 1998 ) values with ours , we get a sample of 32 sight lines ( sample b ) in which the distances range from 150 pc to 5 kpc . we find a mean of 377 ppm with an error in the mean of 10 ppm . finally , we include 5 measurements performed through denser media by cartledge et al . ( 2001 ) : hd27778 , hd 37021 , hd37061 , hd147888 and hd207198 . while the distances of these 5 stars are comparable to the distances found in sample a , the sight lines are different in nature being probably a lot denser . this final sample ( sample c ) is made of 37 sight lines with a wide variety of properties : the total hydrogen column density ranges from 1.5@xmath310@xmath56 @xmath5 to 62@xmath310@xmath56 @xmath5 , log f(h@xmath0 ) from -5.21 to -0.2 , the reddening from 0.09 to 0.64 and the distance to the sun from 150 pc to 5 kpc . table 8 summarizes the oxygen abundance found in the different samples . in particular , we see that in sample c , although the scatter is the largest , the standard deviation is only of the order of 20% around the mean of 362 ppm . from the 19 measurements of the total hydrogen in sample a , we derive an average @xmath46 h@xmath57/e@xmath48 @xmath47 of 6.3@xmath310@xmath4 @xmath5 mag@xmath6 with a standard deviation of 9@xmath310@xmath56 @xmath5 mag@xmath6 ( figure 8) . this value is fully consistent with the known galactic averages @xmath46 h@xmath0/e@xmath48 @xmath47 = 5@xmath310@xmath56 @xmath5 mag@xmath6 ( dufour et al . 1982 ) , @xmath46 h i / e@xmath48 @xmath47 = 4.93@xmath310@xmath4 @xmath5 mag@xmath6 ( diplas & savage 1994 ) implying @xmath46 h@xmath57/e@xmath48 @xmath47 = @xmath46 h i / e@xmath48 @xmath47 + 2 @xmath46 h@xmath0/e@xmath48 @xmath47 = 5.93@xmath310@xmath4 @xmath5 mag@xmath6 . the standard deviation we derive for this ratio is relatively small and is an indication that the total hydrogen content and the dust are well correlated in our sample . finally , we point out that oxygen column density and reddening seem to be comparable tracers of the total hydrogen column density over a wide range of distances and densities . the role of oxygen is fundamental in modern dust models . it is known to be abundant in the cores of silicate dust grains ( hong & greenberg 1980 ) . in particular , the maximum amount of oxygen atoms tied up into grains is limited by the abundance of metals with which oxygen bonds ( o : si : mg : fe ) @xmath35 ( 24:1:1:1 ) . a theoretical upper limit to the number of oxygen atoms that can be incorporated into grains is then calculated assuming that all grains are made of silicates ( cardelli et al . 1996 ) : ( o / h)@xmath58 @xmath59 ppm . it was long believed that this value was too small to reconciliate the gas phase ism oxygen abundance with the solar value . however , recent measurements of the solar oxygen abundance by holweger ( 2001 ) and allende et al . ( 2001 ) have revised the solar abundance downward significantly . allende et al . ( 2001 ) derived a mean solar oxygen abundance of 490 @xmath11 59 ppm while holweger ( 2001 ) derived a similar value ( although with larger errors ) with @xmath60= 545 ppm @xmath11 100 ppm . in the following we adopt a straight average of their values as the solar abundance : @xmath61 ppm . this value is now in good agreement with the maximum total abundance ( gas+dust ) derived using our gas - phase o / h measurements combined : + ( o / h)@xmath62 . thus it would seem that the total interstellar oxygen abundance is now consistent with the best solar oxygen abundance . if we assume that the solar value reflects the total oxygen abundance in the ism and that the total oxygen abundance is constant , we can then estimate the average dust oxygen content for our sight lines . we find @xmath46(o / h)@xmath63 = 109 ppm with a standard deviation of about 60 ppm . since we know the maximum amount of oxygen atoms that can be tied up into silicate grains ( o / h)@xmath58 = 180 ppm ) , then , from our survey , we find that @xmath35 60% of the grains are resilient against destruction by shocks on average , with one extreme sight line showing nearly 100% destruction and others showing no destruction at all . along the sight line toward hd93222 , which probes material surrounding the star ( see 4.1 ) , most of the oxygen atoms have been released in the gas phase . this can be interpreted as the effect of the interaction between the stellar wind and the nearby ism . on the other hand , the three most oxygen depleted sight lines in sample c are from the cartledge et al . ( 2001 ) survey in the dense ism , consistent with an enhanced depletion into dust grains . finally , this is also consistent with the theoretical expectation that only 10 to 15% of the grains are destroyed at each passage of a 100 km s@xmath6 shock in the warm ism ( tielens et al . 1994 , mckee et al . such shocks are believed to be responsible for most of the grain destruction and could explain the observed population of resilient dust grains . these shocks could also explain the variations in oxygen bearing dust . several other indirect pieces of observational evidence support the view that dust grains are long - lived , among them the abundance studies toward the diffuse halo clouds by sembach & savage ( 1996 ) and the detection of dusty structures in the halo of spiral galaxies by howk & savage ( 1999b ) . both observations imply that a fraction of the grains can survive for a long time before being destroyed by thermal sputtering in the halo . meyer et al ( 1998 ) showed that the local oxygen abundance is remarkably constant . our data , probing a wider range of environments at larger distances , show much the same . however , there is a systematic difference of 16% between the two samples that needs to be discussed in the context of a possible enhanced depletion within a few hundred parsecs from the sun . previous studies of the local ism have found some evidences for a global elemental deficit . krypton abundance has been measured toward 10 stars in the local ism by cardelli et al . ( 1997 ) and a depletion of 1/3 was found when compared to the solar abundance . because it is a noble gas , krypton is not expected to be depleted into the dust phase and this observation alone is a strong argument against the sun being an appropriate cosmic standard for the nearby ism . the same has been observed with sulfur , which is not supposed to be depleted in dust grains . the sulfur abundance derived from nearby b stars is about 2/3 of the observed ism abundances beyond 2 kpc ( fitzpatrick & spitzer 1993 , 1995 , 1997 and harris & mas 1986 ) and 2/3 of the solar sulfur abundance . the same depletion pattern is seen for oxygen when the solar value is compared to the nearby b stars while the comparison with distant b stars does not shed much light on the subject because of the large scatter ( smartt & rolleston 1997 ) . more generally , abundances have been measured in nearby b stars and young f and g field stars ( see the summary by snow & witt 1996 ) and are consistently below solar for all measured elements . these star samples are representative of the current local ism and clearly point toward a local elemental deficit . meyer et al . ( 1998 ) have reviewed the possible scenarios to explain such a depletion pattern : presolar nebula enrichment , radial migration of the sun and infall of metal - poor gas . the oxygen abundance we derive from outside the local ism can be used as a test for the latter scenario . as far as the oxygen abundance is concerned , the presolar nebula enrichment and the sun migration might be discarded because of the apparent consistency between the solar value and the total oxygen abundance of the distant ism . the recent infall of metal - poor material seems to be consistent with our present observation that the oxygen gas - phase abundance is slightly higher beyond 1 kpc . this is expected to be the case if the gould belt has been created by an infalling cloud ( comeron & torra 1994 ) . infall is considered as an important component of galactic chemical evolution and is supported by evidence for the infall of low - metalicity clouds onto the milky way ( richter et al . 2001 ) . in this scenario the dilution should have affected all the elements the same way and the local oxygen enhanced depletion is expected to be the same as observed in krypton and sulfur : about one third . however , the difference between our value , = @xmath64 @xmath11 13 ppm beyond 1 kpc and the value found by meyer et al . ( 1998 ) , = @xmath65 @xmath11 15 ppm within 1 kpc , is only one sixth . a possible explanation for a reduced dilution of oxygen in the local ism is the destruction of a fraction of the dust grains that resulted from the impact of a low metalicity cloud , at an estimated velocity of 100 km s@xmath6 ( comeron & torra 1994 ) . however , in this scenario , we should expect that many metals were released in the gas - phase as well . it is beyond the scope of this paper to explore the different infall scenarios that might be consistent with the observations . some of the stars in our sample are sufficiently distant that the measurements might be affected by large scale abundance gradients in the ism . of course the gradient is in part masked by the integrated nature of our absorption lines and by the intrinsic scatter between sight lines . an extreme example of intrinsic scatter between the data points is given by the couple hd93222/hd93205 . both are members of the carina nebula ( garcia & walborn 2000 ) and show intermediate velocity structures in their spectra due to stellar winds interacting with surrounding material . it is well known that in this nebula the spatial variation of na i as well as ca ii can be as large as 50% . although they are within 400 pc of each other , the oxygen abundances we derive differ by as much as 43% . however the typical scatter between our data points is only of the order of 15% and this must be compared to the gradient effect expected in our data . numerous papers have discussed the distribution of heavy elements in spiral galaxies ( see the review by henry & worthey 1999 and references therein ) . the published values by deharveng et al . ( 2000 ) , using h ii regions , and by smartt & rolleston ( 1997 ) , using b stars , give the range of expected values for the oxygen gradient @xmath1 : -0.07@xmath66 dex kpc@xmath6 @xmath67 -0.039@xmath68 dex kpc@xmath6 . it is difficult , if not impossible , to use the velocity of the absorbers to deduce a precise location within the disk assuming the most recent galactic rotation curve . the reason is that , except in a few cases , the expected velocities in the local standard of rest along our sight lines are within the typical scatter ( around 7 km@xmath69 ) . thus , we choose to show the oxygen abundance as a function of the location of the background star , expecting the gradient to show up as well . figure 9 allows the comparison of the extreme @xmath1 values with our data ( excluding the peculiar sight lines hd93222 and hd93205 ) . it turns out that the scatter of the points is largely dominated by the measurement errors while no large scale trend emerges at all . in this paper we have presented a survey of the oxygen gas - phase abundance using stis/_fuse _ data toward 19 stars with the purpose of probing the oxygen abundance far away in the disk . this work follows from previous surveys in the local ism and is possible due to the _ fuse _ detector s sensitivity , which permits measurements of the h@xmath0 content through long pathlengths . a summary of our results is as follow : 1 . we demonstrate that the oxygen abundance shows little variation in various environments and various locations in the disk . we find a mean ratio of = 408 ppm with an error in the mean of 13 ppm . this is 19% higher than the mean previously derived in the local ism by meyer et al . we argue that a recent local infall of metal - poor material is not excluded as an explanation for the difference . 2 . the oxygen abundance derived from our sample combined with previous surveys ( 37 sight lines ) is consistent with both the oxygen abundance of young g & f stars ( @xmath70 ppm ; sofia & meyer 2001 ) or the newly revised solar value ( @xmath71 ppm ; holweger 2001 , allende et al . 2001 ) once dust is taken into account . it is likely that the o i column density can be used as a surrogate tracer of the total neutral hydrogen column density , at least in a statistical sense , in diffuse ism abundance studies . although above we have emphasized the 19% difference between the mean found in this study and the lower value found for the lism by meyer et al . ( 1998 ) , the most striking result from both studies is the lack of large deviations from the mean . the standard deviation in the sample reported here is 15% . if the meyer et al . ( 1998 ) and the cartledge et al . ( 2001 ) data are included , the standard deviation increases to only 20% for a wide variety of diffuse ism conditions . the use of o i as a tracer of the total hydrogen column density could be of use in a variety of chemical abundance studies . as an example , consider the d / h studies , presently under way with the _ fuse _ satellite . in many cases , determining accurate h i column measurements requires using _ hst _ or _ euve _ data , which may not be available . in other cases , because the column densities of d i and h i differ by about five orders of magnitude , accurate measurements of both species may be precluded by special conditions associated with a particular sight line . preliminary studies of d / o ratios using _ fuse _ measurements of d i and o i are encouraging ( moos et al . 2002 , and references therein , hbrard et al . additional studies are also needed to determine the special cases in which there are large deviations in the and identify the special conditions under which such deviations might exist ( e.g hoopes et al . 2003 ) . we are grateful to paule sonnentrucker for sharing with us information on the lines of sight toward hd185418 and hd192639 . we also want to thank dr . ken sembach for fruitful advice . we are pleased to acknowledge comments by dr . stefan cartledge , dr . james lauroesch and dr . david meyer leading to several improvements on this paper . we thank also the referee for his thoughtful remarks . this work is based on data obtained for the guaranteed time team by the nasa - cnes - csa _ fuse _ mission operated by the johns hopkins university . financial support to u. s. participants has been provided in part by nasa contract nas5 - 32985 to johns hopkins university . support for french participation in this study has been provided by cnes . based on observations made with the nasa / esa hubble space telescope , obtained from the data archive at the space telescope science institute , which is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . these observations are associated with proposal 8241 . this work has been done using the profile fitting procedure owens.f developed by m. lemoine and the french _ fuse _ team . abgrall , h. , roueff , e. , launay , f. , roncin , j. y. , & subtil , j. l. 1993a , , 101 , 273 abgrall , h. , roueff , e. , launay , f. , roncin , j. y. , & subtil , j. l. 1993b , , 101 , 323 allende prieto , c. , lambert , d. l. , & asplund , m. 2001 , , 556 , l63 audouze , j. & tinsley , b. m. 1976 , , 14 , 43 biemont , e. & zeippen , c. j. 1992 , , 265 , 850 bohlin , r. c. 1975 , , 200 , 402 cardelli , j. a. , meyer , d. m. , jura , m. , & savage , b. d. 1996 , , 467 , 334 cartledge , s. i. b. , meyer , d. m. , lauroesch , j. t. , & sofia , u. j. 2001 , , 562 , 394 catanzaro , g. , leone , f. , andre , m. , & sonnentrucker , p. 2001 , american astronomical society meeting , 198 , comeron , f. & torra , j. 1994 , , 281 , 35 de boer , k. s. 1981 , , 244 , 848 deharveng , l. , pe~ na , m. , caplan , j. , & costero , r. 2000 , , 311 , 329 diplas , a. & savage , b. d. 1994 , , 93 , 211 dufour , r. j. , shields , g. a. , & talbot , r. j. 1982 , , 252 , 461 garc ' ia , b. & walborn , n. r. 2000 , , 112 , 1549 harris , a. w. & mas hesse , j. m. 1986 , , 308 , 240 h ' ebrard , g. et al . 2002a , , 140 , 103 hbrard , g. et al . 2002b , , 50 , 1169 henry , r. b. c. & worthey , g. 1999 , , 111 , 919 holweger , h. 2001 , aip conf . proc . 598 : joint soho / ace workshop `` solar and galactic composition '' , 23 hong , s. s. & greenberg , j. m. 1980 , , 88 , 194 hoopes , c. g. , sembach , k. r. , h ' ebrard , g. , & moos , h. w. 2003 , in press howk , j. c. , savage , b. d. , & fabian , d. 1999a , , 525 , 253 howk , j. c. & savage , b. d. 1999b , , 517 , 746 howk , j. c. & sembach , k. r. 2000a , , 119 , 2481 howk , j. c. , sembach , k. r. , & savage , b. d. 2000b , , 543 , 278 jenkins , e. b. & tripp , t. m. 2001 , , 137 , 297 johnson , c. e. 1972 , , 5 , 2688 kaltcheva , n. t. & hilditch , r. w. 2000 , , 312 , 753 keenan , f. p. , hibbert , a. , & dufton , p. l. 1985 , , 147 , 89 kimble , r. a. et al . 1998 , , 492 , l83 lemoine , m. et al . 2002 , , 140 , 67 lehner , n. , gry , c. , sembach , k. r. , h ' ebrard , g. , chayer , p. , moos , h. w. , howk , j. c. , & d ' esert , j .- m . 2002 , , 140 , 81 mason , j. j. 1990 , meas.sci.technol , 1 , 596 mckee , c. f. , hollenbach , d. j. , seab , g. c. , & tielens , a. g. g. m. 1987 , , 318 , 674 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 nowak , g. , borst , w. l. , & fricke , j. 1978 , , 17 , 1921 primas , f. , rebull , l. m. , duncan , d. k. , hobbs , l. m. , truran , j. w. , & beers , t. c. 2001 , new astronomy review , 45 , 541 rachford , b. l. et al . 2002 , apj . , 5415 reed , b. c. 2000 , , 119 , 1855 richter , p. , sembach , k. r. , wakker , b. p. , savage , b. d. , tripp , t. m. , murphy , e. m. , kalberla , p. m. w. , & jenkins , e. b. 2001 , , 559 , 318 roy , j .- & kunth , d. 1995 , , 294 , 432 savage , b. d. , massa , d. , meade , m. , & wesselius , p. r. 1985 , , 59 , 397 savage , b. d. & sembach , k. r. 1991 , , 379 , 245 savage , b. d. , sembach , k. r. , & howk , j. c. 2001 , , 547 , 907 sahnow , d. j. et al . 2000 , , 538 , l7 sembach , k. r. & savage , b. d. 1992 , , 83 , 147 sembach , k. r. & savage , b. d. 1996 , , 457 , 211 shull , j. m. et al . 2000 , , 538 , l73 smartt , s. j. & rolleston , w. r. j. 1997 , , 481 , l47 snow , t. p. & witt , a. n. 1996 , , 468 , l65 sonnentrucker , p. , friedman , s. d. , welty , d. e. , york , d. g. , & snow , t. p. 2002 , , 576 , 241 tielens , a. g. g. m. , mckee , c. f. , seab , c. g. , & hollenbach , d. j. 1994 , , 431 , 321 tumlinson , j. et al . 2002 , , 566 , 857 vidal - madjar , a. & ferlet , r. 2002 , , 571 , l169 wells , w. c. & zipf , e. c. 1974 , , 9 , 568 welty , d. e. , hobbs , l. m. , lauroesch , j. t. , morton , d. c. , spitzer , l. , & york , d. g. 1999 , , 124 , 465 weselak , t. , fulara , j. , schmidt , m. r. , & kreowski , j. 2001 , , 377 , 677 wood , b. e. , linsky , j. l. , h ' ebrard , g. , vidal - madjar , a. , lemoine , m. , moos , h. w. , sembach , k. r. , & jenkins , e. b. 2002 , , 140 , 91 woodgate , b. e. et al . 1998 , , 110 , 1183 york , d. g. , spitzer , l. , jenkins , e. b. , bohlin , r. c. , hill , j. , savage , b. d. , & snow , t. p. 1983 , , 266 , l55 ccccccc name & d & z & l & b & @xmath72&spectral type + & [ pc ] & [ pc ] & & & + hd75309 & 2405 & 80 & 265.9 & -1.9 & 0.27&bi iip + hd88115 & 3654 & 350 & 285.3 & -5.5 & 0.16&b1 ib / ii + hd91824 & 2930 & 5 & 285.7 & + 0.1 & 0.27&o6 + hd93205 & 2630 & 32 & 287.6 & -0.7 & 0.45&o3 v + hd93222 & 2900 & 50 & 287.7 & -1.0 & 0.36&o7 iii + hd94493 & 3328 & 70 & 289.0 & -1.2 & 0.20&b0.5 iab / ib + hd99857 & 3070 & 267 & 294.8 & -5.0 & 0.33&b1 ib + hd104705 & 3900 & 20 & 297.4 & -0.3 & 0.26&b0 iii / iv + hd124314 & 1150 & 8 & 312.7 & -0.4 & 0.53&o7 + hd157857 & 2380 & 548 & 13.0 & + 13.3 & 0.49&o7 + hd177989 & 5010&1042 & 17.8 & -12.0 & 0.25&b2 ii + hd185418 & 950 & 33 & 53.6 & -2.0 & 0.43&b0.5 v + hd192639 & 1830 & 48 & 74.9 & + 1.5 & 0.64&o7 + hd202347 & 1300 & 45 & 88.2 & -2.0 & 0.17&b1 v + hd210809 & 3470 & 188 & 99.8 & -3.1 & 0.33&o9 ib + hd210839 & 840 & 38 & 103.8 & + 2.6 & 0.62&o6 iab + hd218915 & 3660 & 436 & 108.1 & -6.9 & 0.29&o9.5 iab + hd224151 & 1360 & 110 & 115.4 & -4.6 & 0.44&b0.5 iii + hd303308 & 2630 & 27 & 287.6 & -0.6 & 0.45&o3 v + lcccc hd75309 & o5c05b010 & 03 - 28 - 99 & 720 & @xmath18 + hd88115 & o54305010 & 04 - 19 - 99 & 1300 & @xmath14 + hd91824 & o5c095010 & 03 - 23 - 99 & 360 & @xmath18 + hd93205 & o4qx01010 & 04 - 20 - 99 & 1200 & @xmath17 + & o4qx01020 & 04 - 20 - 99 & 780 & @xmath17 + hd93222 & o4qx02010 & 12 - 28 - 98 & 1680 & @xmath17 + & o4qx02020 & 12 - 28 - 98 & 1140 & @xmath17 + hd94493 & o54306010 & 04 - 22 - 99 & 1466 & @xmath14 + hd99857 & o54301010 & 02 - 21 - 99 & 1307 & @xmath14 + hd104705 & o57r01010 & 12 - 24 - 98 & 2400 & @xmath17 + hd124314 & o54307010 & 04 - 11 - 99 & 1466 & @xmath14 + hd157857 & o5c04d010 & 06 - 03 - 99 & 720 & @xmath18 + hd177989 & o57r03020 & 05 - 28 - 99 & 2897 & @xmath17 + hd185418 & o5c01q010 & 11 - 12 - 98 & 720 & @xmath18 + hd192639 & o5c08t010 & 03 - 01 - 99 & 1440 & @xmath18 + hd202347 & o5g301010 & 10 - 09 - 99 & 830 & @xmath14 + hd210809 & o5c01v010 & 10 - 30 - 98 & 720 & @xmath18 + hd210839 & o54304010 & 04 - 21 - 99 & 1506 & @xmath14 + hd218915 & 057r05010 & 12 - 23 - 98 & 2018 & @xmath17 + hd224151 & o54308010 & 02 - 18 - 99 & 1496 & @xmath14 + hde303308 & o4qxo4010 & 03 - 19 - 98 & 2220 & @xmath17 + & o4qxo4020 & 03 - 19 - 98 & 1560 & @xmath17 + lcccccc hd75309 & p1022701 & 01 - 26 - 00 & lwrs & 4.7 & 8 & hist + hd88115 & p1012301 & 04 - 04 - 00 & lwrs & 4.5 & 8 & hist + hd91824 & a1180802 & 06 - 02 - 00 & lwrs & 4.6 & 6 & hist + hd93205 & p1023601 & 02 - 01 - 00 & lwrs & 4.7 & 7 & hist + hd93222 & p1023701 & 02 - 03 - 00 & lwrs & 3.9 & 4 & hist + hd94493 & p1024101 & 03 - 26 - 00 & lwrs & 4.4 & 7 & hist + hd99857 & p1024501 & 02 - 05 - 00 & lwrs & 4.3 & 7 & hist + hd104705 & p1025701 & 02 - 05 - 00 & lwrs & 4.5 & 6 & hist + hd124314 & p1026201 & 03 - 22 - 00 & lwrs & 4.4 & 6 & hist + hd157857 & p1027501 & 09 - 02 - 00 & lwrs & 4.0 & 8 & hist + hd177989 & p1017101 & 08 - 28 - 00 & lwrs & 10.3 & 20 & hist + hd185418 & p1162301 & 08 - 10 - 00 & lwrs & 4.4 & 3 & ttag + hd192639 & p1162401 & 06 - 12 - 00 & lwrs & 4.8 & 2 & ttag + hd202347 & p1028901 & 06 - 20 - 00 & lwrs & 0.1 & 1 & hist + hd210809 & p1223101 & 08 - 05 - 00 & lwrs & 5.5 & 10 & hist + hd210839 & p1163101 & 07 - 22 - 00 & lwrs & 6.1 & 10 & hist + hd218915 & p1018801 & 07 - 23 - 00 & lwrs & 5.4 & 10 & hist + hd224151 & p1224101 & 08 - 11 - 00 & lwrs & 6.0 & 12 & hist + hd303308 & p1222601 & 05 - 25 - 00 & lwrs & 6.1 & 9 & hist + & p1222602 & 05 - 27 - 00 & lwrs & 7.7 & 12 & hist + cccccc name & n(h i ) & n(h@xmath0 ) & n(h@xmath73 ) & f(h@xmath13 ) & t@xmath39 + & @xmath74 @xmath75 & @xmath74 @xmath75 & @xmath74 @xmath75 & & ( k ) + hd75309 & 11.2 ( 2.0 ) & 1.5 ( 0.2 ) & 14.2 ( 2.0 ) & 0.21 & 65 + hd88115 & 10.0 ( 2.0 ) & 0.2 ( 0.1 ) & 10.3 ( 2.0 ) & 0.03 & 145 + hd91824 & 11.7 ( 1.7 ) & 0.7 ( 0.2 ) & 13.1 ( 1.7 ) & 0.11 & 65 + hd93205 & 24.0 ( 2.3 ) & 0.6 ( 0.1 ) & 25.3 ( 2.3 ) & 0.05 & 75 + hd93222 & 25.1 ( 3.7 ) & 0.6 ( 0.1 ) & 26.4 ( 3.7 ) & 0.05 & 121 + hd94493 & 12.0 ( 1.2 ) & 1.4 ( 0.2 ) & 14.8 ( 1.3 ) & 0.19 & 53 + hd99857 & 17.4 ( 3.0 ) & 2.7 ( 0.4 ) & 22.8 ( 3.1 ) & 0.24 & 53 + hd104705 & 12.6 ( 2.5 ) & 1.2 ( 0.1 ) & 15.0 ( 2.6 ) & 0.16 & 137 + hd124314&25.7 ( 5.2 ) & 3.3 ( 0.4 ) & 32.3 ( 5.3 ) & 0.20 & 54 + hd157857 & 18.6 ( 2.3 ) & 4.5 ( 0.6 ) & 27.6 ( 2.6 ) & 0.33 & 49 + hd177989 & 9.1 ( 1.8 ) & 1.5 ( 0.2 ) & 12.1 ( 1.8 ) & 0.25 & 88 + hd185418 & 14.1 ( 2.5 ) & 5.1 ( 1.3 ) & 24.3 ( 3.6 ) & 0.42 & 101 + hd192639 & 19.5 ( 3.4 ) & 5.5 ( 1.5 ) & 30.5 ( 4.5 ) & 0.36 & 98 + hd202347 & 8.7 ( 1.8 ) & 0.9 ( 0.2 ) & 10.5 ( 1.8 ) & 0.17 & 116 + hd210809 & 17.8 ( 4.1 ) & 1.3 ( 0.3 ) & 20.4 ( 4.1 ) & 0.13 & 166 + hd210839 & 15.5 ( 1.9 ) & 6.5 ( 1.0 ) & 28.5 ( 2.8 ) & 0.46 & 92 + hd218915 & 14.8 ( 1.4 ) & 1.6 ( 0.2 ) & 18.0 ( 1.5 ) & 0.18 & 56 + hd224151 & 20.9 ( 3.6 ) & 4.1 ( 0.6 ) & 29.1 ( 3.8 ) & 0.28 & 42 + hd303308 & 25.7 ( 4.5 ) & 2.2 ( 0.4 ) & 30.1 ( 4.6 ) & 0.15 & 54 + ccccccc name & w@xmath76 ( 1355 ) & o i ( aod ) & o i ( fit ) & o i ( adopted ) & @xmath77 & @xmath78 + & m & @xmath79 @xmath75 & @xmath79 @xmath75 & @xmath79 @xmath75 & ppm & ppm + hd75309 & 9.0 ( 0.7 ) & 5.3 ( 0.4 ) & 5.2 ( 0.3 ) & 5.2 ( 0.4 ) & 366 ( 60 ) & 151 + hd88115 & 8.8 ( 1.1 ) & 5.1 ( 0.6 ) & 4.4 ( 0.6 ) & 4.7 ( 0.6 ) & 454 ( 117 ) & 62 + hd91824 & 11.0(1.4 ) & 6.2 ( 0.7 ) & 6.5 ( 0.4 ) & 6.4 ( 0.7 ) & 488 ( 84 ) & 29 + hd93205 & 19.5 ( 1.1 ) & 8.5 ( 0.6 ) & 9.4 ( 0.5 ) & 9.0 ( 0.6 ) & 356 ( 40 ) & 161 + hd93222 & 25.8 ( 1.2 ) & 13.0 ( 0.6 ) & 14.0 ( 0.4 ) & 13.5 ( 0.6 ) & 512 ( 75 ) & 5 + hd94493 & 12.2(0.7 ) & 6.6 ( 0.6 ) & 7.5 ( 0.6 ) & 7.0 ( 0.6 ) & 473 ( 57 ) & 44 + hd99857 & 14.1 ( 0.8 ) & 8.0 ( 0.5 ) & 7.6 ( 0.5 ) & 7.8 ( 0.5 ) & 342 ( 51 ) & 175 + hd104705 & 11.2 ( 0.7 ) & 6.4 ( 0.3 ) & 6.3 ( 0.3 ) & 6.4 ( 0.3 ) & 427 ( 74 ) & 90 + hd124314 & 25.3 ( 1.0 ) & 15.3 ( 0.6 ) & 15.4 ( 0.8 ) & 15.3 ( 0.8 ) & 474 ( 81 ) & 43 + hd157857 & 17.7 ( 0.8 ) & 11.6 ( 0.4 ) & 12.1 ( 0.4 ) & 11.9 ( 0.4 ) & 431 ( 43 ) & 86 + hd177989 & 10.5 ( 0.6 ) & 6.1 ( 0.3 ) & 6.0 ( 0.4 ) & 6.1 ( 0.4 ) & 504 ( 84 ) & 13 + hd185418 & 18.0 ( 0.7 ) & 11.4 ( 0.4 ) & 11.0 ( 0.3 ) & 11.2 ( 0.4 ) & 461 ( 70 ) & 56 + hd192639 & 22.1 ( 0.9 ) & 13.5 ( 0.3 ) & 13.5 ( 0.6 ) & 13.5 ( 0.6 ) & 443 ( 69 ) & 74 + hd202347 & 6.8 ( 1.1 ) & 3.8 ( 0.6 ) & 3.8 ( 0.4 ) & 3.8 ( 0.6 ) & 362 ( 85 ) & 155 + hd210809 & 12.5 ( 1.4 ) & 7.1 ( 0.7 ) & 5.6 ( 0.6 ) & 6.3 ( 0.7 ) & 309 ( 71 ) & 208 + hd210839 & 21.6 ( 0.9 ) & 13.4 ( 0.5 ) & 12.3 ( 0.6 ) & 13.0 ( 0.6 ) & 456 ( 49)&61 + hd218915 & 12.0 ( 0.8 ) & 6.8 ( 0.4 ) & 6.5 ( 0.2 ) & 6.6 ( 0.4 ) & 367 ( 37 ) & 150 + hd224151 & 18.5 ( 0.9 ) & 12.2 ( 0.8 ) & 10.7 ( 0.7 ) & 11.4 ( 0.8 ) & 392 ( 58 ) & 125 + hd303308 & 21.5 ( 1.1 ) & 12.7 ( 0.6 ) & 11.8 ( 0.4 ) & 12.2 ( 0.6 ) & 405 ( 65 ) & 112 + ccc stars & kr i & log o / kr + & 10@xmath80 @xmath75 & + hd75309 & 17.4 ( 2.8 ) & 5.47@xmath81 + hd99857 & 19.6 ( 4.1 ) & 5.62@xmath82 + hd104705 & 13.3 ( 4.9 ) & 5.50@xmath83 + hd185418 & 30.1 ( 2.5 ) & 5.57@xmath84 +

this paper discusses multi - skyrmions on the-sphere with variable radius using the rational map ansatz . for baryon number this ansatz produces the lowest energy solutions known so far . by considering the geometry of the model we find an approximate analytic formula for the shape function this provides an insight why skyrmions have a shell - like structure . pacs - number : 12.39.dc pages , 14 figures

hep-th0006147

in this paper we discuss new developments in the @xmath4 skyrme model @xcite and its generalization to the case where the physical space is a @xmath0-sphere of radius @xmath2 . the skyrme model is a nonlinear field theory of mesons whose field configurations are labelled by an integer , the topological charge . this charge can be identified with the baryon number @xmath5 @xcite . static field configurations which minimize the energy of the skyrme model for a given baryon number @xmath5 are called skyrmions . when the theory is quantized , the skyrme model not only describes the proton and the neutron reasonably well @xcite but also larger nuclei @xcite . however , in order to be able to perform the quantization it is important to reach a good understanding of the classical solutions . in @xcite skyrmions were calculated numerically for small baryon number @xmath5 , and it was shown that they have certain discrete symmetries . these symmetries have been confirmed by the rational map ansatz @xcite which also reproduces the energies of the skyrmions with good accuracy . from a mathematical point of view , field configurations in the skyrme model are represented by maps from @xmath6 to @xmath7 . therefore , it is natural to generalize the model such that physical space is a @xmath0-sphere of radius @xmath2 . from a physical point of view , a skyrmion on a @xmath0-sphere describes a finite baryon density @xcite . reducing @xmath2 increases the baryon density . varying @xmath2 the @xmath1 model exhibits phenomena such as localization delocalization transitions and the restoration of chiral symmetry . these results can be compared with skyrme crystal calculations @xcite , but are also interesting in their own right . in this paper we only consider static configurations and their energies . in particular , we generalize the rational map ansatz to describe skyrmions on @xmath1 . the energies of our ansatz are the lowest ones known so far . moreover , these approximations have a well defined limit for @xmath8 , namely , the rational map skyrmions in flat space . geometric considerations lead to an analytic ansatz for the shape function which is completely specified by a small set of parameters . we show that this ansatz agrees well with the numerical solution and captures the behaviour of skyrmions on @xmath1 . in the following section we review the skyrme model on general @xmath0-dimensional manifolds . the first part focuses on the geometrical meaning of the skyrme energy . the next part clarifies the relationship between the geometric formulation in flat space and the standard formulation in terms of the lie group @xmath4 . in section [ rationalmaps ] the model is generalized to the @xmath0-sphere . we describe the rational map ansatz in detail and also recall the doubly axially - symmetric ansatz @xcite . in section [ conformalmap ] we derive an analytic ansatz for the shape function and discuss its symmetries . for special values of @xmath5 the energy can be calculated explicitly as a function of the radius @xmath2 . in section [ numericalresults ] we calculate the shape function numerically . we also compare the numerical shape function to the analytic shape function of the previous section . finally , we use the analytic shape function to approximate skyrmions in @xmath9 . in section [ phasetransitions ] we discuss chiral symmetry and phase transitions . in this section we describe static solutions of the skyrme model on general @xmath0-dimensional manifolds . we follow manton s approach @xcite and present a geometric point of view . first , we introduce the geometric notion of the strain tensor and construct the skyrme energy for general manifolds . then we discuss the properties of the energy density and derive some formulae which will be important in the following sections . finally , we show that in flat space the geometric energy density is equivalent to the standard energy density . a field configuration is a map @xmath10 from a physical space @xmath11 to a target space @xmath12 . both @xmath11 and @xmath12 are @xmath0-dimensional riemannian manifolds which are connected and orientable , and their metrics are @xmath13 and @xmath14 , respectively . we denote by @xmath15 a point in @xmath11 and @xmath16 its image in @xmath12 . choosing dreibeins @xmath17 on @xmath11 and @xmath18 on @xmath12 , we can define the jacobian matrix @xmath19 the matrix @xmath20 is a measure of the deformation induced by the map @xmath10 at the point @xmath15 in @xmath11 . however , it is not unique since a rotation of the dreibeins @xmath18 does not affect the geometrical deformation . this leads us to define the strain tensor @xmath21 which only depends on the metric on @xmath12 . here and in the following we suppress the direct reference to @xmath15 . the strain tensor is symmetric and positive semi - definite is non - degenerate so that @xmath22 is positive definite . however , there are submanifolds of zero baryon density , i.e. @xmath23 . for further discussion see @xcite . ] but it is not invariant under rotation of the dreibeins @xmath24 . in fact , under an orthogonal transformation @xmath25 , @xmath22 transforms into @xmath26 . a well known result from linear algebra is that the characteristic polynomial @xmath27 is invariant under orthogonal transformations . denoting the non - negative eigenvalues of the strain tensor @xmath22 by @xmath28 , @xmath29 , and @xmath30 we obtain the following invariants of @xmath22 : @xmath31 as we will demonstrate in this section , choosing the energy of a field configuration to be the integral of the following sum of these invariants generalizes the skyrme model to arbitrary manifolds @xmath11 and @xmath12 : @xmath32 where the integration measure is @xmath33 . the first three terms are only quadratic in the derivatives and correspond to the `` sigma model term '' . the configurations which are minimizing this term are known as harmonic maps @xcite . geometrically , @xmath10 induces a linear map @xmath34 mapping the unit vector @xmath24 which is formed by the @xmath35th dreibein to the vector @xmath36 . the sum of the squared lengths of these vectors is the `` sigma model term '' . thus , only the lengths of the vectors @xmath37 are important . the geometric meaning of the quartic terms can be understood by considering the area element @xmath38 formed by two dreibeins . this area element is mapped to the following area element in @xmath12 : @xmath39 the sum of the squares of these area elements is proportional to the `` skyrme term '' . this term is also a measure of the angular distortion of the map @xmath34 . so far only the first two invariants of the strain tensor have been used . the third invariant is the square of the determinant of the jacobian , @xmath40 . since the manifolds @xmath11 and @xmath12 are orientable , @xmath41 is globally well defined so that we can set @xmath42 . @xmath43 locally changes the integration measure on @xmath11 into the measure on @xmath12 . therefore , the integral of @xmath43 over @xmath11 is equal to the volume @xmath44 times the degree of the map @xmath10 . the degree of a map is a topological invariant which , roughly speaking , measures how many preimages a point @xmath16 of @xmath12 has in @xmath11 , and hence takes integer values . any field configuration can be labelled by its degree . in the skyrme model the degree of the map @xmath10 is identified with the baryon number @xmath5 , and we obtain @xmath45 we will call classical field configurations which minimize the energy ( [ egeo ] ) skyrmions . it is worth pointing out that there is an alternative expression for equation ( [ egeo ] ) . by `` completing the square '' we can rearrange the @xmath46 in the following way : @xmath47 either the upper or the lower signs are chosen such that the integral over the last term in equation ( [ egeo2 ] ) is non - negative . applying equation ( [ bgeo ] ) , this integral is just @xmath48 . therefore , the energy is bounded below by the so - called faddeev - bogomolny bound @xcite : @xmath49 equation ( [ egeo2 ] ) also gives rise to two sets of three bogomolny equations which are satisfied if and only if the faddeev - bogomolny bound is attained , @xmath50 the only non - trivial solutions of both sets of equations are @xmath51 , and it follows that @xmath52 . therefore , the strain tensor is the identity map , and the map @xmath10 is an isometry . for @xmath53 , this case can only occur if the physical space is a @xmath0-sphere of radius @xmath54 . this has already been proven in @xcite . the relative importance of the `` sigma model term '' and the `` skyrme term '' can be seen from their scaling behaviour . for this purpose we consider a family of metrics @xmath55 where @xmath2 is a constant length scale . the dreibeins @xmath24 are , informally speaking , the square roots of the inverse metric @xmath56 of @xmath11 . therefore , they are proportional to @xmath57 so that the eigenvalues @xmath58 of the strain tensor @xmath59 are proportional to @xmath60 . since the measure @xmath61 scales with @xmath62 , the `` sigma model term '' scales like @xmath2 , whereas the skyrme term scales like @xmath57 . for large radius @xmath2 one might expect the `` sigma model term '' to be dominant . yet , this is not the case because , as we shall see , the skyrmion becomes localized and just as in flat space both of these terms are equally important . however , for small radius @xmath2 the configuration is delocalized , and the `` skyrme term '' will become dominant . in the following , we relate this geometric formulation to the standard skyrme model in flat space . for the remainder of this section @xmath11 will correspond to the flat space @xmath9 and @xmath12 to the lie group @xmath63 . the basic field is the @xmath4 valued field @xmath64 where @xmath65 are the pauli matrices which are hermitian and satisfy the algebra @xmath66 . since @xmath67 is an element of @xmath4 , the fields @xmath68 and @xmath69 obey the constraint @xmath70 . it is worth noting that @xmath69 in ( [ u ] ) play the same role as the coordinates @xmath71 in the discussion above , and @xmath68 is just a function of those coordinates . the static solutions of the skyrme model can be derived by varying the following energy @xcite : @xmath72[r_i , r_j ] \right ) \right ) { \rm d}^3 { \bf x}\ ] ] where @xmath73 is an @xmath74 valued current . a static solution of the variational equations could also be a saddle point . only solutions which minimize the energy are called skyrmions . in order to show that the energies ( [ egeo ] ) and ( [ estandard ] ) are equivalent , we first relate the strain tensor @xmath59 to the current @xmath75 . since @xmath68 is determined by @xmath69 we can calculate the induced metric @xmath14 on the target space @xmath12 in terms of the fields @xmath69 : @xmath76 starting from equation ( [ straintensor ] ) and noting that in flat space the dreibeins can be chosen to be kronecker deltas we obtain the following expression : @xmath77 equation ( [ d2 ] ) follows from ( [ d1 ] ) by using the chain rule and the formula for the metric ( [ metrictau ] ) . conceptually , the step from equation ( [ d2 ] ) to equation ( [ d3 ] ) is very important . the metric @xmath14 is expressed with the group multiplication and the trace which provides a scalar product in @xmath74 . this is the transition from geometric to lie group language . equation ( [ d4 ] ) is a trivial consequence of definition ( [ u ] ) , and the last equation follows from @xmath78 . now , we can show that the two energy densities are equal . since @xmath75 is an @xmath74 current , and therefore anti - hermitian , it can be expressed as @xmath79 . we start with the energy density in ( [ estandard ] ) and expand @xmath75 in terms of pauli matrices : @xmath80 r_i^{\mu^\prime } r_j^{\nu^\prime } [ i \tau_{\mu^\prime},i \tau_{\nu^\prime } ] \right ) \\ \nonumber \\ \label{ed2 } = r_i^\mu r_i^\mu + \frac{1}{4 } { \rm tr } \left(r_i^\mu r_j^\nu \epsilon_{\mu \nu \rho } \tau_\rho r_i^{\mu^\prime } r_j^{\nu^\prime } \epsilon_{\mu^\prime \nu^\prime \rho^\prime } \tau_{\rho^\prime } \right ) \\ \label{ed3 } = d_{i i } + \frac{1}{2 } \left(r_i^\mu r_i^\mu r_j^\nu r_j^\nu - r_i^\mu r_j^\mu r_i^\nu r_j^\nu \right ) \\ \nonumber \\ \label{ed4 } = d_{i i } + \frac{1}{2 } \left(d_{ii } d_{jj } - d_{ij } d_{ji } \right ) \\ \nonumber \\ \label{ed5 } = { \rm tr}~ d + \frac{1}{2 } \left ( { \rm tr}~ d \right)^2 - \frac{1}{2 } { \rm tr}~ d^2 . \end{aligned}\ ] ] equation ( [ ed2 ] ) follows from ( [ ed1 ] ) by inserting the commutation relations of the pauli matrices @xmath81 . in the following equation we use the contraction @xmath82 . in equation ( [ ed4 ] ) we apply the result ( [ d5 ] ) . the last equality follows from the definition of the trace . equation ( [ ed5 ] ) is the energy ( [ egeo ] ) because of ( [ deftrace ] ) . therefore , we have shown the equivalence of the two approaches for @xmath83 . it is worth noting that in the geometric picture the `` skyrme term '' is related to the square of an area element , whereas in the standard approach it depends on the structure constants of the lie algebra @xmath74 . both interpretations lead to generalizations . while in our geometric interpretation it is natural to consider different @xmath0-dimensional manifolds for @xmath11 , and maybe for @xmath12 , the lie algebra approach leads to generalizing the lie group @xmath4 for example to @xmath84 . in the previous section we have established the equivalence of the two energies ( [ egeo ] ) and ( [ estandard ] ) in @xmath9 . in this section we will use ( [ egeo ] ) to generalize the model to @xmath85 such that from now on physical space is a 3-sphere of radius @xmath2 . this includes the original model if we take the limit @xmath86 . we also fix @xmath87 . the skyrme model can not be solved analytically either in flat space , or on @xmath88 , apart from the case if @xmath54 and @xmath89 as mentioned above . however , in flat space there are analytic anstze which give good qualitative and quantitative agreement with exact solutions obtained numerically . by ansatz we mean a test function that minimizes the energy within a given class of test functions . the lower the energy the better we expect the ansatz to approximate the exact solution . in the following we will generalize the rational map ansatz @xcite , which has been very successful in flat space , to @xmath88 . a rational map is a holomorphic function from @xmath90 . treating each @xmath91 as a riemann sphere , with complex coordinates @xmath92 and @xmath93 , respectively , a rational map can be written as @xmath94 where @xmath95 and @xmath96 are polynomials in @xmath92 which are assumed to have no common factors . it has been shown by donaldson @xcite , and also by jarvis @xcite , that there is a one - to - one correspondence between rational maps and monopoles . in flat space it has been found that many solutions of the skyrme equation with baryon number @xmath5 look rather like monopoles with monopole number equal to @xmath5 . therefore , rational maps can be used to approximate skyrmions . a point @xmath15 on @xmath88 is labelled by polar coordinates @xmath97 such that @xmath98 where @xmath99 , @xmath100 $ ] , and @xmath101 $ ] and @xmath102 is the unit vector on @xmath91 . the @xmath0-sphere can be thought of as a collection of @xmath103-spheres with varying radius equal to @xmath104 . with the stereographic projection @xmath105 the @xmath91 can be identified with a riemann sphere using a single complex coordinate @xmath92 . alternatively , we can express the unit vector @xmath106 in terms of @xmath92 : @xmath107 similarly , points in the target @xmath1 can be labelled by @xmath108 where @xmath109 is an angular variable analogous to @xmath99 , and @xmath93 is a complex coordinate . the rational map ansatz simply states that @xmath110 and @xmath111 . this is only consistent if @xmath112 vanishes where @xmath113 does , _ i.e. _ @xmath114 and @xmath115 . the integer @xmath116 is a topological invariant and can not be changed by deforming @xmath117 smoothly . in order to have a good limit for @xmath8 we fix @xmath117 such that @xmath118 and set @xmath119 . in analogy to flat space we define a skyrmion to have @xmath120 , whereas for an anti - skyrmion @xmath121 . since skyrmions and anti - skyrmions are related by reflection we will only consider skyrmions from now on . therefore , our complete boundary conditions are : @xmath122 in contrast to the flat case these boundary conditions do not follow from a regularity argument . they are an artifact of our ansatz and have to be handled with care ( see @xcite for a discussion ) . if we now use the notation of equation ( [ u ] ) we can write the skyrme field in the following way : @xmath123 where @xmath124 is as in equation ( [ normalv ] ) . equation ( [ urat2 ] ) looks quite similar to the well known spherically symmetric hedgehog ansatz @xcite . in fact , the hedgehog ansatz corresponds to the special case @xmath125 . now , we can apply the formulae of the previous section . the rational map ansatz gives rise to the following eigenvalues @xmath58 of the strain tensor ( [ straintensor ] ) : @xmath126 the minus sign in the expression for @xmath127 is a consequence of our definition of positive baryon number in ( [ bc ] ) . one advantage of the rational map ansatz is the decoupling of the radial and the angular strains . starting from formula ( [ bgeo ] ) the baryon number @xmath5 can be written as a product of two integrals , one over @xmath99 and one over @xmath92 and @xmath128 in the following way : @xmath129 the integral over @xmath99 is the integer @xmath130 . the second integral is the pull - back of the area form on the target sphere of the rational map @xmath131 . taking the normalization into account this is just the degree @xmath132 of the rational map @xmath131 . in fact , it can be shown that if @xmath95 has the degree @xmath133 and @xmath96 has the degree @xmath134 then @xmath135 . the energy can now be obtained from formula ( [ egeo ] ) : @xmath136 \frac{2i~{\rm d } z { \rm d } { \bar z } } { ( 1 + |z|^2)^2 } l^3 \sin^2 \mu~{\rm d}\mu.\end{aligned}\ ] ] similarly to the baryon density , the integration over @xmath92 and @xmath137 and the integration over @xmath99 factorizes . therefore , formula ( [ erational0 ] ) can be rewritten as @xmath138 where @xmath132 is the degree of the rational map , and @xmath139 is the following special function on the space of rational maps : @xmath140 to minimize @xmath141 for a given baryon number @xmath142 , one should first minimize @xmath139 over the space of rational maps of degree @xmath132 . this calculation was performed in @xcite , and the result is displayed in table [ tablea1 ] in the appendix . then the profile function @xmath117 is found by solving the following euler - lagrange equation : @xmath143 where @xmath139 now takes the constant value in table [ tablea1 ] . we require @xmath117 to be a solution of ( [ euler ] ) non - singular in @xmath144 $ ] . equation ( [ euler ] ) has to be solved numerically . it has regular singular points at the end points , _ i.e. _ close to these points the solution has the following power law behaviour : @xmath145 for @xmath146 and @xmath147 for @xmath148 . here @xmath149 and @xmath150 and @xmath151 are arbitrary constants . the solution @xmath117 is regular if the exponent is equal to @xmath152 at both end points . equation ( [ erational ] ) has an important discrete symmetry . the transformation @xmath153 transforms a solution of ( [ euler ] ) into a solution which is also compatible with the boundary conditions ( [ bc ] ) . geometrically , @xmath154 is a reflection at the plane through the equator in physical space , whereas @xmath155 is a reflection in target space . this means that a solution which is localized for example at the south pole @xmath156 is transformed to a solution which is localized at the north pole @xmath157 . therefore , for fixed @xmath5 there are two degenerate solutions unless the transformation ( [ fsymmetry ] ) is a symmetry of @xmath117 in which case there is only one symmetric solution . the symmetry ( [ fsymmetry ] ) does not have an analogue in flat space . also note that the transformations @xmath158 and @xmath159 take a skyrmion with baryon number @xmath5 into an anti - skyrmion with baryon number @xmath160 . when we derived the energy of the rational map ansatz we used equation ( [ egeo ] ) . this could be transformed into equation ( [ egeo2 ] ) by `` completing the square '' . having now calculated the integral over the rational map , we can complete the square in a different way and re - express ( [ erational ] ) as @xmath161 { \rm d } \mu \\ \nonumber \\ \label{erat } - 8 \pi(2 n_r+\sqrt{{\cal i } } ) \int_0^\pi f^\prime \sin^2 f~ { \rm d } \mu . \end{aligned}\ ] ] the second integral in ( [ erat ] ) can be evaluated using @xmath162 and @xmath163 . since the first integral in ( [ erat ] ) is positive the second integral provides us with the energy bound @xmath164 the last inequality is valid because @xmath165 which can easily be shown by applying the schwartz inequality . therefore , inequality ( [ bograt ] ) `` improves '' the faddeev - bogomolny bound ( [ bogbound ] ) where @xmath166 and @xmath167 . this rational map bound ( [ bograt ] ) is valid , if the fields obey the rational map ansatz . exact solutions satisfy neither the rational map ansatz nor necessarily the bound ( [ bograt ] ) . we will discuss this bound further in section [ numericalresults ] . before ending this section we review one further ansatz . the symmetry group of @xmath88 , which is @xmath168 , contains an @xmath169 subgroup . et al _ used this symmetry to obtain doubly axially - symmetric anstze for skyrmions of various baryon numbers @xmath5 @xcite . in order to make best use of the symmetry it is convenient to parameterize the @xmath0-sphere with a different set of angles @xmath170 . with these coordinates a point on @xmath88 can be written as @xmath171 where @xmath172 $ ] and @xmath173 $ ] . now we can write an @xmath169 symmetric field configuration @xmath174 as @xmath175 where @xmath176 and @xmath177 are integers and @xmath178 is a shape function . in this ansatz the eigenvalues @xmath46 of the strain tensor ( [ straintensor ] ) are @xmath179 using equation ( [ egeo ] ) we obtain the energy @xmath180 and the baryon number @xmath181 the shape function @xmath182 can be calculated numerically by minimizing the energy @xmath141 subject to the boundary conditions @xmath183 and @xmath184 for various @xmath176 , @xmath177 and @xmath2 . note that these boundary conditions for @xmath182 imply that @xmath185 . in @xcite the energy for a given baryon number @xmath5 was also minimized with respect to the radius @xmath2 . some of the results are displayed in figure [ fig5 ] of section [ numericalresults ] . it was shown that the solutions are only stable as long as @xmath2 is smaller than a @xmath5-dependent critical length @xmath186 . moreover , for large baryon number @xmath5 the minimal energy for the optimal radius @xmath2 is larger than the energy of @xmath5 well separated skyrmions in @xmath9 . therefore , in this situation the ansatz fails badly . however , for small baryon number @xmath5 and small radius @xmath2 the ansatz is quite successful . for @xmath89 it agrees with the known @xmath168 symmetric solution . for @xmath187 it predicts an @xmath169 symmetric solution with a very low energy . given that the exact @xmath187 solution in flat space possesses an @xmath188 symmetry this configuration is likely to be the exact solution on @xmath1 . one particular feature of configurations of the form ( [ o(2)ansatz ] ) is that the following order parameter @xmath189 vanishes : @xmath190 where @xmath191 means the average over the physical @xmath88 . it also turns out that configurations with @xmath192 attain their minimal energy at particularly small values of @xmath2 . moreover , for @xmath192 the solution has a symmetry similar to ( [ fsymmetry ] ) . the transformation @xmath193 transforms solutions of the euler - lagrange equation for the energy ( [ eo(2 ) ] ) into each other . we will need these properties in section [ phasetransitions ] . we discuss next the shape function @xmath117 of the rational map ansatz . since solving equation ( [ euler ] ) numerically provides little insight we derive an analytic shape function . this ansatz approximates the numerical shape function fairly accurately and also confirms geometric ideas . in @xcite manton approximated the shape function of the @xmath89 skyrmion by a conformal map . with this ansatz he showed that a delocalized @xmath89 skyrmion on a @xmath0-sphere is unstable for @xmath194 . in this section we will generalize this idea for higher baryon number @xmath5 . using the rational map ansatz we will derive an ansatz for the shape function which is conformal in an average sense and which we will call quasi - conformal . in the following section , we will show that this quasi - conformal ansatz is a good approximation to the numerically calculated shape function . a skyrmion is a map from physical space @xmath88 labelled by @xmath195 to a target @xmath196 labelled by @xmath108 . a map between two @xmath0-spheres is conformal if their metrics only differ by a conformal factor : @xmath197 we are interested in fields which obey the rational map ansatz , _ i.e. _ @xmath110 and @xmath198 . therefore , we make the following approximations to equation ( [ econformal ] ) . firstly , since @xmath117 is a function of @xmath99 only , we consider a conformal factor @xmath199 which is also only a function of @xmath99 . secondly , we recall that according to equation ( [ brational ] ) the integral over the target @xmath91 is just @xmath200 times the degree @xmath132 of the rational map . therefore , we replace the metric on the target @xmath91 by @xmath132 times the metric on the physical @xmath91 . locally , this is not a very good approximation , but averaged over the whole @xmath91 this reproduces the correct result . since @xmath117 is independent of @xmath92 and @xmath128 it can only detect the averaged value . with these approximations equation ( [ econformal ] ) can be simplified : @xmath201 eliminating @xmath202 from equation ( [ quasiconf ] ) we obtain the following differential equation for @xmath117 : @xmath203 we are interested in solutions which obey the boundary conditions ( [ bc ] ) . it is convenient to replace @xmath112 by @xmath204 before taking the square root : @xmath205 for the negative sign the solution of equation ( [ confeq2 ] ) diverges at the boundary and can therefore be discarded . equation ( [ confeq2 ] ) is solved by separation of variables , and we obtain @xmath206 where @xmath207 is a positive integration constant . a negative @xmath207 would lead to a negative baryon number and is incompatible with the boundary conditions ( [ bc ] ) whereas @xmath208 is just the trivial solution with baryon number @xmath209 . the quasi - conformal shape function ( [ confansatz ] ) satisfies our boundary conditions ( [ bc ] ) if and only if @xmath210 . with equation ( [ b = nfnr ] ) we obtain @xmath211 . the shape function ( [ confansatz ] ) has the following important property . using the identity @xmath212 it is easy to show that for @xmath213 the shape function ( [ confansatz ] ) is invariant under the discrete symmetry ( [ fsymmetry ] ) : @xmath214 this means that for @xmath213 the solution is neither localized at the south pole nor at the north pole . for @xmath215 the solution is delocalized over the whole sphere whereas for @xmath216 it is rather localized at the equator @xmath217 . by taylor expanding equation ( [ confansatz ] ) near @xmath218 we find that the shape function @xmath117 behaves like @xmath219 . recall that for the exact solution of the euler - lagrange equation ( [ euler ] ) we obtained @xmath220 where @xmath221 . so at @xmath222 the shape function ( [ confansatz ] ) and the exact solution have a similar behaviour . the same is true for the other boundary @xmath156 . now , we can substitute the shape function ( [ confansatz ] ) into the equation for the energy ( [ erational ] ) and obtain @xmath223 where @xmath224 and @xmath225 in ( [ eqi1 ] ) and ( [ eqi2 ] ) we have simplified the integrand by substituting @xmath226 . the integrals @xmath227 and @xmath228 can be evaluated in closed form if @xmath229 is an integer . therefore , we will concentrate on @xmath215 , @xmath230 and @xmath231 in the remainder of this section . figure [ fig1 ] shows the shape function ( [ confansatz ] ) and its derivative for @xmath232 and for @xmath233 . figure [ fig1 ] also shows energy density @xmath234 averaged over the @xmath103-sphere , which is the integrand of ( [ erational ] ) . the shape function ( [ confansatz ] ) has symmetry ( [ fsym ] ) for @xmath213 . for @xmath235 the shape function is localized around the north pole @xmath222 , whereas for @xmath236 it is localized around the south pole @xmath237 . this is particularly obvious for its derivative @xmath238 . figure [ fig1 ] also illustrates that for higher baryon number @xmath5 the symmetric solution becomes more and more localized at the equator @xmath217 . furthermore , the energy @xmath234 has been plotted in order to compare the result to the usual skyrme model . @xmath234 corresponds to an average radial density . when the energy is peaked close to @xmath218 this is the usual skyrmion . if the energy is centered around @xmath217 this corresponds to a shell - like configuration . solutions centered around @xmath156 will correspond in the limit @xmath239 to a configuration centered around infinity , with an infinite energy . in order to calculate @xmath234 in figure [ fig1 ] we have to know the values for @xmath139 and @xmath2 . as we shall see in the following , for a given @xmath5 , @xmath240 minimizes the energy ( [ eansatz ] ) for a unique radius @xmath2 . @xmath213 is a solution for many different @xmath2 , but among these solutions @xmath186 takes a special place . for @xmath89 , @xmath234 has been calculated by setting @xmath241 , and @xmath242 corresponding to @xmath243 and @xmath244 . for @xmath213 we have displayed @xmath234 for the critical length @xmath245 . similarly , for @xmath246 , @xmath247 corresponds to @xmath248 and @xmath244 , and we have also displayed @xmath234 for @xmath249 . @xmath139 takes the value @xmath250 of table [ tablea1 ] in the appendix . we will now discuss the energy ( [ eansatz ] ) as a function of @xmath2 . for @xmath89 , and hence @xmath251 , the rational map is the identity map @xmath125 . the integral ( [ cali ] ) can be evaluated explicitly . we obtain @xmath252 so that the energy ( [ eansatz ] ) is @xmath253 it is convenient to set @xmath254 and to calculate the minimum of the energy with respect to @xmath255 . @xmath141 has a single minimum at @xmath256 . for @xmath257 , @xmath258 is not attainable for real @xmath207 , so that the minimum energy occurs at @xmath259 and @xmath213 . in fact , up to a minus sign this is the identity map between the physical @xmath88 and the target @xmath260 . the energy @xmath141 can now be written as @xmath261 for @xmath262 the minimum occurs where @xmath263 equation ( [ quadratic ] ) has two solutions . they are related by the symmetry ( [ fsymmetry ] ) and correspond to solutions localized at one of the poles . their energy is @xmath264 which is lower than ( [ e1k1 ] ) . in the limit @xmath8 the energy becomes @xmath265 , and the shape function turns into @xmath266 where @xmath267 . figure [ fig2 ] shows the energy @xmath141 of the skyrmion as a function of @xmath2 . for small @xmath2 the energy scales like @xmath57 . at the optimal value of the energy @xmath268 the energy reaches its minimum @xmath269 . the skyrmion is symmetric if the radius @xmath2 is below the critical radius @xmath270 . for @xmath271 there are two skyrmions with the same energy . this phenomenon is often called bifurcation . there is still a symmetric solution but it is no longer stable . for @xmath8 the energy tends to the @xmath9 value . for @xmath272 the energy @xmath141 shows similar behaviour . the values @xmath273 , @xmath186 and @xmath274 depend on the baryon number @xmath5 and are good numbers to characterize the behaviour of the solutions . in units of @xmath275 of a @xmath89 skyrmion as a function of the radius @xmath2.[fig2],height=453 ] now , we consider @xmath276 . in this case the optimal rational map has octahedral symmetry and can be written in the following form : @xmath277 we obtain @xmath278 . using expression ( [ eansatz ] ) the energy is @xmath279 where @xmath280 plays the same role as @xmath255 for @xmath89 . we can now minimize the energy with respect to @xmath281 . this gives rise to the cubic equation @xmath282 where @xmath283 and @xmath284 . there is a real solution @xmath285 for all @xmath2 but its exact form is long and not very illuminating . for @xmath286 we have @xmath287 so that @xmath288 and @xmath289 is again the minimum , and the solution is localized around the equator . the critical length can be calculated by substituting @xmath290 into equation ( [ eqcubic ] ) , and we obtain @xmath291 for @xmath213 the energy simplifies : @xmath292 the optimal length can be calculated by minimizing @xmath141 with respect to @xmath2 : @xmath293 for @xmath294 there are two solution such that @xmath295 related by the symmetry ( [ fsymmetry ] ) . the corresponding expression for the energy can be derived explicitly by setting @xmath296 in equation ( [ eb4a ] ) but it is rather lengthy . the limit @xmath8 can be derived from equation ( [ eqcubic ] ) . we are interested in the solution where @xmath281 is of order @xmath2 such that the energy @xmath141 in equation ( [ eb4a ] ) is finite . therefore , we can neglect the last term in ( [ eqcubic ] ) , and the @xmath297 , and obtain @xmath298 . inserting @xmath281 into the expression for the energy ( [ eb4a ] ) we obtain @xmath299 which corresponds to @xmath300 per baryon . for @xmath301 the optimal rational map has @xmath302 symmetry and @xmath303 . it is again possible to introduce an auxiliary variable @xmath304 . however , minimizing the energy leads to a polynomial of degree @xmath305 in @xmath306 . therefore it is no longer possible to calculate the solution explicitly . but the general behaviour remains the same . the solution is symmetric below a critical length @xmath186 which is determined by substituting @xmath307 into the equation for minimizing the energy . we obtain @xmath308 the optimal radius can also be calculated analytically : @xmath309 for @xmath294 there are again two degenerate solutions . in the limit @xmath8 we can balance the terms which are of highest order in @xmath2 . this leads to @xmath310 , and the normalized energy is @xmath311 which is @xmath312 per baryon . in the last part of this section we will discuss the limit @xmath86 for general @xmath132 . setting @xmath267 the quasi - conformal ansatz ( [ confansatz ] ) tends to @xmath313 where @xmath314 is a free parameter . the energy can be written as @xmath315 however , now the integrals @xmath316 and @xmath317 are independent of @xmath314 and depend only on @xmath132 : @xmath318 it is straightforward to minimize the energy ( [ er0 ] ) with respect to @xmath314 . in figure [ fig6 ] and [ fig7 ] , in the following section we have evaluated the minimal energy @xmath319 , and the optimal radius @xmath320 , respectively . the parameter @xmath320 has a natural interpretation as the size of the skyrmion : for @xmath321 the shape function @xmath322 has reached the value @xmath323 . furthermore , when the baryon density is integrated over a ball of radius @xmath320 , then this is just half the total baryon number : @xmath324 it is also worth noting that equations ( [ confansatz ] ) for @xmath213 and ( [ flatansatz ] ) are related by @xmath325 . this is a stereographic projection from a @xmath0-sphere of radius @xmath320 to @xmath9 . minimizing the energy we see that a skyrmion in flat space is related to a skyrmion on the @xmath0-sphere with an optimal radius @xmath273 . in this section we discuss the numerical solution of equation ( [ euler ] ) for the shape function @xmath117 of the rational map ansatz and the properties of the resulting fields , and we compare it to the analytical solutions from the previous section . here , we only consider configurations with @xmath326 such that @xmath327 . a configuration with @xmath328 is discussed in the following section . the numerical solution of ( [ euler ] ) is calculated using a relaxation method . for symmetric initial conditions we obtain the symmetric solution which is a saddle point for @xmath271 . for asymmetric initial conditions we obtain the minimum energy configuration , _ i.e. _ the skyrmion . in figure [ fig3 ] we compare the numerical shape function with the quasi - conformal ansatz ( [ confansatz ] ) for @xmath329 and @xmath330 at the optimal radius @xmath331 . the optimal radius @xmath273 implies @xmath213 in the quasi - conformal ansatz . the numerical result and the quasi - conformal ansatz show good agreement , in particular in the region around the equator @xmath217 . this region is most important for the energy ( [ erational ] ) since both @xmath332 and @xmath333 are large for @xmath334 . near the poles the numerical solution is less steep than the quasi - conformal ansatz as we expect from their exponents : @xmath335 . and @xmath330 . [ fig3],height=453 ] in figure [ fig4 ] we show the energy @xmath141 as a function of @xmath2 for @xmath336 and compare it to our analytic calculations of the previous section . in all the figures the energy @xmath141 is normalized by @xmath275 so that the faddeev - bogomolny bound ( [ bogbound ] ) is equal to @xmath337 . for @xmath89 we plot energy ( [ e1k1 ] ) of the symmetric solution which gives the exact result . for @xmath338 we also plot the energy ( [ e1k2 ] ) . our ansatz predicts the correct critical radius but energy ( [ e1k2 ] ) is slightly too high . for @xmath339 and @xmath340 we only display the energies of the symmetric solutions using formula ( [ eansatz ] ) and setting @xmath213 . the integrals @xmath341 and @xmath342 in ( [ eansatz ] ) are calculated numerically . again there is very good agreement for small @xmath2 , and for @xmath294 the ansatz is also very close to the symmetric solution . for @xmath246 we plot the analytic expression for the symmetric solution ( [ e4k2 ] ) . for @xmath343 we also plot ( [ eb4a ] ) with @xmath296 . the critical radius @xmath344 is slightly higher than @xmath345 , and the energy is also slightly too high . in figure [ fig4 ] we also mark the minimal energy @xmath346 for the optimal radius @xmath347 and for comparison the optimal energy @xmath348 for the doubly axially - symmetric ansatz at its optimal radius @xmath349 . we will discuss these quantities in figure [ fig5 ] . in summary , all the analytical results are in good agreement with the numerical solution for a large range of @xmath2 . this is quite remarkable given that the ansatz for the shape function only depends on one parameter @xmath207 . as a function of the radius @xmath2 for baryon number @xmath350.[fig4],height=604 ] in figure [ fig5 ] we compare the optimal energy of the quasi - conformal map @xmath351 with the energy of the doubly axially - symmetric ansatz @xmath348 for all baryon numbers @xmath352 . for @xmath89 both reproduce the exact result . for @xmath187 the doubly axially - symmetric ansatz is believed to be the exact solution because its symmetry is compatible with the results in flat space . yet , for @xmath353 the energy @xmath351 is lower than @xmath348 . , @xmath348 , @xmath354 and @xmath355 per baryon as a function of the baryon number @xmath5.[fig5],height=453 ] figure [ fig5 ] also shows that the quasi - conformal ansatz always has a lower energy than the numerical solution in @xmath9 . we conclude that @xmath273 for the exact solution will be finite . the horizontal line at @xmath356 in figure [ fig5 ] corresponds to @xmath5 well separated single skyrmions . for @xmath357 the energy of the doubly axially - symmetric ansatz is above this line which is unphysical . note that the energy per baryon in flat space @xmath354 decreases as we increase the baryon number . we would obtain a similar behaviour if we kept the radius @xmath2 fixed . yet , we then have to make the radius @xmath2 large enough so that the largest skyrmion `` fits '' . the physical interpretation is that one multi - skyrmion is more stable than a number of smaller skyrmions . we also display the rational map bound @xmath355 ( [ bograt ] ) . for @xmath187 the energy of the doubly axially - symmetric ansatz is lower than @xmath355 . this confirms our warning that energies of exact solutions do not have to be greater than @xmath355 . however , figure [ fig5 ] shows that the bound follows the general behaviour of @xmath351 and also of @xmath354 . in figure [ fig6 ] we compare the energy of the quasi - conformal map @xmath351 in the limit @xmath8 to the rational map energies @xmath346 and the energy @xmath354 of the exact solution in flat space . @xmath351 is calculated by minimizing the energy of expression ( [ er0 ] ) with respect to @xmath314 . for @xmath358 the energy @xmath351 agrees with the results of section [ conformalmap ] . and the energies @xmath351 and @xmath346 per baryon for @xmath8 as a function of @xmath5.[fig6],height=453 ] figure [ fig6 ] shows that the quasi - conformal ansatz approximates the numerical solution well . the relative difference between @xmath351 and @xmath346 is monotonically decreasing and is less than @xmath359 for @xmath329 . this means that the error of the rational map ansatz with a numerical shape function and the quasi - conformal map ansatz are of the same order of magnitude . , @xmath360 , @xmath186 , and @xmath320 as a function of the baryon number @xmath5.[fig7],height=453 ] in figure [ fig7 ] we compare the different optimal radii . the optimal radius of the quasi - conformal ansatz @xmath360 is always slightly smaller than @xmath347 . in all cases the value of the critical radius @xmath186 is greater than the optimal radius so that the results are consistent . the critical radius can be calculated in two different ways . in the first approach we determine the point at which the solution bifurcates . we use the fact that the symmetric initial condition relaxes into the symmetric solution which is always present , whereas an asymmetric initial condition only relaxes to the symmetric solution when there is no asymmetric solution . this method leads to rather large error bars because it is difficult to decide whether two numerically calculated solutions are different . the second approach is to consider the value of the following integral : @xmath361 this integral is zero if the solution possesses the symmetry ( [ fsymmetry ] ) and is nonzero otherwise . we determine the value of @xmath2 where the integral ( [ intf ] ) first becomes nonzero . both approaches give the same results , but the second one is much more accurate . at the critical value there is a jump in the value of the integral ( [ intf ] ) . this is also a verification of our analytic result that the solution is symmetric for @xmath362 . as mentioned before , for @xmath89 the critical radius is in agreement with the analytic result @xmath363 . for @xmath246 the critical value of the quasi - conformal ansatz @xmath364 in ( [ b4lcrit ] ) is only slightly higher than the critical radius of the rational map @xmath365 . similarly , for @xmath330 we calculated @xmath366 in ( [ b9lcrit ] ) whereas @xmath367 . in figure [ fig7 ] we also find that @xmath320 is always smaller than @xmath360 . the difference between these lengths becomes smaller the larger the baryon number is . this confirms our interpretation stated at the end of the previous section , that the skyrmion in flat space is related to the skyrmion on a @xmath0-sphere with optimal radius . in this section we will describe phase transitions on @xmath1 as @xmath2 is varied . there is no numerical work solving the full set of coupled partial differential equations for @xmath88 as there is in flat space . therefore , we have to rely on our anstze . first , we will discuss two different order parameters . then we will describe particular situations in more detail . in @xmath9 the average @xmath368 is equal to @xmath369 because of the boundary conditions at infinity . however , on @xmath88 it is easy to show that @xmath370 vanishes for @xmath362 when the solution is symmetric with respect to south and north pole . therefore , we can choose @xmath371 as an order parameter . , @xcite . ] it has the necessary property that it is nonzero above the phase transition @xmath271 and identically zero below the transition . as we already pointed out in section [ rationalmaps ] this order parameter also vanishes identically for all doubly axially - symmetric configurations . this is consistent with the fact that for @xmath294 these configurations are bad approximations . another possible order parameter would be @xmath372 . this parameter @xmath373 is motivated by the fact that enhanced chiral symmetry , that is a symmetry that mixes @xmath68 and @xmath71 , could result in a vanishing of @xmath373 . for @xmath89 the situation is well understood . both parameters @xmath189 and @xmath373 vanish at the same time . physically , this means that the localization delocalization transition occurs at the same time as the restoration of chiral symmetry @xcite . in fact , for @xmath374 the solution possesses full chiral symmetry @xmath375 . for @xmath272 the chiral symmetry will be a subgroup of @xmath375 . next we discuss @xmath329 and use both the rational map ansatz and the doubly axially - symmetric ansatz . in figure [ fig8a ] we plot their energies . for @xmath271 the skyrmion is localized at one of the poles . for smaller @xmath2 the skyrmion is localized around the equator . this is the first phase transition and the order parameter is @xmath189 . however , there is a radius @xmath376 where the doubly axially - symmetric ansatz with @xmath377 and @xmath378 becomes the minimal energy solution . as pointed out in section [ rationalmaps ] if @xmath192 the solutions have the symmetry ( [ gsymmetry ] ) . for this special case we checked numerically that the symmetry is actually a symmetry of the shape function @xmath182 : @xmath379 this gives rise to a symmetry between @xmath68 and @xmath380 and leads to the vanishing of the order parameter @xmath373 . therefore , we have a second phase transition at @xmath381 . as a last example we describe a skyrmion with baryon number @xmath382 . for large @xmath2 the @xmath326 solution has minimal energy . however , there is a saddle point solution where two @xmath329 skyrmions are located at opposite poles of the @xmath0-sphere . if @xmath2 is sufficiently small there is a solution with @xmath383 which has lower energy than the one with @xmath326 as shown in figure [ fig8b ] . in the appendix gives an indication when to expect a configuration with @xmath384 to be more stable than the corresponding configuration with @xmath385 . ] this configuration is particularly interesting because both @xmath329 skyrmions have cubic symmetry . for sufficiently small radius @xmath386 these two skyrmions could combine and form a solution with a hypercubic symmetry . this symmetry would be strong enough to restore the symmetry between the @xmath68- and @xmath71-fields . the order parameter @xmath373 would vanish . in this scenario there would be a localization delocalization transition at @xmath186 , whereas the chiral symmetry is further enhanced at @xmath386 . unfortunately , these discrete subgroups of the full @xmath375 symmetry can not be studied with the rational map ansatz . in physical terms we propose the following picture . reducing the radius @xmath2 increases the baryon density . if we start with a @xmath387 skyrmion at large length this is well described by the rational map ansatz . there is a phase transition at the critical radius @xmath186 where the skyrmion becomes delocalized over the @xmath0-sphere and the order parameter @xmath189 vanishes . then , one or more phase transitions follow which make the configuration more and more symmetric , and chiral symmetry is partially restored . note however , the skyrme model is an effective theory which is only valid for low energies . if the radius is too small we expect that further terms in the effective action become physically important . in this paper we have described the rational map ansatz applied to skyrmions on a @xmath0-sphere of radius @xmath2 . we have calculated the energy @xmath141 of a skyrmion as a function of @xmath2 for baryon number @xmath352 and found the following behaviour : for small @xmath2 the energy @xmath141 scales like @xmath57 . @xmath141 has a global minimum at an optimal radius @xmath273 . when the radius @xmath2 is further increased there is a bifurcation point at a critical length @xmath186 : below @xmath186 the skyrmion is symmetric under reflection at the plane through the equator of @xmath1 . above @xmath186 there are two degenerate skyrmions one of which is localized at the north pole whereas the other one is localized at the south pole . for @xmath8 the ansatz tends to the rational map ansatz for @xmath9 , @xcite . for @xmath89 and @xmath54 our ansatz reproduces the known exact solution @xcite . for @xmath187 our results are worse for small @xmath2 than the doubly axially - symmetric ansatz in @xcite . however , for @xmath388 the energies of our solutions are lower than any solutions known to date , including the doubly axially - symmetric solutions . we have also derived an analytic expression for the shape function by imposing that the metrics on the physical @xmath1 and on the target @xmath1 are conformal in an average sense . this quasi - conformal ansatz for the shape function depends only on one parameter @xmath207 . by varying @xmath207 we obtain very good agreement with the numerical results for a large range of @xmath2 , even for flat space . it is worth noting that the relative error of the quasi - conformal ansatz decreases with increasing baryon number . we have shown that the solution is localized at the equator if @xmath213 . for @xmath389 we have shown that the skyrmion really is symmetric below a critical value @xmath390 , _ i.e. _ @xmath213 gives the minimal energy solution . one particular property of the quasi - conformal shape function is that it becomes more and more localized around the equator as the baryon number increases . to be more precise , the derivative of the shape function , which is connected to the baryon density , has a peak at the equator . since this ansatz was derived with the assumption that the map between the metrics is `` as conformal as possible '' we have found a geometric explanation of why skyrmions on the @xmath0-sphere are shell - like . in flat space the map can not be conformal . however , one `` half '' of the skyrmion in flat space resembles `` half '' a skyrmion on a @xmath0-sphere with optimal radius . this fits well with the observation in section [ numericalresults ] that the parameter @xmath320 which is a measure of the size of the skyrmion agrees well with the optimal length @xmath347 . this line of thought can be carried even further . skyrmions on the @xmath0-sphere might be a reasonable model for nuclei , once they are quantized . the main advantage of this model is that on the @xmath0-sphere one - loop corrections are expected to play a far less important role than in flat space . in particular , if the skyrmion lives on the @xmath0-sphere with the optimal radius we expect these corrections to be so small that the predictions of the skyrme model could be compared with experiment . the main motivation for this claim are the promising results in @xcite for the @xmath89 skyrmion on the @xmath0-sphere , and the fact that they predict a different value for the skyrme coupling which is in agreement with one - loop calculations performed in @xcite and the skyrme coupling therein . the author is grateful to n s manton for suggesting the problem , and also wants to thank the participants of the damtp soliton seminar for fruitful discussions . the author thanks pparc for a research studentship and the studienstiftung des deutschen volkes for a phd scholarship . table [ tablea1 ] shows the minimal value of the integral @xmath139 in equation ( [ cali ] ) as a function of @xmath132 , as has been calculated in @xcite . `` approx '' is the energy of the rational map ansatz for flat space skyrmions where the corresponding variational equation for the shape function has been solved numerically . the column `` true '' shows the value of the energy obtained by a numerical solution of the full set of partial differential equations . the next column `` sym '' shows the symmetries of the skyrmions . the numerical solutions and the rational map ansatz give the same symmetry . note that the respective values for @xmath330 in table [ tablea1 ] are taken from @xcite . there is evidence that for @xmath330 the symmetry and the other respective values in the table are different from the ones cited in @xcite . the last two columns are the optimal length @xmath273 and the optimal energy @xmath393 which are displayed in figure [ fig7 ] and [ fig6 ] , respectively . as described in chapter [ numericalresults ] , @xmath393 has been obtained by using the rational map of table [ tablea1 ] and calculating the shape function numerically . following @xcite we also include the values for @xmath394 . in this case the solution in flat space has marginally lower energy than @xmath346 . however , we still expect that the rational map ansatz is a good approximation even in this situation . .values of @xmath139 , the energies in flat space ( approx , true ) , the symmetry ( sym ) , and the optimal length @xmath273 and energy @xmath346 on @xmath1 . all values are given for baryon number @xmath327 for @xmath352 and @xmath395.[tablea1 ] [ cols="^,^,^,^,^,^,^ " , ] table [ tablea2 ] can be used to estimate which combination of @xmath130 and @xmath132 might be more stable for a given baryon number @xmath5 . it turns out that for @xmath382 the bound is lower for @xmath383 than for @xmath326 . it is also probable that the @xmath396 skyrmion has lower energy than the @xmath397 skyrmion . abcdefgh t h r skyrme 1961 a nonlinear field theory _ proc . soc . _ * a260 * 127 . g adkins , c nappi and e witten 1983 static properties of nucleons in the skyrme model _ nucl . phys . _ * b228 * 552 . f meier and h walliser 1997 quantum corrections to baryon properties in chiral soliton models _ phys . rep . _ * 289 * 383 . r a leese , n s manton and b j schroers 1995 attractive channel skyrmions and the deuteron _ nucl . phys . _ * b442 * 228 . p irwin 2000 zero mode quantization of multi - skyrmions _ phys . _ * d61 * 114024 . r a battye and p m sutcliffe 1997 symmetric skyrmions _ phys . lett . _ * 79 * 363 . c j houghton , n s manton and p m sutcliffe 1998 rational maps , monopoles and skyrmions _ nucl . phys . _ * b510 * 507 . s manton 1987 geometry of skyrmions _ commun . math . phys . _ * 111 * 469 . l castillejo , p s j jones , a d jackson , j j m verbaarshot , a jackson 1989 dense skyrmion systems _ nucl . * a501 * 801 . m kugler , s shtrikman 1988 a new skyrmion crystal _ phys.lett . _ * b208 * 491 . m kugler , s shtrikman 1989 skyrmion crystals and their symmetries _ phys . rev . _ * d40 * 3421 . a d jackson , n s manton and a wirzba 1989 new skyrmion solutions on a three sphere _ nucl . phys . _ * a495 * 499 . c j houghton and s krusch 2001 folding in the skyrme model _ j. math . phys . _ * 42 * 4079 . j eells and j j sampson 1964 harmonic mappings of riemannian manifolds _ am . _ * 86 * 109 . l d faddeev 1976 some comments on the many dimensional solitons _ lett . * 1 * 289 . n s manton and p j ruback 1986 skyrmions in flat space and curved space _ phys . * b181 * ( 1986 ) 137 . s k donaldson 1984 nahm s equations and the classification of monopoles _ commun . phys . _ * 96 * 387 . s jarvis 2000 a rational map for euclidean monopoles via radial scattering _ j. reine angew . math _ * 524 * 17 . m loss 1987 the skyrme model on riemannian manifolds _ lett . math . _ * 14 * 149 . s - t hong 1998 the static properties of hypersphere skyrmions _ phys . lett . _ * b417 * 211 . r a battye and p m sutcliffe 2000 skyrmions , fullerenes and rational maps _ rev . _ * 14 * 29 . [ lastref ]

we study instabilities occurring in the electron system whose fermi surface has flat regions on its opposite sides . such a fermi surface resembles fermi surfaces of some high- superconductors . in the framework of the parquet approximation , we classify possible instabilities and derive renormalization - group equations that determine the evolution of corresponding susceptibilities with decreasing temperature . numerical solutions of the parquet equations are found to be in qualitative agreement with a ladder approximation . for the repulsive hubbard interaction , the antiferromagnetic ( spin - density - wave ) instability dominates , but when the fermi surface is not perfectly flat , the-wave superconducting instability takes over .

cond-mat9609118

an interacting electron gas in one dimension has many unusual properties , such as the spin - charge separation , the power law of correlation functions , and the linear dependence of the electron relaxation rate on temperature and frequency ( see ref.@xcite for a review ) . these one - dimensional ( 1d ) results are well established , in many cases exactly , by applying a variety of mathematical methods including the bethe ansatz , the bosonization , and the parquet , or the renormalization group . to distinguish the exotic behavior of the 1d electron gas from a conventional fermi - liquid behavior , haldane introduced a concept of the so - called luttinger liquid @xcite . the discovery of high-@xmath0 superconductivity renewed interest in the luttinger - liquid concept . anderson suggested that a two - dimensional ( 2d ) electron gas behaves like the 1d luttinger liquid , rather than a conventional fermi liquid @xcite . it is difficult to verify this claim rigorously , because the methods that prove the existence of the luttinger liquid in 1d can not be applied directly to higher dimensions . the bethe ansatz construction does not work in higher dimensions . the bosonization in higher dimensions @xcite converts a system of interacting electrons into a set of harmonic oscillators representing the electron density modes . this procedure replaces the exact @xmath2 commutation relations @xcite with approximate boson commutators , which is a questionable , uncontrolled approximation . on the other hand , the parquet method , although not being as exact as the two other methods , has the advantage of being formulated as a certain selection rule within a standard many - body diagram technique ; thus , it can be applied to higher dimensions rather straightforwardly . the parquet method has much in common with the renormalization - group treatment of fermi liquids @xcite . the 1d electron gas has two types of potential instabilities : the superconducting and the density - wave , which manifest themselves through logarithmic divergences of the corresponding one - loop susceptibilities with decreasing temperature . within the parquet approach , a sum of an infinite series of diagrams , obtained by adding and inserting the two basic one - loop diagrams into each other , is calculated by solving a system of nonlinear differential equations , which are nothing but the renormalization - group equations @xcite . this procedure was developed for the first time for meson scattering @xcite and later was successfully applied to the 1d electron gas @xcite , as well as to the kondo problem @xcite and the x - ray absorption edge problem @xcite . by considering both the superconducting and the density - wave instabilities on equal footing and adequately treating their competition , the parquet approximation differs from a conventional ladder ( or mean - field ) approximation , commonly applied in higher dimensions , where only one instability is taken into account . under certain conditions in the 1d case , the superconducting and density - wave instabilities may cancel each other , giving rise to a non - trivial metallic ground state at zero temperature , namely the luttinger liquid . in this case , the parquet derivation shows that the electron correlation functions have a power - law structure , which is one of the characteristic properties of the luttinger liquid @xcite . one may conclude that the competition between the superconducting and density - wave instabilities is an important ingredient of the luttinger liquid theory . in a generic higher - dimensional case , where density - wave instability does not exist or does not couple to superconducting instability because of corrugation of the fermi surface , the parquet approach is not relevant . nevertheless , there are a number of higher - dimensional models where the parquet is applicable and produces nontrivial results . these include the models of multiple chains without single - electron hopping @xcite and with single - electron hopping but in a magnetic field @xcite , as well as the model of an isotropic electron gas in a strong magnetic field @xcite . in all of these models , the electron dispersion law is 1d , which permits to apply the parquet method ; at the same time , the interaction between electrons is higher - dimensional , which makes a nontrivial difference from the purely 1d case . the particular version of the parquet method used in these cases is sometimes called the `` fast '' parquet , because , in addition to a `` slow '' , renormalization - group variable , the parquet equations acquire supplementary , `` fast '' variables , which label multiple electron states of the same energy . taking into account these considerations , it seems natural to start exploring a possibility of the luttinger liquid behavior in higher dimensions by considering a model that combines 1d and higher - dimensional features . this is the model of an electron gas whose fermi surface has flat regions on its opposite sides . the flatness means that within these regions the electron dispersion law is 1d : the electron energy depends only on the one component of momentum that is normal to the flat section . on the other hand , the size of the flat regions is finite , and that property differentiates the model from a purely 1d model , where the size is infinite , since nothing depends on the momenta perpendicular to the direction of a 1d chain . a particular case of the considered model is one where the 2d fermi surface has a square shape . this model describes 2d electrons on a square lattice with the nearest - neighbor hopping at the half filling . it is a simplest model of the high-@xmath0 superconductors . the model has already attracted the attention of theorists . virosztek and ruvalds studied the `` nested fermi liquid '' problem within a ladder or mean - field approximation @xcite . taking into account the 1d experience , this approach may be considered questionable , because it does not treat properly the competition between the superconducting and the density - wave channels . houghton and marston @xcite mapped the flat parts of the fermi surface onto discrete points . such an oversimplification makes all scattering processes within the flat portion equivalent and artificially enhances the electron interaction . mattis @xcite and hlubina @xcite used the bosonization to treat the interaction between the electron density modes and claimed to solve the model exactly . however , mapping of the flat fermi surface onto quantum chains and subsequent bosonization by luther @xcite indicated that the treatment of mattis and hlubina is insufficient , because the operators of backward and umklapp scattering on different quantum chains require a consistent renormalization - group treatment . luther did not give solution to this problems , as well as he missed the interaction between the electrons located on four different quantum chains . in the present paper , we solve the model consistently , using the fast parquet approach , where all possible instabilities occurring in the electron system with the flat regions on the fermi surface are treated simultaneously . this approach was applied to the problem earlier @xcite in order to explain the antiferromagnetism of chromium . in the present paper , we advance the study further by including the order parameters of the odd symmetry , missed in @xcite , performing detailed numerical calculations , and investigating the effect of a curvature of the fermi surface . to simplify numerical calculations and to relate to the high-@xmath0 superconductors , we consider the 2d case , although the method can be straightforwardly generalized to higher dimensions as well . we find that the presence of the boundaries of the flat portions of the fermi surface has a dramatic effect on the solutions of the parquet equations . even if the initial vertex of interaction between electrons does not depend on the momenta along the fermi surface ( which are the `` fast '' variables ) , the vertex acquires a strong dependence on these variables upon renormalization , which greatly reduces the feedback coupling between the superconducting and density - wave channels relative to the 1d case . instead of the two channels canceling each other , the leading channel , which is the spin - density - wave ( sdw ) in the case of the repulsive hubbard interaction , develops its own phase transition , inducing on the way a considerable growth of the superconducting @xmath1-wave susceptibility . at the same time , the feedback from the superconducting to the sdw channel , very essential in the 1d case , is found negligible in the 2d case . these results are in qualitative agreement with the picture of the antiferromagnetically - induced @xmath1-wave superconductivity , which was developed within a ladder approximation for the flat fermi surface in @xcite and for a generic nested hubbard model in @xcite . recent experiments strongly suggest that the high-@xmath0 superconductivity is indeed of the @xmath1-wave type @xcite . on the other hand , our results disagree with refs.@xcite . the origin of the discrepancy is that the bosonization arbitrarily replaces the exact @xmath2 commutation relations @xcite by approximate boson commutators ; thus , the renormalization of the electron - electron interaction , which is an important part of the problem , becomes neglected . in addition to having the flat sides , the square fermi surface also has sharp corners , where the saddle points of the electron dispersion law , which produce the van hove singularity in the density of states , are located . the presence of the van hove singularity at the fermi level enhances the divergence of the superconducting and density - wave loops to the square of the temperature logarithm . the fast parquet problem was formulated in this case in ref . @xcite , where the contribution from the flat sides , being less divergent than the contribution from the saddle points , was neglected . the present paper completes the study by considering a fermi surface with the flat sides and rounded corners , that is , without saddle points at the fermi level . our physical conclusions for both models are in qualitative agreement . as photoemission experiments @xcite demonstrate ( see also @xcite ) , many of the high-@xmath0 superconductors indeed have flat regions on their fermi surfaces . hence , some of the results of this paper may be applicable to these materials . however , the primary goal of our study is to elucidate general theoretical concepts rather than to achieve detailed description of real materials . in order to distinguish the new features brought into the problems by introducing higher dimensions , we present material in an inductive manner . in sec . [ sec : spinless ] , we recall the derivation of the parquet equations in the simplest case of 1d spinless electrons . in sec . [ sec : spin1d ] , we generalize the procedure to the case of 1d electrons with spin @xcite . then , we derive the parquet equations in the 2d case in sec . [ sec:2d ] and solve them numerically in sec . [ sec : numerical ] . the paper ends with conclusions in sec . [ sec : conclusion ] . let us consider a 1d electron gas with a fermi energy @xmath3 and a generic dispersion law @xmath4 , where @xmath5 is the energy and @xmath6 is the momentum of the electrons . as shown in fig.[fig:1d ] , the fermi surface of this system consists of two points located at @xmath7 , where @xmath8 is the fermi momentum . assuming that the two points are well separated , let us treat the electrons whose momenta are close to @xmath9 as two independent species and label them with the index @xmath10 . in the vicinity of the fermi energy , the dispersion laws of these electrons can be linearized : @xmath11 where the momenta @xmath6 are counted from the respective fermi points @xmath9 for the two species of the electrons , @xmath12 are the corresponding fermi velocities , and the energy @xmath5 is counted from the chemical potential @xmath3 . first , let us consider the simplest case of electrons without spin . the bare hamiltonian of the interaction between the @xmath10 electrons , @xmath13 , can be written as @xmath14 where @xmath15 is the bare vertex of interaction , and the operators @xmath16 and @xmath17 create and destroy the @xmath10 electrons . the tendencies toward the superconducting or density - wave ( @xmath18 ) instabilities in the system are reflected by the logarithmic divergences of the two one - loop diagrams shown in fig.[fig : loops ] , where the solid and dashed lines represent the green functions @xmath19 and @xmath20 of the @xmath21 and @xmath22 electrons , respectively . the two diagrams in fig . [ fig : loops ] differ in the mutual orientation of the arrows in the loops . in the matsubara technique , the integration of the green functions over the internal momentum @xmath6 and energy @xmath23 produces the following expressions for the two diagrams : @xmath24 where the upper sign corresponds to the superconducting and the lower to the density - wave susceptibility . in eq . ( [ 1loop ] ) , @xmath25 is the temperature , @xmath26 is the external energy passing through the loop , and @xmath27 is the external momentum for the superconducting loop and the deviation from @xmath18 for the density - wave loop . with logarithmic accuracy , the value of the integral ( [ 1loop ] ) is determined by the upper and lower cutoffs of the logarithmic divergence . in eq . ( [ 1loop ] ) , the upper and lower cutoffs are written approximately , up to numerical coefficients of the order of unity , whose logarithms are small compared to @xmath28 . the variable @xmath29 , introduced by eq . ( [ 1loop ] ) , plays a very important role in the paper . since @xmath29 is the logarithm of the infrared cutoff , the increase of @xmath29 represents renormalization toward low temperature and energy . the two primary diagrams of fig . [ fig : loops ] generate higher - order corrections to the vertex of interaction between electrons , @xmath30 , as illustrated in fig . [ fig : sample ] . in this figure , the dots represent the bare interaction vertex @xmath15 , whereas the renormalized vertex @xmath30 is shown as a circle . the one - loop diagrams in fig . [ fig : sample ] are the same as in fig.[fig : loops ] . the first two two - loop diagrams in fig.[fig : sample ] are obtained by repeating the same loop twice in a ladder manner . the last two diagrams are obtained by inserting one loop into the other and represent coupling between the two channels . the diagrams obtained by repeatedly adding and inserting the two basic diagrams of fig . [ fig : loops ] in all possible ways are called the parquet diagrams . the ladder diagrams , where only the addition , but not the insertion of the loops is allowed , represent a subset of the more general set of the parquet diagrams . selection of the parquet diagrams is justified , because , as one can check calculating the diagrams in fig . [ fig : sample ] , they form a series with the expansion parameter @xmath31 : @xmath32 . if the bare interaction vertex @xmath15 is small and the temperature is sufficiently low , so that @xmath33 is big , one can argue @xcite that nonparquet diagrams may be neglected , because their expansion parameter @xmath15 is small compared to the parquet expansion parameter @xmath31 . every diagram in fig . [ fig : sample ] , except the bare vertex @xmath15 , can be divided into two disconnected pieces by cutting one solid and one dashed line , the arrows of the cut lines being either parallel or antiparallel . the sum of those diagrams where the arrows of the cut lines are parallel ( antiparallel ) is called the superconducting ( density - wave ) `` brick '' . thus , the vertex @xmath30 can be decomposed into the bare vertex @xmath15 , the superconducting brick @xmath34 , and the density - wave brick @xmath35 : @xmath36 eq . ( [ vertex : spinless ] ) is illustrated in fig.[fig : spinlessvertex ] , where the bricks are represented as rectangles whose long sides , one being a solid and another a dashed line , represent the lines to be cut . in a general case , the vertices and the bricks depend on the energies and momenta ( @xmath37 , and @xmath38 ) of all incoming and outgoing electrons . equations for the bricks can be found in closed form in the case where all their arguments are approximately equal within the logarithmic accuracy , that is , the ratios of the arguments and of their linear combinations are of the order of unity @xcite . practically , this means that all vertices and bricks are considered to be functions of the single renormalization - group variable @xmath29 , defined in eq.([1loop ] ) . it was proved in @xcite that the two pieces obtained by cutting a brick are the full vertices of interaction , as illustrated graphically in fig . [ fig : spinlessbricks ] . analytically , the equations for the bricks are [ integral ] @xmath39 the two vertices @xmath30 in the r.h.s . of eqs . ( [ integral ] ) represent the two pieces obtained from a brick by cutting , whereas the integrals over @xmath40 represent the two connecting green functions being integrated over the internal momentum and energy of the loop . the value of the renormalized vertex @xmath41 changes as the integration over @xmath40 progresses in eqs . ( [ integral ] ) . in agreement with the standard rules of the diagram technique @xcite , a pair of the parallel ( antiparallel ) lines in fig.[fig : spinlessbricks ] produces a negative ( positive ) sign in the r.h.s . of eq . ( [ c ] ) [ ( [ z ] ) ] . eqs . ( [ integral ] ) can be rewritten in differential , renormalization - group form : [ differential ] @xmath42 combining eqs . ( [ differential ] ) with eq.([vertex : spinless ] ) , we find the renormalization equation for the full vertex @xmath30 : [ rg : spinless ] @xmath43 we see that the two terms in the r.h.s . of eq . ( [ cancellation ] ) , representing the tendencies toward density - wave and superconducting instabilities , exactly cancel each other . in a ladder approximation , where only one term is kept in the r.h.s . , the result would be quite different , because @xmath44 would diverge at a finite @xmath29 indicating an instability or generation of a pseudogap in the system . in order to study possible instabilities in the system , we need to calculate corresponding generalized susceptibilities . for that purpose , let us add to the hamiltonian of the system two fictitious infinitesimal external fields @xmath45 and @xmath46 that create the electron - electron and electron - hole pairs : @xmath47 . \label{hext}\end{aligned}\ ] ] now we need to introduce triangular vertices @xmath48 and @xmath49 that represent the response of the system to the fields @xmath45 and @xmath46 . following the same procedure as in the derivation of the parquet equations for the bricks @xcite , we find the parquet equations for the triangular vertices in graphic form , as shown in fig . [ fig : spinlesstriangle ] . in that figure , the filled triangles represent the vertices @xmath48 and @xmath49 , whereas the dots represent the fields @xmath50 and @xmath46 . the circles , as in the other figures , represent the interaction vertex @xmath30 . analytically , these equations can be written as differential equations with given initial conditions : [ triangular ] @xmath51 we will often refer to the triangular vertices @xmath52 as the `` order parameters '' . indeed , they are the superconducting and density - wave order parameters induced in the system by the external fields @xmath45 and @xmath46 . if , for a finite @xmath53 ( @xmath54=sc , dw ) , a vertex @xmath55 , which is proportional to @xmath53 , diverges when @xmath56 , this indicates that a _ spontaneous _ order parameter appears in the system , that is , the order parameter may have a finite value even when the external field @xmath53 is zero . the external fields are introduced here only as auxiliary tools and are equal to zero in real systems . we also note that the two terms in the r.h.s . of eq . ( [ hext ] ) are not hermitially self - conjugate ; thus , the fields @xmath53 are the complex fields . consequently , the order parameters @xmath55 are also complex , so , generally speaking , @xmath52 and @xmath57 do not coincide . according to eqs . ( [ rg : spinless ] ) , @xmath58 , so eqs.([triangular ] ) have the following solution : [ triangular : solutions ] @xmath59 now we can calculate the susceptibilities . the lowest order corrections to the free energy of the system due to the introduction of the fields @xmath45 and @xmath46 , @xmath60 and @xmath61 , obey the parquet equations shown graphically in fig.[fig : spinlesssusceptibility ] and analytically below : [ freeenergy ] @xmath62 substituting expressions ( [ triangular : solutions ] ) into eqs.([freeenergy ] ) and dropping the squares of @xmath45 and @xmath46 , we find the susceptibilities : [ susceptibilities ] @xmath63/2 g , \\ \chi_{\rm dw}(\xi)&=&\bigm[\exp(2g\xi)-1\bigm]/2g.\end{aligned}\ ] ] according to eqs . ( [ susceptibilities ] ) , when the interaction between electrons is repulsive ( attractive ) , that is , @xmath15 is positive ( negative ) , the density - wave ( superconducting ) susceptibility increases as temperature decreases ( @xmath64 and @xmath65 ) : @xmath66 susceptibilities ( [ powerlaw ] ) have power dependence on the temperature and energy , which is one of the characteristic properties of the luttinger liquid . the susceptibilities are finite at finite temperatures and diverge only at zero temperature , in agreement with the general theorem @xcite that phase transitions are impossible at finite temperatures in 1d systems . mathematically , the absence of divergence at finite @xmath29 is due to the cancellation of the two terms in the r.h.s . of eq . ( [ cancellation ] ) and subsequent nonrenormalization of @xmath44 . this nontrivial 1d result can be obtained only within the parquet , but not the ladder approximation . now let us consider 1d electrons with spin . in this case , there are three vertices of interaction , conventionally denoted as @xmath67 , @xmath68 , and @xmath69 , which represent backward , forward , and umklapp scattering , respectively @xcite . umklapp scattering should be considered only when the change of the total momentum of the electrons in the interaction process , @xmath70 , is equal to the crystal lattice wave vector , which may or may not be the case in a particular model . in this paper , we do not consider the vertex @xmath71 , which describes the interaction between the electrons of the same type ( @xmath21 or @xmath22 ) , because this vertex does not have logarithmic corrections . the bare hamiltonian of the interaction , @xmath13 , can be written as @xmath72 \biggm\ } , \label{interaction}\end{aligned}\ ] ] where the coefficients @xmath73 denote the bare ( unrenormalized ) values of the interaction vertices @xmath74 . the operators @xmath75 and @xmath76 create and destroy electrons of the type @xmath77 and the spin @xmath78 . the spin structure of the interaction hamiltonian is dictated by conservation of spin . we picture the interaction vertices in fig . [ fig : interaction ] , where the solid and dashed lines represent the @xmath21 and @xmath22 electrons . the thin solid lines inside the circles indicate how spin is conserved : the spins of the incoming and outgoing electrons connected by a thin line are the same . according to the structure of hamiltonian ( [ interaction ] ) , the umklapp vertex @xmath69 describes the process where two + electrons come in and two electrons come out , whereas the complex conjugate vertex @xmath79 describes the reversed process . the three vertices of interaction contain six bricks , as shown schematically in fig . [ fig : vertices ] : [ vertices ] @xmath80 where @xmath81 and @xmath82 are the superconducting bricks , and @xmath83 , @xmath84 , @xmath85 , and @xmath86 are the density - wave bricks . in fig.[fig : vertices ] , the thin solid lines inside the bricks represent spin conservation . the umklapp vertex has two density - wave bricks @xmath85 and @xmath86 , which differ in their spin structure . parquet equations for the bricks are derived in the same manner as in sec . [ sec : spinless ] by adding appropriate spin structure dictated by spin conservation . it is convenient to derive the equations graphically by demanding that the thin spin lines are continuous , as shown in fig . [ fig : bricks ] . corresponding analytic equations can be written using the following rules . a pair of parallel ( antiparallel ) lines connecting two vertices in fig.[fig : bricks ] produces the negative ( positive ) sign . a closed loop of the two connecting lines produces an additional factor @xmath87 due to summation over the two spin orientations of the electrons . [ bricks ] @xmath88 , \\ \frac{dz_{ii}(\xi)}{d\xi } & = & 2\gamma_3(\xi)\:\gamma_2(\xi).\end{aligned}\ ] ] combining eqs . ( [ vertices ] ) and ( [ bricks ] ) , we obtain the well - known closed equations for renormalization of the vertices @xcite : [ rg1d ] @xmath89.\end{aligned}\ ] ] in the presence of spin , the electron operators in eq.([hext ] ) and , correspondingly , the fields @xmath53 and the triangular vertices @xmath55 acquire the spin indices . thus , the superconducting triangular vertex @xmath90 becomes a vector : @xmath91 parquet equations for the triangular vertices are given by the diagrams shown in fig . [ fig : spinlesstriangle ] , where the spin lines should be added in the same manner as in fig . [ fig : bricks ] . the superconducting vertex obeys the following equation : @xmath92 where the matrix @xmath93 is @xmath94 linear equation ( [ mtrx ] ) is diagonalized by introducing the singlet , @xmath95 , and the triplet , @xmath96 , superconducting triangular vertices : [ sc ] @xmath97 which obey the following equations : @xmath98 \;{\cal t}_{\rm ssc(tsc)}(\xi ) . \label{scop}\ ] ] in eq . ( [ scop ] ) the sign @xmath22 and the index ssc correspond to the singlet superconductivity , whereas the sign @xmath21 and the index tsc correspond to the triplet one . in the rest of the paper , we use the index sc where discussion applies to both ssc and tsc . now let us consider the density - wave triangular vertices , first in the absence of umklapp . they form a vector @xmath99 which obeys the equation @xmath100 with the matrix @xmath101 eq . ( [ dwmtrx ] ) is diagonalized by introducing the charge- , @xmath102 , and the spin- , @xmath103 , density - wave triangular vertices : [ dw ] @xmath104 which obey the following equations : [ dwop ] @xmath105\;{\cal t}_{\rm cdw}(\xi ) , \\ \frac{d{\cal t}_{\rm sdw}(\xi)}{d\xi } & = & \gamma_2(\xi)\;{\cal t}_{\rm sdw}(\xi).\end{aligned}\ ] ] when the umklapp vertices @xmath69 and @xmath79 are introduced , they become offdiagonal matrix elements in eqs . ( [ dwop ] ) , mixing @xmath102 and @xmath103 with their complex conjugates . assuming for simplicity that @xmath69 is real , we find that the following linear combinations diagonalize the equations : @xmath106 and the equations become : [ dwop+- ] @xmath107 \;{\cal t}_{{\rm cdw}\pm}(\xi ) , \\ \frac{d{\cal t}_{{\rm sdw}\pm}(\xi)}{d\xi } & = & [ \gamma_2(\xi)\pm\gamma_3(\xi)]\;{\cal t}_{{\rm sdw}\pm}(\xi).\end{aligned}\ ] ] if the external fields @xmath53 are set to unity in the initial conditions of the type ( [ triangular ] ) for all triangular vertices @xmath54 = ssc , tsc , cdw@xmath10 , and sdw@xmath10 , then the corresponding susceptibilities are equal numerically to the free energy corrections of the type ( [ freeenergy ] ) : @xmath108 eqs . ( [ rg1d ] ) , ( [ scop ] ) , ( [ dwop+- ] ) , and ( [ chii ] ) were solved analytically in ref . @xcite , where a complete phase diagram of the 1d electron gas with spin was obtained . now let us consider a 2d electron gas with the fermi surface shown schematically in fig . [ fig:2dfs ] . it contains two pairs of flat regions , shown as the thick lines and labeled by the letters @xmath109 and @xmath110 . such a fermi surface resembles the fermi surfaces of some high-@xmath0 superconductors @xcite . in our consideration , we restrict the momenta of electrons to the flat sections only . in this way , we effectively neglect the rounded portions of the fermi surface , which are not relevant for the parquet consideration , because the density - wave loop is not divergent there . one can check also that the contributions of the portions @xmath109 and @xmath110 do not mix with each other in the parquet manner , so they may be treated separately . for this reason , we will consider only the region @xmath109 , where the 2d electron states are labeled by the two momenta @xmath6 and @xmath111 , the latter momentum being restricted to the interval @xmath112 $ ] . in our model , the energy of electrons depends only on the momentum @xmath6 according to eq . ( [ eps ] ) and does not depend on the momentum @xmath111 . we neglect possible dependence of the fermi velocity @xmath113 on @xmath111 ; it was argued in ref . @xcite that this dependence is irrelevant in the renormalization - group sense . in the 2d case , each brick or vertex of interaction between electrons acquires extra variables @xmath114 , @xmath115 , and @xmath116 in addition to the 1d variables @xmath37 , and @xmath38 . these two sets of variables play very different roles . the green functions , which connect the vertices and produce the logarithms @xmath29 , depend only on the second set of variables . thus , following the parquet approach outlined in the previous sections , we dump all the @xmath117 and @xmath118 variables of a vertex or a brick into a single variable @xmath29 . at the same time , the @xmath114 , @xmath115 , and @xmath116 variables remain independent and play the role of indices labeling the vertices , somewhat similar to the spin indices . thus , each vertex and brick is a function of several variables , which we will always write in the following order : @xmath119 . it is implied that the first four variables satisfy the momentum conservation law @xmath120 , and each of them belongs to the interval @xmath112 $ ] . the assignment of the variables @xmath114 , @xmath115 , @xmath116 , and @xmath121 to the ends of the vertices and bricks is shown in fig.[fig : vertices ] , where the labels @xmath122 ( @xmath123 ) should be considered now as the variables @xmath124 . to shorten notation , it is convenient to combine these variable into a single four - component vector @xmath125 so that the relation between the vertices and the bricks can be written as [ 2dgammas ] @xmath126 after this introduction , we are in a position to write the parquet equations for the bricks . these equations are shown graphically in fig . [ fig : bricks ] , where again the momenta @xmath122 should be understood as @xmath124 . analytically , the equations are written below , with the terms in the same order as in fig . [ fig : bricks ] : [ 2dbricks ] @xmath127 where [ 2dk ] @xmath128 and the tilde and the bar operations are defined as [ tildebar ] @xmath129 in eqs . ( [ 2dbricks ] ) , we introduced the operation @xmath130 that represents the integration over the internal momenta of the loops in fig . [ fig : bricks ] . it denotes the integration over the intermediate momentum @xmath131 with the restriction that both @xmath131 and @xmath132 , another intermediate momentum determined by conservation of momentum , belong to the interval @xmath112 $ ] . for example , the explicit form of the first term in the r.h.s . of eq . ( [ c1 ] ) is : @xmath133 eqs . ( [ 2dbricks ] ) and ( [ 2dgammas ] ) with definitions ( [ k ] ) , ( [ 2dk ] ) , and ( [ tildebar ] ) form a closed system of integrodifferential equations , which will be solved numerically in sec . [ sec : numerical ] . the initial conditions for eqs.([2dbricks ] ) and ( [ 2dgammas ] ) are that all the @xmath34 and @xmath35 bricks are equal to zero at @xmath134 . parquet equations for the superconducting triangular vertices can be found in the 2d case by adding the @xmath111 momenta to the 1d equations ( [ scop ] ) . the equations are shown graphically in fig.[fig : spinlesstriangle ] , where the momenta @xmath135 and @xmath136 should be interpreted as @xmath111 and @xmath137 : @xmath138 where @xmath139 and the operator @xmath130 denotes the integration over @xmath140 with the restriction that both @xmath141 and @xmath142 belong to the interval @xmath112 $ ] . the @xmath143 signs in front of @xmath67 in eq . ( [ fsc ] ) correspond to the singlet and triplet superconductivity . as discussed in sec . [ sec : spinless ] , the triangular vertex @xmath144 is the superconducting order parameter , @xmath137 and @xmath111 being the @xmath145-components of the total and the relative momenta of the electrons in a cooper pair . indeed , the vertex @xmath144 obeys the linear equation shown in fig . [ fig : spinlesstriangle ] , which is the linearized gorkov equation for the superconducting order parameter . as the system approaches a phase transition , the vertex @xmath144 diverges in overall magnitude , but its dependence on @xmath111 for a fixed @xmath137 remains the same , up to a singular , @xmath29-dependent factor . the dependence of @xmath146 on @xmath111 describes the distribution of the emerging order parameter over the fermi surface . the numerically found behavior of @xmath144 is discussed in sec.[sec : numerical ] . due to the particular shape of the fermi surface , the vertices of interaction in our 2d model have two special symmetries : with respect to the sign change of all momenta @xmath111 and with respect to the exchange of the @xmath21 and @xmath22 electrons : [ symmetry ] @xmath147 where in eq . ( [ * ] ) we assume for simplicity that @xmath69 is real . as a consequence of ( [ symmetry ] ) , eqs . ( [ 2dscop ] ) are invariant with respect to the sign reversal of @xmath111 in @xmath146 at a fixed @xmath137 . the following combinations of the triangular vertices form two irreducible representations of this symmetry , that is , they are independent and do not mix in eqs.([2dscop ] ) : @xmath148 the triangular vertices @xmath149 describe the superconducting order parameters that are either symmetric or antisymmetric with respect to the sign change of @xmath111 . when @xmath150 is extended over the whole 2d fermi surface ( see fig . [ fig:2dfs ] ) , it acquires the @xmath151-wave symmetry , whereas @xmath152 the @xmath1-wave symmetry . the symmetrized vertices @xmath149 obey the same eqs . ( [ 2dscop ] ) as the unsymmetrized ones . the equations for the density - wave triangular vertices are obtained in a similar manner : [ csdwopa ] @xmath153 where @xmath154 the @xmath10 signs in the subscripts of @xmath52 in eqs.([csdwopa ] ) and in front of @xmath69 in eqs.([fcdw])([fsdw ] ) refer to the umklapp symmetry discussed in sec . [ sec : spin1d ] , whereas the @xmath10 signs in the superscripts of @xmath52 refer to the symmetry with respect to sign reversal of @xmath111 , discussed above in the superconducting case . the @xmath111-antisymmetric density waves are actually the waves of charge current and spin current @xcite , also known in the so - called flux phases @xcite . once the triangular vertices @xmath155 are found , the corresponding susceptibilities @xmath156 are calculated according to the following equation , similar to eq . ( [ chii ] ) : @xmath157 where the integration over @xmath111 is restricted so that both @xmath158 belong to the interval @xmath112 $ ] . using functions ( [ fsc ] ) , ( [ fcdw ] ) , and ( [ fsdw ] ) and symmetries ( [ symmetry ] ) , we can rewrite eqs . ( [ 2dbricks ] ) in a more compact form . for that purpose , we introduce the ssc , tsc , cdw , and sdw bricks that are the linear combinations of the original bricks : [ newbricks ] @xmath159 where the tilde operation is defined in eq . ( [ tilde ] ) . then , eqs . ( [ 2dbricks ] ) become : [ newrg ] @xmath160 the parquet equations in the form ( [ newrg ] ) were obtained in ref . @xcite . it is instructive to trace the difference between the parquet equations ( [ newrg ] ) and the corresponding ladder equations . suppose that , for some reason , only one brick , say @xmath161 , among the six bricks ( [ newbricks ] ) is appreciable , whereas the other bricks may be neglected . using definitions ( [ 2dgammas ] ) and ( [ fsc ] ) , we find that eq . ( [ sc1 ] ) becomes a closed equation : @xmath162 where @xmath163 eq . ( [ fscrg ] ) is the ladder equation for the singlet superconductivity . when the initial value @xmath164 of the vertex @xmath165 is positive , eq ( [ fscrg ] ) has a singular solution ( @xmath166 at @xmath56 ) , which describes a phase transition into the singlet superconducting state at a finite temperature . repeating this consideration for every channel , we construct the phase diagram of the system in the ladder approximation as a list of necessary conditions for the corresponding instabilities : [ ladderphasediagram ] @xmath167 the difference between the ladder and the parquet approximations shows up when there are more than one appreciable bricks in the problem . then , the vertex @xmath165 contains not only the brick @xmath161 , but other bricks as well , so eqs . ( [ newrg ] ) get coupled . this is the case , for example , for the 1d spinless electrons , where the bricks @xmath34 and @xmath35 are equally big , so they cancel each other in @xmath30 ( see sec . [ sec : spinless ] ) . the numerical procedure consists of three consecutive steps ; each of them involves solving differential equations by the fourth - order runge kutta method . first , we solve parquet equations ( [ 2dgammas ] ) and ( [ 2dbricks ] ) for the interaction vertices , which are closed equations . then , we find the triangular vertices @xmath155 , whose equations ( [ 2dscop ] ) and ( [ csdwopa ] ) involve the interaction vertices @xmath168 through eqs . ( [ fsc ] ) , ( [ fcdw ] ) , and ( [ fsdw ] ) . finally , we calculate the susceptibilities @xmath156 from eqs . ( [ 2dchii ] ) , which depend on the triangular vertices @xmath155 . we select the initial conditions for the interaction vertices to be independent of the transverse momenta @xmath169 : @xmath170 . the momentum - independent interaction naturally appears in the hubbard model , where the interaction is local in real space . in this chapter , the results are shown mostly for the repulsive hubbard model without umklapp : @xmath171 ( figs.[fig : gammadata][fig : phasediagram110 ] ) , or with umklapp : @xmath172 ( figs . [ fig : chi111][fig : phasediagram111 ] ) , where @xmath15 is proportional the hubbard interaction constant @xmath173 . the absolute value of @xmath15 ( but not the sign of @xmath15 ) is not essential in our calculations , because it can be removed from the equations by redefining @xmath29 to @xmath174 . after the redefinition , we effectively have @xmath175 in the initial conditions . the actual value of @xmath176 matters only when the logarithmic variable @xmath177 is converted into the temperature according to the formula @xmath178 . the initial independence of @xmath179 on @xmath169 does not imply that this property is preserved upon renormalization . on the contrary , during renormalization , @xmath180 develops a very strong dependence on @xmath181 and may even change sign in certain regions of the @xmath181-space . we illustrate this statement in fig . [ fig : gammadata ] by showing typical dependences of @xmath182 and @xmath183 on the average momentum @xmath184 of the incoming electrons at @xmath185 and @xmath186 after some renormalization ( @xmath187 ) . in figs.[fig : gammadata][fig : tdwdata ] , the upper and lower limits on the horizontal axes are the boundaries @xmath188 of the flat region on the fermi surface , which are set to @xmath189 without loss of generality . one can observe in fig . [ fig : gammadata ] that the electron - electron interaction becomes negative ( attractive ) at large @xmath190 , even though initially it was repulsive everywhere . mathematically , the dependence of @xmath180 on @xmath169 arises because of the finite limits of integration , @xmath112 $ ] , imposed on the variables @xmath131 and @xmath132 in eqs . ( [ 2dbricks ] ) . for example , in eq.([c1 ] ) , when @xmath184 equals zero , @xmath131 may change from @xmath191 to @xmath192 while @xmath132 stays in the same interval . when @xmath193 , @xmath131 has to be confined to a narrower interval @xmath194 $ ] to ensure that @xmath195 stays within @xmath112 $ ] . this difference in the integration range subsequently generates the dependence of @xmath180 on @xmath190 and , more generally , on @xmath169 . since many channels with different geometrical restrictions contribute to @xmath196 in eqs . ( [ 2dbricks ] ) , the resultant dependence of @xmath180 on the four - dimensional vector @xmath169 is complicated and hard to visualize . because of the strong dependence of @xmath180 on @xmath169 , it is not possible to describe the 2d system by only three renormalizing charges @xmath197 , @xmath198 , and @xmath199 , as in the 1d case . instead , it is absolutely necessary to consider an infinite number of the renormalizing charges @xmath180 labeled by the continuous variable @xmath169 . this important difference was neglected in ref . @xcite , where the continuous variable @xmath169 was omitted . having calculated @xmath180 , we solve eqs.([2dscop ] ) and ( [ csdwopa ] ) for the triangular vertices ( the order parameters ) @xmath200 , which depend on both the relative ( @xmath111 ) and the total ( @xmath137 ) transverse momenta . we find numerically that the order parameters with @xmath201 diverge faster than those with @xmath202 . this is a natural consequence of the integration range restrictions discussed above . for this reason , we discuss below only the order parameters with zero total momentum @xmath201 . we select the initial conditions for the symmetric and antisymmetric order parameters in the form : @xmath203 in figs . [ fig : tscdata ] and [ fig : tdwdata ] , we present typical dependences of the superconducting and density - wave order parameters on the relative momentum @xmath111 at the same renormalization `` time '' @xmath187 as in fig . [ fig : gammadata ] . the singlet antisymmetric component ( @xmath204 ) dominates among the superconducting order parameters ( fig . [ fig : tscdata ] ) , whereas the symmetric sdw order parameter ( @xmath205 ) is the highest in the density - wave channel ( fig . [ fig : tdwdata ] ) . having calculated the triangular vertices @xmath52 , we find the susceptibilities from eq . ( [ 2dchii ] ) . the results are shown in fig . [ fig : chi110 ] . the symmetric sdw has the fastest growing susceptibility @xmath206 , which diverges at @xmath207 . this divergence indicates that a phase transition from the metallic to the antiferromagnetic state takes place at the transition temperature @xmath208 . a similar result was obtained in ref.@xcite by analyzing the convergence radius of the parquet series in powers of @xmath31 . in the ladder approximation , the sdw instability would take place at @xmath209 , as follows from eqs . ( [ fsdw ] ) and ( [ sw1 ] ) . since @xmath210 , the transition temperature @xmath211 , calculated in the parquet approximation , is lower than the temperature @xmath212 , calculated in the ladder approximation : @xmath213 . the parquet temperature is lower , because competing superconducting and density - wave instabilities partially suppress each other . thus far , we considered the model with ideally flat regions on the fermi surface . suppose now that these regions are only approximately flat . that is , they can be treated as being flat for the energies higher than a certain value @xmath214 , but a curvature or a corrugation of the fermi surface becomes appreciable at the smaller energies @xmath215 . because of the curvature , the fermi surface does not have nesting for @xmath215 ; thus , the density - wave bricks in the parquet equations ( [ 2dbricks ] ) stop to renormalize . formally , this effect can be taken into account by introducing a cutoff @xmath216 , so that the r.h.s . ( [ z1])([zii ] ) for the density - wave bricks are replaced by zeros at @xmath217 . at the same time , eqs . ( [ c1 ] ) and ( [ c2 ] ) for the superconducting bricks remain unchanged , because the curvature of the fermi surface does not affect the superconducting instability with @xmath201 . the change of the renormalization equations at @xmath218 is not a completely rigorous way @xcite to take into account the fermi surface curvature ; however , this procedure permits obtaining explicit results and has a certain qualitative appeal . for a more rigorous treatment of the corrugated fermi surface problem see ref . @xcite . in fig . [ fig : chicutoff ] , we show the susceptibilities calculated using the cutoff procedure with @xmath219 . the density - wave susceptibilities remain constant at @xmath217 . at the same time , @xmath220 diverges at @xmath221 indicating a transition into the singlet superconducting state of the @xmath1-wave type . thus , if the sdw instability is suppressed , the system is unstable against formation of the @xmath1-wave superconductivity . this result is in agreement with the conclusions of refs . @xcite . from our numerical results , we deduce that the dependence of @xmath222 on @xmath223 is linear : @xmath224 with @xmath225 , as shown in the inset to fig . [ fig : phasediagram110 ] . converting @xmath29 into energy in this relation , we find a power law dependence : @xmath226 eq . ( [ tcr1 ] ) demonstrates that increasing the cutoff energy @xmath214 decreases the temperature of the superconducting transition , @xmath227 . such a relation can be qualitatively understood in the following way . there is no bare interaction in the superconducting @xmath1-wave channel in the hubbard model , so the transition is impossible in the ladder approximation . the growth of the superconducting @xmath1-wave correlations is induced by the growth of the sdw correlations , because the two channels are coupled in the parquet equations ( [ newrg ] ) . if @xmath214 is high , the sdw correlations do not have enough renormalization - group `` time '' @xmath29 to develop themselves because of the early cutoff of the density - wave channels ; thus , @xmath227 is low . hence , decreasing @xmath228 increases @xmath227 . however , when @xmath214 becomes lower than @xmath229 , the sdw instability overtakes the superconducting one . corresponding phase diagram is shown in fig.[fig : phasediagram110 ] . generally speaking , the phase diagram plotted in the energy variables , as opposed to the logarithmic variables @xmath29 , may depend on the absolute value of the bare interaction constant @xmath176 . in fig . [ fig : phasediagram110 ] , we placed the points for the several values of @xmath15 = 0.3 , 0.4 , and 0.5 ; the phase boundary does not depend much on the choice of @xmath15 . the phase diagram of fig . [ fig : phasediagram110 ] qualitatively resembles the experimental one for the high-@xmath0 superconductors , where transitions between the metallic , antiferromagnetic , and superconducting states are observed . the value of @xmath214 may be related to the doping level , which controls the shape of the fermi surface . taking into account the crudeness of our approximations , detailed agreement with the experiment should not be expected . we perform the same calculations also for the hubbard model with umklapp scattering ( @xmath230 ) . as one can see in fig.[fig : chi111 ] , where the susceptibilities are shown , the umklapp process does not modify the qualitative picture . the leading instability remains the sdw of the symmetric type , which is now also symmetric with respect to the umklapp scattering , whereas the next leading instability is the singlet @xmath1-wave superconductivity . the sdw has a phase transition at @xmath231 , which is close to the ladder result @xmath232 . some of the susceptibilities in fig . [ fig : chi111 ] coincide exactly , which is a consequence of a special su(2)@xmath233su(2 ) symmetry of the hubbard model at the half filling @xcite . the phase diagram with the energy cutoff ( fig . [ fig : phasediagram111 ] ) is similar to the one without umklapp ( fig . [ fig : phasediagram110 ] ) , but the presence of the umklapp scattering decreases the transition temperature of the @xmath1-wave superconductivity . an important issue in the study of the 1d electron gas is the so - called @xmath15-ology phase diagram , which was constructed for the first time by dzyaloshinskii and larkin @xcite . they found that , in some regions of the @xmath234 space , the 1d electron system develops a charge or spin gap , which is indicated by divergence of @xmath235 with increasing @xmath29 . in the region where none of the gaps develops , the luttinger liquid exists . it is interesting whether such a region may exist in our 2d model . to study the phase diagram of the 2d system , we repeat the calculations , systematically changing relative values of @xmath236 , @xmath237 , and @xmath238 . from the physical point of view , the relative difference of @xmath236 , @xmath237 , and @xmath238 roughly mimics dependence of the interaction vertex on the momentum transfer . as an example , we show the susceptibilities in the case where @xmath239 , @xmath240 , and @xmath241 in fig.[fig : chi210 ] . in this case , the leading instabilities are simultaneously the triplet superconductivity of the symmetric type ( tsc+ ) and the spin - density wave . for all studied sets of @xmath242 , we find that the leading instabilities calculated in the parquet and the ladder approximations always coincide . ( we do not introduce the energy cutoff here . ) thus , the parquet effects do not modify the @xmath15-ology phase diagram of the 2d model derived in the ladder approximation , even though the transition temperatures in the parquet approximation are always lower than those obtained in the ladder approximation . in that sense , the parquet corrections are much less important in the 2d case than in the 1d case . from the mathematical point of view , this happens because a leading divergent brick develops a strong dependence on the transverse momenta @xmath169 and acquires the so - called mobile pole structure @xcite : @xmath243 the name `` mobile pole '' is given , because the position of the pole in @xmath29 in eq . ( [ movingpole ] ) , @xmath244 , strongly depends on the momenta @xmath169 . it was shown in refs.@xcite that , because of the mobility of the pole , the leading channel decouples from the other channels , and the parquet description effectively reduces to the ladder one , as described at the end of sec . [ sec:2d ] . the phase diagram of the 2d system in the ladder approximation is given by eqs.([ladderphasediagram ] ) . it follows from eqs.([ladderphasediagram ] ) that every point in the @xmath234 space has some sort of instability . thus , the luttinger liquid , defined as a nontrivial metallic ground state where different instabilities mutually cancel each other , does not exist in the 2d model . generally speaking , other models may have different types of solutions of the fast parquet equations , such as immobile poles @xcite or a self - similar solution @xcite , the latter indeed describing some sort of a luttinger liquid . in our study of a 2d model with the van hove singularities @xcite , we found a region in the @xmath15-space without instabilities , where the luttinger liquid may exist @xcite . however , we find only the mobile - pole solutions in the present 2d model . in this paper we derive and numerically solve the parquet equations for the 2d electron gas whose fermi surface contains flat regions . the model is a natural generalization of the 1d electron gas model , where the luttinger liquid is known to exist . we find that , because of the finite size of the flat regions , the 2d parquet equations always develop the mobile pole solutions , where the leading instability effectively decouples from the other channels . thus , a ladder approximation is qualitatively ( but not necessarily quantitatively ) correct for the 2d model , in contrast to the 1d case . whatever the values of the bare interaction constants are , the 2d system always develops some sort of instability . thus , the luttinger liquid , defined as a nontrivial metallic ground state where different instabilities mutually cancel each other , does not exist in the 2d model , contrary to the conclusions of refs.@xcite . in the case of the repulsive hubbard model , the leading instability is the sdw , i.e. , antiferromagnetism @xcite . if the nesting of the fermi surface is not perfect , the sdw correlations do not develop into a phase transition , and the singlet superconductivity of the @xmath1-wave type appears in the system instead . these results may be relevant for the high-@xmath0 superconductors and are in qualitative agreement with the findings of refs.@xcite . in the bosonization procedure @xcite , a higher - dimensional fermi surface is treated as a collection of flat patches . since the results of our paper do not depend qualitatively on the size of the flat regions on the fermi surface , the results may be applicable , to some extent , to the patches as well . precise relation is hard to establish because of the infinitesimal size of the patches , their different orientations , and uncertainties of connections between them . on the other hand , the bosonization procedure seems to be even better applicable to a flat fermi surface , which consists of a few big patches . mattis @xcite and hlubina @xcite followed that logic and claimed that the flat fermi surface model is exactly solvable by the bosonization and represents a luttinger liquid . the discrepancy between this claim and the results our paper indicates that some conditions must restrict the validity of the bosonization approximations . luther gave a more sophisticated treatment to the flat fermi surface problem by mapping it onto multiple quantum chains @xcite . he found that the bosonization converts the interaction between electrons into the two types of terms , roughly corresponding to the two terms of the sine - gordon model : the `` harmonic '' terms @xmath245 and the `` exponential '' terms @xmath246 , where @xmath247 is a bosonization phase . the harmonic terms can be readily diagonalized , but the exponential terms require a consistent renormalization - group treatment . if the renormalization - group equations were derived in the bosonization scheme of @xcite , they would be the same as the parquet equations written in our paper , because the renormalization - group equations do not depend on whether the boson or fermion representation is used in their derivation @xcite . long time ago , luther bosonized noninteracting electrons on a curved fermi surface @xcite ; however , the interaction between the electrons remained intractable because of the exponential terms . the recent bosonization in higher dimensions @xcite managed to reformulate the problem in the harmonic terms only . this is certainly sufficient to reproduces the landau description of sound excitations in a fermi liquid @xcite ; however , it may be not sufficient to derive the electron correlation functions . the validity of the harmonic approximation is hard to trace for a curved fermi surface , but considerable experience has been accumulated for the flat fermi surface models . in the model of multiple 1d chains without single - electron tunneling between the chains and with forward scattering between different chains , the bosonization produces the harmonic terms only , thus the model can be solved exactly @xcite . however , a slight modification of the model by introducing backward scattering between different chains @xcite or interaction between four different chains @xcite adds the exponential terms , which destroy the exact solvability and typically lead to a cdw or sdw instability . even if no instability occurs , as in the model of electrons in a high magnetic field @xcite , the fast parquet method shows that the electron correlation functions have a complicated , nonpower structure , which is impossible to obtain within the harmonic bosonization . further comparison of the fast parquet method and the bosonization in higher dimensions might help to establish the conditions of applicability of the two complementary methods . f. d. m. haldane , in _ perspectives in many - particle physics _ , proceedings of the international school of physics `` enrico fermi , '' course 121 , varenna , 1992 , edited by r. schrieffer and r. a. broglia ( north - holland , new york , 1994 ) . d. v. khveshchenko , r. hlubina , and t. m. rice , phys . b * 48 * , 10766 ( 1993 ) ; d. v. khveshchenko and p. c. e. stamp , phys . rev . lett . * 71 * , 2118 ( 1993 ) ; phys . b * 49 * , 5227 ( 1994 ) ; d. v. khveshchenko , phys . rev . b * 52 * , 4833 ( 1995 ) [ cond - mat/9409118 ] . a. houghton , h .- j . kwon , and j. b. marston , phys . b * 50 * , 1351 ( 1994 ) [ cond - mat/9310043 ] ; a. houghton , h .- j . kwon , j. b. marston , and r. shankar , j. phys . cond . mat . * 6 * , 4909 ( 1994 ) [ cond - mat/9312067 ] ; h .- j . kwon , a. houghton , and j. b. marston , phys . * 73 * , 284 ( 1994 ) [ cond - mat/9401041 ] ; phys . b * 52 * , 8002 ( 1995 ) [ cond - mat/9501067 ] . p. kopietz , j. hermisson , and k. schonhammer , phys . b * 52 * , 10877 ( 1995 ) [ cond - mat/9304014 ] ; p. kopietz , phys . rev . b * 53 * , 12761 ( 1996 ) [ cond - mat/9503090 ] ; p. kopietz and g. e. castilla , phys . * 76 * , 4777 ( 1996 ) [ cond - mat/9603158 ] . i. t. diatlov , v. v. sudakov , and k. a. ter - martirosian , zh . . fiz . * 32 * , 767 ( 1957 ) [ sov . jetp * 5 * , 631 ( 1957 ) ] ; v. v. sudakov , dokl . nauk sssr * 111 * , 338 ( 1956 ) [ sov . doklady * 1 * , 662 ( 1957 ) ] . i. e. dzyaloshinskii , zh . . fiz . * 93 * , 1487 ( 1987 ) [ sov . jetp * 66 * , 848 ( 1987 ) ] ; i. e. dzyaloshinskii and v. m. yakovenko , zh . . fiz . * 94 * , 344 ( 1988 ) [ sov . jetp * 67 * , 844 ( 1988 ) ] ; i. e. dzyaloshinskii and v. m. yakovenko , int . j. mod b * 2 * , 667 ( 1988 ) . v. n. prigodin and yu . a. firsov , pisma zh . * 25 * , 90 ( 1977 ) [ jetp lett . * 25 * , 79 ( 1977 ) ] ; zh . . fiz . * 76 * , 736 ( 1979 ) [ sov . jetp * 49 * , 369 ( 1979 ) ] ; zh . . fiz . * 76 * , 1602 ( 1979 ) [ sov . jetp * 49 * , 813 ( 1979 ) ] .

we give a correspondence between graphs with a given degree sequence and fillings of ferrers diagrams by nonnegative integers with prescribed row and column sums . in this setting ,-crossings and-nestings of the graph become occurrences of the identity and the antiidentity matrices in the filling . we use this to show the equality of the numbers of-noncrossing and-nonnesting graphs with a given degree sequence . this generalizes the analogous result for matchings and partition graphs of chen , deng , du , stanley , and yan , and extends results of klazar to . moreover , this correspondence reinforces the links recently discovered by krattenthaler between fillings of diagrams and the results of chen et al .

math0602195

let @xmath2 be a graph on @xmath3 $ ] ; unless otherwise stated , we allow multiple edges and isolated vertices , but no loops . two edges @xmath4 and @xmath5 are a _ crossing _ if @xmath6 and they are a _ nesting _ if @xmath7 . if we draw the vertices of @xmath2 on a line and represent the corresponding edges by arcs above the line , crossings and nestings have the obvious geometric meaning . a graph without crossings ( respectively , nestings ) is called _ noncrossing _ ( resp . , _ nonnesting _ ) . klazar @xcite proves the equality between the numbers of noncrossing and nonnesting simple graphs , counted by order , and also between the numbers of noncrossing and nonnesting graphs without isolated vertices , counted by size . the purpose of this paper is to study analogous results for sets of @xmath0 pairwise crossing and @xmath0 pairwise nested edges . a _ @xmath0-crossing _ is a set of @xmath0 edges every two of them being a crossing , that is , edges @xmath8 @xmath9 such that @xmath10 . @xmath0-nesting _ is a set of @xmath0 edges pairwise nested , that is , @xmath8 @xmath9 such that @xmath11 . a graph with no @xmath0-crossing is called _ @xmath0-noncrossing _ and a graph with no @xmath0-nesting is called _ @xmath0-nonnesting_. the largest @xmath0 for which a graph @xmath2 has a @xmath0-crossing ( respectively , a @xmath0-nesting ) is denoted @xmath12 ( resp . , @xmath13 ) . the aim of this paper is to show that the number of @xmath0-noncrossing graphs equals the number of @xmath0-nonnesting graphs , counted by order , size , and degree sequences . this problem was originally posed by martin klazar and we learned of it at the homonolo 2005 workshop @xcite ; the case where the number of vertices of the graph is @xmath14 was proved by a. pr ( unpublished ) . our main result ( theorem [ thm : kcrossknest ] ) states that the numbers of @xmath0-noncrossing and @xmath0-nonnesting graphs with a given degree sequence are the same . chen et al . @xcite prove the equality of the numbers of @xmath0-noncrossing and @xmath0-nonnesting graphs for two subclasses of graphs , namely for perfect matchings and for partition graphs , also counted by degree sequences ( under a different but equivalent terminology ) . a perfect matching is a graph where each vertex has degree one , and a partition graph is a graph that is a disjoint union of monotone paths , that is , where each vertex has at most one edge to its right and at most one to its left . the latter correspond in a natural way to set partitions , hence the result can be stated in terms of these . the paper @xcite also contains other identities and enumerative results on @xmath0-noncrossing and @xmath0-nonnesting matchings and partitions . krattenthaler @xcite deduces most of these from his more general results on fillings of ferrers diagrams . in this paper we also use fillings of diagrams to prove results about graphs . the difference is that whereas in @xcite , and also in @xcite , the results about graphs follow from general theorems by restricting the shape of the diagram , here we show that the results about graphs are in fact equivalent to those about fillings with arbitrary shapes . the main idea is to encode graphs by fillings of ferrers diagrams in such a way that @xmath0-crossings and @xmath0-nestings are easy to recognize . a @xmath0-noncrossing ( @xmath0-nonnesting ) graph becomes a filling of a diagram that avoids the identity ( antiidentity ) matrix of order @xmath0 , and the degree sequence of the graph can be recovered from the shape of the diagram and the row and column sums of the filling . then proving that there are as many @xmath0-noncrossing as @xmath0-nonnesting graphs is equivalent to showing that the numbers of fillings avoiding these two matrices are the same . this idea generalizes easily to other subgraphs in addition to crossing and nestings , and allows us to show that the study of fillings of ferrers diagrams with forbidden configurations is equivalent to the study of graphs avoiding certain subgraphs , in the sense defined in section [ sec : degree ] . the structure of the paper is as follows . in section [ sec : general ] we show that the equality of the numbers of @xmath0-noncrossing and @xmath0-nonnesting graphs counted by size and order is already in the literature , although not explicitly stated in this form . we introduce some notation on pattern avoiding fillings of ferrers diagrams and we rephrase results of krattenthaler @xcite and jonsson and welker @xcite in terms of @xmath0-noncrossing and @xmath0-nonnesting graphs . section [ sec : degree ] introduces a new correspondence between graphs and fillings of diagrams that keeps track of degree sequences . then we discuss why , from the perspective of pattern avoiding , graphs and fillings of diagrams are equivalent objects . in particular , showing that the number of @xmath0-noncrossing graphs with a fixed degree sequence equals the number of such @xmath0-nonnesting graphs is equivalent to proving a result on fillings of diagrams with restrictions on the row and column sums . our proof is an adaptation of the one in @xcite to allow arbitrary entries in the filling , and this is the content of section [ sec : proof ] . we conclude with some remarks and open questions . we start by setting some notation on fillings of ferrers diagrams . let @xmath15 be an integer partition . the _ ferrers diagram of shape @xmath16 _ ( or simply a _ diagram _ ) is the arrangement of square cells , left - justified and from top to bottom , having @xmath17 cells in row @xmath18 , for @xmath18 with @xmath19 . for a ferrers diagram @xmath20 of shape @xmath16 with rows indexed from top to bottom and columns from left to right , a _ filling _ @xmath21 of @xmath20 consists of assigning a nonnegative integer to each cell of the diagram . we say that a cell is _ empty _ if it has been assigned the integer @xmath22 . let @xmath23 be an @xmath24 @xmath25 matrix . we say that the filling _ contains @xmath23 _ if there is a selection of rows @xmath26 and columns @xmath27 of @xmath20 such that if @xmath28 then the cell @xmath29 of @xmath20 is nonempty and moreover the cell @xmath30 is in the diagram ( in other words , we require that the matrix @xmath23 is fully contained in @xmath20 ) . we say that the filling _ avoids @xmath23 _ if there is no such selection of rows and columns . if a filling @xmath21 contains @xmath23 , by an _ occurrence _ of @xmath23 we mean the set of cells of @xmath20 that correspond to the @xmath31 s in @xmath23 . we are mainly concerned about diagrams avoiding the identity matrix @xmath32 and the antiidentity matrix @xmath33 ; the latter is the matrix with @xmath31 s in the main antidiagonal and @xmath22 s elsewhere . as an example of these concepts , figure [ fig : diagram1 ] shows a filling of a diagram of shape @xmath34 that contains the matrices @xmath35 and @xmath36 but avoids @xmath37 . ( for clarity , we omit the zeros corresponding to the empty cells . ) and @xmath36 but avoids @xmath37 . right : the graph determined by the filling has a @xmath38-nesting and several @xmath39-crossings , but no @xmath38-crossing.,width=340 ] studying fillings of diagrams avoiding matrices is a natural generalization of pattern avoiding permutations , as explained in @xcite . we explore two types of connections between graphs and fillings of diagrams . the first one is straightforward , being essentially the adjacency matrix , and it has been used in @xcite to derive results on @xmath0-noncrossing maximal graphs and @xmath0-noncrossing and @xmath0-nonnesting matchings and partitions . suppose @xmath2 is a graph on @xmath3 $ ] and consider a diagram @xmath40 of shape @xmath41 . then if there are @xmath42 edges joining vertices @xmath18 and @xmath43 , with @xmath44 , fill the cell of column @xmath18 and row @xmath45 with @xmath46 . let this filling of the diagram be called @xmath47 . obviously the sum of the entries of @xmath47 is the number of edges of @xmath2 and the number of vertices is just one plus the number of rows of @xmath40 . if @xmath2 is a simple graph , then @xmath47 is a @xmath25 filling . if the edges @xmath48 are a @xmath0-nesting of @xmath2 , then @xmath47 contains the @xmath49 identity matrix @xmath50 in columns @xmath51 and rows @xmath52 . similarly , if @xmath2 contains a @xmath0-crossing , then @xmath47 contains the antiidentity matrix @xmath53 ( the condition @xmath54 guarantees that the matrix is indeed contained in the diagram ) . krattenthaler @xcite derives many of the results of chen et al . @xcite for matchings and partitions by specializing to @xmath47 his results on fillings of diagrams avoiding large identity or antiidentity matrices . his theorem 13 gives a generalization to arbitrary graphs which is implicitly included in the remark after it ; we explicitly state his result here ( see also the comment after theorem [ thm : kcrossknest ] in the next section ) . the following is a weaker version of ( * ? ? ? * theorem 13 ) [ thm : thm13 ] for any diagram @xmath20 and any integer @xmath55 , consider fillings of @xmath20 with nonnegative integers adding up to @xmath55 . then for each @xmath56 , the number of such fillings that do not contain the identity matrix @xmath50 equals the number of fillings that do not contain the antiidentity matrix @xmath53 . by restricting to @xmath57 we immediately get the following . [ cor : kcrossknestmulti ] the number of @xmath0-noncrossing graphs with @xmath58 vertices and @xmath55 edges equals the number of @xmath0-nonnesting such graphs . actually , from the statement of ( * ? ? ? * theorem 13 ) one gets a stronger result . for this we need to introduce weak @xmath0-crossings and weak @xmath0-nestings . the edges @xmath48 are a _ weak @xmath0-crossing _ if @xmath59 ; similarly , they are a _ weak @xmath0-nesting _ if @xmath60 @xmath61 . let @xmath62 ( respectively , @xmath63 ) be the largest @xmath0 for which @xmath2 has a weak @xmath0-crossing ( resp . , weak @xmath0-nesting ) . then the following is a corollary of the full version of ( * ? ? ? * theorem 13 ) . [ cor : weak ] the number of graphs with @xmath58 vertices and @xmath55 edges with @xmath64 and @xmath65 equals the number of such graphs with @xmath66 and @xmath67 . ideally , one would like to have an analogous result proving the symmetry of the distribution of @xmath12 and @xmath13 for all graphs . this is known to be true for matchings and partition graphs ( * ? ? ? * theorem 1 and corollary 4 ) ; actually , for these graphs weak crossings ( respectively , weak nestings ) are the same as crossings ( resp . , nestings ) . for simple graphs , the result would follow if problem 2 in @xcite has a positive answer for the diagram @xmath40 . the bijection used to prove ( * ? ? ? * theorem 13 ) does not preserve the values of the entries of the filling , so we can not deduce from it the corresponding result for simple graphs . however , this follows from a result of jonsson and welker . they deal with fillings not of diagrams , but of stack polyominoes . stack polyomino _ consists of taking a diagram , reflecting it through the vertical axis , and gluing it to another ( unreflected ) diagram . the content of a stack polyomino is the multiset of the lengths of its columns . the definitions of fillings and containment of matrices in stack polyominoes are analogous to those for diagrams . the following is corollary 6.5 of @xcite . ( the particular case where @xmath55 below is maximal was proved in @xcite . ) [ thm : pfaffians ] the number of @xmath25 fillings of a stack polyomino with @xmath55 nonzero entries that avoid the matrix @xmath50 depends only on the content of the polyomino and not on the ordering of the columns . by a simple reflection argument we get the following for the triangular diagram @xmath40 of shape @xmath41 : the number of @xmath25 fillings of @xmath40 with @xmath55 non - zero entries and that avoid the matrix @xmath50 is the same as those that avoid the matrix @xmath53 . hence we have the following in terms of graphs . [ cor : simple ] the number of @xmath0-noncrossing simple graphs on @xmath58 vertices and @xmath55 edges equals the number of such @xmath0-nonnesting simple graphs . in the next section we deal with graphs with a fixed degree sequence . for this we need to consider diagrams of arbitrary shapes , since the correspondence between graphs and fillings is no longer restricted to the triangular diagram @xmath40 . the _ left - right degree sequence _ of a graph on @xmath3 $ ] is the sequence @xmath68 , where @xmath69 ( resp . , @xmath70 ) is the left ( resp . , right ) degree of vertex @xmath18 ; by the left ( resp . , right ) degree of @xmath18 we mean the number of edges that join @xmath18 to a vertex @xmath43 with @xmath71 ( resp . , @xmath72 ) . obviously @xmath73 is the degree of vertex @xmath18 ( loops are not allowed ) . for instance , if @xmath74 and @xmath75 for all @xmath18 , then the graph is either a matching or a partition graph , perhaps with some isolated vertices . if a graph @xmath2 has @xmath76 as its left - right degree sequence , we say that @xmath2 is a graph _ on @xmath76_. a useful way of thinking of left - right degree sequences is drawing for each vertex @xmath18 , @xmath69 half - edges going left and @xmath70 half - edges going right . then a graph is just a way of matching these half - edges ; recall that we allow multiple edges . for completeness we mention here that a sequence @xmath77 is the left - right degree sequence of some graph on @xmath3 $ ] if and only if @xmath78.\ ] ] this and the next section are devoted to proving that for each left - right degree sequence @xmath76 there are as many @xmath0-noncrossing graphs on @xmath76 as @xmath0-nonnesting . we stress that the fact that we allow multiple edges is essential , since if we restrict to simple graphs the result does not hold . for instance , one can check that there is one simple nonnesting graph with left - right degree sequence @xmath79 , but no such noncrossing simple graph . however , it turns out that there is a bijection between @xmath0-noncrossing and @xmath0-nonnesting simple graphs that preserves left degrees ( or right degrees , but not both simultaneously ) . this follows from the following result of rubey ( * ? ? ? * theorem 4.2 ) applied to the filling @xmath47 by noting that the sum of the entries in row @xmath45 of @xmath47 corresponds to the left degree of vertex @xmath43 . rubey s result is for moon polyominoes , but we state the version for stack polyominoes . ( a weaker version of this result was proved by jonsson ( * ? ? ? * corollary 26 ) . ) [ thm : cor26 ] for any stack polyomino @xmath80 with @xmath81 rows and for any sequence @xmath82 of nonnegative integers , the number of @xmath25 fillings of @xmath80 that avoid @xmath50 and have @xmath83 nonzero entries in row @xmath18 depends only on the content of @xmath80 and not on the ordering of the columns . by the same reflection argument as at the end of section [ sec : general ] we obtain the following corollary . [ cor : simplemax ] let @xmath84 be a sequence of nonnegative integers . then the number of @xmath0-noncrossing simple graphs on @xmath3 $ ] with vertex @xmath18 having left degree @xmath69 for @xmath85 is the same as the number of such @xmath0-nonnesting simple graphs . the main result of this paper says that by allowing multiple edges we can simultaneously fix left and right degrees . [ thm : kcrossknest ] for any left - right degree sequence @xmath76 , the number of @xmath0-noncrossing graphs on @xmath76 equals the number of @xmath0-nonnesting graphs on @xmath76 . this result generalizes to arbitrary graphs some of the results of @xcite , which are only for partition graphs but also taking into account degree sequences ( with different terminology ) . to approach theorem [ thm : kcrossknest ] we could use again the filling @xmath47 of the previous section and fix the sums of the entries in each row and column . in this setting theorem [ thm : kcrossknest ] is again implicitly included in the remark after ( * ? ? ? * theorem 13 ) by keeping track of the changes in the partitions involved in the proof of that theorem . ( i am grateful to christian krattenthaler for this observation . ) however , our approach consists of encoding graphs not by the triangular diagram @xmath40 but by an arbitrary diagram whose shape depends on the degree sequence . by doing this we actually show that not only results on @xmath0-noncrossing and @xmath0-nonnesting graphs can be deduced from results on fillings of ferrers diagrams avoiding @xmath50 and @xmath53 , but that actually these two families of results are completely equivalent . moreover , we have an analogous assertion for arbitrary matrices ( see theorem [ thm : equivalence ] ) . we start with an easy lemma that follows immediately from the fact that the edges in a @xmath0-crossing or a @xmath0-nesting must be vertex - disjoint . [ lem : split ] the number of @xmath0-noncrossing ( resp . , @xmath0-nonnesting ) graphs with left - right degree sequence @xmath86 is the same as that of those with left - right degree sequence @xmath87 for any @xmath18 with @xmath88 . hence it is enough to prove theorem [ thm : kcrossknest ] for left - right degree sequences whose elements @xmath89 are such that either @xmath69 or @xmath70 is @xmath22 . we call these graphs _ left - right _ graphs ; note though that we do not require that the degrees of the vertices alternate between right and left . the case where both left and right degrees are @xmath22 corresponds to an isolated vertex . we now describe a bijection between left - right graphs and fillings of ferrers diagrams of arbitrary shape ; this bijection has the property that the left - right degree sequence of the graph can be recovered from the shape and filling of the diagram . let @xmath2 be a left - right graph . if the degree of vertex @xmath18 is of the form @xmath90 we say that @xmath18 is _ opening _ , and if it is of the form @xmath91 we say that @xmath18 is _ closing_. an isolated vertex is both opening and closing . let @xmath92 be the closing vertices of @xmath2 and let @xmath93 be the opening ones . for each closing vertex @xmath18 , let @xmath94 be the number of vertices @xmath43 with @xmath71 that are opening . we consider a diagram @xmath95 of shape @xmath96 , and if there are @xmath46 edges going from the opening vertex @xmath97 to the closing vertex @xmath98 , we fill the cell in column @xmath81 and row @xmath99 with the integer @xmath46 ( see figure [ fig : diagram2 ] ) . thus graphs with left degrees @xmath100 and right degrees @xmath101 correspond to fillings of this diagram with nonnegative entries such that the sum of the entries in row @xmath18 is @xmath69 and the sum of the entries in column @xmath43 is @xmath102 . conversely , any filling of a diagram arises in this way . indeed , given a filling @xmath21 of a diagram @xmath20 , the shape of @xmath20 gives the ordering of the opening and closing vertices of the graph , the row and column sums give the left and right degrees ( it is easy to see that they must satisfy equation ( [ eq : deg ] ) ) , and the entries of the filling give the edges of the graph . given a graph @xmath2 , we denote by @xmath103 the filling of @xmath95 corresponding to @xmath2 . similarly , given a filling @xmath21 of a diagram , we denote by @xmath104 the left - right graph corresponding to this filling . in this setting , it is immediate to check that again @xmath0-crossings of @xmath2 correspond to occurrences of @xmath50 in @xmath103 and @xmath0-nestings to occurrences of @xmath53 . of a diagram with row sums @xmath105 and column sums @xmath106 and the corresponding graph @xmath104 . , width=340 ] by a _ diagram with prescribed row and column sums _ we mean a diagram and two sequences @xmath107 and @xmath108 of nonnegative integers such that the only fillings allowed for this diagram are those where the row and column sums are given by the sequences @xmath107 and @xmath108 . given two matrices @xmath23 and @xmath109 , we say that they are _ equirestrictive _ if for all diagrams @xmath20 with prescribed row and column sums , the number of fillings of @xmath20 that avoid @xmath23 equals the number of fillings of @xmath20 that avoid @xmath109 . with this notation , theorem [ thm : kcrossknest ] is an immediate consequence of the following result , the proof of which is the content of the next section . [ thm : avoidk ] the identity matrix @xmath50 and the antiidentity matrix @xmath53 are equirestrictive . before moving to the proof of theorem [ thm : avoidk ] , let us make some remarks and point out some consequences of the proof . we start by further exploring the bijection between left - right graphs and fillings of diagrams . let @xmath2 and @xmath110 be graphs on @xmath3 $ ] and @xmath111 $ ] , respectively , with @xmath112 . for the rest of this section we assume that @xmath110 is simple ( but @xmath2 can have multiple edges as usual ) . we say that @xmath2 contains @xmath110 if there is an order - preserving injection @xmath113\rightarrow [ n]$ ] such that if @xmath4 is an edge of @xmath110 then @xmath114 is an edge of @xmath2 . for instance , a @xmath0-noncrossing graph is a graph that does not contain the graph on @xmath115 $ ] with edges @xmath116 . a @xmath25 matrix @xmath23 with @xmath81 rows and @xmath117 columns can also be viewed as a filling of the diagram of shape @xmath118 . by the correspondence between graphs and fillings of diagrams described above , we have that @xmath23 gives a graph @xmath119 with @xmath117 opening vertices and @xmath81 closing vertices and such that all opening vertices appear before the closing vertices . let us call such a graph a _ split graph _ , a particular case being the graph of a @xmath0-crossing or a @xmath0-nesting . as a consequence of the previous discussion we have that in terms of containment of substructures ( matrices or split graphs ) , fillings of diagrams and graphs are equivalent objects . [ thm : bijection ] for any split graph @xmath110 there is a matrix @xmath120 such that a left - right graph @xmath2 contains @xmath110 if and only if the filling @xmath103 contains @xmath120 . and conversely , for each matrix @xmath23 there is a split graph @xmath121 such that a filling @xmath21 of a diagram @xmath20 contains @xmath23 if and only if the graph @xmath104 contains @xmath121 . observe now that lemma [ lem : split ] can be generalized by substituting `` @xmath0-noncrossing graphs '' with `` graphs that do not contain the split graph @xmath110 '' . hence the following . [ thm : equivalence ] let @xmath110 and @xmath122 be two split graphs . then for any left - right degree sequence @xmath76 there are as many graphs on @xmath76 avoiding @xmath110 as graphs on @xmath76 avoiding @xmath122 if and only if for each diagram with prescribed row and column sums there are as many fillings avoiding @xmath120 as fillings avoiding @xmath123 . following the notation for matrices , we say that two split graphs @xmath110 and @xmath122 are _ equirestrictive _ if for any left - right degree sequence @xmath76 , there are as many graphs on @xmath76 avoiding @xmath110 as graphs on @xmath76 avoiding @xmath122 . all the split graphs that are known to be equirestrictive are obtained from the graph of a @xmath0-crossing or a @xmath0-nesting by using proposition [ prop : avoidextension ] from the next section . this proposition states that if @xmath23 and @xmath109 are equirestrictive matrices , then for any other matrix @xmath124 the matrices @xmath125 defined by blocks are also equirestrictive . this has the following implications in terms of graphs . given a split graph @xmath110 on @xmath111 $ ] , a _ @xmath126-crossing _ is a graph on @xmath127 such that the graph induced by the vertices @xmath128\cup \{k+h+1,\ldots,2k+h\}$ ] is a @xmath0-crossing , the graph induced by @xmath129 is @xmath110 , and there are no other edges . a _ @xmath126-nesting _ is defined similarly . then by combining theorems [ thm : avoidk ] and [ thm : equivalence ] and proposition [ prop : avoidextension ] we deduce the following . [ cor : kh ] for any split graph @xmath110 and any nonnegative integer @xmath0 , @xmath126-crossings and @xmath126-nestings are equirestrictive . observe that if we take @xmath110 to be an @xmath130-nesting , a @xmath126-nesting is a @xmath131-nesting , so it follows that a @xmath0-nesting , a @xmath0-crossing , and any combination of a @xmath117-crossing over " a @xmath132-nesting are equirestrictive . however , it is not true that @xmath117-nestings over @xmath132-crossings are equirestrictive , not even within matchings , as observed in the remark after theorem 1 of @xcite . this implies also that there is no analogous version of proposition [ prop : avoidextension ] where @xmath124 is the top - left block and @xmath23 and @xmath109 are the bottom - right blocks of the matrix . finally , we comment on the results known for @xmath25 fillings of diagrams with row and column sums equal to @xmath31 . our correspondence translates these results into results for matchings and partition graphs , as we next explain . in the literature , two permutation matrices @xmath23 and @xmath109 are called _ shape - wilf - equivalent _ if for each diagram @xmath20 with row and column sums set to @xmath31 , the number of fillings avoiding @xmath23 equals the number of fillings avoiding @xmath109 . ( in view of this notation , we could have chosen the name graph - wilf - equivalent instead of equirestrictive . ) let @xmath133 be a @xmath134 permutation matrix . the split graph corresponding to @xmath133 is a matching ( these are sometimes called _ permutation matchings _ ) . now if two permutation matrices @xmath133 and @xmath135 are shape - wilf - equivalent , then by straightforward application of theorem [ thm : equivalence ] we have that for all graphs whose left and right degrees are one , the number of graphs avoiding the matching @xmath136 equals the number of graphs avoiding the matching @xmath137 . since graphs with left and right degrees one are exactly partition graphs , it turns out that shape - wilf - equivalence is equivalent to the matchings @xmath136 and @xmath137 being equirestrictive among partition graphs , counted by left - right degree sequences . there are not many pairs of permutation matrices known to be shape - wilf - equivalent . backelin , west , and xin @xcite show that @xmath50 and @xmath53 are shape - wilf - equivalent ; in graph theoretic terms , this gives another alternative proof of the equality between @xmath0-noncrossing and @xmath0-nonnesting partition graphs from @xcite . let us mention here that krattenthaler @xcite deduces both the result of chen et al . and that of backelin , west , and xin from his theorem 3 , but for the first one he sets @xmath57 and for the second he restricts the number of non - empty cells in the filling ( and takes arbitrary shapes ) . since these two apparently unrelated results are in fact equivalent , it is obvious that they must follow from the same theorem , but it is interesting that they do in different ways . another observation is that by lemma [ lem : split ] , and its generalization to split graphs , if we know that two split graphs are equirestrictive within matchings , then they are so within partition graphs . for instance , a bijective proof of the equality of the numbers of @xmath0-noncrossing and @xmath0-nonnesting matchings would immediately give a bijection for @xmath0-noncrossing and @xmath0-nonnesting partition graphs . in addition to the matrices @xmath32 and @xmath33 and the ones that follow from proposition [ prop : avoidextension ] , the only other pair of matrices known to be shape - wilf - equivalent are ( see @xcite ) @xmath138 the graph theoretic version of this result has been independently proved by jelnek @xcite . it is not known to us if @xmath139 and @xmath140 are also equirestrictive , or more generally if there is a pair of shape - wilf - equivalent permutation matrices that are not equirestrictive . lastly , let us mention that all the discussion of this section can be carried out with almost no changes to the case where the matrix we want to avoid in the filling can have arbitrary nonnegative entries ; this corresponds to avoiding split graphs with multiple edges . the interested reader will have no problems in filling in the details . this section is devoted to the proof of theorem [ thm : avoidk ] . we show that we can adapt to our setting the proof of @xcite , which is for shape - wilf - equivalence , that is , row and column sums equal to @xmath31 ; we include the details for the sake of completeness . ( actually , @xcite contains two proofs of the analogous of our theorem [ thm : avoidk ] for shape - wilf - equivalence ; the proof we adapt is the first one . ) this bijection has been further studied in @xcite . here we show that it extends , in a quite straightforward way , to arbitrary fillings . this gives a result stronger than theorem [ thm : avoidk ] , the consequences of which in graph theoretic terms have already been pointed out at the end of the previous section . let us also mention that theorem [ thm : avoidk ] can also be proved using the techniques of @xcite . from now on @xmath141 denotes a diagram with prescribed row and column sums . when we say that a cell is above ( or below , to the right , to the left ) of another cell we always mean strictly . if we say that a cell is weakly above ( below , etc . ) we mean not above ( not below , etc . ) if @xmath124 and @xmath142 are two matrices , by @xmath143 $ ] we mean the matrix having @xmath124 and @xmath142 as blocks , that is , @xmath144 [ prop : avoidextension ] let @xmath23 and @xmath109 be a pair of equirestrictive matrices and let @xmath124 be any matrix . then the matrices @xmath145 $ ] and @xmath146 $ ] are also equirestrictive . let @xmath21 be a filling of the diagram @xmath141 that avoids @xmath145 $ ] . let @xmath147 be the set of cells @xmath148 of @xmath20 such that the cells to the right and below @xmath148 contain the matrix @xmath124 . @xmath147 is a diagram , since if @xmath148 is in @xmath147 all the cells weakly above and weakly to the left of it are also in @xmath147 . now set the row and column sums of @xmath147 according to the restriction of @xmath21 to @xmath147 , call it @xmath149 , giving a diagram @xmath150 . now @xmath149 is a filling of @xmath150 that avoids @xmath23 , so by assumption there is a bijection between such fillings and the ones that avoid @xmath109 . change the entries of @xmath21 corresponding to @xmath147 to obtain a filling of @xmath141 that avoids @xmath146 $ ] . the bijection in the other direction goes just in the same way . let @xmath151 be the matrix @xmath152 $ ] . the proof of the following proposition takes the rest of this section . [ prop : jt_ft ] for all @xmath117 , @xmath151 and @xmath33 are equirestrictive . we get as a corollary a stronger version of theorem [ thm : avoidk ] . [ cor : it_jt ] for all @xmath117 , @xmath153 $ ] and @xmath154 $ ] are equirestrictive . by proposition [ prop : avoidextension ] it is enough to show that @xmath32 and @xmath33 are equirestrictive . the proof is by induction on @xmath117 ; clearly @xmath155 and @xmath156 are equirestrictive . by proposition [ prop : jt_ft ] , it is enough to show that @xmath157 and @xmath158 are equirestrictive , and this follows by the induction hypothesis combined with proposition [ prop : avoidextension ] . a sketch of the proof of proposition [ prop : jt_ft ] is as follows . we first define two maps between fillings that transform occurrences of @xmath151 into occurrences of @xmath33 , and conversely , and use them to define to algorithms that transform a filling avoiding @xmath151 into a filling avoiding @xmath33 , and conversely . the fact that these two algorithms are inverses of each other follows from a series of lemmas . for any filling @xmath21 , given two occurrences @xmath159 and @xmath160 of @xmath33 in @xmath21 , we say that @xmath159 _ precedes _ @xmath160 if the first entry in which they differ , from left to right , is either higher in @xmath159 or it is at the same height and the one in @xmath159 is to the left . so two occurrences are either equal or comparable . the order for the occurrences of @xmath151 goes the other way around , i.e. , we look at the first entry in which they differ , from right to left , and the lower entries have preference , and if they are at the same height , the one more to the right goes first . let @xmath21 be a filling with the first occurrence of @xmath33 in rows @xmath161 and columns @xmath162 . let @xmath163 be the result of substracting @xmath31 from each cell @xmath164 , @xmath165 and adding @xmath31 to each cell @xmath166 , @xmath167 and to cell @xmath168 . since row and column sums have not been altered , @xmath163 is a filling of @xmath141 . so we have changed an occurrence of @xmath33 to an occurrence of @xmath151 . define @xmath169 as the inverse procedure , that is , @xmath169 takes a filling of the diagram , looks for the first occurrence of @xmath151 , and replaces it by an occurrence of @xmath33 . we define the algorithms @xmath170 and @xmath171 in the following way . algorithm @xmath170 starts with a filling avoiding @xmath151 and applies @xmath172 successively until there is no occurrence of @xmath33 . the result ( provided the algorithm finishes ) is a filling that avoids @xmath33 . similarly , algorithm @xmath171 starts with a filling avoiding @xmath33 and applies @xmath169 until there are no occurrences of @xmath151 left . we claim that @xmath170 and @xmath171 are inverse of each other . we prove this through a series of analogous lemmas . it is enough to prove the following claims . * that both algorithms end . ( lemmas [ lem : nojtabove ] and [ lem : noftbelow2 ] . ) * that @xmath173 for all @xmath58 . ( lemma [ lem : psipicksb ] . ) * that @xmath174 for all @xmath58 . ( lemma [ lem : phipicksa ] . ) in order to prove these claims , we need to investigate some properties of the maps @xmath172 and @xmath169 . we start by studying the map @xmath172 . let us first introduce some notation . let @xmath21 be a filling of the diagram and let @xmath175 be the cells of the first @xmath33 in @xmath21 , listed from left to right ; say they are @xmath176 . so in each cell @xmath177 there is a positive integer , possibly greater than one . let @xmath178 be the cells @xmath179 and @xmath168 ; hence , @xmath178 are the cells corresponding to the occurrence of @xmath151 that is created after applying @xmath172 to @xmath21 . so cell @xmath180 is in the same row as @xmath181 and in the same column as @xmath177 , for @xmath18 with @xmath182 . consider now the following two paths of cells determined by @xmath175 and @xmath178 ( see figure [ fig : abe ] ) . the path @xmath124 starts at the leftmost cell in the row of @xmath183 , continues to the right until it reaches the column of @xmath184 , then takes this column up until it hits cell @xmath184 , then turns right until reaching the column of @xmath185 , goes up until @xmath185 , then turns right again , and so on , until it reaches cell @xmath186 , at which point continues up until the top of the diagram . the path @xmath142 is defined in a similar manner . it starts at the leftmost cell of the row of cell @xmath187 , and goes right until it hits @xmath187 . then it turns up until the row of @xmath188 , where it turns and continues to the right until hitting @xmath188 . then it goes up until the row of @xmath189 , and then turns to the right until @xmath189 , and so on , until reaching @xmath190 , at which point it goes up until reaching the top of the diagram . since @xmath175 are the first occurrence of @xmath33 , the cells that are both to the right of @xmath142 and to the left of @xmath124 are empty , or , in other words , this region of the diagram avoids @xmath156 . we denote this region by @xmath191 . the choice of the first @xmath33 also imposes some other less trivial bounds on the longest @xmath192 s that can be found in some other areas determined by @xmath191 . note that in the next lemma the area left of @xmath191 includes the path @xmath142 . and @xmath191,width=340 ] [ lem : leftofe ] with the above notation , the following hold for any filling @xmath21 and for the corresponding @xmath163 . * for all @xmath18 with @xmath182 , there is no @xmath192 below @xmath180 and to the left of @xmath191 . * for all @xmath18 with @xmath182 , there is no @xmath193 above and to the right of @xmath180 and to the left of @xmath191 . * for all @xmath194 with @xmath195 , the rectangle determined by @xmath180 and @xmath196 contains no @xmath197 to the left of @xmath191 ; that is , there is no @xmath197 below @xmath196 , above @xmath180 , to the right of @xmath180 , and to the left of @xmath191 . the arguments below apply to both @xmath21 and @xmath163 since they do not use the entries in cells @xmath180 . * assume there was such a @xmath192 . then this @xmath192 together with @xmath198 would form a @xmath33 contradicting the choice of @xmath183 . * suppose there was such a @xmath193 . then @xmath199 followed by this @xmath193 form a @xmath33 that contradicts the choice of @xmath181 . * again , if there was such a @xmath197 , combined with @xmath200 and @xmath201 , it would create a @xmath33 contradicting the choice of @xmath177 . [ lem : nojtabove ] there is no @xmath33 in @xmath163 in the rows above @xmath183 we argue by contradiction . let @xmath2 be an occurrence of @xmath33 in @xmath163 . since @xmath172 picked @xmath183 as the topmost cell being the left - bottom cell of a @xmath33 , @xmath2 must use at least one of the cells @xmath202 . the idea is to substitute these cells @xmath180 , and possibly others , by some of the cells @xmath177 , to find an occurrence of @xmath33 in @xmath21 in the rows above @xmath183 , hence contradicting the choice of @xmath183 . now for each cell @xmath203 which belong to @xmath2 , find the largest integer @xmath43 such that all cells of @xmath2 above @xmath203 and weakly below @xmath204 lie left of @xmath191 . in this way it is possible to find two sequences @xmath205 and @xmath206 with the following properties : * @xmath207 , @xmath208 , and @xmath209 for all @xmath0 ; * @xmath210 is in @xmath2 ; * if @xmath211 is in @xmath2 , then @xmath212 for some @xmath0 ; * all cells of @xmath2 above @xmath210 and weakly below @xmath213 are to the left of @xmath191 , and @xmath214 is the largest integer with this property . now we show that we can replace the cells of @xmath2 that fall left of @xmath191 and are contained in the rectangles determined by @xmath210 and @xmath215 by some of the @xmath177 , giving an instance of @xmath33 contained in @xmath21 and above @xmath183 . we need to distinguish two cases , according to whether @xmath216 or not . assume first that @xmath217 . for each @xmath0 , consider the rectangles determined by @xmath210 and @xmath215 . by lemma [ lem : leftofe].(iii ) , there are at most @xmath218 elements of @xmath2 in this rectangle and to the left of @xmath191 . replace these cells , together with @xmath210 , by a ( possibly proper ) subset of @xmath219 . after doing this for each @xmath0 , we still have an occurrence of @xmath33 starting above @xmath183 , but now it is contained in the original filling @xmath21 , contradicting the hypothesis . now assume that @xmath216 . for @xmath220 , do the same substitutions as in the previous case ; for @xmath221 , we have by lemma [ lem : leftofe].(ii ) that there are at most @xmath222 cells of @xmath2 left of @xmath191 and above @xmath210 . replace these cells and @xmath210 by a subset of @xmath223 . again we obtain an occurrence of @xmath33 in @xmath21 that starts above @xmath183 , a contradiction . this lemma alone shows that algorithm @xmath170 terminates . indeed , after one application of @xmath172 all the cells in the row of @xmath183 and to the left of @xmath183 are empty ( because of the choice of @xmath183 ) , and the cell @xmath183 has decreased its value by one . so the leftmost cell of the first occurrence of @xmath33 in @xmath163 is either @xmath183 , or it is to the right of @xmath183 , or it is below @xmath183 . but since the value in cell @xmath183 decreases and cells to the left of @xmath183 stay empty , eventually there will be no occurrence of @xmath33 whose leftmost cell is @xmath183 . so the selection of @xmath33 s goes from top to bottom and from left to right , so for some @xmath58 the filling @xmath224 is free of @xmath33 s . it is not the case that if we apply @xmath172 to an arbitrary filling @xmath21 of @xmath141 we have that @xmath225 . but algorithm @xmath170 starts with a filling that avoids @xmath151 and the successive applications of @xmath172 create occurrences of @xmath151 from top to bottom and from left to right . we need to show that in this situation after each application of @xmath172 , the first occurrence of @xmath151 is precisely the one created by @xmath172 . the next lemmas are devoted to proving this . [ lem : noftbelow ] if @xmath21 contains no @xmath151 with at least one square below @xmath183 , then @xmath163 contains no such @xmath151 . the proof is similar to the one of the previous lemma . let @xmath2 be an occurrence of @xmath151 in @xmath163 with at least one cell below @xmath183 . since @xmath21 had no such occurrence , @xmath2 contains at least one of the cells @xmath180 . the bottom - right cell of @xmath2 is below @xmath183 , and it can not be to the right of @xmath226 , otherwise this cell together with @xmath227 would form an @xmath151 in @xmath21 . by an argument similar to the one in the previous lemma , we change all cells @xmath180 of @xmath2 , and possibly others , to some of the cells @xmath177 , so that at the end we have an occurrence of @xmath228 that together with the bottom - right cell of @xmath2 gives an occurrence of @xmath151 that contradicts the hypothesis . for each @xmath180 that is in @xmath2 , look for the smallest @xmath43 such that all cells in @xmath2 that are left of @xmath180 and weakly to the right of @xmath229 are left of @xmath191 . by doing this we find integers @xmath230 and @xmath206 with the following properties : * @xmath231 , @xmath232 , and @xmath233 for all @xmath0 ; * @xmath210 is in @xmath2 for all @xmath0 with @xmath234 ; * @xmath214 is the smallest integer such that all cells of @xmath2 that are left of @xmath210 and weakly to the right of @xmath235 are to the left of @xmath191 ; * if @xmath211 is in @xmath2 , then @xmath236 for some @xmath0 . we have to distinguish whether @xmath237 or not . assume first @xmath238 . since by lemma [ lem : leftofe].(iii ) there are at most @xmath239 cells of @xmath2 in the rectangle determined by @xmath215 and @xmath210 , these cells , together with @xmath210 , can be replaced by a ( possibly proper ) subset of @xmath240 . by doing this for all @xmath0 , we have an occurrence of @xmath228 in @xmath21 that together with the right - bottom cell of @xmath2 contradicts the hypothesis . if @xmath241 , then we do the same substitutions for all @xmath242 ; for @xmath221 , we have by lemma [ lem : leftofe].(i ) that there are at most @xmath243 cells of @xmath2 left of @xmath191 and below @xmath244 , so we can substitute those and @xmath244 by @xmath245 . after these substitutions , the result is again an occurrence of @xmath151 in @xmath21 that contains a cell below @xmath183 , contradicting the hypothesis . the following is easy but we state it for the sake of completeness . [ lem : noftright ] if @xmath21 contains no @xmath151 with a cell to the right of @xmath186 and below @xmath184 , then @xmath163 contains no such @xmath151 . again we argue by contradiction . suppose @xmath2 is an @xmath151 in @xmath163 that contains a cell to the right of @xmath186 and below @xmath184 . this cell together with @xmath246 gives an occurrence of @xmath151 in @xmath21 that contradicts to the assumption . [ lem : ftfromphi ] for each @xmath0 with @xmath247 , there is no @xmath53 in @xmath163 above @xmath183 and to the left of and below @xmath248 . let @xmath2 be an occurrence of such a @xmath53 . if @xmath2 contains none of @xmath249 , then @xmath2 followed by @xmath250 forms a @xmath33 in @xmath21 that is above @xmath183 , and this contradicts the choice of @xmath183 . hence , @xmath2 uses some @xmath180 for @xmath251 . by an argument analogous to that of the proof of lemma [ lem : nojtabove ] , we can substitute the cells @xmath180 that are in @xmath2 and possibly others by some @xmath177 s so that we get an occurrence of @xmath252 in @xmath21 that is below @xmath248 and above @xmath183 . this followed by @xmath250 , gives an @xmath33 in @xmath21 that contradicts the choice of @xmath183 . the following lemma is just a combination of the previous and induction ; it implies that the inverse of algorithm @xmath170 is @xmath171 . [ lem : psipicksb ] * if @xmath21 does not contain any occurrence of @xmath151 below @xmath183 , then the first occurrence of @xmath151 in @xmath163 is @xmath178 . * if @xmath21 is a filling that avoids @xmath151 , then @xmath173 . for the first statement , let @xmath253 be the first occurrence of @xmath151 in @xmath163 , with the elements ordered from left to right . recall that @xmath178 is an occurrence of @xmath151 in @xmath163 ; we need to show that @xmath254 for all @xmath18 . by lemma [ lem : noftbelow ] , @xmath255 is in the same row as @xmath256 . by lemma [ lem : noftright ] , @xmath255 can not be to the right of @xmath186 , hence @xmath257 . now use induction on @xmath258 . suppose we know @xmath259 . it is enough now to show that @xmath260 lies in the same row as @xmath180 , since all the cells to the right of @xmath180 but left of @xmath261 lie in @xmath191 , which we know contains only empty cells . but now lemma [ lem : ftfromphi ] guarantees that there is no @xmath192 below @xmath180 , to the left of @xmath261 , and above @xmath256 , as required . for the second statement , it follows by lemma [ lem : noftbelow ] and induction on @xmath58 that the filling @xmath262 contains no @xmath151 whose lowest cell is below the lowest cell of the first occurrence of @xmath33 . hence the previous statement applied to @xmath224 gives immediately that @xmath173 . so the inverse of algorithm @xmath170 is @xmath171 . now we only need to prove the converse . the proof follows exactly the same steps and we content ourselves by stating and proving the corresponding lemmas . actually in this case some proofs are slightly simpler . we keep the notation as above . let @xmath21 be now a filling of @xmath141 and let @xmath178 be the first occurrence of @xmath151 and let @xmath263 be the occurrence of @xmath33 in @xmath264 created after applying @xmath169 to @xmath21 . consider again the region @xmath191 as defined above . by the choice of @xmath178 as the first occurrence of @xmath151 in @xmath21 , all the cells of @xmath191 are again empty . [ lem : rightofe ] for all @xmath194 with @xmath265 , the rectangle determined by @xmath177 and @xmath266 contains no @xmath197 to the right of @xmath191 in either @xmath21 or @xmath264 ; that is , there is no @xmath197 below @xmath266 , above @xmath177 , to the left of @xmath266 , and to the right of @xmath191 . suppose there was such a @xmath197 . then @xmath267 , followed by this @xmath197 and then followed by @xmath268 gives an occurrence of @xmath151 in @xmath21 that contradicts the choice of @xmath204 . [ lem : noftbelow2 ] there is no @xmath151 in @xmath264 with at least one cell in a row below @xmath183 . suppose there is such an @xmath151 . its right - bottom cell is below @xmath183 and also weakly to the left of @xmath190 , since otherwise @xmath202 and this cell would form an @xmath151 contradicting the choice of @xmath256 . let @xmath2 be this occurrence of @xmath151 except the right - bottom cell . @xmath2 must contain some of the cells @xmath175 . as in the previous lemmas , the idea is to substitute the @xmath177 in @xmath2 together with other cells by some of the @xmath180 so that we obtain an occurrence of @xmath151 in @xmath21 contradicting the choice of @xmath256 . find integers @xmath205 and @xmath269 with the following properties : * @xmath270 , @xmath208 , and @xmath271 for all @xmath0 ; * @xmath272 is in @xmath2 for all @xmath0 with @xmath234 ; * @xmath214 is the largest integer such that all cells of @xmath2 that are to the right of @xmath272 and weakly to the left of @xmath273 are to the right of @xmath191 ; * if @xmath274 is in @xmath2 , then @xmath212 for some @xmath0 . now , by lemma [ lem : rightofe ] , there are at most @xmath218 elements of @xmath2 in the rectangle determined by @xmath272 and @xmath275 . together with @xmath272 , they account for at most @xmath276 elements of @xmath2 ; substitute them for a subset of @xmath277 . doing this for all @xmath0 , we get an occurrence of @xmath151 in @xmath21 that contains a cell below @xmath183 , hence contradicting the choice of @xmath178 as the first @xmath151 in @xmath21 . [ lem : nojtabove2 ] if @xmath21 contains no @xmath33 that is above @xmath183 , then @xmath264 contains no such @xmath33 . let @xmath2 be such a @xmath33 ; @xmath2 must contain some of the cells @xmath177 . find integers @xmath205 and @xmath269 with the following properties : * @xmath278 , @xmath279 , and @xmath280 for all @xmath0 ; * @xmath272 is in @xmath2 for all @xmath0 with @xmath234 ; * @xmath214 is the smallest integer such that all cells of @xmath2 that are below @xmath272 and weakly above @xmath281 are to the right of @xmath191 ; * if @xmath274 is in @xmath2 , then @xmath282 for some @xmath0 . as in the proof of the previous lemma , it is possible to substitute the elements of @xmath2 contained in the rectangles determined by @xmath272 and @xmath275 , plus the cell @xmath272 , by ( a subset of ) the elements @xmath283 . these substitutions give a @xmath33 in @xmath21 that is above @xmath183 , contrary to the hypothesis . [ lem : nojtleft ] if @xmath21 contains no @xmath33 with a cell to the left of @xmath183 and below @xmath184 , then neither does @xmath264 . if this were the case , the leftmost cell of this @xmath33 together with @xmath202 would give a @xmath33 contradicting the hypotheses [ lem : jtfrompsi ] if @xmath21 contains no @xmath33 above @xmath183 , there is no @xmath284 in @xmath264 above @xmath285 such that the lowest cell of this @xmath284 is weakly to the left of @xmath285 . suppose @xmath2 is an occurrence of such a @xmath284 . @xmath2 must contain some of the cells @xmath286 , otherwise @xmath287 followed by @xmath2 would form a @xmath33 contradicting the hypothesis . find integers @xmath205 and @xmath269 with the following properties : * @xmath278 , @xmath279 , and @xmath288 for all @xmath0 ; * @xmath272 is in @xmath2 for all @xmath0 with @xmath234 ; * @xmath214 is the smallest integer such that all cells of @xmath2 that are below @xmath272 and weakly above @xmath281 are to the right of @xmath191 ; * if @xmath274 is in @xmath2 , then @xmath282 for some @xmath0 . as before , the rectangle determined by @xmath272 and @xmath275 contains at most @xmath239 cells of @xmath2 ; these cells , together with @xmath272 , can be replaced by a subset of @xmath283 . after all these substitutions we get an occurrence of @xmath284 in @xmath21 that combined with @xmath287 gives an occurrence of @xmath33 in @xmath21 contradicting the hypothesis . [ lem : phipicksa ] * if @xmath21 does not contain any occurrence of @xmath33 above @xmath256 , then the first occurrence of @xmath33 in @xmath163 is @xmath175 . * if @xmath21 is a filling that avoids @xmath33 , then @xmath174 . for the first statement , let @xmath289 be the first occurrence of @xmath33 in @xmath264 , with cells listed from left to right . we want to show that @xmath290 for all @xmath18 with @xmath291 . by lemma [ lem : nojtabove2 ] , @xmath292 is in the same row as @xmath183 , and by lemma [ lem : nojtleft ] it is weakly to the right of @xmath183 , hence @xmath293 . now we proceed by induction on @xmath18 . suppose @xmath294 . by lemma [ lem : jtfrompsi ] we have that the only @xmath193 in @xmath264 that is weakly above and weakly to the left of @xmath181 is @xmath295 , hence @xmath296 , as needed . for the second statement , by induction and lemma [ lem : nojtabove2 ] we get that @xmath297 satisfies the hypothesis of part ( i ) , hence it follows that @xmath174 . in his paper @xcite , krattenthaler speaks of a bigger picture " that would englobe several recent results on pattern avoiding fillings of diagrams . we believe that our correspondence between graphs and fillings of diagrams also belongs to this picture and that it may shed some light in the understanding of it . we have shown that for each statement in pattern avoiding fillings there is a statement about graphs avoiding certain split graphs . so we can claim that in some sense the resources available to attack either problem have doubled . an example of this are the repeated " results in the literature mentioned at the end of section [ sec : degree ] . for completeness , we mention here a result by bousquet - mlou and steingrmsson @xcite that can be cast in terms of @xmath0-noncrossing and @xmath0-nonnesting graphs . they restrict to diagrams with self - conjugate shape and row and column sums are set to @xmath31 , and they only consider symmetric @xmath25 fillings ( that is , symmetric with respect to the main diagonal of the diagram ) . for these fillings , they show that @xmath32 and @xmath33 are equirestrictive . in terms of matchings , this says that for each left - right degree sequence , the number of @xmath0-noncrossing symmetric matchings is the same as the number of @xmath0-nonnesting ones , where a matching on @xmath298 $ ] is symmetric if it equals its reflection through the vertical axis that goes between vertices @xmath58 and @xmath299 . similar results for symmetric graphs can be deduced from ( * ? ? ? * theorem 15 ) . let us finish by going back to our initial motivation of studying @xmath0-noncrossing and @xmath0-nonnesting graphs . even if our main question has been answered positively , it is fair to say that it has not been solved in the most satisfactory way ; ideally we would like to find a bijective proof in graph theoretic terms . note that due to its roundabout character , our proof of theorem [ thm : avoidk ] does not give a clear bijection , neither in terms of graphs nor of fillings . also the proofs of corollaries [ cor : kcrossknestmulti ] , [ cor : simple ] , and [ cor : simplemax ] do not provide bijections in graph theoretic terms . a bijective proof of theorem [ thm : kcrossknest ] for @xmath300 has recently been found by jelnek , klazar , and de mier @xcite . other interesting questions related to @xmath0-crossings and @xmath0-nestings of graphs include , as mentioned before , to determine whether the pairs @xmath301 are symmetrically distributed among all graphs . this is already known for matchings and partition graphs @xcite . one would also hope for a wide generalization of theorem [ thm : kcrossknest ] stating that the number of graphs with @xmath302 @xmath0-crossings and @xmath81 @xmath0-nestings equals the number of graphs with @xmath81 @xmath0-crossings and @xmath302 @xmath0-nestings . again , the case @xmath300 is known for matchings @xcite and partition graphs @xcite . unfortuntely , for @xmath303 this is not true even for matchings ; for instance , marc noy @xcite checked that there are more matchings with six edges and only one @xmath38-crossing than with only one @xmath38-nesting . i am very grateful to sergi elizalde , vt jelnek , martin klazar , martin loebl , and marc noy for many fruitful and stimulating discussions and for pointing out several useful references .

the spiral magnetic order in cubic mnsi - type crystals is considered using the model of classical heisenberg ferromagnetics with an extra interaction of the dzyaloshinskii moriya ( dm ) type between neighboring atoms . for all shortest mn - mn bonds the dm vectors are expressed via the-vector for an arbitrary chosen bond , in accordance with the space group of these crystals and bond directions . the-vectors have the same length and 24 different directions so that the components of the-vectors are changing similarly to the components of the corresponding bond vectors . after minimization of the magnetic energy it is found how the wave vector of magnetic helices depends on : for any helix direction where is the exchange interaction constant and-vector is in the units of the lattice constant . the wave number determines both the sign and strength of _ global _ spiraling whereas _ locally _ , within a unit cell , the helical order can be strongly frustrated so that the twist angles between neighboring ferromagnetic layers may be even of different signs . conical deformations of helices caused by an arbitrary directed external magnetic field is also considered within the same model . the critical field of helix unwinding is found and it is shown that even in the unwound state there remains a residual periodic splay of magnetic moments with the splay angle proportional to which can be measured by diffraction methods . the third component of the-vector , , appears only in higher approximations . it is also demonstrated how the usually used continuous picture of moment distribution can be obtained from the discrete one in a coarse grain approximation .

1109.6163

the spiral arrangements appear in very different systems of condense matter physics and biology @xcite providing an important example of a primitive self - organization . sometimes this self - organization results in rather intricate structures ( may be the first staircases to more complicated biological patterns ) like the blue phases in chiral liquid crystals @xcite or the skyrmion textures in chiral magnetics with @xmath8 crystal structure ( mnsi , mnge , fe@xmath9co@xmath10si ) @xcite . there is a remarkable similarity in the physics of those so different systems @xcite . therefore it is interesting and challenging to look for local interactions responsible for macroscopic spiral ordering and , in particular , to find relationships between parameters of the local interaction and the final structures . for instance , the first question is what determines the period and handedness of helices ? most frequently , the helical arrangements are described phenomenologically , just by adding chiral terms like @xmath11 to the free energy of the system where vector @xmath12 describes a local system anisotropy . the period and handedness of helices are determined by the absolute value and sign of the pseudoscalar coefficient @xmath13 ; of course this term should be allowed by the system symmetry ( lack of inversion is required ) . in the chiral liquid crystals , the value of @xmath13 is believed to be determined by steric interactions between neighboring chiral molecules but any quantitative theory is still absent because of baffling complexity of the problem . in cubic magnetic crystals of the mnsi type , the helical ordering is induced by the dzyaloshinskii moriya ( dm ) interaction between neighboring magnetic moments @xmath14 and @xmath15 @xcite . this interaction @xcite can exist both in centrosymmetric and non - centrosymmetric crystals but only in the latter case it results in long - range helical structures . the most simple form of the dm term in the interaction energy is written as @xmath16 $ ] where pseudovector coupling coefficients @xmath17 originate from superexchange interaction via neighboring atoms . it was shown by moriya that vectors @xmath17 are determined by asymmetry of local atomic arrangements . recently it was carefully investigated , using synchrotron radiation and polarized neutron small - angle diffraction , what is the relationship between the handedness of the mnsi crystal structure and the sign of magnetic helix @xcite . it was unequivocally demonstrated that the structural chirality rigorously determines the magnetic chirality of these compounds . the interplay between structural chirality is also important for spin chirality in multilayer structures @xcite . however , no analytical formula related @xmath13 and @xmath17 has been found yet even for the simple case of interaction between magnetic moments of neighboring atoms . this problem is indeed even more complicated because the itinerant nature of magnetism in mnsi crystals . detailed numerical studies of the dm interaction in mnsi - type crystals are performed by hopkinson and kee @xcite and their results will be discussed below in comparison with the results of the present paper . a generic feature , intrinsic to complicated spacial structures , is _ frustration _ of their local ordering : the configurations with the best local energy can not be spread for all the system . in the chiral liquid crystals and mnsi heliomagnetics , the most favorable double - twist local arrangements are restricted by topological constrains existing in two - dimensional and three - dimensional spaces @xcite . nevertheless , the stable three - dimensional blue phases are possible in a narrow range of temperatures whereas the two - dimensional system of spin vertices can be stabilized by external magnetic field . however , in complicated crystals like la@xmath18cuo@xmath19 , there is an additional mechanism of frustrations @xcite related with different orientations of @xmath17 for different inter - spin bonds . we have found that this type of frustration is also intrinsic to the mnsi structure : at the atomic scale , spin rotation in the magnetic helices can be strongly nonmonotonic because of different orientations of @xmath17 for different bonds . as a result , unequal righthand and lefthand twists usually coexist in each of the helices . for instance , the coefficient @xmath13 can become zero even for nonzero @xmath17 . in this paper we first find a simple analytical expression related the helix period and handedness with the dm vectors of the shortest mn - mn bonds . in this consideration , we adopt a usually used approximation that the exchange interaction is stronger than dm interaction and twist angles between neighboring atoms are correspondingly small . it is demonstrated that for general orientation of dm vectors the mnsi helical structure is frustrated so that the twist angles between neighboring ferromagnetic layers have different signs . then , using the same approximations , conical distortions of helices in external magnetic fields are analyzed and the critical field of helix unwinding is determined . finally , the parameters of the known phenomenological description of mnsi helices are related to the dm vectors of the microscopic theory . in a simple model of classical heisenberg ferromagnetics with an extra interaction of the dzyaloshinskii moriya ( dm ) type , the energy of the system is expressed as a sum of pair interactions between magnetic atoms and the interaction of individual atoms with an external magnetic field @xmath20 : @xmath21 ) - g\mu_b \sum_i \mathbf{h } \cdot \mathbf{s}_i , \ ] ] where the first summation is taken over all pairs @xmath22 of magnetic atoms , @xmath23 is the exchange interaction constant , @xmath17 are the dzyaloshinskii moriya vectors , and @xmath24 is an effective magnetic moment . hereafter we will use the approximation of classical spins with ferromagnetic exchange coupling between nearest neighbors , @xmath25 , and suppose that @xmath26 for all magnetic atoms . using the nearest neighbor approximation for the mnsi - type crystals we can rewrite eq . ( [ eq : energy1 ] ) as @xmath27 ) - g\mu_b \sum_i \mathbf{h } \cdot \mathbf{s}_i , \ ] ] where the first sum ( @xmath28 ) is taken over all mn atoms of the crystal and the second sum ( @xmath29 ) is over six nearest mn neighbors of @xmath28-th atom , i.e. over six bonds connecting each mn atom with its nearest neighbors . it should be emphasized that vectors @xmath17 have different orientations for different bonds and relations between different @xmath17 can be found from symmetry reasons . in the unit cell of mnsi - type crystals , the magnetic atoms are at the @xmath30 positions with coordinates @xmath31 , @xmath32 , @xmath33 , @xmath34 @xcite . there are 24 crystallographically equivalent bonds between nearest neighbors , and dm vectors for all bonds can be obtained from dm vector of any bond using symmetry transformations of the point group @xmath35 of mnsi crystal augmented with the inversion ; the latter appears owing to the permutation of atoms : @xmath36 . the procedure is , in principle , very simple : we choose an arbitrary vector @xmath37 as dm vector of the `` first '' bond , say @xmath38 , connecting mn atom at @xmath31 with mn atom at @xmath39 . it should be an arbitrary vector because no symmetry operation is associated with the bond . then , for this cubic space group , any shortest bond between mn atoms can be obtained from the `` first '' one using cyclic permutations ( i.e. threefold rotations ) and/or sign inversions ( i.e. twofold rotations and inversions ) of its vector components . the components of the corresponding dm vector are expressed through @xmath40 , @xmath41 and @xmath42 by exactly the same cyclic permutations and/or sign inversions . the list of all the bonds and their @xmath43-vectors is given in table [ tabled ] . all @xmath17 have the same length : @xmath44 . it was shown @xcite that the dm term should be accompanied by two extra terms : @xmath45 these terms are of the same order of magnitude as the dm term because for small dm interaction , @xmath46 , small angles between @xmath14 and @xmath15 are of order @xmath47 . in eq . ( [ eq : energys ] ) , with the same accuracy we can put @xmath48 for a small domain containing all 12 different interatomic bonds and averaging of @xmath49 over these 12 bonds gives @xmath50 . ( [ eq : energys ] ) can be written as @xmath51 this term is isotropic and can be included into exchange term of eq . ( [ eq : energy1 ] ) . we see that in cubic crystals the extra terms given by eq . ( [ eq : energys ] ) are not important if @xmath46 . having a list of dm vectors one can calculate and minimize numerically the magnetic energies ( [ eq : energy1 ] ) and ( [ eq : energy2 ] ) for rather big sets of atoms with distinct moment configurations , including those similar to skyrmions . however in this paper we will use only analytic calculations for detailed consideration of helical structures . let us start with the simple case of free helices which provides a minimum of the magnetic energies ( [ eq : energy1 ] ) and ( [ eq : energy2 ] ) for @xmath52 . as it is known from numerous experiments , well below the curie point the magnetic moments of mn atoms form a simple helix . however , in the general case , the helices may be different for four different types of equivalent atomic positions in the unit cell ( i.e. for positions with four different orientations of the local threefold axes ) , and this is confirmed by numerical simulations . then , the spin structure can be described by the superposition of four helices @xmath53 so that the condition @xmath54 is obviously obeyed . here @xmath55 is an atomic position , @xmath56 and @xmath57 are perpendicular unit vectors , @xmath2 is a wave vector . upper index @xmath58 enumerates four atomic positions in the `` first '' unit cell : @xmath31 , @xmath59 , @xmath60 , @xmath61 , correspondingly . we express all the atomic positions in the units of the cubic unit cell so that both @xmath55 and @xmath2 vectors are dimensionless . notice that all four spirals given by eq . ( [ eq : s1 ] ) have the same wave vector whereas in the skyrmion - like structures there is a set of different wave vectors . it can be easily seen that the spins of sublattice @xmath62 rotate in the plane perpendicular to vector @xmath63 , \ ] ] whose direction generally speaking can differ from that of wave vector @xmath2 . now , using eq . ( [ eq : s1 ] ) we can rewrite expression ( [ eq : energy2 ] ) for energy ( at @xmath52 ) as @xmath64 ) \cos\mathbf{k } \cdot \mathbf{r}_i \cos\mathbf{k } \cdot \mathbf{r}_j \\ & + ( -j \mathbf{n}_{cos}^{t(i ) } \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{cos}^{t(i ) } \times \mathbf{n}_{sin}^{t(j ) } ] ) \cos\mathbf{k } \cdot \mathbf{r}_i \sin\mathbf{k } \cdot \mathbf{r}_j \\ & + ( -j \mathbf{n}_{sin}^{t(i ) } \cdot \mathbf{n}_{cos}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^{t(i ) } \times \mathbf{n}_{cos}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_i \cos\mathbf{k } \cdot \mathbf{r}_j \\ & + ( -j \mathbf{n}_{sin}^{t(i ) } \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^{t(i ) } \times \mathbf{n}_{sin}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_i \sin\mathbf{k } \cdot \mathbf{r}_j \ } , \end{array}\ ] ] where upper index @xmath65 designates `` type '' of @xmath28-th atom or its belonging to one of four sublattices , @xmath66 . the sines and cosines products in eq . ( [ eq : ehelix1 ] ) can be expressed through sines and cosines of arguments @xmath67 and @xmath68 , with @xmath69 being the distance between @xmath28-th and @xmath29-th atoms , @xmath70 . it should be noted here that dependence on atomic positions is present in these trigonometric functions only . after summation over all unit cells of the crystal all the functions of argument @xmath67 vanish because @xmath2 is not a reciprocal lattice vector , whereas the functions of argument @xmath68 do not depend on the unit cells to which @xmath28-th and @xmath29-th atoms belong . thus we can rewrite expression ( [ eq : ehelix1 ] ) for energy as @xmath71 ) \cos\mathbf{k } \cdot \mathbf{r}_{ij } \\ & + ( -j \mathbf{n}_{cos}^i \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{cos}^i \times \mathbf{n}_{sin}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_{ij } \\ & - ( -j \mathbf{n}_{sin}^i \cdot \mathbf{n}_{cos}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^i \times \mathbf{n}_{cos}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_{ij } \\ & + ( -j \mathbf{n}_{sin}^i \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^i \times \mathbf{n}_{sin}^{t(j ) } ] ) \cos\mathbf{k } \cdot \mathbf{r}_{ij } \ } , \end{array}\ ] ] where @xmath72 is the number of unit cells in the crystal , and first sum is now taken over four magnetic atoms within one unit cell , so @xmath73 . ( [ eq : ehelix2 ] ) seems to be simple enough for possible minimization . it depends on 15 variables only : three angles characterizing orientation of triad @xmath74 for each of four mn sublattices , and three components of the wave vector @xmath2 . the exchange interaction constant @xmath4 , dm vectors @xmath17 and distances @xmath69 between neighboring atoms are supposed to be invariable ; we do not consider phenomena ( e.g. magnetostriction ) changing these parameters . below , instead of numerical minimization , we will look for approximate analytical solutions relying on a small parameter . indeed , it is known from the experimental data that the spin - orbit interaction is usually much smaller than the exchange one , so that @xmath46 . in the mnsi crystal , strong exchange interaction leads to the ferromagnetic ordering , whereas small dm interaction results in small canting of neighboring magnetic moments . small values of @xmath75 and @xmath76 ( the latter is supposed to be of the order of @xmath75 ) allows us to perform approximate minimization of the energy . in particular , we can take advantage of the fact , that in this approximation the triads @xmath77 have close orientations for four mn sublattices . let us suppose that all triads @xmath78 are close to triad @xmath79 , with @xmath80 being a vector , which can differ from @xmath81 ( @xmath82 is a wave number ; the sign defines the chirality of the helix ) , and @xmath83 being an orthonormal basis of the plane perpendicular to @xmath80 , @xmath84 = \mathbf{n}_3 $ ] . it is useful to introduce vector @xmath85 for considering all partial helices on an equal footing and for possible future consideration of anisotropy effects . the deviation of triad @xmath78 from @xmath79 can be considered as a sum of two small rotations : first one on the axis @xmath80 by the angle @xmath86 , and another one by the angle @xmath87 . then , the vectors @xmath88 , @xmath89 , and @xmath90 can be represented as @xmath91 with @xmath92 being the rotation matrix by the angle @xmath93 . the substitution of eq . ( [ eq : basis2 ] ) into eq . ( [ eq : ehelix2 ] ) gives @xmath94 + [ ( r^i \mathbf{n}_2 ) \times ( r^{t(j ) } \mathbf{n}_2 ) ] \ } \cos(\omega_i - \omega_{t(j ) } - \mathbf{k } \cdot \mathbf{r}_{ij } ) \\ & + \mathbf{d}_{ij } \cdot \ { [ ( r^i \mathbf{n}_2 ) \times ( r^{t(j ) } \mathbf{n}_1 ) ] - [ ( r^i \mathbf{n}_1 ) \times ( r^{t(j ) } \mathbf{n}_2 ) ] \ } \sin(\omega_i - \omega_{t(j ) } - \mathbf{k } \cdot \mathbf{r}_{ij } ) \ } . \end{array}\ ] ] the seeming dependence on arbitrary vectors @xmath95 and @xmath96 can be eliminated with use of following two equations [ eq : useful1 ] @xmath97 then , @xmath98 the phases @xmath99 are presented in expression ( [ eq : ehelix4 ] ) only as differences , which is in agreement with the fact , that they are defined up to a constant term associated with the origin choice . the matrix @xmath100 can be expressed through the coordinates of the vector @xmath101 : @xmath102 with @xmath103 . then , we can rewrite the energy up to second order terms on @xmath104 : @xmath105 with @xmath106 @xmath107 as we can see , the energy breaks up into two contributions depending on @xmath99 and @xmath101 and related to each other only by direction of the axis @xmath80 . the routine minimization of @xmath108 as a function of independent variables @xmath109 , @xmath110 and @xmath111 ( performed with wolfram research mathematica ) gives @xmath112 and @xmath113 it is natural that @xmath108 does not depend on @xmath114 , because real atomic coordinates are not involved in energy definition ( [ eq : energy1 ] ) . from the physical point of view it is more interesting that @xmath108 is also independent of orientation of wave vector @xmath2 relative to coordinate axes . this fact is in a good accordance with experimental evidence of the possibility of easy reorientation of helix axis in mnsi by magnetic field , surface anchoring or other effects . the minimization of the expression ( [ eq : eps1 - 2 ] ) on @xmath5 gives us the following value for the wave number depending on dm vector @xmath43 and rotation axis orientation relative to @xmath2 @xmath115 and the part @xmath108 of unit cell magnetic energy can be finally expressed as @xmath116 thereby the elasticity of the helix is @xmath117 . it is evident from eq . ( [ eq : eps1 - 3 ] ) that @xmath108 has minimum , when @xmath80 coincides with @xmath118 , but it should be noted that , when @xmath80 deviates from @xmath118 by a small angle @xmath119 , the energy obtains only a small positive addition of the order of @xmath120 . thus , there is a soft mode associated with deviation of rotation axis from helix direction . this deviation gives a cycloidal component to the helix . ( [ eq : eps2 - 1 ] ) for @xmath121 contains vectors @xmath93 only within differences , so , in analogy with @xmath99 , the vectors @xmath93 are defined up to a constant vector perpendicular to @xmath80 . however , unlike the case of @xmath99 , where a common term added to all phases results only in redefining of the origin , the constant vector added to all vectors @xmath93 turns rotation axes for all four sublattices by the same angle . in fact , this turning has been already taken into account in the dependence of @xmath108 on @xmath122 . we should also remind that eq . ( [ eq : eps2 - 1 ] ) has been obtained in the assumption that @xmath119 . in order to avoid the ambiguity of definition of vectors @xmath93 , we can use additional condition @xmath123 . the routine minimization of @xmath121 on independent variables @xmath124 , @xmath125 and @xmath126 ( performed with wolfram research mathematica ) gives @xmath127 and @xmath128 the full energy of the helix can be now written as @xmath129 where the contribution @xmath130 per unit cell is associated with ferromagnetic ordering of spins ( 12 is the number of mn - mn bonds per unit cell ) , the term @xmath131 , combining @xmath121 with a part of @xmath108 , describes the energy of divergence between sublattices ( the energy of canting ) , and remaining terms represent the twist energy of the helix . the minimum of the energy ( [ eq : ehelix6 ] ) is achieved , when @xmath132 . then , @xmath133 where @xmath134 it is also useful to write the full divergence vectors for sublattices using the formula @xmath135 then , from eqs . ( [ eq : omega1 ] ) and ( [ eq : gamma1 ] ) follows @xmath136 or , using the additional condition @xmath137 , @xmath138 where @xmath139 , @xmath140 , @xmath141 and @xmath142 are the unit vectors directed along 3-fold symmetry axes corresponding to four sublattices : @xmath143 finally , we can express vectors @xmath144 , @xmath145 and @xmath146 for each of the sublattices : @xmath147 = \mathbf{n}_1 + \frac{\sqrt{3}}{4 } \frac{d_x+d_z}{j } [ \mathbf{e}_t \times \mathbf{n}_1 ] + \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_t ) \mathbf{n}_2 , \\ \mathbf{n}_{sin}^t = \mathbf{n}_2 + [ \vec{\psi}_t \times \mathbf{n}_2 ] = \mathbf{n}_2 + \frac{\sqrt{3}}{4 } \frac{d_x+d_z}{j } [ \mathbf{e}_t \times \mathbf{n}_2 ] - \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_t ) \mathbf{n}_1 , \\ \mathbf{n}_{rot}^t = \mathbf{n}_3 + [ \vec{\psi}_t \times \mathbf{n}_3 ] = \mathbf{n}_3 + \frac{\sqrt{3}}{4 } \frac{d_x+d_z}{j } [ \mathbf{e}_t \times \mathbf{n}_3 ] . \end{array } \right.\ ] ] thus , in the common case each sublattice has its own rotation axis different from @xmath148 as it shown in fig . [ fig1](a ) . the divergence of the axes is defined by eqs . ( [ eq:3n ] ) . the substitution of eqs . ( [ eq:3n ] ) into eq . ( [ eq : s1 ] ) gives the spin of an individual mn atom @xmath149 + \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_{t(i ) } ) \mathbf{n}_2 \right ) \cos \mathbf{k } \cdot \mathbf{r}_i \\ & + \left ( \mathbf{n}_2 + \frac{\sqrt{3}}{4 } \frac{d_x+d_z}{j } [ \mathbf{e}_{t(i ) } \times \mathbf{n}_2 ] - \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_{t(i ) } ) \mathbf{n}_1 \right ) \sin \mathbf{k } \cdot \mathbf{r}_i , \end{array}\ ] ] the question can arise : why parameter @xmath114 appears in eq . ( [ eq : s3 ] ) , in spite of its absence in the initial eq . ( [ eq : energy2 ] ) ? in fact , @xmath114 is brought in by eq . ( [ eq : s1 ] ) depending on real atomic positions . indeed , parameter @xmath114 is implicitly involved with atomic coordinates @xmath150 , and expressing @xmath55 as @xmath151 with @xmath152 being a period of the crystal or the origin of a unit cell , and @xmath153 being an atomic position within the cell , and using the condition @xmath154 , we can rewrite eq . ( [ eq : s3 ] ) as @xmath155 + \left\ { \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_{t(i ) } ) + \mathbf{k } \cdot \mathbf{r}^\prime_{t(i ) } \right\ } \mathbf{n}_2 \right ) \cos \mathbf{k } \cdot \mathbf{p}_i \\ & + \left ( \mathbf{n}_2 + \frac{\sqrt{3}}{4 } \frac{d_x+d_z}{j } [ \mathbf{e}_{t(i ) } \times \mathbf{n}_2 ] - \left\ { \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{k } \cdot \mathbf{e}_{t(i ) } ) + \mathbf{k } \cdot \mathbf{r}^\prime_{t(i ) } \right\ } \mathbf{n}_1 \right ) \sin \mathbf{k } \cdot \mathbf{p}_i , \end{array}\ ] ] where parameter @xmath114 vanishes in curly brackets after substitution of corresponding @xmath156 and @xmath157 . [ fig2 ] shows spin helix structures calculated with the use of eq . ( [ eq : s3 ] ) for directions @xmath158 , @xmath159 and @xmath160 of spiral axis in the case of @xmath132 . the wave number @xmath76 is chosen to be equal to @xmath161 , which is close to the experimental value for mnsi crystal . the canting @xmath43-vector component is taken to be rather large , @xmath162 , in order to emphasize nonmonotonic frustrated behavior of helices . in this case @xmath163 and the main contribution in a canting is due to the terms with @xmath164 $ ] . the extended sinusoids with phase shift of @xmath165 represent spin projections on directions @xmath95 and @xmath96 , the low ones do spin components along the wave vector , which appear due to the canting . the coloured circles correspond to magnetic atoms in four sublattices of the crystal . being all different in the common case , the sinusoids connected with different sublattices can coincide for certain directions of @xmath2 reflecting the symmetry . for instance , in the case of @xmath166 ( fig . [ fig2](a ) ) the extended sinusoids coincide , which correspond to sublattices 1 , 3 and 2 , 4 ; but the circles of different colors do not coincide because the atoms belonging to different sublattices lie in different planes perpendicular to @xmath2 . on the contrary , in the case of @xmath167 ( fig . [ fig2](b ) ) coincide both the extended sinusoids and related circles corresponding to sublattices 2 , 3 , 4 , because their sites lie in the same atomic planes perpendicular to @xmath2 . both two cases , ( a ) and ( b ) , due to their symmetry ( ( 001 ) is the axis @xmath168 and ( 111 ) is the axis @xmath169 of the crystal ) are characterized by only two extended sinusoids , which defines the view of the black saws representing the angle @xmath170 between spins in the successive ferromagnetic planes , @xmath171 . the symmetry is also reflected in the low sinusoids corresponding to the spin components parallel to @xmath2 . when @xmath172 ( fig . [ fig2](c ) ) there are three extended sinusoids , which is also reflected in the view of the cooresponding black saw ( fig . [ fig2](d ) ) . in this case , coincide the extended sinusoids corresponding to sublattices 2 , 3 and the low ones corresponding to sublattices 1 , 4 . it may seem from fig . [ fig2](c ) that sublattices 1 and 4 have the same projections on @xmath2 , but it is evident from fig . [ fig2](d ) that it is not so . indeed , the atoms of types 1 and 4 lie in close but different planes ; their proximity is due to the approximate equality @xmath173 . the `` saws '' in the figures show a frustration which is the intrinsic feature of the helices due to the spin canting . owing to the frustration the spins in successive ferromagnetic planes can rotate inversely to twist direction of the helix . let us return to the case of nonzero magnetic field @xmath20 . numerous experimental data , as well as the results obtained with the use of phenomenological theory , show that in this case the magnetic moments of mn atoms form a conical helix . the peculiarity of such structure is that the field of magnetic moment posesses a constant component directed along @xmath20 . then , the spin structure can be described by the following expression @xmath174 which satisfies the condition @xmath54 . here @xmath175 is an angle characterizing the ratio between constant and rotating components of the magnetic moments . the substitution of eq . ( [ eq : s2 ] ) in the expression ( [ eq : energy2 ] ) and summation over all the unit cells of the crystal give @xmath176 ) \sin\xi_{i } \sin\xi_{t(j ) } \\ & + \frac{n_{cell}}{4 } \sum_{i=1}^{4 } \sum_{j=1}^{6 } \ { ( -j \mathbf{n}_{cos}^{i } \cdot \mathbf{n}_{cos}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{cos}^{i } \times \mathbf{n}_{cos}^{t(j ) } ] ) \cos\mathbf{k } \cdot \mathbf{r}_{ij } \\ & + ( -j \mathbf{n}_{cos}^{i } \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{cos}^{i } \times \mathbf{n}_{sin}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_{ij } \\ & - ( -j \mathbf{n}_{sin}^{i } \cdot \mathbf{n}_{cos}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^{i } \times \mathbf{n}_{cos}^{t(j ) } ] ) \sin\mathbf{k } \cdot \mathbf{r}_{ij } \\ & + ( -j \mathbf{n}_{sin}^{i } \cdot \mathbf{n}_{sin}^{t(j ) } + \mathbf{d}_{ij } \cdot [ \mathbf{n}_{sin}^{i } \times \mathbf{n}_{sin}^{t(j ) } ] ) \cos\mathbf{k } \cdot \mathbf{r}_{ij } \ } \cos\xi_{i}\cos\xi_{t(j ) } \\ & - n_{cell } g\mu_b \sum_{i=1}^{4 } \mathbf{h } \cdot \mathbf{n}_{rot}^{i } \sin\xi_{i } . \end{array}\ ] ] using eq . ( [ eq : basis2 ] ) we can rewrite the energy up to second order terms on @xmath104 : @xmath177 the last term has been obtained supposing that @xmath178 , which allows us to use approximation @xmath179 . the first term in curly brackets has zero order on @xmath180 and achieves the minumum , when @xmath181 then , we can rewrite the energy as @xmath182 where @xmath108 and @xmath121 are the same as in the previous section . the minimum of the energy ( [ eq : ehelixh3 ] ) is achieved , when @xmath183 and @xmath132 . then , @xmath184 or , after minimization on @xmath175 , @xmath185 and @xmath186 the helix vanishes , when @xmath187 , which corresponds to magnetic field @xmath188 ( this field , unwinding the helix , is usually called @xmath189 @xcite ) . thus , in the approximation used here , the wave vector @xmath2 is directed along the applied magnetic field . the spin structure for each sublattice obtains a permanent component along its rotation axis @xmath146 , which is different from direction @xmath190 , see fig . [ fig1](b ) . the helices become conical , and as the magnetic field increases up to @xmath191 , the permanent components grow proportionally to the field value , eq . ( [ eq : sinxi ] ) , whereas the rotated components decrease correspondingly . it should be noted that if the field is less than @xmath191 , it does not affect the axes @xmath146 and they have the same directions , defined by eq . ( [ eq : psi3 ] ) , as in the absence of magnetic field . in the phenomenological theory , the energy of chiral magnetics is usually written as @xmath192 \right ) d\mathbf{r } , \ ] ] where @xmath193 is the local magnetic moment changing smoothly in space . below it will be shown how this equation and phenomenological coefficients @xmath194 and @xmath13 arise from microscopic consideration based on eq . ( [ eq : energy2 ] ) . let us suppose that the typical scale at which a considerable change of magnetic moment occures is much bigger than the size of the unit cell , and any two neighboring spins are rotated relative to each another by a small angle of order @xmath180 . let @xmath14 and @xmath15 be the neighboring spins and @xmath195 , or @xmath196 + \frac{1-\cos\zeta_{ij}}{\zeta_{ij}^2 } \vec{\zeta}_{ij } ( \vec{\zeta}_{ij } \cdot \mathbf{s}_i ) , \ ] ] where @xmath197 . then , @xmath198 @xmath199 \approx \vec{\zeta}_{ij } - \mathbf{s}_i ( \mathbf{s}_i \cdot \vec{\zeta}_{ij } ) = \vec{\zeta}_{ij\perp } , \ ] ] where @xmath200 is the part of vector @xmath201 perpendicular to unit vector @xmath14 . if spins vary slowly along the crystal , we can take a domain , which is much bigger than a unit cell , but small enough as compared with the scale at wich a considerable change of spin occures . the spins within this domain are almost unidirectional . consequently , all the vectors @xmath200 nearly lie in a plane . let us neglect this `` nearly '' and suppose that all @xmath200 belong to the plane perpendicular to some mean spin @xmath202 . in addition , staying in the domain , we can discard the parts of @xmath201 parallel to @xmath202 and suppose that @xmath203 . another simplification can be achieved with taking into account that a spin does not change upon round a close chain of atoms . this means that the sum of small angles @xmath201 corresponding to the bonds of the chain should be equal to zero . for instance , for the triangle formed by spins @xmath204 , @xmath205 and @xmath206 we obtain the following condition @xmath207 which is evidently satisfied , if @xmath208 , where @xmath209 and @xmath210 are perpendicular to @xmath202 vectors associated with @xmath28-th and @xmath29-th atoms . with the approximations accepted , using the expressions ( [ eq : sisj2 ] ) and ( [ eq : sisj3 ] ) , the energy ( [ eq : energy2 ] ) in the absence of magnetic field can be now rewritten as @xmath211 the minimization of the energy on the vector @xmath209 associated with @xmath28-th atom and belonging to the plane perpendicular to @xmath202 gives @xmath212 or @xmath213 where symbol `` @xmath214 '' designates the projection of a vector to the plane perpendicular to @xmath202 , @xmath215 is one of the vectors ( [ eq : e ] ) . the last therm in the right part of ( [ eq : zeta2 ] ) reflects the evident fact that the sum of dm vectors associated with all the bonds of an atom should be parallel to the 3-fold axis to which the atomic position belongs . let us introduce a new vector @xmath216 connected with the old one by equation @xmath217 where @xmath218 is a constant . then , @xmath219 where it has been taken into account that an atom belonging to one of four sublattices has two neighbors at each of three remaining ones , so @xmath220 choosing @xmath221 we can now write @xmath222 @xmath223 and , after the substitution of eq . ( [ eq : zeta5 ] ) into eq . ( [ eq : energy3 ] ) , @xmath224 where @xmath225 the first part of eq . ( [ eq : energy4 ] ) is analogous to the expression ( [ eq : energy3 ] ) , but with modificated dm vectors . it is important here that the combination @xmath226 varies from bond to bond with the use of the same symmetry transformations of the point group @xmath227 as vector @xmath17 and , consequently , @xmath228 . the vectors @xmath228 are characterized by condition @xmath229 , so they have only two independent components . for example , the modificated dm vector of the bond @xmath230 is @xmath231 the second part of eq . ( [ eq : energy4 ] ) contains the terms like @xmath232 depending on the direction of @xmath202 . the vectors @xmath233 and @xmath234 vary from bond to bond with the use of the symmetry transformations of the point group @xmath35 , and the averaging over 12 bonds in the unit cell , as well as for the case of the averaging over the sphere , gives @xmath235 with @xmath236 being the rotation matrices of the group @xmath35 . thereby , the mean contribution of the second part of eq . ( [ eq : energy4 ] ) to the energy of a bond is @xmath237 and we can now rewrite the expression ( [ eq : energy4 ] ) for the energy as @xmath238 let us introduce now smooth continuous vector fields @xmath239 , @xmath240 , @xmath241 and @xmath242 , connecting with @xmath216 by the equation @xmath243 we can suppose that functions @xmath244 change slowly enough with coordinates and postulate the homogeneity condition @xmath245 reflecting the fact that the divergence between sublattices is preserved over big distances . then , we can write eq . ( [ eq : zeta6 ] ) for 1st atom in a cell in the explicit form as @xmath246 the last equation is correct not only for @xmath247 but also for arbitrary @xmath55 due to homogeneity condition . so , we can write eq . ( [ eq : hat2 ] ) and analogous equations for other atoms in a cell , supposing that all @xmath248 are taken at the same point : @xmath249 the system ( [ eq : hat3 ] ) is degenerate on variables @xmath250 , @xmath251 , @xmath252 , @xmath253 and satisfied , when @xmath254 let us allocate two parts from expression ( [ eq : energy5 ] ) for the energy associated with a unit cell , @xmath255 and @xmath256 with the use of eq . ( [ eq : hat4 ] ) we can rewrite these parts as @xmath257 and @xmath258 the second part of ( [ eq : edm2 ] ) vanishes , because @xmath259 using @xmath260 the first part of ( [ eq : eexch2 ] ) can be calculated , @xmath261 the second part of ( [ eq : eexch2 ] ) can be expressed as @xmath262 where summation is taken over all elements @xmath263 of the point symmetry group @xmath227 , and we have used that @xmath264 and @xmath265 so , finally , @xmath266 similarly , eq . ( [ eq : edm2 ] ) can be rewritten as @xmath267 or @xmath268 it should be noted that , whereas eqs . ( [ eq : eexch2 ] ) and ( [ eq : edm2 ] ) contain parameter @xmath114 , the resulting expressions ( [ eq : eexch3 ] ) and ( [ eq : edm3b ] ) do not depend on @xmath114 . this looks like a coincidence , but it is not so . indeed , the initial statement of the problem do not include atomic coordinates ( e.g. , see eq . ( [ eq : energy1 ] ) ) , and therefore the coordinates will not appear in results ( at least until strong changes of the coordinates changing the topology of the net of nearest neighbors ) . the substitution of eqs . ( [ eq : eexch3 ] ) and ( [ eq : edm3b ] ) into eq . ( [ eq : energy5 ] ) gives @xmath269 with @xmath270 and @xmath271 given by the following equations @xmath272 @xmath273 the summation over all unit cells of the crystal has been replaced with the integration over the crystal volume . let us suppose that the vector field @xmath274 is connected with a spin field @xmath275 : @xmath276 , \ ] ] where @xmath277 is a constant vector , @xmath278 . then , @xmath279 and @xmath280 \approx - \frac{\partial \hat{\zeta}_n}{\partial x_n } .\ ] ] so , we can rewrite eq . ( [ eq : energy6 ] ) for the energy as @xmath281 \right ) d\mathbf{r } + n_{cell } j \left ( -12 - 2\left ( \frac{d_x+d_z}{j } \right)^2 \right ) , \ ] ] the first part of which coincides with the usually used energy in the phenomenological theory with the elastic constants @xmath270 and @xmath271 related to the the constants of microscopic interactions by eqs . ( [ eq : j1 ] ) and ( [ eq : d1 ] ) . the wave number for the helix structure can be now found using the phenomenological approach , @xmath282 which coincides with eq . ( [ eq : k2 ] ) . thus , we have passed from the discrete spin distribution of the microscopic theory to the continuous spin field @xmath275 of the coarse grain theory . in order to complete our reasoning we can also consider the way back from the continuous approximation to dicrete spin distibution . let @xmath283 be a continuous spin field from the phenomenological theory . within a small domain of the crystal we can associate with @xmath275 a vector field @xmath274 using the transformation inverse to ( [ eq : szeta1 ] ) : @xmath284 .\ ] ] here @xmath277 is some mean value of @xmath275 , which is constant over the domain . then , @xmath285 \approx [ \hat{\mathbf{s } } \times \frac{\partial \hat{\mathbf{s}}}{\partial r_n } ] .\ ] ] let us define continuous fields @xmath250 , @xmath251 , @xmath252 and @xmath253 for four sublattices of magnetic atoms by formula @xmath286 obviously obeying eq . ( [ eq : hat4 ] ) . thus @xmath274 is the average of these four fields , @xmath287 then , using eqs . ( [ eq : hat1 ] ) and ( [ eq : zeta5 ] ) we can express vector @xmath209 for individual atom trough @xmath274 , @xmath288 where @xmath289 the substitution of eq . ( [ eq : zeta7 ] ) into the equation @xmath290\ ] ] gives us the following expression for the spin of individual magnetic atom @xmath291 + \frac{\sqrt{3}}{8 } ( 1 - 8x ) ( \mathbf{e}_{t(i ) } \cdot \vec{\nabla } ) \hat{\mathbf{s } } ( \mathbf{r}_i ) .\ ] ] this equation is valid for any smooth continuous field @xmath283 . for one - dimensional helices , the first term in eq . ( [ eq : sdiscrete ] ) , which is common for all sublattices , describes spin rotation in the plane perpendicular to the wave vector @xmath2 , the second one defines a spin component parallel to @xmath2 , and second and third ones taken together do a shift of the phase of the rotation for each sublattice . for instance , if one chooses @xmath275 in the form of @xmath292 then @xmath293 has the form ( [ eq : s3 ] ) . as it is shown above , in the approximation linear on @xmath180 , only two components of dm vectors are significant for the considered magnetic structure of mnsi : @xmath294 and @xmath295 . the third component , @xmath7 , appears in upper approximations . the pseudo - scalar constant @xmath296 used in the phenomenological theory is found to be proportional to the first component , @xmath294 . in fact , we could expect that only this combination of @xmath43-vector coordinates played a role in the simplest microscopic approximation . surprisingly , it has been found that another component , @xmath295 , is also very important both because of its contribution into magnetic energy and its influence on spins orientations . this effect is very similar to `` frozen '' optical phonon displacements of atoms in crystals . whereas the former component is responsible for the long scale chiral magnetic structure , the latter leads to a canting of spins within a unit cell and this canting could be in principle measured experimentally using diffraction methods @xcite . for instance , for the magnetic fields stronger than @xmath191 , the residual splay of spins results in non - zero value of the neutron scattering amplitude for reflections of @xmath297 type , forbidden both for nuclear scattering and for ferromagnetic scattering owing to @xmath168 screw axes . an interesting result about the symmetry of the problem can be revealed considering the inverse structure , where magnetic atoms occupy positions @xmath298 rather than @xmath299 . the crystal thus obtained has the same space group , @xmath1 , but both the crystals are related as right- and left - hand structures , via the inversion transformation . it can be easily seen from the expression ( [ eq : energy1 ] ) that energy does not change over this transformation , if the spins in corresponding positions are equal and so do dzyaloshinskii moriya vectors associated with corresponding bonds . this means that dm vectors are really pseudovectors . also it is well understood @xcite that the inversion should change the chirality of the structure and , consequently , the sign of phenomenological constant @xmath271 and wave number @xmath5 . however it is not so evident from eq . ( [ eq : k3 ] ) that @xmath5 changes its sign over inversion , because @xmath43 is a pseudovector rather than a vector . we should rewrite the wave number in an invariant form , including an explicit dependence on the bond direction , in order to make transparent the change of chirality upon inversion : @xmath300 the pseudovector @xmath301 does not change its sign under inversion but @xmath302 does . notice that in this form the sign of @xmath5 does not depend on our choice which of atoms is 1 or 2 . it is interesting to compare our results with those ones obtained by hopkinson and kee @xcite using mean - field and real - space rigid rotor minimization methods . according to this work , for both methods , a sizable spin helicity can be obtained only when the dm vectors lie _ parallel _ to the corresponding mn - mn bonds . correspondingly , three components of dm vector are considered : the first one , parallel to the bond , the second one , perpendicular to the triangle of bonds , to which this bond is belonging , and the third one , lying in the plane of the triangle , but perpendicular to the bond . for instance , for the bond @xmath303 these three directions can be written in explicit form as @xmath304 , @xmath305 and @xmath306 , correspondingly , or , using @xmath307 , as @xmath308 , @xmath309 and @xmath310 . the first component is responsible for the helix structure , and corresponding direction , @xmath308 , is very close to direction @xmath311 for @xmath294 component found in our work ( the angle between them is about 3.4 degrees ) . the second component , which does not ca nt spins @xcite , coincides with our `` neutral '' component @xmath7 directed along the corresponding 3-fold axis . the third component leads to spin canting and rather small spin spiraling , and its direction , @xmath312 , is very close to direction @xmath313 for @xmath295 component . obviously , those numerical results almost coincide with our analytical solutions . an advantage of the analytical solution is that it does not rely on the exact bond directions and atomic coordinates ; that is natural from the theoretical point of view because the atomic parameter @xmath114 does not figure in eq . [ eq : energy1 ] . moreover , different model considerations predict that the superexchange dm vectors are either perpendicular @xcite or almost perpendicular @xcite to the corresponding bonds . if this is also the case in mnsi , we could expect pronounced frustration of the helices . in any case , it is obvious that modern _ ab initio _ calculations of the dyzaloshinskii - moriya vectors would be very important for mnsi - type crystals . in conclusion , it is found which component of the dyzaloshinskii - moriya vector determines the long - range spiral ordering . the most interesting point is that the local structure of helices can be rather complicated owing to symmetry - driven frustrations . we hope that qualitatively all the considered features will appear also in more sophisticated models taking into account the itinerant nature of magnetism and thermal and quantum properties of spiral ordering in mnsi . we are grateful to s. v. grigoriev and s. v. maleyev for useful discussions . this work was supported by two basic research programs of the presidium of the russian academy of sciences : `` thermophysics and mechanics of extreme energy actions and the physics of a strongly compressed substance '' and `` resonant x - ray diffraction and topography in forbidden reflections '' . 99 a. a. kornyshev , d. j. lee , s. leikin , and a. wynveen , rev . phys . , * 79 * , 943 ( 2007 ) . v. a. belyakov and v. e. dmitrienko , sov . , * 28 * , 535 ( 1985 ) ; usp . nauk , * 146 * , 369 ( 1985 ) . d. c. wright and n. d. mermin , rev . phys . * 61 * , 385 , ( 1989 ) . a. n. bogdanov and d. a. yablonskii , sov . jetp * 68 * , 101 ( 1989 ) . u. k. rler , a. n. bogdanov , and c. pfleiderer , nature * 442 * 797 ( 2006 ) . x. z. yu , n. kanazawa , y. onose , k. kimoto , w. z. zhang , s. ishiwata , y. matsui , and y. tokura , nature mat . * 10 * , 106 ( 2011 ) . s. tewari , d. belitz , and t. r. kirkpatrick , phys . lett . * 96 * , 047207 ( 2006 ) . o. nakanishi , a. yanase , a. hasegawa , and m. kataoka , solid state commun . * 35 * , 995 ( 1980 ) . p. bak and m. h. jensen , j. phys . c : solid state phys . * 13 * , l881 ( 1980 ) . i. e. dzyaloshinsky , zh . . fiz . * 32 * , 1547 ( 1957 ) ; sov . jetp , * 5 * , 1259 ( 1957 ) . i. dzyaloshinsky , j. phys . solids * 4 * , 241 ( 1958 ) . t. moriya , phys . lett . * 4 * , 228 ( 1960 ) . t. moriya , phys . rev . * 120 * , 91 ( 1960 ) . s. v. grigoriev , d. chernyshov , v. a. dyadkin , v. dmitriev , s. v. maleyev , e. v. moskvin , d. menzel , j. schoenes , and h. eckerlebe , phys . rev . lett . * 102 * , 037204 ( 2009 ) . s. v. grigoriev , d. chernyshov , v. a. dyadkin , v. dmitriev , e. v. moskvin , d. lamago , th . wolf , d. menzel , j. schoenes , s. v. maleyev , and h. eckerlebe , phys . b * 81 * , 012408 ( 2010 ) . a. n. bogdanov and u. k. rler , phys . rev . lett . * 87 * , 037203 ( 2001 ) . s. v. grigoriev , d. lott , yu . o. chetverikov , a. t. d. grnwald , r. c. c. ward , and a. schreyer , phys . b * 81 * , 195432 ( 2010 ) . j. m. hopkinson and h .- y . kee , phys . rev . b * 79 * , 014421 ( 2009 ) . l. shekhtman , o. entin - wohlman , and a. aharony , phys . lett . * 69 * , 836 ( 1992 ) . t. yildirim , a. b. harris , a. aharony , and o. entin - wohlman , phys . b * 52 * , 10 239 ( 1995 ) . international tables for crystallography , vol . a : space - group symmetry , t. hahn ( ed . ) . 1989 . kluwer academic publishers , dordrecht , the netherlands . s. v. grigoriev , s. v. maleyev , a. i. okorokov , yu . o. chetverikov , and h. eckerlebe , phys . b * 73 * , 224440 ( 2006 ) . v. e. dmitrienko , v. a. chizhikov ( unpublished ) . f. keffer , phys . rev . * 126 * , 896 ( 1962 ) . l. shekhtman , a. aharony , and o. entin - wohlman , phys . b * 47 * , 174 ( 1993 ) . v. v. mazurenko and v. i. anisimov , phys . b * 71 * , 184434 ( 2005 ) . m. i. katsnelson , y. o. kvashnin , v. v. mazurenko , and a. i. lichtenstein , phys . rev . b * 82 * , 100403 ( 2010 ) . .[tabled ] the bonds between neighboring mn atoms in mnsi crystal and corresponding dzyaloshinskii moriya vectors . the listed 12 bonds are connected by symmetry transformations of the point group @xmath35 . other 12 bonds appear owing to permutation of @xmath28-th and @xmath29-th atoms ( @xmath314 , @xmath315 ) . [ cols=">,^,^,^,^",options="header " , ] ( a ) organization of a free helix with an arbitrary wave vector @xmath2 ( no magnetic field ) . the spins of each of four sublattices rotate around their own rotation directions @xmath146 ( see eq . ( [ eq:3n ] ) ) , which differ slightly from that of @xmath2 . in the considered model , the energy of the free helix does not depend on direction of @xmath2 whereas experimentally the free helix is oriented either along @xmath316 $ ] or @xmath317 $ ] direction owing to the crystal field anisotropy . ( b ) conical helix spin structure in an external magnetic field . the spins of each of four sublattices have a component directed along corresponding vector @xmath146 ; its value is proportional to @xmath318 for @xmath319 and is equal to 1 for @xmath320 ; @xmath2-vector is directed along the magnetic field.,title="fig:",width=302 ] ( a ) organization of a free helix with an arbitrary wave vector @xmath2 ( no magnetic field ) . the spins of each of four sublattices rotate around their own rotation directions @xmath146 ( see eq . ( [ eq:3n ] ) ) , which differ slightly from that of @xmath2 . in the considered model , the energy of the free helix does not depend on direction of @xmath2 whereas experimentally the free helix is oriented either along @xmath316 $ ] or @xmath317 $ ] direction owing to the crystal field anisotropy . ( b ) conical helix spin structure in an external magnetic field . the spins of each of four sublattices have a component directed along corresponding vector @xmath146 ; its value is proportional to @xmath318 for @xmath319 and is equal to 1 for @xmath320 ; @xmath2-vector is directed along the magnetic field.,title="fig:",width=302 ] ) for the wave vectors @xmath2 directed along @xmath158 , @xmath159 and @xmath160 crystal axes , correspondingly . the wave number @xmath76 is chosen to be equal to @xmath161 , which is close to the experimental value for mnsi crystal where the helix period is approximately 40 unit cells . the canting @xmath43-vector component is taken to be rather large , @xmath162 , to emphasize nonmonotonic frustrated behavior of helices . three spin components along directions @xmath95 , @xmath96 and @xmath132 are presented by points with four colors corresponding to four sublattices . the black saw represents the angle @xmath170 between spins in the successive ferromagnetic planes , @xmath171 . ( d ) an enlarged part of the saw restricted by a small box in ( c ) ; the line is drown to guide the eyes.,title="fig:",width=302 ] ) for the wave vectors @xmath2 directed along @xmath158 , @xmath159 and @xmath160 crystal axes , correspondingly . the wave number @xmath76 is chosen to be equal to @xmath161 , which is close to the experimental value for mnsi crystal where the helix period is approximately 40 unit cells . the canting @xmath43-vector component is taken to be rather large , @xmath162 , to emphasize nonmonotonic frustrated behavior of helices . three spin components along directions @xmath95 , @xmath96 and @xmath132 are presented by points with four colors corresponding to four sublattices . the black saw represents the angle @xmath170 between spins in the successive ferromagnetic planes , @xmath171 . ( d ) an enlarged part of the saw restricted by a small box in ( c ) ; the line is drown to guide the eyes.,title="fig:",width=302 ] ) for the wave vectors @xmath2 directed along @xmath158 , @xmath159 and @xmath160 crystal axes , correspondingly . the wave number @xmath76 is chosen to be equal to @xmath161 , which is close to the experimental value for mnsi crystal where the helix period is approximately 40 unit cells . the canting @xmath43-vector component is taken to be rather large , @xmath162 , to emphasize nonmonotonic frustrated behavior of helices . three spin components along directions @xmath95 , @xmath96 and @xmath132 are presented by points with four colors corresponding to four sublattices . the black saw represents the angle @xmath170 between spins in the successive ferromagnetic planes , @xmath171 . ( d ) an enlarged part of the saw restricted by a small box in ( c ) ; the line is drown to guide the eyes.,title="fig:",width=302 ] ) for the wave vectors @xmath2 directed along @xmath158 , @xmath159 and @xmath160 crystal axes , correspondingly . the wave number @xmath76 is chosen to be equal to @xmath161 , which is close to the experimental value for mnsi crystal where the helix period is approximately 40 unit cells . the canting @xmath43-vector component is taken to be rather large , @xmath162 , to emphasize nonmonotonic frustrated behavior of helices . three spin components along directions @xmath95 , @xmath96 and @xmath132 are presented by points with four colors corresponding to four sublattices . the black saw represents the angle @xmath170 between spins in the successive ferromagnetic planes , @xmath171 . ( d ) an enlarged part of the saw restricted by a small box in ( c ) ; the line is drown to guide the eyes.,title="fig:",width=302 ]

abstract rigid bodies collision maps in dimension two , under a natural set of physical requirements , can be classified into two types : the standard specular reflection map and a second which we call , after broomhead and gutkin , _ no - slip . _ this leads to the study of _ no - slip billiards_planar billiard systems in which the moving particle is a disc ( with rotationally symmetric mass distribution ) whose translational and rotational velocities can both change at each collision with the boundary of the billiard domain . in this paper we greatly extend previous results on boundedness of orbits ( broomhead and gutkin ) and linear stability of periodic orbits for a sinai - type billiard ( wojtkowski ) for no - slip billiards . we show among other facts that : ( i ) for billiard domains in the plane having piecewise smooth boundary and at least one corner of inner angle less than , no - slip billiard dynamics will always contain elliptic period- orbits ; ( ii ) polygonal no - slip billiards always admit small invariant open sets and thus can not be ergodic with respect to the canonical invariant billiard measure ; ( iii ) the no - slip version of a sinai billiard must contain linearly stable periodic orbits of period and , more generally , we provide a curvature threshold at which a commonly occurring period- orbit shifts from being hyperbolic to being elliptic ; ( iv ) finally , we make a number of observations concerning periodic orbits in a class of polygonal billiards .

1612.03355

no - slip billiard systems have received so far very little attention despite some interesting features that distinguish them from the much more widely studied standard billiards . these non - standard types of billiards are discrete time systems in dimension @xmath2 ( after taking the natural poincar section of a flow ) representing a rotating disc with unit kinetic energy that moves freely in a billiard domain with piecewise smooth boundary . although not hamiltonian , these systems are nevertheless time reversible and leave invariant the canonical billiard measure . they also exhibit dynamical behavior that is in sharp contrast with standard billiards . for example , they very often contain period-@xmath1 orbits having small elliptic islands . these regions exist amid chaos that appears , in numerical experiments , to result from the usual mechanisms of dispersing and focusing . we are aware of only two articles on this subject prior to our @xcite : one by broomhead and gutkin @xcite showing that no - slip billiard orbits in an infinite strip are bounded , and another by wojtkowski , characterizing linear stability for a special type of period-@xmath1 orbit . in this paper we extend their results as will be detailed shortly , and develop the basic theory of no - slip billiards in a more systematic way . in this section we explain the organization of the paper and highlight our main new results . section [ preliminaries ] gives preliminary information and sets notation and terminology concerning rigid collisions . it specializes the general results from @xcite ( stated in that paper in arbitrary dimension for bodies of general shapes and mass distributions ) to discs in the plane with rotationally symmetric mass distributions . the main fact is briefly summarized in proposition [ classification ] . although the classification into specular and no - slip collisions is the same as in @xcite , our approach is more differential geometric in style and may have some conceptual advantages . for example , we derive this classification ( in @xcite ) from an orthogonal decomposition of the tangent bundle @xmath3 restricted to the boundary @xmath4 ( orthogonal relative to the kinetic energy riemannian metric in the system s configuration manifold @xmath5 ) into physically meaningful subbundles . this orthogonal decomposition is explained here only for discs in the plane . by a planar _ no - slip billiard _ system we mean a mechanical system in @xmath6 in which one of the colliding bodies , which may have arbitrary shape , is fixed in place , whereas the second , moving body is a disc with rotationally symmetric mass distribution ; post - collision velocities ( translational and rotational ) are determined from pre - collision velocities via the no - slip collision map and between consecutive collisions the bodies undergo free motion . contrary to the standard case , the moving particle s mass distribution influences the collision properties ( via an angle parameter which is denoted @xmath7 throughout the paper ) . the main definitions and notations concerning no - slip billiard systems , in particular the notions of _ reduced phase space _ , _ velocity phase space _ , and the _ product , eigen- _ and _ wavefront _ frames , are introduced in section [ noslip billiards definitions ] . a special feature of no - slip billiards around which much of the present study is based , is the ubiquitous occurrence of period-@xmath1 trajectories . the general description of these trajectories is given in section [ periodic section ] . in section [ sec : differential ] we obtain the differential of the no - slip billiard map and show the form it takes for period-@xmath1 trajectories . for general collision systems as considered in @xcite ( satisfying energy and momenta conservation , time reversibility , involving impulsive forces that can only act at the contact point between colliding bodies ) , the issue of characterizing smooth invariant measures still needs much further investigation , although it is shown there that the canonical ( liouville ) measure is invariant if a certain field of subspaces defined in terms of the collision maps at each collision configuration @xmath8 is parallel with respect to the kinetic energy metric . this is the case for planar no - slip billiards , so that the standard billiard measure is still invariant . ( note , however , that the configuration manifold is now @xmath9-dimensional . ) a detailed proof of this fact , in addition to comments on time reversibility are given in section [ measure reversible ] . dynamics proper begins with sections [ wedge section ] and [ higher order polygon ] , which are concerned with no - slip billiard systems in wedge - shaped regions and polygons . here we generalize the main result from @xcite . in that paper it is shown that orbits of the no - slip billiard system in an infinite strip domain are bounded . by extending and refining this fact to wedge regions we obtain local stability for periodic orbits in no - slip polygonal billiards . this is theorem [ polygon stability ] . we also give in section [ higher order polygon ] an exhaustive description of periodic orbits in wedge billiards . finally , in section [ curved section ] we consider linear stability of period-@xmath1 orbits in the presence of curvature . our results here extend those of @xcite for no - slip sinai billiards . wojtkowski makes in @xcite the following striking observation : for a special period-@xmath1 orbit in a sinai billiard ( corresponding in our study to angle @xmath10 ) there is a parameter defined in terms of the curvature of the circular scatterer that sets a threshold between elliptic and hyperbolic behavior . this is based on an analysis of the differential of the billiard map at the periodic trajectory . here we derive similar results for period-@xmath1 orbits in general . although the analysis is purely linear , we observe the occurrence of elliptic islands and stable behavior in systems that are the no - slip counterpart of fully chaotic standard billiards . the fast transition between stability and chaos near the threshold set by the curvature parameter obtained from the linear analysis is also very apparent numerically . as already observed by wojtkowski in @xcite , proving local stability in the presence of curvature would require a difficult kam analysis ( for reversible , but not hamiltonian systems ; cf . for a much more general treatment of the material of this section ( not restricted to discs and valid in arbitrary dimension ) see @xcite . see also @xcite . let @xmath11 be the standard coordinates in @xmath6 . a mass distribution on a body @xmath12 is defined by a finite positive measure @xmath13 on @xmath12 . we assume without loss of generality that the first moments @xmath14 vanish . let @xmath15 be the total mass : @xmath16 . the second moments of @xmath13 ( divided by @xmath15 ) will be denoted by @xmath17 . when @xmath18 is a disc of radius @xmath19 centered at the origin of @xmath6 , it will be assumed that @xmath13 is rotationally symmetric , in which case the symmetric matrix @xmath20 is scalar : @xmath21 . also in this case , @xmath22 , where @xmath23 corresponds to all mass being concentrated at the origin and the upper bound corresponds to having all mass concentrated on the circle of radius @xmath19 . for the uniform mass distribution , @xmath24 . it will be useful to introduce the parameter @xmath25 . the moment of inertia of a disc of radius @xmath19 is given in terms of @xmath26 by @xmath27 . for the uniform mass distribution on the ball @xmath28 . in general , @xmath29 . from now on @xmath18 will be the disc of radius @xmath19 in @xmath6 centered at the origin . a _ configuration _ of the moving disc is an image of @xmath18 under a euclidean isometry ; it is parametrized by the coordinates of the center of mass @xmath30 of the image disc and its angle of orientation @xmath31 . introducing the new coordinate @xmath32 and denoting by @xmath33 the velocity vector in configuration space , the kinetic energy of the body takes the form @xmath34 , where @xmath35 is the standard euclidean norm . we consider now a system of two discs with radii @xmath36 , denoted by @xmath37 when in their reference configuration , that is , when centered at the origin of @xmath6 . the configuration manifold @xmath5 of the system , which is the set of all non - overlapping images of @xmath38 under euclidean isometries of the plane , is then the set @xmath39 where @xmath40 and @xmath41 indicates the coordinate projection of @xmath42 in @xmath6 . given mass distributions @xmath43 , @xmath44 , the kinetic energy of the system at a state @xmath45 is @xmath46 the manifold @xmath5 becomes a riemannian manifold with boundary by endowing it with the _ kinetic energy metric _ @xmath47 where the dot means ordinary inner product in @xmath48 . let @xmath49 and @xmath50 denote the unit tangent and normal vectors at the contact point @xmath51 of the bodies in a boundary configuration @xmath8 as indicated in figure [ frame_definition ] . we introduce the orthonormal frame @xmath52 of @xmath48 as defined in the figure . if the system is at a state @xmath45 , then @xmath53 will have velocity @xmath54 when regarded as a point in body @xmath38 in the given configuration @xmath55 . one easily obtains @xmath56 e_2 + v_1\cdot e_3 e_3 , \ \ v_2(q)=\left[v_2\cdot e_2 - \gamma_2^{-1}v_2\cdot e_1\right ] e_2 + v_2\cdot e_3 e_3.\ ] ] the unit normal vector to @xmath4 at the configuration @xmath57 pointing towards the interior of @xmath5 will be denoted @xmath58 . note that this is defined with respect to the kinetic energy metric . explicitly , letting @xmath59 , @xmath60 we also define the _ no - slip _ subspace @xmath61 of the tangent space to @xmath4 at @xmath62 : @xmath63 and the orthogonal complement to @xmath61 in @xmath64 , which we write as @xmath65 . the orthogonal direct sum of the latter and the line spanned by @xmath58 was denoted by @xmath66 in @xcite and called the _ impulse _ subspace . these two spaces are given by @xmath67 and @xmath68 figure [ orthogonal ] shows typical vectors in these orthogonal subspaces . we can now state in the present very special case one of the main results of @xcite . see also @xcite . in the interior of @xmath5 , the motion of the system , in the absence of potential forces , is geodesic relative to the kinetic energy metric . in dimension two this amounts simply to constant linear and angular velocities . for the motion to be fully specified it is necessary to find for any given @xmath69 at @xmath8 such that @xmath70 ( representing the bodies velocities immediately before a collision ) a @xmath71 such that @xmath72 ( representing the bodies velocities immediately after ) . this correspondence should be given by a map @xmath73 . we call such correspondence a _ collision map _ @xmath74 at @xmath8 . [ classification ] linear collision maps @xmath75 at @xmath8 describing energy preserving , ( linear and angular ) momentum preserving , time reversible collisions having the additional property that impulsive forces between the bodies can only act at the single point of contact ( denoted by @xmath53 above ) are given by the linear orthogonal involutions that restrict to the identity map on @xmath61 and send @xmath58 to its negative . thus @xmath74 is fully determined by its restriction to @xmath65 , where it can only be ( in dimension @xmath1 ) the identity or its negative . we refer to @xcite for a more detailed explanation of this result . the key point for our present purposes is that , in addition to the standard reflection map given by specular reflection with respect to the kinetic energy metric , there is only one ( under the stated assumptions ) other map , which we refer to as the _ no - slip collision map . _ the explicit form of the no - slip collision map @xmath74 can thus be obtained by first decomposing the pre - collision vector @xmath76 according to the orthogonal decomposition @xmath77 and then switching the sign of the @xmath66 component to obtain @xmath71 . if @xmath78 is the orthogonal projection to @xmath66 , then @xmath79 setting @xmath80=[(v_1,v_2)]:=\left(v_1\cdot e_1 , v_1\cdot e_2 , v_1\cdot e_3 , v_2\cdot e_1 , v_2\cdot e_2 , v_2\cdot e_3\right)^t,$ ] the pre- and post - collision velocities are related by @xmath81 = [ c_q ] [ v^-]$ ] where @xmath82=\left(\begin{array}{cccccc}1-\frac{\delta}{m_1\gamma_1 ^ 2 } & \frac{\delta}{m_1\gamma_1 } & 0 & -\frac{\delta}{m_1\gamma_1\gamma_2 } & -\frac{\delta}{m_1\gamma_1 } & 0 \\\frac{\delta}{m_1\gamma_1 } & 1-\frac{\delta}{m_1 } & 0 & \frac{\delta}{m_1\gamma_2 } & \frac{\delta}{m_1 } & 0 \\0 & 0 & 1-\frac{2 m_2}{m } & 0 & 0 & \frac{2m_2}{m } \\-\frac{\delta}{m_2\gamma_1\gamma_2 } & \frac{\delta}{m_2\gamma_2 } & 0 & 1-\frac{\delta}{m_2\gamma_2 } & -\frac{\delta}{m_2\gamma_2 } & 0 \\-\frac{\delta}{m_2\gamma_1 } & \frac{\delta}{m_2 } & 0 & -\frac{\delta}{m_2\gamma_2 } & 1-\frac{\delta}{m_2 } & 0 \\0 & 0 & \frac{2m_1}{m } & 0 & 0 & 1-\frac{2 m_1}{m}\end{array}\right ) . \endgroup\ ] ] here @xmath59 and @xmath83 + \frac1{m_2}\left[1+\frac1{\gamma_2 ^ 2}\right]\right\}^{-1}.$ ] we record two special cases . first suppose that the two discs have the same mass distribution and @xmath84 . then @xmath82= \left(\begin{array}{cccccc}\frac{\gamma^2}{1+\gamma^2 } & \frac{\gamma}{1+\gamma^2 } & 0 & -\frac{1}{1+\gamma^2 } & -\frac{\gamma}{1+\gamma^2 } & 0 \\\frac{\gamma}{1+\gamma^2 } & \frac{1}{1+\gamma^2 } & 0 & \frac{\gamma}{1+\gamma^2 } & \frac{\gamma^2}{1+\gamma^2 } & 0 \\0 & 0 & 0 & 0 & 0 & 1 \\-\frac{1}{1+\gamma^2 } & \frac{\gamma}{1+\gamma^2 } & 0 & \frac{\gamma^2}{1+\gamma^2 } & -\frac{\gamma}{1+\gamma^2 } & 0 \\-\frac{\gamma}{1+\gamma^2 } & \frac{\gamma^2}{1+\gamma^2 } & 0 & -\frac{\gamma}{1+\gamma^2 } & \frac{1}{1+\gamma^2 } & 0 \\0 & 0 & 1 & 0 & 0 & 0\end{array}\right ) \endgroup\ ] ] the second case of interest assumes that one body , say @xmath85 is fixed in place . it makes sense to pass to the limit in which the mass and moment of inertia of @xmath85 approach infinity and its velocity ( which does not change during collision process as a quick inspection of @xmath74 shows ) is set equal to @xmath23 . in this case only the velocity of @xmath86 changes and we write @xmath87 and @xmath88 . then @xmath89 where vectors are expressed in the frame @xmath90 . the ( lower right block of the ) matrix @xmath91 $ ] is now @xmath92= \left(\begin{array}{ccr}-\frac{1-\gamma^2}{1+\gamma^2 } & -\frac{2\gamma}{1+\gamma^2 } & 0 \\-\frac{2\gamma}{1+\gamma^2 } & \ \ , \ , \frac{1-\gamma^2}{1+\gamma^2 } & 0 \\0 & 0 & -1\end{array}\right)=\left(\begin{array}{ccc}-\cos\beta & -\sin\beta & 0 \\-\sin\beta & \ \ , \cos\beta & 0 \\0 & 0 & -1\end{array}\right ) . \endgroup\ ] ] as noted earlier , @xmath93 can take any value between @xmath23 and @xmath94 ( equivalently , @xmath95 ) in dimension @xmath1 ; thus it makes sense to define the angle @xmath7 as we did above . when the mass distribution is uniform , @xmath28 and @xmath96 , @xmath97 . notice that these expressions still hold regardless of the shape of the fixed body @xmath85 . in what follows we denote the above @xmath9-by-@xmath9 matrix by @xmath98 $ ] . we focus attention on the last case indicated at the end of section [ preliminaries ] , which we call a planar no - slip _ billiard _ : the billiard table is the complement of the fixed body @xmath85 , now any set in @xmath6 with non - empty interior and piecewise smooth boundary , and the billiard ball is the disc @xmath86 of radius @xmath19 with a rotationally symmetric mass distribution . the configuration manifold is then the set @xmath5 consisting of all @xmath99 for which the distance between @xmath100 and @xmath85 is at least @xmath19 . we define @xmath101 . the projection @xmath102 maps @xmath5 onto the _ billiard table _ , denoted @xmath103 . the boundary of @xmath5 is @xmath104 and the frame @xmath105 at @xmath8 is as indicated in figure [ frame ] . we also view this @xmath62-dependent frame as the orthogonal map @xmath106 that sends the standard basis vectors @xmath107 of @xmath48 to @xmath108 , for @xmath109 . this allows us to write @xmath110 for each @xmath62 . due to energy conservation , the norms of velocity vectors are not changed during collision or during the free motion between collisions ; we restrict attention to vectors of unit length ( in the kinetic energy metric , which agrees with the standard euclidean metric in @xmath48 under the choice of coordinate @xmath111 for the disc s angle of orientation ) . throughout the paper the notations @xmath112 and @xmath113 are both used for the standard inner product in @xmath48 , the choice being a matter of typographical convenience . ( recall that the kinetic energy metric has been reduced to the standard inner product in @xmath48 by our definition of the variable @xmath114 . ) the _ phase space _ of the billiard system is @xmath115 we refer to vectors in @xmath116 as _ post - collision _ velocities ; we similarly define the space @xmath117 of _ pre - collision _ velocities . the _ billiard map _ @xmath118 , whose domain is a subset of @xmath119 , is the composition of the free motion between two points @xmath120 in @xmath4 and the billiard map @xmath121 at the endpoint . thus @xmath122 is given by @xmath123 where @xmath124 . ( we assume that the shape of the billiard table @xmath103 is such that @xmath118 makes sense and is smooth for all @xmath125 in some big subset of @xmath119 , say open of full lebesgue measure . ) let @xmath126 the first map in this composition , which we denote by @xmath127 , is parallel translation of @xmath128 from @xmath62 to @xmath129 , and the second map , denoted @xmath130 , applies the no - slip reflection map to the translated vector , still denoted @xmath128 , at @xmath129 . hence @xmath131 . taking into account the rotation symmetry of the moving disc , we may for most purposes ignore the angular coordinate ( but not the angular velocity ! ) and restrict attention to the _ reduced phase space_. this is defined as @xmath132 , where an element @xmath133 of the unit disc represents the velocity vector at @xmath134 ( pointing into the billiard region ) given by @xmath135 by _ velocity phase space _ we mean this unit disc . figure [ frame ] summarizes these definitions and figure [ example ] shows what an orbit segment looks like in each space . on the left of the latter figure is shown the two dimensional projection of the orbit segment defined in the @xmath9-dimensional space @xmath5 . the rotation symmetry that justifies passing from the @xmath2-dimensional phase space to the @xmath9-dimensional reduced phase space may be formally expressed by the identity @xmath136 keep in mind that @xmath137 is a parallel vector field ( independent of @xmath62 ) . in particular , @xmath137 is invariant under the differential map : @xmath138 for all @xmath139 . in addition to the orthonormal frames @xmath140 it will be useful to introduce a frame consisting of eigenvectors of the collision map @xmath74 . we define @xmath141 see figure [ eigenframe ] . then @xmath142 yet a third orthonormal frame will prove useful later on in our analysis of period-@xmath1 trajectories . let @xmath143 . then @xmath144 is the orthonormal frame at @xmath62 such that @xmath145 note that @xmath146 and @xmath147 span the @xmath1-space perpendicular to @xmath128 . [ frames ] for any given @xmath143 we refer to @xmath148 as the _ product frame _ , the _ eigenframe _ , and the _ wavefront frame _ , respectively . it appears to be a harder problem in general to show the existence of periodic orbits for no - slip billiards than it is for standard billiard systems in dimension @xmath1 , despite numerical evidence that such points are common . a few useful observations can still be made for specific shapes of @xmath103 . we begin here with the general description of period-@xmath1 orbits . the reader should bear in mind that , when we represent billiard orbits in figures such as [ near_periodic ] , we often draw their projections on the plane , even though periodicity refers to a property of orbits in the @xmath9-dimensional reduced phase space . let @xmath139 be the initial state of a periodic orbit of period @xmath1 , @xmath149 , and @xmath150 the time of free flight between collisions . then , clearly , @xmath151 so that @xmath152 and @xmath153 . because @xmath128 and @xmath154 ( respectively , @xmath155 ) are eigenvectors for different eigenvalues of the orthogonal map @xmath74 ( respectively , @xmath156 ) , @xmath128 is perpendicular to both @xmath154 and @xmath155 . it follows from ( [ u in e ] ) that @xmath157 . thus the projection of @xmath137 to @xmath158 is proportional to @xmath159 . by the definition of the wavefront vector @xmath146 ( and the angle @xmath160 , cf . figure [ eigenframe ] ) we have @xmath161 now observe that @xmath162 is perpendicular to @xmath159 . it follows from this remark and a glance at figure [ eigenframe ] ( to determine the orientation of the vectors ) that @xmath163 notice , in particular , that @xmath128 is a positive multiple of @xmath164 . an elementary calculation starting from this last observation gives @xmath128 in terms of the product frame : @xmath165}{\sqrt{1-\cos^2(\beta/2)\cos^2\phi}}.\ ] ] a more physical description of the velocity @xmath128 of a period-@xmath1 orbit is shown in figure [ period2 ] in terms of the moment of inertia @xmath166 . equally elementary computations yield the collision map @xmath74 in the wavefront frame at @xmath62 , for a period-@xmath1 state @xmath139 . we register this here for later use . to shorten the equations we write @xmath167 and @xmath168 . @xmath169 the following easily obtained inner products will also be needed later . @xmath170 figure [ sinai ] shows ( two copies of the fundamental domain of ) the configuration manifold of the no - slip sinai billiard . here the billiard table is the complement of a circular scatterer in a two - dimensional torus and @xmath5 is the cartesian product of the latter with a one - dimensional torus . notice that there is a whole one - parameter family of initial conditions giving period-@xmath1 orbits , parametrized by the angle @xmath160 . we obtain infinitely many such families by choosing different pairs of fundamental domains . we will return to the no - slip sinai billiard in section [ curved section ] . mostly , in this section , we write @xmath112 instead of @xmath113 for the standard inner product of @xmath48 . let @xmath171 be a smooth curve in @xmath4 such that @xmath172 and @xmath173 . define @xmath174 where @xmath175 is the space of antisymmetric @xmath176 matrices ( the lie algebra of the rotation group ) and @xmath177 is the product frame . as @xmath137 is a parallel field and @xmath178 is antisymmetric we have @xmath179 except possibly for @xmath180 and @xmath181 . denoting by @xmath182 directional derivative of vector fields along @xmath183 at @xmath62 , @xmath184= \left\langle e_2(q ) , d_x e_3\right\rangle= \left\langle e_2(q ) , x\right\rangle \left\langle e_2(q ) , d_{e_2(q)}e_3\right\rangle\ ] ] since @xmath185 . the inner product @xmath186 is the geodesic curvature of the boundary of @xmath103 at @xmath187 , where @xmath187 is the base point of @xmath62 in @xmath188 . thus @xmath189 where @xmath190 given vector fields @xmath191 , we define @xmath192 as the map @xmath193 [ curvlemma ] the directional derivative of @xmath130 along @xmath194 is @xmath195 where @xmath196 , @xmath197=2\cos(\beta/2)\left(\begin{array}{ccc}0 & 0 & \ \ , \sin(\beta/2 ) \\0 & 0 & -\cos(\beta/2 ) \\\sin(\beta/2 ) & -\cos(\beta/2 ) & 0\end{array}\right)\ ] ] and @xmath198 was defined above in ( [ collision_map ] ) . furthermore , @xmath199 and @xmath200 notice that @xmath201 thus @xmath202 therefore , @xmath203 \mathcal{c } \sigma^{-1 } - \sigma \mathcal{c}\left[\sigma^{-1 } d_x\sigma\right ] \sigma^{-1}= \sigma[\omega(x),\mathcal{c}]\sigma^{-1}.\ ] ] the first claimed expression for @xmath204 is now a consequence of equation [ curvature ] . a simple computation also gives , for any given @xmath205 , @xmath206 yielding the second expression for @xmath204 . it is convenient to define the following two projections . let @xmath207 . the space @xmath208 decomposes as a direct sum @xmath209 where @xmath210 ( recall that @xmath211 . ) we refer to these as the _ horizontal _ and _ vertical _ subspaces of @xmath208 . we use the same symbols to denote the projections @xmath212 and @xmath213 defined by @xmath214 notice that for @xmath207 and @xmath215 @xmath216 also observe that @xmath217 , where @xmath218 is the second vector in the wavefront frame ( cf . definition [ frames ] ) and @xmath219 is the orthogonal projection of @xmath128 to the plane perpendicular to @xmath137 . thus , denoting by @xmath220 the angle between @xmath219 and @xmath221 ( this is the same @xmath160 as in figures [ eigenframe ] , [ period2 ] and [ sinai ] ) @xmath222 let @xmath8 , @xmath223 , @xmath224 , @xmath225 , @xmath226 . define @xmath227 clearly @xmath228 is defined on all of @xmath48 , not only on @xmath229 , but we are particularly interested in its restriction to the latter subspace . let @xmath139 be a point contained in a neighborhood of @xmath119 where @xmath118 is defined and differentiable . set @xmath230 . we wish to describe @xmath231 . let @xmath232 be a differentiable curve in @xmath119 with @xmath233 and define @xmath234 then @xmath235 where @xmath236 and @xmath237 . from the equality @xmath238 it follows that @xmath239 consequently , @xmath240 and @xmath241 satisfy @xmath242 and @xmath243v = c_{\tilde{q } } y + \kappa(\tilde{q } ) \langle e_2(\tilde{q } ) , \tilde{x}\rangle \mathcal{o}_{\tilde{q } } v\ ] ] where we have used lemma [ curvlemma ] . from the same lemma , @xmath244 thus @xmath245 as already noted , @xmath246 . by using the projection @xmath247 introduced earlier we may identify @xmath248 with the sum @xmath249 . in this way @xmath250 is regarded as a map from @xmath249 to @xmath251 [ differential ] let @xmath122 be the billiard map , @xmath143 and @xmath252 , where @xmath253 , and @xmath254 . under the identification of the tangent space @xmath255 with @xmath249 as indicated just above , we may regard the differential @xmath250 as a linear map from @xmath249 to @xmath251 also recall from ( [ cross ] ) the definition of @xmath256 . then @xmath257 is given by @xmath258 where @xmath259 and @xmath219 is the orthogonal projection of @xmath128 to @xmath260 . this is a consequence of the preceding remarks and definitions . [ cor ] if @xmath139 is periodic of period @xmath1 , then @xmath152 , @xmath261 , and the map of proposition [ differential ] reduces to @xmath262 where @xmath263 , @xmath259 . clearly , @xmath152 , whence @xmath261 and @xmath264 . also notice that @xmath265 whenever @xmath266 . the corollary follows . it will be seen below that the no - slip billiard map does not preserve the natural symplectic form on @xmath119 , so these mechanical systems are not hamiltonian . nevertheless , the canonical billiard measure derived from the symplectic form ( the liouville measure ) is invariant and the system is time reversible , so some of the good features of hamiltonian systems are still present . ( see , for example , @xcite where a kam theory is developed for reversible systems . ) recall that the invertible map @xmath118 is said to be _ reversible _ if there exists an involution @xmath267 such that @xmath268 in order to see that the no - slip billiard map @xmath118 is reversible we first define the following maps : @xmath269 , where @xmath150 is the time of free motion of the trajectory with initial state @xmath270 , so that @xmath271 ; the collision map @xmath272 given by @xmath273 ; and the flip map @xmath274 where @xmath8 and @xmath205 . recall that @xmath131 . now set @xmath275 . it is clear ( since @xmath74 is an involution by proposition [ classification ] ) that @xmath276 and that @xmath277 . therefore , @xmath278 notice that if @xmath279 is a reversible isomorphism of a vector space @xmath280 with time reversal map @xmath281 ( so that @xmath282 ) then for any eigenvalue @xmath283 of @xmath284 associated to eigenvector @xmath133 , @xmath285 is also an eigenvalue for the eigenvector @xmath286 , as easily checked . these simple observations have the following useful corollary . let @xmath287 be a periodic point of period @xmath288 of the no - slip billiard system and let @xmath283 be an eigenvalue of the differential map @xmath289 corresponding to eigenvector @xmath133 . then @xmath285 is also an eigenvalue of @xmath290 corresponding to eigenvector @xmath291 , where @xmath267 is the composition of the collision map @xmath130 and the flip map @xmath292 . furthermore , @xmath137 ( see definition [ frames ] ) is always an eigenvector of @xmath250 and all its powers , corresponding to the eigenvalue @xmath94 . we now turn to invariance of the canonical measure . the canonical @xmath94-form @xmath31 on @xmath119 is defined by @xmath293 for @xmath143 and @xmath294 . its differential @xmath295 is a symplectic form on @xmath296 and @xmath297 is the canonical volume form on this same set . in terms of horizontal and vertical components of vectors in @xmath298 , the symplectic form is expressed as @xmath299 where @xmath300 . an elementary computation shows that the measure on @xmath119 associated to @xmath301 is given by @xmath302 where @xmath303 , @xmath304 is the area measure on @xmath4 , and @xmath305 is the area measure on the hemisphere @xmath306 . the canonical @xmath2-form @xmath301 on @xmath119 transforms under the no - slip billiard map as @xmath307 . in particular , the associated measure @xmath308 , shown explicitly in equation [ liouville ] , is invariant under @xmath118 . let @xmath133 be a vector field on @xmath4 and introduce the one - form @xmath309 on @xmath119 given by @xmath310 for @xmath139 and @xmath311 . taking @xmath133 to be each of the vector fields @xmath312 we obtain the @xmath94-forms @xmath313 and @xmath314 . as @xmath315 and @xmath316 , we have @xmath317 the no - slip collision map @xmath130 acts on @xmath318 as follows : for @xmath319 , @xmath320 it follows that @xmath321 we now compute the differentials @xmath322 for @xmath323 . observe that @xmath324 , where @xmath325 is the function on @xmath119 defined by @xmath326 and @xmath327 is the pull - back under the projection map @xmath328 of the @xmath94-form @xmath329 on @xmath4 given by @xmath330 . thus @xmath331 a simple calculation gives @xmath332 the vector field @xmath333 is parallel on @xmath4 . in fact , its derivative in direction @xmath194 only has component in the normal direction , given by @xmath334 omitting the dependence on @xmath62 , we have @xmath335 another simple calculation gives @xmath336 so @xmath337 . explicitly , @xmath338 where @xmath339 notice that @xmath340 is the area form on @xmath4 . a convenient way to express @xmath341 is as follows . define the @xmath94-form @xmath342 on @xmath119 by @xmath343 , where @xmath344 , and the function @xmath345 . these extra bits of notation now allow us to write @xmath346 the main conclusion we wish to derive from these observations is that @xmath347 . this is the case because , as @xmath348 , we must have @xmath349 and @xmath350 . therefore , @xmath351 finally , @xmath352 the forms @xmath295 and @xmath301 are invariant under the geodesic flow and under the map it induces on @xmath119 . as @xmath118 is the composition of this map and @xmath130 , the proposition is established . we set the following conventions for a wedge table with corner angle @xmath353 . see figure [ wedge ] . ( this is the same @xmath160 that has appeared before in previous figures . ) the boundary planes of the configuration manifold are denoted @xmath354 and @xmath355 . the orthonormal vectors of the constant product frame on plane @xmath356 are @xmath357 for @xmath44 where @xmath358 let @xmath359 be the constant orthogonal map such that @xmath360 , where @xmath107 , @xmath109 , is our notation for the standard basis vectors in @xmath48 . let @xmath361 be the eigenvectors of the no - slip reflection map associated to the plane @xmath356 and set @xmath362 . for easy reference we record their matrices here : @xmath363 the initial velocity @xmath128 for the period-@xmath1 trajectory points in the direction of @xmath364 and is given by @xmath365 this periodic trajectory connects the points @xmath366 and @xmath367 . any such pair of points can be written as @xmath368 where @xmath369 , @xmath370 . in what follows we assume without loss of generality that @xmath371 and @xmath372 . thus @xmath373 let @xmath374 . the collision maps @xmath375 , @xmath44 are given by the matrices @xmath376 where @xmath198 was defined in [ collision_map ] . we now introduce coordinates on @xmath377 as follows . let @xmath378 and define @xmath379 by @xmath380 regarding @xmath381 as @xmath382 , we may then write @xmath383 clearly , the billiard map is not defined on all of @xmath384 since those initial velocities not pointing towards the other plane will escape to infinity , but we are interested in the behavior of the map on a neighborhood of the periodic point @xmath385 , @xmath386 . the question of interest here is whether some open neighborhood of @xmath387 remains invariant under the billiard map . it is easily shown that the coordinates of the state @xmath387 ( of the period-@xmath1 orbit at the plane @xmath356 ) are @xmath388 where @xmath389 let @xmath390 be the billiard map restricted to @xmath377 expressed in the coordinate system defined by @xmath391 . thus @xmath392 on their domains @xmath393 . we now find the explicit form of @xmath394 . define @xmath395 and orthogonal matrices @xmath396 and @xmath397 , both in @xmath398 . also define @xmath399 observe that @xmath400 . it is easily shown that @xmath401 in particular , @xmath402 . let @xmath403 be defined by @xmath404 notice that @xmath405 . we now have @xmath406 for easy reference we record @xmath407 and @xmath408 using the notation @xmath409_3:=z\cdot\epsilon_3 $ ] and elementary computations based on the above gives : the return map in the coordinate system defined by @xmath410 has the form @xmath411_3 v(y ) , sa_1^tsa_1 y\right)\ ] ] where @xmath412_3 a_1^tsa_1y- [ a_1^tsa_1 y]_3y}{[a_1y]_3 [ a_1^tsa_1 y]_3}.\ ] ] this vector satisfies : @xmath413_3=0 $ ] and @xmath414 . in particular , @xmath415 whenever @xmath416 is in the domain of @xmath417 . in order to study this return map in a neighborhood of @xmath416 we use spherical coordinates about the axis @xmath418 : @xmath419 where @xmath420 form an orthonormal frame . see figure [ directions ] . ( notice the typographical distinction between the corner angle @xmath160 of the wedge domain and the spherical coordinate @xmath421 . ) let @xmath422 and define @xmath423 thus we may write @xmath424 . since the rotation @xmath425 fixes @xmath418 , it acts on @xmath426 as @xmath427 for some constant angle @xmath31 . it follows that @xmath428 the following proposition summarizes these observations and notations . [ r ] for points @xmath429 in a neighborhood of @xmath418 we adopt spherical coordinates relative to the axis @xmath430 , so that @xmath431 where @xmath432 see figure [ directions ] . we also use the notations @xmath409_3:=z\cdot\epsilon_3 $ ] , @xmath433 , and @xmath434 . let @xmath435 be the @xmath1-step return map as defined above , whose domain contains a neighborhood of @xmath416 for all @xmath381 . then @xmath436 for all @xmath114 and @xmath437 for an angle @xmath31 , depending only on the wedge angle @xmath353 and the characteristic angle @xmath7 of the no - slip reflection , such that @xmath438 where @xmath439 . writing @xmath440 we have @xmath441_3}{[y_1]_3 } \frac { ( i+s_1)w-\frac{[(i+s_1)w]_3 y_1}{[y_1]_3 } + \tan\psi \frac{[w]_3 s_1w- [ s_1w]_3 w}{[y_1]_3}}{1 - \tan\psi\left(\frac{[(a_1+s_1)w]_3}{[y_1]_3 } - \tan\psi\frac{[a_1w]_3[s_1w]_3}{[y_1]_3 ^ 2}\right)}\\ \phi & = \varphi + \theta\\ \psi & = \psi \end{cases}\ ] ] denoting @xmath442 , we further have @xmath443 since @xmath444 remains constant under iterations of the return map @xmath445 , we regard @xmath444 as a fixed parameter . we are interested in small values of @xmath446 . notice that @xmath447_3:=(a_1 z)\cdot\epsilon_3=z\cdot(a_1^t \epsilon_3)=\mu_0\cdot z$ ] , where @xmath448 write @xmath449 , so @xmath450 then the proposition shows that @xmath19 has the form @xmath451 where the vector @xmath452 can be made arbitrarily ( uniformly ) small by choosing @xmath444 ( or @xmath453 ) sufficiently close to @xmath23 . observe from the explicit form @xmath454_3}\frac { r\left((i+s_1)w-\frac{[(i+s_1)w]_3 y_1}{[y_1]_3}\right ) + r^2 \frac{[w]_3 s_1w- [ s_1w]_3 w}{[y_1]_3}}{1 - r\frac{[(a_1+s_1)w]_3}{[y_1]_3 } + r^2\frac{[a_1w]_3[s_1w]_3}{[y_1]_3 ^ 2}}\ ] ] that @xmath455 so that @xmath183 is indeed in @xmath6 . [ inv ] the quantity @xmath456 satisfies the coboundary relation @xmath457 where @xmath458 in fact , the transformation @xmath19 on the @xmath9-dimensional space @xmath459 , obtained by fixing a value of @xmath444 ( hence of @xmath453 ) , leaves invariant the measure @xmath460 where @xmath461 is a positive constant ( only dependent on the fixed parameters @xmath462 ) and @xmath463 is the standard area measure on @xmath6 . the canonical invariant measure on @xmath464 has the form @xmath465 , where @xmath466 is the area measure on @xmath467 . for a fixed value of @xmath444 we obtain an invariant measure on @xmath468 of the form @xmath469 . using the form of @xmath470 given by ( [ y ] ) , one obtains @xmath471 this shows that , up to a multiplicative constant , the invariant measure @xmath472 has the indicated form . equation ( [ coboundary ] ) is an easy consequence of the invariance of @xmath472 with respect to @xmath19 . by using the coordinate system @xmath473 on @xmath6 , the area measure is @xmath474 and , as observed at the end of proposition [ r ] , the transformation @xmath19 maps the fibers of the projection @xmath475 to fibers preserving the length measure on fibers . thus we obtain a transformation @xmath476 on @xmath477 preserving the measure @xmath478 where @xmath479 has the stated expression . using the quotient coordinates @xmath480 and @xmath160 and writing @xmath481 we obtain @xmath482 in particular , @xmath483 the invariant measure is @xmath484 where @xmath479 is the density given in proposition [ inv ] . it is now immediate that @xmath485 this shows that all the iterates of @xmath486 remain uniformly close to the initial point for small values of @xmath444 . also notice that @xmath487 . this means that if @xmath488 remains bounded , the length coordinate along the base of @xmath354 also must be similarly bounded . from this we conclude : [ local_stability ] assume the notation introduced at the beginning of this section . for all @xmath489 , @xmath44 , and any neighborhood @xmath490 of the period-@xmath1 state @xmath491 , there exists a small enough neighborhood @xmath492 of @xmath493 the orbits of whose points remain in @xmath494 . because any ( bounded ) polygonal billiard shape must have a corner with angle less than @xmath0 , the following corollary holds . polygonal no - slip billiards can not be ergodic for the canonical invariant measure . the analysis of the previous section is based on the existence of period-@xmath1 orbits in wedge - shaped no - slip billiard tables . existence of periodic orbits of higher periods is in general difficult to establish , although one such result for wedge domains will be indicated below in this section . we first point out a generalization of corollary [ local_stability ] to perturbations of periodic orbits in general polygon - shaped domains . figure [ near_periodic ] illustrates the type of stability implied by the following theorem [ polygon stability ] . [ polygon stability ] periodic orbits in no - slip polygon - shaped billiard domains are locally stable . that is , given an initial state @xmath495 for a period-@xmath496 orbit in such a billiard system , and for any neighborhood @xmath490 of @xmath497 , there exists a small enough neighborhood @xmath492 of @xmath125 the orbits of whose elements remain in @xmath490 . the idea is essentially the same as used in the proof of proposition [ inv ] and corollary [ local_stability ] . we only indicate the outline . by a choice of convenient coordinates around the periodic point , it is possible to show that the @xmath496-th iterate of the billiard map @xmath118 , denoted @xmath498 , can be regarded as a map from an open subset of @xmath468 into this latter set , having the form @xmath499 for a certain angle @xmath31 , where @xmath500 is a linear transformation independent of @xmath114 . rotation invariance implies that @xmath19 must satisfy the invariance property @xmath501 for a vector @xmath502 . from this we define a map @xmath503 on ( a subset of ) the quotient @xmath477 , @xmath504 . furthermore , denoting by @xmath505 the coordinates in this quotient space , invariance of the canonical measure implies invariance of a measure @xmath472 on this quotient having the form @xmath506 invariance is with respect to the quotient map @xmath507 for some function @xmath508 . this function must then take the form @xmath509 . iterates of @xmath503 will then behave like the corresponding map for the wedge domain , defined prior to theorem [ local_stability ] . we turn now to the question of existence of periodic orbits of higher ( necessarily even ) periods for wedge shapes . clearly , a necessary condition is that the angle @xmath31 introduced in proposition [ r ] ( see also figure [ directions ] ) be rational . for orbits that do not eventually escape to infinity , this is also a sufficient condition , as a simple application of poincar recurrence shows . ( see @xcite . ) moreover , as @xmath31 is only a function of @xmath439 , which is given by ( proposition [ r ] ) @xmath510 where @xmath7 is the characteristic angle of the system ( a function of the mass distribution on the disc ) and @xmath353 is the corner angle of the wedge domain , if a higher order periodic orbit exists for a given @xmath511 , all bounded orbits have the same period . we give a few examples for the uniform mass distribution , for which @xmath512 solving [ delta ] for @xmath513 , for @xmath514 , choosing first the negative square root , gives @xmath515 a few examples are shown in figure [ periodpq1 ] . notice that there are no restrictions on the values of @xmath516 and @xmath62 . the following proposition is a consequence of these remarks . for any positive even integer @xmath496 there exists a wedge domain for which the no - slip billiard has period-@xmath496 orbits . more specifically , all bounded orbits of the no - slip billiard in a wedge domain with corner angle @xmath517 satisfying equation [ phipq ] are periodic with period @xmath518 . solving [ delta ] for @xmath513 , for @xmath514 , but choosing now the positive square root , gives @xmath519 this makes sense so long as @xmath520 , which greatly restricts the choices of @xmath516 and @xmath62 . a few examples in this case are shown in figure [ periodpq ] . it is interesting to observe that all orbits of the equilateral triangle are periodic with period @xmath2 or @xmath521 . ( see figure [ equilateral ] and @xcite for the proof . ) we do not know of any other no - slip billard domain all of whose orbits are periodic . a final observation concerning polygonal no - slip billiards is suggested by plots of their velocity phase portraits . a typical such plot is shown in figure [ moon ] . it is apparent that the orbits drawn all seem to lie on a stable neighborhood of some periodic orbit , and this pattern is seen at all scales that we have explored , but we do not yet have a clear topological dynamical description of this observation . we now turn to the problem of characterizing stability of period-@xmath1 orbits for no - slip billiard domains whose boundary may have non - zero geodesic curvature . here we only address _ linear _ rather than local stability as we did before for polygonal billiards . in other words , we limit ourselves to the problem of determining when the differential of the billiard map @xmath250 at a period-@xmath1 collision state @xmath139 is elliptic or hyperbolic , and precise thresholds ( where it is parabolic ) . to go from this information to local stability would require developing a kam theory for no - slip billiards in the model of @xcite , something we do not do in this paper . let @xmath139 be periodic with period @xmath1 for the no - slip billiard map and consider the differential @xmath522 . then either all the eigenvalues of @xmath523 are real , of the form @xmath524 or , if not all real , they are @xmath525 where @xmath526 . this is a consequence of the following observations . first , we know that @xmath307 , where @xmath301 is the canonical symplectic form ( cf . section [ measure reversible ] ) . therefore , the product of the eigenvalues of @xmath523 counted with multiplicity is @xmath94 . the vector @xmath527 , where @xmath137 is the first vector in the product frame and @xmath528 is the first vector in the wavefront frame , is an eigenvector for eigenvalue @xmath94 of @xmath250 due to rotation symmetry , as already noted . if we regard @xmath250 as a self - map of @xmath249 as in the corollary to proposition [ differential ] then we should use instead the vector @xmath529 . ( recall that @xmath528 is collinear with the orthogonal projection of @xmath137 to @xmath158 . ) in addition , by reversibility of @xmath118 , if @xmath283 is an eigenvalue of @xmath523 , then @xmath285 is one also , and since @xmath523 is a real valued linear map , the complex conjugates @xmath530 and @xmath531 are also eigenvalues . as the dimension of the linear space is @xmath2 , if one of the eigenvalues , @xmath283 , is not real , it must be the case that @xmath532 and we are reduced to the case @xmath525 with @xmath533 . if all eigenvalues are real , and @xmath534 is one eigenvalue , then we are reduced to the case @xmath535 . to proceed , it is useful to express the differential map of corollary [ cor ] in somewhat different form . first observe , in the period-@xmath1 case ( in which @xmath538 and @xmath539 ) , that @xmath540 ( see section [ periodic section ] . ) now define the rank-@xmath94 operator @xmath541 then @xmath542 when the geodesic curvature satisfies @xmath543 we obtain a simplification in the criterion for ellipticity , as will be seen shortly . with this special case in mind we define the linear map @xmath544 on @xmath158 by @xmath545 , @xmath546 notice that @xmath547 . then @xmath548 the same notation @xmath549 will be used for the map on @xmath249 given by @xmath550 . notice that @xmath551 since @xmath552 . it follows that @xmath553 in particular , when @xmath543 , we have @xmath554 and @xmath555 . therefore , rather than computing the trace of @xmath556 , we need only consider the easier to compute trace of @xmath557 . a straightforward calculation gives the trace of these maps , which we record in the next lemma . these traces can now be computed using equations ( [ cwavefront ] ) and ( [ inner ] ) . the matrices expressing @xmath563 in the wavefront basis of @xmath158 are given as follows . for convenience we write @xmath564 where @xmath565 and @xmath566 are defined in corollary [ cor ] . @xmath567_w=\left(\begin{array}{cc}1 - 2c^2 c^2_\phi & -2 c c_\phi \varrho \\-2cc_\phi\varrho & -1 + 2 c^2 c^2_\phi\end{array}\right ) , \ \ [ r]=\left(\begin{array}{cr}1 & 0 \\0 & -1\end{array}\right ) , \ \ [ \theta]_w=2 c \frac{c_\psi}{c_\phi}\left(\begin{array}{cc}0 & \varrho \\0 & -cc_\phi\end{array}\right).\ ] ] let @xmath568 be the distance between the projections of @xmath62 and @xmath129 on plane the billiard table , @xmath219 the projection of @xmath128 on the same plane and @xmath150 , as before , the time between consecutive collisions . from @xmath569 it follows that @xmath570 . we then obtain @xmath571 -2\kappa \bar{d } \cos^2(\beta/2)\cos\phi\ ] ] and @xmath572 ^ 2\right.\\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ -(\kappa(q)+\kappa(\tilde{q}))\cos^2(\beta/2)\cos\phi \left[1 - 2\cos^2(\beta/2)\cos^2\!\phi\right]\bar{d}\\ & \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \left.+ \kappa(q)\kappa(\tilde{q } ) \cos^4(\beta/2)\cos^2\ ! \phi\ , { \bar{d}}^2\right\ } \end{split}\end{aligned}\ ] ] observe that in the special case in which @xmath543 we have @xmath573- 2\kappa \cos^2(\beta/2)\cos\phi\ , \bar{d}\right\}^2\ ] ] [ general corner ] suppose that the billiard domain has a piecewise smooth boundary with at least one corner having inner angle less than @xmath0 . then , arbitrarily close to that corner point , the no - slip billiard has ( linearly ) elliptic period-@xmath1 orbits . period-@xmath1 orbits exist arbitrarily close to the corners of a piecewise smooth billiard domain as figure [ corner ] makes clear . for period-@xmath1 orbits near a corner the above expression for @xmath574 gives for small @xmath568 @xmath575 ^ 2 + o(\bar{d})<4.\ ] ] this implies that @xmath576 and the theorem follows from corollary [ trace corollary ] . theorem [ general corner ] ( and numerical experiments ) strongly suggests that such no - slip billiards will aways admit small invariant open sets and thus can not be ergodic with respect to the canonical billiard measure . we illustrate numerically the transition between elliptic and hyperbolic in the special case of equal curvatures at @xmath62 and @xmath129 . define @xmath577 . when @xmath578 ( equivalently , the curvature is positive ) , the critical value of @xmath579 is @xmath580 the condition for the periodic point to be elliptic is @xmath581 . when @xmath582 , the critical value of @xmath579 is @xmath583 and the condition for ellipticity is @xmath584 . consider the example of the no - slip sinai billiard . ( see figures [ example ] and [ sinai ] . ) we examine small perturbations of the periodic orbit corresponding to the angle @xmath10 . figure [ cutoff ] suggests a transition from chaotic to more regular type of behavior for a radius between @xmath585 and @xmath586 . in reality the critical radius for the @xmath10 periodic orbits is exactly @xmath587 . so the observed numbers are smaller . we should bear in mind that the periodic points are not isolated , but are part of a family parametrized by @xmath160 . as @xmath160 increases , the critical parameter @xmath588 changes ( for the uniform mass distribution , where @xmath589 ) according to the expression @xmath590 . given in terms of the radius of curvature , @xmath591 . solving for the critical @xmath19 yields @xmath592 . thus for a period-@xmath1 trajectory having a small but non - zero @xmath160 , the critical radius is less than @xmath587 . it is then to be expected that the experimental critical value of @xmath19 , for orbits closed to that having @xmath10 will give numbers close to but less than @xmath587 . moreover , as @xmath593 approaches @xmath23 when @xmath160 approaches @xmath594 , we obtain the following proposition which , together with experimental evidence indicates that the no - slip sinai billiard is not ergodic . in this case , the critical transition from hyperbolic to elliptic , for the horizontal periodic orbit at middle height shown in the figure , happens for the disc . a transition behavior similar to that observed for the sinai billiard seems to occur near the period-@xmath1 orbits shown in figure [ contact ] . the number indicated below each velocity phase portrait is the angle of each circular arc . thus , for example , the disc corresponds to angle @xmath0 ; smaller angles give shapes like that on the left in figure [ contact ] . the cut - off angle at which the indicated periodic orbit becomes elliptic is @xmath0 . notice , however , that the experimental value for this angle is greater than @xmath0 . just as in the sinai billiard example , we should keep in mind that the periodic orbits are not isolated ; in this case , the bias would be towards greater values of the angle . c. cox , r. feres , _ differential geometry of rigid bodies collisions and non - standard billiards_. discrete and continuous dynamical systems - a , 33 ( 2016 ) no . 11 , 60656099 . c. cox , r. feres , _ no - slip billiards in dimension @xmath1_. ( arxiv:1602.01490 )

the ferrimagnetic spinel oxide znfeo combines high curie temperature and spin polarization with tunable electrical and magnetic properties , making it a promising functional material for spintronic devices . we have grown epitaxial znfeo thin films ( ) on mgo(001 ) substrates with excellent structural properties both in pure ar atmosphere and an ar / o mixture by laser molecular beam epitaxy and systematically studied their structural , magnetotransport and magnetic properties . we find that the electrical conductivity and the saturation magnetization can be tuned over a wide range ( and/f.u . at room temperature ) by zn substitution and/or finite oxygen partial pressure during growth . our extensive characterization of the films provides a clear picture of the underlying physics of the spinel ferrimagnet znfeo with antiparallel fe moments on the and sublattice : ( i ) zn substitution removes both fe moments from the sublattice and itinerant charge carriers from the sublattice , ( ii ) growth in finite oxygen partial pressure generates fe vacancies on the sublattice also removing itinerant charge carriers , and ( iii ) application of both zn substitution and excess oxygen results in a compensation effect as zn substitution partially removes the fe vacancies . both electrical conduction and magnetism is determined by the density and hopping amplitude of the itinerant charge carriers on the sublattice , providing electrical conduction and ferromagnetic double exchange between the mixed - valent fe/fe ions on the sublattice . a decrease ( increase ) of charge carrier density results in a weakening ( strengthening ) of double exchange and thereby a decrease ( increase ) of conductivity and the saturation magnetization . this scenario is confirmed by the observation that the saturation magnetization scales with the longitudinal conductivity . the combination of tailored znfeo films with semiconductor materials such as zno in multi - functional heterostructures seems to be particularly appealing .

0808.3642

spintronic materials and devices are in the focus of current research activities @xcite . regarding functional spintronic materials , transition metal oxides are promising candidates , since they show a rich variety of electrical and magnetic properties specifically interesting for applications in spintronics . evidently , useful spintronic devices such as magnetoresistive elements based on the tunneling magnetoresistance @xcite or spin transistors @xcite require ferromagnetic materials with high curie temperature @xmath12 well above room temperature and large spin polarization @xmath13 of the charge carriers at the fermi level @xmath14 . moreover , advanced applications often require materials with magnetic properties which can be deliberately tuned by external control parameters such as an applied electric field @xcite , elastic stress @xcite , or light @xcite . in this respect , fe@xmath15o@xmath2 ( magnetite ) is a highly promising candidate . first , it has both a high curie temperature of @xmath16k and according to band structure calculations is expected a half - metal @xcite corroborated by spin - resolved photoelectron spectroscopy@xcite . second , recent experiments indicated that both the electronic and magnetic properties of fe@xmath15o@xmath2 thin films can be nicely tailored in solid solution systems of fe@xmath17o@xmath2 ( @xmath18 ) @xcite . third , it has been demonstrated that the magnetic anisotropy of fe@xmath15o@xmath2 can be tuned by elastic stress imposed by a piezoelectric actuator@xcite . taken all together , due to their interesting , versatile , and tunable properties magnetite thin films and heterostructures became the focus of recent research activities . before discussing the various possibilities for tuning and tailoring the electronic and magnetic properties of magnetite we briefly review its basic structural and magnetic properties . magnetite is known to have an inverse spinel structure as shown in fig . [ fig : spinel](a ) . the @xmath7 sites ( 8 per unit cell ) , which are surrounded by oxygen tetrahedra , are occupied by trivalent fe@xmath9 ions ( @xmath19 ) , whereas on the octahedrally coordinated @xmath8 sites ( 16 per unit cell ) there is an alternating arrangement of fe@xmath10 ( @xmath20 ) and fe@xmath11 ( @xmath19 ) ions . therefore , the sum formula of magnetite can be expressed as [ fe@xmath21@xmath22[fe@xmath23fe@xmath24@xmath25o@xmath2 . the density of itinerant charge carriers is determined by the density of the @xmath26 spin - down electron on the @xmath8 site , i.e. by the density of fe@xmath10 . the magnetic exchange in magnetite is governed by a combination of antiferromagnetic superexchange ( se ) and ferromagnetic double exchange ( de ) interactions . there are three antiferromagnetic se interactions @xmath27 ( @xmath7-o-@xmath7 ) , @xmath28 ( @xmath8-o-@xmath8 ) , and @xmath29 ( @xmath7-o-@xmath8 ) between the fe@xmath23 ions on the @xmath7 and @xmath8 sites mediated by the oxygen ( o ) ions . in addition , there is a ferromagnetic de interaction mediated by the itinerant spin - down @xmath26 electrons hopping between the mixed - valent fe ions on the @xmath8 sites ( cf . [ fig : spinel ] ) . owing to hund s rule coupling , the spins of these itinerant electrons are antiferromagnetically coupled to the localized spins formed by the @xmath30 spin - up electrons . as pointed out by nel @xcite , in the simplest model ferrimagnetism in fe@xmath15o@xmath2 with high @xmath12 is obtained without any de interaction for large @xmath31 forcing an antiparallel alignment of the moments on the @xmath7 and @xmath8 sites ( see fig . [ fig : spinel](b ) ) . since the antiparallel fe@xmath9 and fe@xmath11 moments compensate each other , a saturation magnetization of @xmath32/formula unit ( f.u . ) is expected from the remaining fe@xmath10 ( @xmath33 ) moments . later on , the simple nel model has been extended by yafet and kittel @xcite . they proposed a more elaborate model in which the @xmath8 sublattice is subdivided into two fe@xmath10 and fe@xmath11 sublattices . it was shown that on weakening @xmath29 and strengthening @xmath28 , the @xmath8 site magnetic moments are no longer rigidly parallel to the @xmath7 site moments . the stronger @xmath8-o-@xmath8 se interaction results in spin canting expressed by a finite yafet - kittel angle and thus a reduction of the saturation magnetization . more recent models show that a detailed modelling of the magnetic properties of magnetite is only possible by taking into account the ferromagnetic de interaction of the @xmath8 sublattice competing with the antiferromagnetic se @xcite . on the tetrahedrally coordinated @xmath7 site and mixed - valent fe@xmath10 ( @xmath34 , s=2)/fe@xmath11 ( @xmath35 ) ions on the octahedrally coordinated @xmath8 site . zn substitution replaces fe@xmath9 on the @xmath7 site by nonmagnetic zn@xmath36 ( @xmath37 ) ions . also shown is the occupation of the fe@xmath10 and fe@xmath11 @xmath30 states with @xmath26 and @xmath38 symmetry separated by the crystal field splitting due to the octahedral oxygen surrounding . the @xmath26 spin - down electron can hop between fe@xmath10 and fe@xmath11 ions with its spin coupled anti - parallel to the local moment formed by the spin - up electrons owing to hund s rule coupling . ( b ) - ( e ) shows the @xmath7 and @xmath8 sublattice configuration for zn@xmath0fe@xmath1o@xmath2 for ( b ) @xmath39 , ( c ) @xmath40 , ( d ) @xmath41 , and ( e ) @xmath42 . ] there are several possibilities to tailor the magnetic properties of magnetite . first , fe@xmath9 on the @xmath7 site can be replaced by an isovalent nonmagnetic ion . in this way antiferromagnetically coupled moments on the @xmath7 sites are removed without affecting the magnetic exchange on the @xmath8 sublattices . hence , an increase of the saturation magnetization is expected at low substitution level . however , on increasing the @xmath7 site substitution the dilution of the @xmath7 site moments reduces @xmath29 , whereas @xmath28 stays about constant . then the moments on the @xmath7 and @xmath8 site do no longer need to be rigidly antiparallel , resulting in a finite yafet - kittel angle ( see fig . [ fig : spinel](c ) and ( d ) ) . this angle increases with increasing substitution , resulting in a reduction of the saturation magnetization . due to the isovalent substitution the charge carrier density should stay unaffected . however , the hopping between the mixed - valent ( fe@xmath10/fe@xmath11 ) ions on the @xmath8 sublattice is reduced due to the spin canting , resulting in a lower electrical conductivity and reduced de interaction . to our knowledge a nonmagnetic isovalent @xmath7 site substitution in fe@xmath15o@xmath2 has not been reported so far . most likely this can be attributed to the fact that suitable trivalent ions usually prefer the octahedral coordination on the @xmath8 site and thus can not be substituted solely on the tetrahedral @xmath7 site . second , as done in our work , fe@xmath9 on the @xmath7 site can be replaced by a nonmagnetic divalent ion such as zn@xmath36 ( @xmath37 ) @xcite . it is known that zn@xmath36 preferably occupies the tetrahedrally coordinated @xmath7 site in the inverse spinel structure @xcite . again some of the antiferromagnetically coupled moments on the @xmath7 sites are removed , leading to an initial increase of the saturation magnetization at low zn substitution @xcite . however , in the same way as discussed before , a finite yafet - kittel angle and , in turn , a reduction of the saturation magnetization is expected going to larger substitution levels @xcite . moreover , substitution of fe@xmath9 by zn@xmath36 on the @xmath7 site also reduces the amount of fe@xmath10 on the @xmath8 site due to charge neutrality . that is , the amount of itinerant charge carriers mediating the de on the @xmath8 sublattice is reduced as shown by photoemission spectroscopy @xcite . thus , with increasing zn substitution the electrical conductivity is reduced both by a reduction of the density of itinerant charge carriers and their hopping amplitude due to spin canting . third , fe vacancies can be introduced by preparing magnetite samples in excess oxygen @xcite . since fe vacancies are formally equivalent to the presence of excess of o@xmath43 ions , charge neutrality again requires a reduced ( increased ) amount of fe@xmath10 ( fe@xmath11 ) ions on the @xmath8 site . this strengthens antiferromagnetic se and weakens ferromagnetic de on the @xmath8 sublattice . in turn , this results in a finite yafet - kittel angle and a reduced saturation magnetization . furthermore , a reduced electrical conductivity is expected both by a reduced carrier density and hopping amplitude . our brief discussion shows that fe based magnetic oxides with spinel structure are interesting materials . they do not only have high curie temperature and high spin polarization but also offer various opportunities to tailor their electrical ( charge carrier density , electrical conductivity ) and magnetic properties ( saturation magnetization , curie temperature ) . such tunable ferromagnetic materials are desired for spintronic devices operating at room temperature . for example , tuning the electrical conductivity of ferromagnetic materials with high spin polarization is promising for spin injection devices as one can reduce and deliberately choose the conductivity mismatch between semiconductors and ferromagnetic spin injectors . to this end , recently the growth of epitaxial fe@xmath15o@xmath2/zno heterostructures has been demonstrated @xcite . here , we report on the growth as well as the structural , magnetic and magnetotransport properties of epitaxial thin film samples of zn@xmath0fe@xmath1o@xmath2 with @xmath44 deposited in different oxygen partial pressure . we show that their saturation magnetization and electrical conductivity can be tuned over a wide range both by zn substitution and varying oxygen partial pressure during growth . furthermore , our systematic study provides clear evidence that their electrical conductivity is closely related to the overall magnetization . we also carefully consider the presence of anti - phase boundaries ( apbs ) in the thin film samples which could affect both their electrical conductivity and their saturation magnetization @xcite . however , we have clear evidence that apbs do not play a dominant role in this study . in particular , they can not explain the change of the magnetic and magnetotransport properties on varying the zn content and the deposition atmosphere . in contrast , these properties can be consistently explained solely by disorder and spin canting on the @xmath8 sublattice , tending to localize the itinerant charge carriers on the @xmath8 sublattice . epitaxial thin films of zn@xmath0fe@xmath1o@xmath2 with @xmath45 and thicknesses between 40 nm and 60 nm were deposited from stoichiometric targets on single crystalline , ( 001 ) oriented mgo substrates by laser molecular beam epitaxy ( laser - mbe)@xcite . the exact film thickness was determined by x - ray reflectometry . both bulk fe@xmath15o@xmath2 having an inverse spinel structure ( @xmath46 ) and mgo ( @xmath47 ) are cubic with lattice constants of @xmath48 @xcite and @xmath49 , respectively , resulting in a small lattice mismatch of @xmath50 . the energy density of the krf excimer laser ( @xmath51 nm ) at the target was 3.1j/@xmath52 and the laser repetition rate 2hz . at the same total pressure of @xmath53 mbar , two sets of samples were deposited in two different growth atmospheres . in pure ar , thin films with @xmath39 , 0.1 , 0.5 , and 0.9 were grown at a substrate temperature of @xmath54c . in an ar / o@xmath4 ( 99:1 ) mixture , thin films with @xmath41 were deposited at @xmath54c and with @xmath39 , 0.33 , and 0.9 at @xmath55c . the growth process was monitored _ in - situ _ by reflection high energy electron diffraction ( rheed)@xcite . we observed four rheed intensity oscillations per unit cell , indicating a block - by - block growth mode with four charge neutral blocks . this is already known for fe@xmath15o@xmath2@xcite and also holds when substituting zn up to zn@xmath56fe@xmath57o@xmath2 . a more detailed description of the thin film deposition for fe@xmath15o@xmath2 on mgo is given elsewhere@xcite . the structural properties of the zn@xmath0fe@xmath1o@xmath2 films were analyzed by high resolution x - ray diffractometry using a bruker axs d8 discover four - circle diffractometer . our detailed analysis reveals a very high epitaxial quality of all thin film samples . in @xmath58 $ ] direction , @xmath59-@xmath60 scans display no impurity phases - in particular from other iron or zinc oxides . we note that we may not be able to distinguish between fe@xmath15o@xmath2 and @xmath61-fe@xmath4o@xmath15 ( maghemite ) as they both share the same inverse spinel lattice@xcite . it is known that excess oxygen results in the formation of fe vacancies and in the extreme case to formation of fe@xmath15o@xmath2/@xmath61-fe@xmath4o@xmath15 solid solutions @xcite . the fe@xmath15o@xmath2/@xmath61-fe@xmath4o@xmath15 thermodynamic equilibrium line was determined to @xmath62 with @xmath63bar @xcite . unfortunately , both calculations and experimental studies of phase and point defect equilibria always apply to thermal equilibrium situations which certainly are not appropriate for the laser - mbe growth process occuring far from equilibrium . nevertheless , the high energy of the particles in the laser plume can be associated to an effective growth temperature much larger than the substrate temperature favoring the stability of fe@xmath15o@xmath2 @xcite . therefore , for the deposition temperature and oxygen partial pressure used in our experiments we are sufficiently far away from the fe@xmath15o@xmath2/@xmath61-fe@xmath4o@xmath15 equilibrium line @xcite . axis lattice parameter of zn@xmath0fe@xmath1o@xmath2 films grown on mgo(001 ) in pure ar atmosphere ( squares ) and an ar / o@xmath4 mixture ( circles ) with zn content @xmath64 . the bulk @xmath65 axis lattice constant of fe@xmath15o@xmath2 is marked by the dashed horizontal line . ( b ) reciprocal space map around the ( 204 ) reflection of the mgo(001 ) substrate . the ( 408 ) reflection from the zn@xmath66fe@xmath67o@xmath2 film appears at the same in - plane scattering vector @xmath68 marked by the dashed vertical line . ] as shown by fig . [ fig : x - ray ] , the ( 004 ) reflection from zn@xmath0fe@xmath1o@xmath2 can be clearly observed together with satellites due to laue oscillations . these oscillations demonstrate that the thin films are coherently strained and have a small surface roughness . the rocking curves of the ( 004 ) or ( 008 ) reflections ( insets of fig . [ fig : x - ray ] ) show a full width at half maximum ( fwhm ) of @xmath69 . this demonstrates the excellent structural quality of the zn@xmath0fe@xmath1o@xmath2 thin film samples with very low mosaic spread , comparable to epitaxial fe@xmath15o@xmath2 films @xcite . in the @xmath59-@xmath60 scan , the ( 004 ) reflection from fe@xmath15o@xmath2 is positioned at a slightly larger angle than the ( 002 ) reflection from the mgo substrate ( see fig . [ fig : x - ray](a , b ) ) , indicating @xmath70 . with increasing zn concentration , the ( 004 ) reflection moves to lower angles and crosses the mgo(002 ) reflection at @xmath41 ( see fig . [ fig : x - ray](e , f ) ) . at @xmath71 , the position of the ( 004 ) reflection from the zn@xmath0fe@xmath1o@xmath2 film is shifted to an angle below the ( 002 ) reflection from mgo ( see fig . [ fig : x - ray](g , h ) ) . this corresponds to @xmath72 . the absolute value and variation of the @xmath65 axis lattice parameter of the zn@xmath0fe@xmath1o@xmath2 films with zn content @xmath64 are related to two effects , namely epitaxial coherency strain and the larger ionic radius of tetrahedrally coordinated zn@xmath36 of 0.6 compared to the radius of only 0.49 of fe@xmath23 on the tetrahedrally coordinated @xmath7 site @xcite . first , for @xmath39 the lattice mismatch of @xmath73 between fe@xmath15o@xmath2 film and mgo substrate leads to a tensile in - plane epitaxial coherency strain and , in turn , to a slight reduction of the @xmath65 axis lattice parameter of the fe@xmath15o@xmath2 film below the bulk value ( cf . [ fig : rsm](a ) ) . the fully coherent growth of the zn@xmath0fe@xmath1o@xmath2 thin films is demonstrated by reciprocal space maps around the ( 204 ) reflection of the mgo substrate ( see fig . [ fig : rsm](b ) ) . the ( 408 ) reflection of the film is located exactly at the same in - plane scattering vector @xmath68 as the ( 204 ) reflection of the substrate . this clearly demonstrates that the in - plane lattice constant of the film perfectly matches twice the one of the mgo substrate . second , for @xmath74 the unit cell volume of zn@xmath0fe@xmath1o@xmath2 is expected to increase about linearly with increasing @xmath64 due to the larger ionic radius of zn@xmath36 . however , since the zn@xmath0fe@xmath1o@xmath2 films grow coherently on the mgo(001 ) substrate without any relaxation up to the maximal thickness of 60 nm studied in our experiments , the in - plane lattice constant stays unchanged . only the out - of - plane lattice constant is found to vary with increasing @xmath64 . depending on whether the in - plane strain is tensile or compressive , a reduction or expansion of the unit cell in @xmath58 $ ] direction , respectively , is found ( fig . [ fig : rsm](a ) ) , resulting in a tetragonal distortion of the cubic lattice . the @xmath65 axis parameter increases from a value below ( @xmath39 , tensile strain ) to a value above the bulk value ( @xmath75 , compressive strain ) . the expansion of the out - of - plane lattice parameter with increasing @xmath64 has been reported also for ( 111)-oriented films grown on al@xmath4o@xmath15(0001 ) substrates @xcite . however , we note that those films are relaxed due to the very large lattice mismatch of 8% between film and substrate . therefore , both results can not be compared directly . an interesting result shown in fig . [ fig : rsm ] is the fact that the @xmath65 axis lattice parameters of the zn@xmath0fe@xmath1o@xmath2 films grown in pure ar atmosphere are larger than those of the films grown in an ar / o@xmath4 mixture and also show a slightly weaker increase with zn content in the range @xmath76 . this most likely is caused by the formation of fe vacancies when growing the films in finite oxygen partial pressure @xcite . the presence of fe vacancies can be viewed as an internal negative pressure effect leading to a reduced cell volume . the weaker increase of the @xmath65 axis parameter for the films grown in ar / o@xmath4 suggests a partial compensation of the fe vacancies by zn substitution . this is in agreement with the transport data discussed below . we note , however , that this result has to be considered with some care . the problem is that the close vicinity of film and substrate reflections in the @xmath77-@xmath60 scans does not allow an unambiguous derivation of the film @xmath65 axis parameter with small error bars . in summary , the structural analysis demonstrates that our zn@xmath0fe@xmath1o@xmath2(001 ) thin films are coherently strained and show a very small mosaic spread . zn substitution results in an increase of the unit cell volume , causing the in - plane epitaxial strain to change from tensile to compressive on increasing the zn content @xmath64 . growth in excess oxygen results in a reduction of the unit cell volume most likely due to the formation of fe vacancies . the latter most likely are partially compensated by additional zn substitution . for magnetite the electrical conduction above the verwey transition is believed to be determined by the hopping of the spin - down electrons between the mixed - valent fe@xmath10 and fe@xmath11 ions on the @xmath8 sublattice . from the hopping amplitude @xmath78ev an electron conduction bandwidth @xmath79ev is expected . however , experimental data can not be explained within a simple band theory . they can be better ascribed to some thermally activated motion of charge carriers . the reason is that the bare hopping amplitude and bandwidth is strongly reduced due to the translation of a polaronic lattice deformation associated with electron motion@xcite . if the reduced bandwidth is smaller than the energy of polarization phonons ( typically 70mev in magnetite ) , small polaron hopping is expected@xcite . we note that the hopping takes place between fe@xmath10 and fe@xmath11 ions on the @xmath8 sublattice . below the curie temperature , the spins of these ions are aligned parallel in the ideal case . however , as discussed above in reality there is some spin canting on the @xmath8 sublattice . this has to be taken into account by an appropriate spin correlation factor@xcite leading to a spin dependent part of the activation energy . since the spin alignment is improved by applying a magnetic field , this spin - dependent part of the activation energy can be reduced by the external field resulting in a large negative magnetoresistance as shown below . for magnetotransport measurements , the films were patterned into typically @xmath80 m wide and @xmath81 m long hall bars directed in the @xmath82 direction by photolithography and ar ion beam milling @xcite . the longitudinal resistivity @xmath83 has been measured as a function of temperature @xmath84 and applied magnetic field @xmath85 using a standard four - probe technique . the films with @xmath71 were found to be insulating and are not further discussed in the following . we note that measurements of the anomalous and the ordinary hall effect have been reported previously @xcite and are not discussed here . the @xmath86 curves of representative samples with @xmath87 are shown in fig . [ fig : rho ] . despite a similar shape of all @xmath86 curves , it is evident that the absolute magnitude of @xmath83 sensitively depends on the zn substitution @xmath64 and the growth atmosphere . for both samples sets grown in pure ar atmosphere and an ar / o@xmath4 ( 99:1 ) mixture , @xmath86 increases by more than two orders of magnitude on decreasing @xmath84 from 375k down to 90k . plotting @xmath88 on a reciprocal @xmath84 scale as shown in fig . [ fig : rho](b ) results in about linear curves over a wide temperature range except for the @xmath39 and @xmath89 samples grown in pure ar atmosphere . the kink in @xmath86 of the @xmath39 sample results from the verwey transition discussed below . [ fig : rho](b ) suggests that the longitudinal resistivity of the samples with @xmath90 follows a simple activated behavior @xmath91 with the boltzmann constant @xmath92j / k and activation energies @xmath93 ranging between 61 and 84mev . however , an unambiguous determination of the detailed transport process is difficult . as indicated by fig . [ fig : rho](c ) , the @xmath86 data also can be equally well fitted by small polaron hopping @xcite , @xmath94 , with the potential barrier for polaron hopping , @xmath95 , ranging between 64 and 95mev . actually , small polaron hopping yields a slightly better fit for the samples grown in pure ar atmosphere , whereas the simple activated transport better fits the transport data of the samples grown in an ar / o@xmath4 mixture . the derived activation energies are typical for zinc ferrites@xcite . we also note that a similar transport behavior with similar activation energies is reported for the mixed - valent manganites , where charge transport is determined by the hopping between mixed - valent mn@xmath23 and mn@xmath96 ions@xcite . as a function of temperature @xmath84 for epitaxial zn@xmath0fe@xmath1o@xmath2 thin films , grown in pure ar atmosphere ( solid lines ) or an ar / o@xmath4 ( 99:1 ) mixture ( dashed lines ) on mgo(001 ) substrates . in ( b ) and ( c ) @xmath83 and @xmath97 of the same samples are plotted versus a reciprocal temperature scale . fitting the linear parts of the curves in ( b ) and ( c ) by @xmath98 and @xmath99 , respectively , gives the listed activation energies @xmath93 and @xmath95 . ] we next discuss the observability of the _ _ verwey__@xcite transition in the measured @xmath86 curves . in bulk material , this metal - insulator transition , associated with a cubic to monoclinic structural phase transition , occurs at @xmath100k and in the classic picture is argued to arise from a charge ordering of the fe@xmath10 and fe@xmath11 on the @xmath8 sublattice in a process reminiscent to wigner crystallization @xcite . whereas at @xmath101 the extra electrons from fe@xmath10 ions can hop to neighboring fe@xmath11 ions on the corner sharing tetrahedral network on the @xmath8 sublattice , giving rise to electrical conduction , this process is frozen in at @xmath102 due to coulomb repulsion @xcite . however , more recently it became evident that also elastic and orbital interactions play a significant role . this leads to a renewed controversial discussion of the nature of the verwey transition @xcite . it is evident from fig . [ fig : rho ] that a pronounced change in the @xmath86 curves due to the verwey transition is observed only for the fe@xmath15o@xmath2 film ( @xmath39 ) grown in pure ar atmosphere . for this film there is no jump but a clear kink in the @xmath86 curve at a slightly reduced temperature of about 115k . for all other samples no clear signature of the verwey transition is observable this can be understood in a straightforward way . as discussed above , both substitution of zn and introducing fe vacancies by growing the films in finite oxygen partial pressure results in deviations from equal numbers of fe@xmath10 and fe@xmath11 on the @xmath8 sublattice . this can be viewed as disorder on the @xmath8 sublattice , tending to smear out the verwey transition . the fact that the observability of the verwey transition seems to sensitively depend on deviations from perfect stoichiometry already has been reported in literature @xcite and is further supported by our results . the other way round , we can argue that the absence of any clear signature of the verwey transition for the fe@xmath15o@xmath2 films substituted by zn or grown in finite oxygen partial pressure provides evidence for the presence of an unequal number of fe@xmath10 and fe@xmath11 ions on the @xmath8 sublattice . we next discuss the @xmath83 values of the zn@xmath0fe@xmath1o@xmath2 ( @xmath103 ) films grown in pure ar atmosphere . the strong increase of resistivity with decreasing @xmath84 can be understood in terms of a thermally activated hopping type transport mechanism of the itinerant electrons on the @xmath8 sublattice . for @xmath39 , the room temperature @xmath83 value ( @xmath104 m ) corresponds well with literature data for fe@xmath15o@xmath2 thin films@xcite or single crystals@xcite . in this context , we note that the presence of anti - phase boundaries ( apbs ) in fe@xmath15o@xmath2 thin film samples may have a significant influence on the measured longitudinal resistivity @xcite . however , this is not the case for our samples . first , our fe@xmath15o@xmath2 thin film grown in ar atmosphere has a room - temperature resistivity value close to that of fe@xmath15o@xmath2 single crystals ( @xmath105 m ) considered as apb - free @xcite . second , comparing the resistivity value to those of fe@xmath15o@xmath2 thin films with different apb density , we estimate that the volume fraction of the apb phase in our thin film is less than 20% @xcite . third , it is known that the formation and the density of apbs sensitively depends on the used substrates @xcite or the film thickness @xcite . however , in our study always the same substrate and film thickness is used . since also zn substitution is not expected to severely affect the apb density , we can safely assume that all samples grown in the same atmosphere have a similar apb density . with increasing @xmath64 , as discussed in the introduction , zn@xmath36 preferably occupies the tetrahedrally coordinated @xmath7 site in the inverse spinel structure@xcite . hence , the substitution of fe@xmath9 by zn@xmath36 reduces the amount of fe@xmath10 as @xmath106_{a } \left[{\rm fe}^{3+}_{1}{\rm fe}^{2+}_{1}\right]_{b } { \rm o}^{2-}_4 \rightarrow \hspace*{20 mm } \nonumber \\ \left[{\rm zn}^{2+}_x{\rm fe}^{3+}_{1-x}\right]_{a } \left[{\rm fe}^{3+}_{1+x}{\rm fe}^{2+}_{1-x}\right]_{b } { \rm o}^{2-}_4 . \label{eq : zn - substitution}\end{aligned}\ ] ] that is , the density of itinerant electrons on the @xmath8 sublattice is reduced , resulting in an increase of the resistivity with increasing @xmath64 . this is in perfect agreement with our observation and literature data @xcite . we note however , that the increase of resistivity is not only caused by a reduction of the carrier density but also by several other effects . first , deviations from the 1:1 balance of fe@xmath10 and fe@xmath11 on the @xmath8 sublattice causes disorder , tending to localize the itinerant charge carriers . second , zn substitution dilutes the @xmath7 sublattice , thereby weakening the antiferromagnetic exchange @xmath29 between the @xmath7 and @xmath8 sublattices . as discussed above , this results in spin canting on the @xmath8 sublattice , reducing the hopping amplitude . this amplitude is maximum for a parallel alignment of the local moments due to strong antiparallel hund s rule coupling of the spins of the itinerant electrons . both effects are expected to result in an increase of the activation energy for the hopping transport with increasing @xmath64 in agreement with our observations . the @xmath86 curves of the zn@xmath0fe@xmath1o@xmath2 ( @xmath107 ) films grown in an ar / o@xmath4 mixture show a similar overall temperature dependence , however , with resistivity values that are much higher than those measured for the films grown in pure ar atmosphere . furthermore , the resistivity values decrease with increasing @xmath64 in contrast to what is observed for the films grown in ar . we first discuss the origin of the much higher resistivity value ( @xmath108 m at 300k ) of the fe@xmath15o@xmath2 ( @xmath39 ) film grown in finite oxygen partial pressure compared to that ( @xmath109 m at 300k ) of the film grown in pure ar . since growth of magnetite in excess oxygen is known to create fe vacancies x@xmath110 @xcite , equivalent to two ( three ) missing electrons per fe@xmath36 ( fe@xmath23 ) vacancy , the requirement of charge neutrality shifts the 1:1 balance between fe@xmath10 and fe@xmath11 towards fe@xmath11 as @xmath111_{a } \left[{\rm fe}^{3+}_{1}{\rm fe}^{2+}_{1}\right]_b { \rm o}^{2-}_4 \rightarrow \hspace*{20 mm } \nonumber \\ \left[{\rm fe}^{3+}_1\right]_{a } \left[{\rm fe}^{3+}_{1 + 2\delta}{\rm fe}^{2+}_{1 - 3\delta}{\rm x}^0_\delta \right]_{b } { \rm o}^{2-}_4 . \label{eq : understoichiometry}\end{aligned}\ ] ] evidently , fe vacancies in the same way as zn substitution result in a reduction of fe@xmath10 . hence , the density of itinerant electrons on the @xmath8 sublattice decreases and , in turn , the resistivity increases with increasing @xmath112 . this is exactly what is observed in our experiments and reported in literature @xcite . actually , for zn@xmath0fe@xmath1o@xmath2 ( @xmath113 ) single crystals wang _ _ report a compositional correspondence @xmath114 regarding the measured resistivity data @xcite . again , with the same arguments given above for the films grown in pure ar , a reliable evaluation of the various mechanisms ( localization , spin canting ) responsible for the increase in resistivity is not possible . we also note that according to eq.([eq : understoichiometry ] ) growth in high oxygen partial pressure ( not the case in our experiments ) may lead to the formation of the cubic @xmath61-fe@xmath4o@xmath15 ( maghemite ) phase in the extreme case of @xmath115@xcite . zn substitution in our magnetite films grown in an ar / o@xmath4 mixture results in a decrease of resistivity in contrast to what is observed for the films grown in pure ar atmosphere . on first sight , this is astonishing and seems to be in conflict with the above discussion , since zn@xmath36 again preferably occupies the tetrahedrally coordinated @xmath7 lattice sites@xcite , substituting fe@xmath23 and thereby reducing carrier density . that is , one would expect an increase of resistivity with increasing @xmath64 . however , since zn substitution also results in an increase of the unit cell volume due to the larger ionic radius of zn@xmath36 compared to fe@xmath23 , it is expected that zn substitution removes part of the fe vacancies . this is intuitive since now fe can be more easily incorporated into the expanded lattice . then , it is expected that zn@xmath36 does not only remove an electron by substituting fe@xmath9 on an @xmath7 site , but also adds two / three electrons by removing a vacancy on a fe@xmath10/fe@xmath11 site . that is , by zn substitution also x@xmath110 is replaced by fe@xmath36 or fe@xmath23 . evidently , in total this leads to an effective increase of the carrier density with increasing @xmath64 in agreement with our experimental data in the range up to @xmath41 . the partial removal of the fe vacancies by zn substitution is in agreement with the x - ray data of fig . [ fig : rsm](a ) . in summary , the @xmath86 curves of the zn@xmath0fe@xmath1o@xmath2 films grown under different oxygen partial pressure show a similar shape , originating from a hopping type transport mechanism , but strongly differing absolute resistivity values . these differences can be consistently explained by the change of the itinerant charge carrier density on the @xmath8 sublattice and their hopping amplitude by either zn substitution or the generation of fe vacancies due to growth at finite oxygen partial pressure . however , applying both zn substitution and excess oxygen results in an increase of the carrier density as zn substitution is removing part of the fe vacancies generated by excess oxygen . it would be interesting to directly check the change of the carrier density by measurements of the hall effect . however , we can not unambiguously separate the small ordinary from the large anomalous hall contribution and any attempt to do so would result in large errors of the derived carrier density @xcite . we note that the observed dependence of the resistivities on the zn substitution levels @xmath64 can not be simply explained by the assumption of different apb densities in the different thin film samples . it is unlikely that with increasing @xmath64 the apb density increases for samples of the same thickness when grown in ar atmosphere while it decreases for those prepared in the ar / o@xmath4 mixture . moreover , the samples show a universal scaling relation of the anomalous hall conductivity with the longitudinal conductivity indicating a negligibly small apb resistivity as already pointed out earlier @xcite . in this subsection we address the magnetoresistance ( mr ) of the zn@xmath0fe@xmath1o@xmath2 films . the mr effect , @xmath116 / \rho_{xx}(0)$ ] is shown in fig . [ fig : mreffect](a ) at several temperatures for the @xmath89 samples grown in pure ar . as shown in fig . [ fig : mreffect](b ) , very similar @xmath117 curves with similar absolute values of @xmath118 at room temperature are found for the other samples , although the absolute values of their resistivities vary by more than two orders of magnitude ( cf . [ fig : rho ] ) . this interesting observation can be consistently explained in the framework of thermally activated hopping of the itinerant electrons on the @xmath8 sublattice . due to the strong on - site hund s rule coupling , the spins of the itinerant spin - down electrons are coupled antiparallel to the localized spins formed by the spin - up electrons . therefore , the activation energy for the hopping process is given by the sum of a spin independent energy ( @xmath93 or @xmath95 ) and a spin dependent contribution @xmath119 . the latter depends on the spin canting of the local moments on the @xmath8 sublattice and disappears for a perfect parallel alignment . since in the presence of a finite canting of the local moments an applied magnetic field tends to improve the alignment , a magnetic field dependent total activation energy @xmath120 is obtained . then , no matter whether the transport is by thermally activated hopping , @xmath121 $ ] , or small polaron hopping , @xmath122 $ ] , the mr effect is obtained as @xmath123 with @xmath124 . evidently , with the assumptions made , @xmath117 is independent of the magnitude of the prefactors ( @xmath125 or @xmath126 ) and the activation energies ( @xmath93 or @xmath95 ) , which may strongly vary for samples with different @xmath64 and grown in different atmosphere . moreover , similar @xmath117 curves with comparable absolute values are obtained , if the magnetic field induced change @xmath127 of the spin - dependent part of the activation energy is about the same for all samples . since the transport in zn@xmath0fe@xmath1o@xmath2 by hopping of the @xmath26 spin - down electrons between mixed - valent fe@xmath10/fe@xmath11 ions is equivalent to the hopping of the @xmath38 electrons between mixed - valent mn@xmath23/mn@xmath96 ions in the doped manganites , we can adopt models developed for the mr effect in the perovskite manganites @xcite . according to these models the magnetic field dependent change of the activation energy for hopping between two lattice sites @xmath128 and @xmath129 can be expressed as @xmath130 where @xmath131 and @xmath132 are the local moments on the lattice sites @xmath128 and @xmath129 , @xmath133 and @xmath134 denote the sample average at zero and finite magnetic fields , respectively , and @xmath135 is a constant that may depend on temperature . the local moments @xmath131 and @xmath132 can be associated either with the localized spins of the individual fe ions or the moments of small spin clusters . note that these spin clusters also can be considered as spin polarons , that is , itinerant electrons dressed by a small cloud of parallel aligned localized spins . evidently , @xmath136 is independent of @xmath85 , if the local moments are already perfectly aligned in zero magnetic field ( ideal ferromagnet ) . in this case , no mr effect is expected . however , if there is a finite canting of the neighboring moments , applying a magnetic field results in a reduction of this spin canting and thereby in @xmath137 . according to eq.([eq : mreffect1 ] ) this results in a finite negative mr effect with a magnitude determined by the value of @xmath138 . the magnetic field dependence of @xmath139 can be estimated by keeping in mind that in the ferromagnetic state both the molecular field and the applied magnetic field support a parallel alignment of the local moments . since the molecular field is usually much larger than the applied field , the deviations of the directions of the local moments from the average direction are small . in this case @xmath127 is found to follow the brillouin function @xmath140 giving@xcite @xmath141 with the vacuum permeability @xmath142vs / am . here , @xmath143 is the average value of the local moments and @xmath144 a constant , which may show a weak temperature dependence . assuming @xmath145 , what certainly is a reasonable assumption at room temperature , we can approximate eq.([eq : mreffect1 ] ) by @xmath146 this shows that the magnetic field dependence of the mr effect measured at constant temperature should follow the brillouin function . as shown in fig . [ fig : mreffect](b ) , this is indeed the case . the data of all samples can be well fitted by eq.([eq : mreffect4 ] ) . fitting the data gives the @xmath143 values of the local moments . they range between 70 to @xmath147 and 60 to @xmath148 , corresponding to about 18 to 30 and 15 to 18 fe@xmath36 ions , for the samples grown in pure ar and an ar / o@xmath4 mixture , respectively . here , @xmath149 is bohr s magneton . evidently , the derived moments are larger than the localized spins of the individual fe ions . they can be considered as the moments of small ferromagnetic clusters with perfectly parallel spins , in which the electrons can move freely , but with small misalignment between neighboring clusters . since there are eight fe@xmath36 ions per unit cell , their size ranges between about two and four unit cells . we also note that eq.([eq : mreffect4 ] ) does not only describe the magnetic field dependence of the mr effect but also nicely explains the decrease of the mr effect with increasing temperature . in summary , the mr effect of the zn@xmath0fe@xmath1o@xmath2 films in the same way as the @xmath86 data can be well described within a model based on thermally activated hopping of itinerant charge carriers or small polarons between the mixed - valent fe@xmath10/fe@xmath11 ions on the @xmath8 sublattice . the activation energy contains a magnetic field dependent part , which depends on the misalignment of the neighboring spin moments . within this model , @xmath150 is expected in good agreement with the experimental data . in particular , the similar @xmath117 curves for samples with strongly different @xmath83 values and the decrease of the mr effect with increasing temperature are nicely reproduced . we have seen that zn substitution and the creation of fe vacancies in magnetite results in changes of the carrier density and hopping amplitudes of the itinerant charge carriers on the @xmath8 sublattice . following our discussion in the introduction , this is expected to have significant influence on the magnetic properties . therefore , we also systematically analyzed the magnetic properties of the zn@xmath0fe@xmath1o@xmath2 films . the magnetic characterization was performed with the unpatterned films using squid magnetometry with magnetic fields @xmath151 up to 7 t applied in the film plane . at room temperature , the @xmath152 loops show ferromagnetic behavior for all samples ( fig . [ fig : mh ] ) . however , the measured saturation magnetization @xmath153 as well as the remanent magnetization @xmath154 and coercive field was found to strongly depend on both the zn substitution level @xmath64 and the growth atmosphere . versus magnetic field @xmath85 applied in the film plane for zn@xmath0fe@xmath1o@xmath2 thin films grown in ( a ) pure ar atmosphere and ( b ) in an ar / o@xmath4 ( 99:1 ) mixture . the insets show the saturation magnetization @xmath153 ( full symbols ) and the remanence @xmath154 ( open symbols ) in units of @xmath149/f.u . as a function of the zn substitution level @xmath64 . the lines are guides to the eye . ] we start our discussion with the stoichiometric fe@xmath15o@xmath2 film grown in pure ar atmosphere , serving as a reference . figure [ fig : mh](a ) shows that for this film the highest values for the room temperature saturation magnetization ( @xmath155/f.u . ) and remanence ( @xmath156/f.u . ) are obtained . we note that we do not obtain the theoretically expected value of @xmath157/f.u . this is typical for thin film samples due to the presence of apbs @xcite . however , the value measured for our film is among the highest reported in literature and corresponds well with the volume fraction of the apb phase of 20% estimated above . we next discuss the evolution of the saturation magnetization of the zn@xmath0fe@xmath1o@xmath2 films grown in pure ar atmosphere with increasing @xmath64 . as shown in the inset of fig . [ fig : mh](a ) , both @xmath153 and @xmath154 were found to decrease with increasing @xmath64 . recalling eq.([eq : zn - substitution ] ) , the substitution of fe@xmath9 ( @xmath35 ) by zn@xmath36 ( @xmath37 ) results in a decrease of the magnetization on the @xmath7 sublattice . that is , since the compensating magnetization of the @xmath7 sublattice is removed , one would expect an increase of the total magnetization of the ferrimagnet . however , zn substitution also weakens @xmath29 by diluting the @xmath7 site moments . furthermore , it converts fe@xmath10 ( @xmath158 ) into fe@xmath11 on the @xmath8 sublattice , resulting in a reduction of the itinerant charge carrier density . this weakens the ferromagnetic double exchange , competing with antiferromagnetic superexchange interaction on the @xmath8 sublattice . as discussed in the introduction , taken together this leads to an increase of the spin canting on the @xmath8 sublattice with increasing @xmath64 ( cf . [ fig : spinel ] ) , explaining the observed reduction of the total magnetization of the ferrimagnetic zn@xmath0fe@xmath1o@xmath2 films . we note that recently takaobushi _ et al . _ reported a monotonic increase of the saturation magnetization from about @xmath159/f.u . to @xmath160/f.u . at 10k in the range @xmath161 and explained this observation by the assumption that the @xmath7 sublattice magnetization is reduced by zn substitution @xcite . comparison of the low temperature saturation magnetization of the @xmath39 films ( @xmath162/f.u . when grown in ar in our case , @xmath159/f.u . in ref . @xcite ) clearly shows that the samples can not be directly compared . the reason is that the films in ref . @xcite have been grown at finite oxygen partial pressure of @xmath163 mbar . in this case , the effect of fe vacancies has to be taken into consideration as follows . plotted versus temperature @xmath84 for a zn@xmath164fe@xmath165o@xmath2 film grown in an ar / o@xmath4 ( 99:1 ) mixture . the data were taken after field cooling at 7 t on increasing the temperature at zero applied magnetic field . the inset schematically shows the hump in the @xmath166 curve as the result of two opposite sublattice magnetizations with different @xmath84 dependence . ] the magnetic properties of the zn@xmath0fe@xmath1o@xmath2 films grown in an ar / o@xmath4 ( 99:1 ) mixture are shown in fig . [ fig : mh](b ) . there are pronounced differences to the films grown in pure ar atmosphere . first , for @xmath39 both @xmath153 and @xmath154 are much lower than for the films grown in pure ar . second , @xmath153 and @xmath154 is found to increase with increasing @xmath64 for @xmath167 and then to decrease again for @xmath71 as shown in the inset of fig . [ fig : mh](b ) . the coercive field decreases from about 55mt to 8mt on increasing @xmath64 from 0 to 0.9 . the @xmath168 dependence can be straightforwardly explained by the presence of fe vacancies in films grown in finite oxygen partial pressure and their removal by zn substitution . for @xmath39 , the fe vacancies shift the fe@xmath10/fe@xmath11 1:1 balance towards fe@xmath11 , removing itinerant charge carriers and weakening the ferromagnetic double exchange at the expense of antiferromagnetic superexchange on the @xmath8 sublattice . this results in spin canting on the @xmath8 sublattice , giving rise to a reduced saturation magnetization . on zn substitution , part of the fe vacancies are removed . this shifts the fe@xmath10/fe@xmath11 ratio back towards a 1:1 balance . the related increase of the itinerant charge carrier density and de interaction on the @xmath8 sublattice reduces the spin canting with increasing @xmath64 . this explains the increase of the room temperature saturation magnetization in the range @xmath169 from @xmath170/f.u . up to @xmath171/f.u . at @xmath41 . at 10k , @xmath153 increases from @xmath172/f.u . at @xmath39 up to @xmath173/f.u . at @xmath41 . for even larger @xmath64 , the fe@xmath9 moments on the @xmath7 sublattice are strongly diluted , resulting in a strong weakening of the antiferromagnetic superexchange @xmath29 between the @xmath7 and @xmath8 sublattices . furthermore , zn substitution may overcompensate the fe vacancies again shifting the fe@xmath10/fe@xmath11 balance towards fe@xmath11 . taken together , this explains the decrease of @xmath153 above the critical value of @xmath174 , which of course depends on the oxygen partial pressure during growth and the corresponding amount of fe vacancies . the observed increase of @xmath153 with increasing @xmath64 for zn@xmath0fe@xmath1o@xmath2 films grown in an ar / o@xmath4 mixture is consistent with the result of takaobushi _ _ et al.__@xcite , who reported an increase of @xmath175 from @xmath159/f.u . to @xmath160/f.u . on increasing @xmath64 from 0 to even 0.9 . however , our systematic study strongly suggests that this increase is most likely caused by the removal of fe vacancies rather than due to the removal of fe@xmath9 moments on the @xmath7 sites as argued in ref . @xcite . figure [ fig : mt ] shows the temperature dependence of the remanent magnetization @xmath154 for a zn@xmath164fe@xmath165o@xmath2 film grown in an ar / o@xmath4 mixture . in agreement with the @xmath86 data of fig . [ fig : rho ] , there is no indication for the _ verwey _ transition in @xmath176 around 120k . this is the case for all samples , except for the fe@xmath15o@xmath2 film grown in pure ar . this observation is not only consistent with our transport data but also with magnetization data from other groups , indicating a smearing or suppression of the verwey transition by an amount of zn substitution as small as @xmath177 @xcite . as shown in fig . [ fig : mt ] , the measured @xmath176 curve first decreases with increasing @xmath84 , goes through a minimum at @xmath178k , then increases up to @xmath179k , before it decreases again towards room temperature . such a behavior is often observed in ferrimagnets and can be explained on the basis of nel s two sublattice model @xcite . within this model the origin of the hump at 275k is a different @xmath84 dependence of the two sublattice magnetizations @xmath180 and @xmath181 due to different effective molecular fields on the @xmath7 and @xmath8 sites . in our case the effective molecular field on the @xmath7 sublattice is expected to be smaller due to the partial substitution of fe@xmath9 by nonmagnetic zn@xmath36 . as a result , the combined magnetization @xmath182 should show an upward hump in agreement with the experimental result . note that the observation of a hump , which is directed upwards , provides direct experimental evidence for @xmath7 site substitution of zn@xmath36 in the inverse spinel structure @xcite . for the @xmath183 sample , no hump could be observed , most likely due to a too small imbalance in the effective molecular fields at the smaller substitution level . plotted versus the longitudinal conductivity @xmath184 for epitaxial zn@xmath0fe@xmath1o@xmath2 thin films with different zn substitution @xmath87 grown in pure ar atmosphere ( squares ) or an ar / o@xmath4 ( 99:1 ) mixture ( circles ) . ] in conclusion , the magnetic properties of the zn@xmath0fe@xmath1o@xmath2 films grown under different oxygen partial pressure strongly depend on whether the films were grown in pure ar atmosphere or at finite oxygen partial pressure . fully consistent with the transport data these differences can be explained by the following effects : ( i ) for films grown in pure ar , zn substitution on the @xmath7 site removes fe@xmath9 moments from the @xmath7 sublattice and due to charge neutrality itinerant electrons from the @xmath8 sublattice . the resulting weakening of both the antiferromagnetic @xmath7-o-@xmath8 se and the ferromagnetic de on the @xmath8 sublattice results in an increasing spin canting on the @xmath8 sublattice with increasing @xmath64 . this causes a reduction of the saturation magnetization with increasing @xmath64 . ( ii ) for films grown in oxygen partial pressure , the magnetic properties are strongly influenced by the presence of fe vacancies . they reduce the density of itinerant electrons on the @xmath8 sublattice thereby weakening the ferromagnetic de at the expense of antiferromagnetic se . this causes spin canting on the @xmath8 sublattice and , in turn , a reduced saturation magnetization . the effect of additional zn substitution is the partial removal of the fe vacancies . this effectively increases the density of itinerant electrons on the @xmath8 sublattice thereby strengthening the ferromagnetic de . the result is an increase of the saturation magnetization with increasing @xmath64 . again , the observed dependence of the saturation magnetization on the zn substitution levels @xmath64 can not be simply explained by different apb densities in the different thin film samples . it is very unlikely that the apb density increases with increasing @xmath64 for samples of the same thickness when grown in ar atmosphere while for those prepared in the ar / o@xmath4 mixture it first decreases for @xmath185 and then increases again for @xmath186 . we finally link the magnetic properties to the transport data by plotting the saturation magnetization @xmath153 versus the electrical conductivity @xmath187@xcite for films with various @xmath64 and grown in ar or ar / o@xmath4 ( fig . [ fig : spezialplot ] ) . evidently , there is a clear correlation between @xmath153 and @xmath184 . samples with high @xmath153 have large @xmath184 and vice versa . again , this correlation can be straightforwardly explained within the scenario presented above . charge transport is dominated by the hopping of the itinerant @xmath26 electrons between the mixed - valent fe@xmath10/fe@xmath11 ions on the @xmath8 sublattice . due to the strong on - site hund s rule coupling the spin of the hopping spin - down electron is aligned anti - parallel to the local moment of the spin - up electrons ( cf . [ fig : spinel ] ) . therefore , the hopping amplitude is significantly suppressed if the moments of neighboring @xmath8 sites are not parallel . that is , any spin canting on the @xmath8 sublattice reducing the saturation magnetization also results in a reduction of the electrical conductivity . we have grown epitaxial zn@xmath0fe@xmath1o@xmath2 thin films ( @xmath3 ) with excellent structural properties both in pure ar atmosphere and an ar / o@xmath4 mixture using laser molecular beam epitaxy . we show that the electrical conductivity and the saturation magnetization can be tuned over a wide range ( @xmath6/f.u . and @xmath5 at room temperature ) by zn substitution and/or finite oxygen partial pressure during growth . this demonstrates that zn@xmath0fe@xmath1o@xmath2 can be used as a material with tailored electrical and magnetic properties . our comprehensive study shows that electrical conduction , magnetotransport , and the magnetic properties of zn@xmath0fe@xmath1o@xmath2 are dominated by the density and thermally activated hopping of the itinerant @xmath26 electrons between the mixed - valent fe@xmath10/fe@xmath11 ions , providing electrical conduction and mediating a ferromagnetic de interaction on the @xmath8 sublattice . this scenario is confirmed by the observation that the saturation magnetization scales with the longitudinal conductivity and the field dependence of the mr effect . our combined systematic analysis of the magnetotransport and magnetic properties also provides a clear picture of the effect of zn substitution and growth in finite oxygen partial pressure . first , zn substitution removes both fe@xmath9 moments from the @xmath7 sublattice and itinerant charge carriers from the @xmath8 sublattice . second , growth in finite oxygen partial pressure generates fe vacancies , also removing itinerant charge carriers from the @xmath8 sublattice . hence , in both cases a reduction of the charge carrier density and a weakening of the ferromagnetic de on the @xmath8 sublattice is obtained . this results in an increase of resistivity and a reduction of the saturation magnetization due to spin canting on the @xmath8 sublattice . third , applying both zn substitution and growth in oxygen at the same time does not result in an additive effect . in contrast , a compensation effect is observed , since zn substitution removes part of the fe vacancies . to sum up , one can say that the high curie temperature together with the electrical and magnetic properties that can be tuned over a wide range make zn@xmath0fe@xmath1o@xmath2 a promising functional material for spintronic devices . the combination of zn@xmath0fe@xmath1o@xmath2 with semiconductor materials such as zno in multi - functional heterostructures @xcite will be particularly appealing . we thank andreas erb for the preparation of polycrystalline target materials for the laser - mbe process and thomas brenninger for continuous technical support . financial support by the german science foundation within the priority programs 1157 and 1285 ( project nos . gr 1132/13 & 14 ) and the german excellence initiative via the _ nanosystems initiative munich ( nim ) _ is gratefully acknowledged . d.v . and m.s.r.r . thank the daad for financial support . j. takaobushi , m. ishikawa , s. ueda , e. ikenaga , j .- kim , m. kobata , y. takeda , y. saitoh , m. yabashi , y. nishino , d. miwa , k. tamasaku , t. ishikawa , i. satoh , h. tanaka , k. kobayashi , and t. kawai , phys . b * 76 * , 205108 ( 2007 ) . d. t. margulies , f. t. parker , f. e. spada , r. s. goldman , j. li , r. sinclair , and a. e. berkowitz , phys . b * 53 * , 9175 ( 1996 ) ; see also d. t. margulies , f. t. parker , m. l. rudee , f. e. spada , j. n. chapman , p. r. aitchison , and a. e. berkowitz , phys . lett . * 79 * , 5162 ( 1997 ) . r. gross , j. klein , b. wiedenhorst , c. hfener , u. schoop , j. b. philipp , m. schonecke , f. herbstritt , l. alff , yafeng lu , a. marx , s. schymon , s. thienhaus , and w. mader , proc . spie * 4058 * , 278 ( 2000 ) . fleet , acta cryst . b * 37 * , 917 ( 1981 ) . j. garca , g. subas , m. g. proietti , j. blasco , h. renevier , j. l. hodeau , and y. joly , phys . b * 63 * , 054110 ( 2001 ) ; see also j. garca , g. subas , m. g. proietti , h. renevier , y. joly , j. l. hodeau , j. blasco , m. c. sanchez , and j. f.berar , phys * 85 * , 578 ( 2000 ) .

we present experimental results on slow flow properties of a granular assembly confined in a vertical column and driven upwards at a constant velocity v. for monodisperse assemblies this study evidences at low velocities ( ) a stiffening behaviour i.e. the stress necessary to obtain a steady sate velocity increases roughly logarithmically with velocity . on the other hand , at very low driving velocity ( ) , we evidence a discontinuous and hysteretic transition to a stick - slip regime characterized by a strong divergence of the maximal blockage force when the velocity goes to zero . we show that all this phenomenology is strongly influenced by surrounding humidity . we also present a tentative to establish a link between the granular rheology and the solid friction forces between the wall and the grains . we base our discussions on a simple theoretical model and independent grain / wall tribology measurements . we also use finite elements numerical simulations to confront experimental results to isotropic elasticity . a second system made of polydisperse assemblies of glass beads is investigated . we emphasize the onset of a new dynamical behavior , i.e. the large distribution of blockage forces evidenced in the stick - slip regime .

cond-mat0302581

granular flows are presently at the focus of many attentions @xcite . the classical model for granular media stability was proposed at the end of the xviiith century by ch . a. de coulomb who revealed a strong analogy between the failure properties of a granular assembly and the phenomenology of solid on solid friction . modern developments have elaborated sophisticated empirical approaches around this fundamental idea @xcite but so far there is no deep physical understanding nor rigorous derivation describing the passage from the granular level description to a set of evolution equations involving macroscopic quantities such as stress , strain or packing fraction . furthermore , there is an additional difficulty to apprehend complex behaviors such as aging under stress or to account for the influence of external parameters such as surrounding humidity which effects are often noticed in practice . note that similar questions are still under active consideration in the field of tribology @xcite , but in the case of granular assemblies , a supplementary difficulty lies in the fragile character of granular structures which can be easily modified under the action of external constraints . dynamical behavior of slowly driven granular materials was investigated by different groups both in compression and/or in shearing experiments @xcite . here we present an experimental situation of the same type , but in a quite different geometry . we investigate the rheology of a granular assembly confined in a cylindrical column and pushed vertically from the bottom . the resistance to vertical motion as well as the blocking / unblocking transitions reveals a phenomenology possibly shared by many confined granular assemblies . note that this column configuration may help to understand several practical situations like pipe flows @xcite , compaction under stress or dense granular paste extrusion . a previous investigation of the same display was made in 2d @xcite and also in 3d @xcite as a preliminary report , and more recently see refs @xcite . those contributions have shown a rich phenomenology partly sorted by the solid friction properties of the grains and the boundaries . the importance of surrounding humidity was also evidenced @xcite as it would strongly influence the rheology of the granular column . in this report , we push further the investigation as we change the column material , the beads characteristics and size dispersity , under various humidity conditions . we also propose to compare our data with the outcome of a simple numerical model of isotropic elasticity . in the classical situation of newtonian fluid pushed by a piston ( typically in a syringe ) , one would obtain a relation between pressure and flow rate , which is characterized by a fluid constitutive parameter : the viscosity . for an isotropic elastic medium the resistance to pushing would depend on the material poisson s ratio and wall frictional properties @xcite ( we will detail this question further in the text ) . for a granular material , the situation is a priori more complex since the piling structure can be modified so as to adapt to the external constraints . friction at the walls may also create internal granular recirculation flows @xcite . therefore it is an important but difficult matter to clarify the rheology of this system by sorting the respective influence of grain - boundary friction versus bulk structural changes . along this line we started with a simple situation of weakly frictional steel beads with frictional boundaries . a second situation is studied where the granular material is made of rugous polydisperse glass beads . most experiments are performed on dry , non cohesive and monodisperse steel beads of diameter @xmath2 mm piled into a vertical duralumin or brass cylinder of diameter @xmath3 mm . we also perform a series of experiments on a miscellany of glass beads of three diameters ( 1.5 mm , 2 mm and 3 mm ) with equal volume of each kind , in a pmma cylinder . the column is closed at the bottom by a movable brass piston avoiding contact with the column ( diameter mismatch is @xmath4 mm ) . a force probe of stiffness @xmath5 @xmath6 is located under the piston and is pushed at a constant driving velocity @xmath7 ( between @xmath8 @xmath9 and @xmath10 @xmath11 ) via a stepping motor ( see fig . [ fig1 ] ) . the resistance force @xmath12 encountered by the piston is measured as a function of time . we also monitor the relative humidity ( @xmath13 ) and the surrounding temperature . we work in the range @xmath14 , as well as in dry air ( @xmath15 ) and humid air ( @xmath16 ) . actually , except for the dry and humid situation , we do not regulate this last parameter ( @xmath13 ) but we record its values close to the experimental set - up . we obtain dry air by having a weak air flux ( @xmath17 ) flow into a cylinder filled with silica . we obtain humid air by making a weak air flux ( @xmath17 ) bubble through water . then , for the duration of an experiment , weak fluxes of humid or dry air are set to flow through the column from top to bottom , maintaining a constant and homogeneous level of humidity . note that a stationary relative humidity is reached in a few minutes . we start our experiments when this stationary level is attained . we verified that the weak air flux does not perturb the system : we notice no evolution of the force when the flux is stopped as long as relative humidity remains unchanged . temperature is kept at @xmath18c . we actually find no correlation between the force fluctuations and the temperature variations in this range . note that the brass and duralumin cylinders have very different surface properties ( fig . [ fig2 ] ) . the duralumin cylinder is rough cast , and its roughness is @xmath19 nm . the brass cylinder was machine - turned , and has mean roughness @xmath20 m with undulations of wavelength @xmath21 m and depth @xmath22 m . by @xmath23 , with @xmath24 steps ; vertical resolution is 100 nm . the horizontal axis on the graphes is the symmetry axis of the columns.,width=264 ] in order to compare precisely rheological properties of granular assemblies to solid friction properties , we built a special device ( called the slider , see fig . [ fig1 ] inset ) designed to study the tribology of bead / wall contact . this device is set to apply a constant normal load ( @xmath25 @xmath26 ) on three steel beads sliding vertically on the cylinder s wall . then , the dynamical evolution of the resistance force encountered by the piston pushing a granular material can be compared to the slider s friction resistance driven in the same conditions . two granular systems will be studied . first a model assembly of monodisperse steel spheres and second a polydisperse assembly of rough glass beads . the largest part of this report is devoted to the monodisperse assembly . we now report on the simplest situation i.e. monodisperse low frictional steel beads in a duralumin or brass cylinder . we observe two distinct dynamical stationary regimes ( fig . [ fig3 ] ) : for high driving velocities , the motion is characterized by a steady - sliding and a constant pushing force ; for low velocities , the system undergoes a dynamic instability characterized by a stick - slip motion . ( 380 g ) of steel beads in the duralumin cylinder vs. displacement @xmath27 of the translation stage . a : in the stick - slip regime ( @xmath28 ) . b : in the steady - sliding regime ( @xmath29).,width=264 ] we first notice that as far as averaged packing fraction and pushing force are concerned , whatever the initial state of preparation is , the same stationary regime is reached ( fig . [ fig4 ] ) . when a dense packing ( @xmath30 ) is prepared by rain - filling , the force in the steady sliding regime increases up to a maximum , then decreases slowly for a piston displacement of about 3 mm . at this point where a stationary value @xmath31 of the force is attained , the average packing fraction @xmath32 is @xmath33 . in the case of an initially loose packing ( @xmath34 ) prepared by using an inner cylinder slowly removed after filling , the force in the steady sliding regime increases monotonically for a 3 mm displacement before the same stationary value @xmath31 of force is reached . this state is also characterized by a @xmath35 packing fraction . for all preparations , we always get a stationary regime characterized by the same pushing force @xmath31 , which depends on driving velocity @xmath7 , relative humidity @xmath13 and packing s height @xmath36 , as well as the same packing fraction ( @xmath37 ) , independent of @xmath7 , @xmath13 and @xmath36 . the stick - slip regime displays a similar phenomenology : the piling reaches a stationary packing fraction of @xmath37 ; in the transitory regime , the pushing force for dense packing displays a maximum and for a loose packing shows a monotonous increase . in section [ iii.a.1 ] we propose an explanation for the phenomenology of the transitory regimes . ; b ) @xmath29.,width=264 ] in the following , we present the phenomenology observed in the steady - sliding and the stick - slip regime , when the stationary state is attained . for each regime , we first focus on the mechanical properties and propose a model accounting for friction at the walls and bulk properties ; we then study and analyse the rheological properties . in a third section , we point out ambiguity of dependence on relative humidity . we then analyse our results in a standard solid on solid friction framework ( the dieterich - ruina model ) . finally , we study the transition from steady - sliding to stick - slip . for a given height of beads in the column , at given driving velocity @xmath7 and relative humidity @xmath13 , the force in the steady sliding regime is constant at about @xmath38 ( fig . [ fig5]a ) , and its distribution around mean value @xmath31 is nearly gaussian . m during 5 mm of sliding in the steady - sliding regime of fig . [ fig3 ] at @xmath29 . b : distribution of the maximum and minimum resistance forces @xmath39 and @xmath40 in the stick - slip regime of fig . [ fig3 ] at @xmath28 during 5 mm of sliding.,width=264 ] next , we study the mean pushing force @xmath31 behavior as a function of the packing height , the driving velocity and the relative humidity . for a vertically pushed granular assembly , the driving force exerted by the piston is screened due to friction with the walls . to evaluate this effect , the mean resistance force @xmath31 in the steady - sliding regime is measured as a function of the packing height ( see fig . [ fig6 ] ) . the resistance force @xmath31 increases very rapidly with the packing s height @xmath36 . following the standard janssen screening picture , this strong resistance to motion is due to the leaning of the granular material on the walls created by the horizontal redirection of vertical stress in association with solid friction at the sidewalls . it means that we may relate horizontal and vertical stresses averaged on a slice at height @xmath41 by an effective relation : _ zz(z)=k_rr(z)where @xmath42 is called janssen s redirection constant . at the walls , we suppose a sliding of the granular material at a velocity @xmath43 ( the driving velocity ) ; the shearing stress is then _ rz(z)=_d(v_0)_rr(z)where @xmath44 is the dynamic coefficient of friction between the beads and the cylinder s wall at a velocity @xmath43 . the constant @xmath45 is introduced in order to differentiate between pushing and pulling experiments . when @xmath46 the granular material is moving upwards and friction is fully mobilized downwards ( our pushing experiment ) and when @xmath47 the granular is moving downwards and friction is fully mobilized upwards . the force @xmath48 exerted by the grains on the piston can be derived from equilibrium equations for all slices , thus we obtain : @xmath49 where @xmath50 is the mass density of the granular material , @xmath51 is the cylinder radius and @xmath52 the acceleration of gravity . the length @xmath53 is the effective screening length . it is easily seen from ( [ 1 ] ) that when @xmath54 , any slight change in @xmath55 or @xmath42 is exponentially amplified with a drastic influence on the pushing force @xmath31 . in the steady state regime , the experimental data obtained for a given pushing velocity @xmath7 can be fitted by relation ( [ 1 ] ) by adjusting only one parameter i.e. p_+1=k_d(v_0)where @xmath56 is the dynamic coefficient of friction at velocity @xmath57.we see on fig . [ fig6 ] that the data are well fitted by eq . ( [ 1 ] ) , and , for a relative humidity @xmath58 , for steel beads in the duralumin cylinder , we obtain @xmath59 at @xmath60 and @xmath61 at @xmath62 . we are now able to propose an interpretation of the features observed for different initial packing fractions within janssen s framework . since janssen s coefficient @xmath42 is an increasing function of the packing fraction , as it was shown previously @xcite , the resistance force of an initially dense packing displays a maximum , as a dense packing leans more efficiently on walls than a loose one ; the pushing force then decreases while the packing loosens ( i.e. while @xmath42 decreases ) . on the other hand , for an initially loose packing the resisting force in the transitory regime will increase while the packing densifies ( i.e. while @xmath42 increases).for an initially dense packing , we extract a parameter @xmath63 from a fit of the pushing force maximum value in the transitory regime with formula ( [ 1 ] ) : we obtain @xmath64 at @xmath65 and @xmath64 at @xmath66 ; as this experiment was not repeated sufficiently , we do not have uncertainties on these values . nevertheless , we notice that the @xmath63 value is roughly 25% higher than the @xmath67 stationary value , whatever the velocity is . these results are consistent with a @xmath42 dependence on compacity @xmath32 derived in @xcite for an assembly of monodisperse glass beads , i.e. @xmath68 ; for initial compacity @xmath69 and steady - state compacity @xmath35 , this empirical formula actually leads to @xmath70 . therefore , in the janssen framework , we are led to attribute the difference between maximum force and stationary force to a difference in stress redirection ( i.e. in @xmath42 ) due to a difference in compacity . the force history in the transitory regime would then just reflect the compacity history . as a check of consistency , we perform the following dynamical experiment . first , the granular column is pushed upwards in order to mobilize the friction forces downwards and far enough to reach the steady state compacity . starting from this situation , the friction forces at the walls are reversed by moving the piston downwards at a constant velocity @xmath71 , until a stationary regime is attained . note that this stationary regime is characterized by the same compacity @xmath72 as in the pushing situation . following relation ( [ 1 ] ) , this procedure would imply a change of @xmath73 from @xmath74 to @xmath75 , and consequently , the dynamical force on the piston should decrease from @xmath76 to @xmath77 . in fig . [ fig7 ] the pushing force @xmath77 is measured for different packing heights @xmath36 . the fit of experimental results with eq . ( [ 1 ] ) gives @xmath78 which is @xmath79 larger than @xmath80 . this difference , though small , can be observed out of uncertainties , and is systematic . it can not be due to a slight change in compacity @xmath32 as from relation @xmath81 , we would expect a 2% variation in compacity between the pushing and the pulling experiment , which would be observed ; we actually measured @xmath82 . according to janssen s picture , this would imply that vertical stress redirection is more efficient in the downward pulling situation . we believe this is a clear evidence of a granular structuration effects but its also shows that this effect is not dominant : it affects only @xmath79 of the average mechanical parameter @xmath42 . we have shown in another report that in a janssen experiment , an isotropic homogeneous elastic material can also be characterized by stress redirection properties @xcite . for a granular column at high depth an effective relation : _ zz = k_el_rris obtained with k_el=_p/(1-_p)where @xmath83 is the material s poisson coefficient . this actually leads to a curve similar to janssen s saturation curve as long as friction at the walls is small ( typically less than @xmath4 ) . + in order to get the isotropic homogeneous elasticity prediction for the pushing experiment , we perform a series of numerical simulations using finite element method @xcite . the column is modelled as an isotropic elastic medium . we vary the friction at the walls @xmath84 , the young modulus @xmath85 and the poisson coefficient @xmath83 . we impose a rigid , either perfectly stick or perfectly slip bottom . we find no appreciable difference between these two previous cases . the condition @xmath86 is imposed everywhere at the walls ( for the pulling situation , we impose @xmath87 ) . the cylinder is modelled as a duralumin elastic medium . as long as the young modulus @xmath85 of the elastic medium is less than @xmath88 mpa , which is usually the case for granular media , we find no dependence of the results on @xmath85 . we verified that in all the simulations we performed , there is no traction in the elastic medium , so that this could be a fair modelling for a granular material . we find no appreciable difference between the elastic prediction ( fig . [ fig8 ] ) and the curve given by eq . ( [ 1 ] ) with @xmath89 . therefore , regarding the dependence of the stationary state force @xmath31 on the height of beads , our system can not be distinguished from an elastic medium . note that finite element simulations show that the presence of a rigid bottom implies that the effective janssen s parameter @xmath90 extracted from janssen s scaling for the pulling situation is higher than @xmath91 @xcite , whereas for the pushing @xmath92 ( as can be seen on fig . [ fig8 ] : the fit of the elastic curve with @xmath89 is good ) . details on this numerical work will be given elsewhere @xcite . actually , if we adjust the elastic predictions for pushing and pulling experiments with an elastic material of poisson coefficient @xmath93 , eq . ( [ 1 ] ) yields a janssen s constant @xmath90 for the pushing which is about @xmath94 lower than @xmath90 for the pulling . this is qualitatively ( though not quantitatively ) in agreement with the experimental results . then isotropic elasticity can be a good framework only if we neglect the existence of bulk structuration effects inducing differences in the effective poisson coefficient of the material between the pulling and the pushing . note that in this case , an isotropic modelling of the granular material is somehow questionable . in a previous study @xcite , it was found that the janssen picture has a general tendency to slightly underestimate the stress below a granular column for a homogeneous packing of glass beads . we have showed in another report that it is no more true as the friction at walls is very well controlled . therefore , this model , though elementary , seems a fair base for analysis and provides an analytical expression from which constitutive rheological parameters can be extracted . a central question is still that the fitting parameters @xmath95 extracted from the model does not allow to distinguish between @xmath84 and @xmath42 separately . in the following , we will show experiments which aim at sorting out the relative contributions of wall - bead interactions ( i.e. @xmath84 ) and bulk properties ( i.e. @xmath42 ) which have an influence on the rheological properties when velocity and relative humidity are changed . the mean resistance force @xmath31 in the steady - sliding regime increases strongly when the velocity and the relative humidity are increased . in the duralumin column ( fig . [ fig9 ] ) , the resistance force @xmath31 at velocity @xmath29 is @xmath96 higher for relative humidity @xmath97 than for @xmath98 ; near the transition to the stick - slip regime ( @xmath99 ) , dependence of @xmath31 on @xmath13 is much less important , and there is actually no difference in @xmath31 values for @xmath13 varying between @xmath100 and @xmath101 . another important feature is that the increase of @xmath31 with velocity is sharper when @xmath13 is higher . in the brass column [ fig10 ] ) , the mean resistance force @xmath31 also increases with relative humidity @xmath13 , but now the variation coefficient with velocity ( the slope on fig . [ fig10 ] ) does not seem to depend on @xmath13 . moreover , for a velocity @xmath102 , the @xmath31 increase is only @xmath103 from dry to humid ( @xmath16 ) atmosphere . in the steady sliding regime for height @xmath104 ( 380 g ) of steel beads in the duralumin cylinder , for various relative humidities @xmath13 : @xmath105 ( diamonds ) , @xmath106 ( open inverted triangles ) , @xmath107 ( triangles ) , @xmath100 ( open circles ) , and @xmath108 ( squares ) . b ) coefficient @xmath109 extracted from a fit of @xmath110 with formula ( [ 1]).,width=264 ] in the steady sliding regime for height @xmath104 ( 380 g ) of steel beads in the brass cylinder , for various relative humidities @xmath13 : @xmath111 ( squares ) , @xmath107 ( open circles ) , and @xmath108 ( triangles ) . b ) coefficient @xmath109 extracted from a fit of @xmath110 with formula ( [ 1]).,width=264 ] the fit of @xmath31 with janssen s formula adapted to the pushing case of eq . ( [ 1 ] ) gives parameter @xmath112 for different relative humidities @xmath13 . the fundamental differences of behavior in the steady sliding regime between the duralumin column and the brass column for a same granular material , at same density , suggest that the variations in @xmath13 and velocity have an influence mainly on the coefficient of friction @xmath84 at the walls , and little on mechanical properties of the granular material ( i.e. @xmath42 ) . therefore , in the following analysis , we will consider @xmath42 as a constant at first order of approximation . [ fig9 ] shows that dynamical parameter @xmath113 is globally less important in a dry atmosphere than in the ambient atmosphere . in ambient atmosphere ( @xmath114 ) , a change in @xmath13 seems to change only the variation coefficient of @xmath113 with velocity : i.e. the increase of @xmath113 with velocity is sharper when @xmath13 is higher . this phenomenon could be interpreted as a viscous contribution of the water condensed at the contacts . note that this phenomenology contrasts with the observations of riedo et al . @xcite who find , for solid on solid nanoscopic sliding friction measurements , that in all the systems they study , of various wettability , the effect of a humidity increase is to add a negative value to the coefficient of friction force dependence on logarithm of velocity . table [ tablestrengtheningdural ] shows the results of a rough logarithmic fit of data with @xmath115 . we use the parameter @xmath116 as a standard reference to the dieterich - ruina model @xcite . [ c]|c|c|relative humidity @xmath13 & @xmath117 + @xmath108 & @xmath118 + @xmath100 & @xmath119 + @xmath107 & @xmath120 + @xmath106 & @xmath121 + @xmath105 & @xmath122 + we observe that the increase rate @xmath123 of @xmath113 with the logarithm of velocity is multiplied by @xmath124 when @xmath13 increases from @xmath100 to @xmath105 . in order to compare these results with those obtained in a solid on solid friction experiment , we need to evaluate janssen s parameter @xmath42 . we extract @xmath42 from a classical janssen experiment @xcite and we obtain : @xmath125 for a mean aging time of @xmath126 seconds . from the measurement of static friction , using the sliding angle of a three bead tripod , we extract @xmath127.we suppose @xmath42 is constant and is not affected by any variation in relative humidity nor velocity . the dynamical coefficient of friction obtained in a solid on solid friction experiment with the slider ( fig . [ fig1 ] inset ) is then compared on fig . [ fig11 ] with the one extracted from a fit of the resistance force with formula ( [ 1 ] ) . note that the uncertainty on @xmath84 extracted from formula ( [ 1 ] ) is mainly systematic ( coming from uncertainties on @xmath42 , which come themselves from uncertainty on static friction measurements ) . we observe that the dynamical coefficients of friction measured in a solid on solid experiment increase slightly more strongly with velocity than the one extracted from the resistance force of the granular material . but importantly , the increase of @xmath84 with relative humidity @xmath13 is the same in both cases . + in this analysis we assumed a constant value for janssen s coefficient @xmath42 . we can also imagine a slight decrease of @xmath42 with increasing velocity so that the solid on solid @xmath84 would match the one extracted from the pushing experiment . however , we will remain prudent as the mean pressure contact on beads in the solid on solid experiment is much higher than in the granular column , which may cause quantitative differences between friction properties . more precisely , at the bottom of the column , for a pushing force @xmath128n , the mean contact force per bead is about @xmath129mn ; a hertz contact would give mean contact pressure @xmath130mpa . for a contact force per bead of @xmath131n ( the slider case ) , a hertz contact would give @xmath132mpa . in the brass column , we observe the same phenomenology as in the duralumin one , i.e. an increase of @xmath113 with velocity and relative humidity @xmath133 . the effect of a change in @xmath13 is yet less important than in the duralumin column , as the coefficient @xmath113 increases only by @xmath103 at velocity @xmath102 when @xmath13 increases from @xmath134 to @xmath111 . a major difference with the duralumin column is that in the brass column , a variation in @xmath13 seems to induce a variation of friction but hardly affects the coefficient of variation with velocity ( the slope ) . table [ tablestrengtheninglaiton ] shows the results of a logarithmic fit of the data with the function : @xmath115 . we observe indeed that the slope @xmath123 of @xmath113 with the logarithm of velocity does not practically vary with relative humidity @xmath13 , and is much weaker than in the duralumin column . [ c]|c|c|relative humidity @xmath13 & @xmath123 + @xmath108 & @xmath135 + @xmath107 & @xmath135 + @xmath111 & @xmath136 + the comparison with the slider , and the differences between the duralumin and bras columns , suggest that the rheological properties in the steady - sliding regime are dominated by the friction properties at the walls . we now investigate the stick - slip regime . for a height @xmath137 ( 380 g ) of steel beads in the duralumin column . b : force vs. velocity of the granular material for the slip event of fig . a , width=264 ] in the stick - slip regime , for a given height of beads in the column and at given driving velocity @xmath7 and relative humidity @xmath13 , the maximum force before a slip and the minimum force at the end of a slip are constant within @xmath38 ( fig . [ fig5]b ) . their distribution around mean values @xmath138 and @xmath139 is nearly gaussian . + the dynamical evolution @xmath140 of the pushing force during the slip phase can be translated into a function of the granular material instantaneous velocity . motion of the granular material at velocity @xmath7 induces a variation @xmath141 of resistance force during time @xmath142 , where @xmath143 is the force probe stiffness and @xmath43 the driving velocity ; we thus obtain @xmath144 . we see ( fig . [ fig12 ] ) that the acceleration and deceleration phases are nearly symmetric . this phenomenology is currently observed in solid friction experiments . this is in contrast with previous plane shearing experiments of granular materials @xcite in which the deceleration phase would occur in two steps , first fast and then slow . however , a major difference is that in the latter case , the displacements during slippage are millimetric ( a grain size ) whereas in our experiment , they are micrometric ( @xmath145 m on fig . [ fig12 ] ) . as a consequence , the evolution of forces observed during a slip seems to reflect the slippage of grain contacts at the walls . the maximum velocities obtained during the slip phase are of a few hundred @xmath11 . we did not study systematically the variation of instantaneous velocity with bead height nor with driving velocity . in the following , we report on the @xmath138 and @xmath139 properties as a function of the packing height , the driving velocity and the relative humidity . in the stick - slip domain , the mean maximum and minimum resistance forces @xmath138 and @xmath139 are measured as functions of the packing height ( see fig . [ fig13 ] ) , but now we choose to perform our experiments not at a constant velocity @xmath7 , but at constant stick time @xmath146 and we explain why in the following . experimentally we find that the resistance forces @xmath138 and @xmath139 increase exponentially with the packing s height @xmath36 . the data can actually be fitted by a formula of type eq . ( [ 1 ] ) , but the parameter @xmath67 extracted may be different and should not depend on @xmath56 as the granular material is not sliding during stick , and does not slide at driving velocity @xmath43 during slip . the model can actually be simply modified for this situation . in the stick - slip regime , when the resistance force @xmath140 is equal to @xmath138 , the material just starts slipping , which means that static friction is fully mobilized at this instant for each contact . so we can write @xmath147 everywhere at the walls , where @xmath148 is static friction between beads and the cylinder s walls . static friction coefficients are known to evolve in time @xcite , we then have to include aging effects . as the granular material was at rest in the column since the previous slip , aging time in this situation is the time of stick : @xmath149 so the final formula for @xmath138 in the janssen framework is : @xmath150 where the screening length is now @xmath151 . here again , @xmath54 for a pushing experiment ( i.e. @xmath152 ) , @xmath153 for a janssen experiment . the choice of imposing a constant stick - time @xmath146 instead of constant driving velocity @xmath43 for the experiments in the stick - slip regime is now justified by the dependence of @xmath148 on @xmath146 , whereas @xmath84 depends on @xmath43 . in the steady state regime , the experimental data obtained for a given stick - time @xmath146 can be fitted by relation ( [ 3 ] ) by adjusting only one parameter i.e. @xmath154 where @xmath155 is the static coefficient of friction for an aging time @xmath146 . for the data of fig . [ fig13 ] obtained for a relative humidity @xmath98 with a mean stick time of @xmath156s , we find @xmath157 from @xmath138 ( @xmath158 in eq . ( [ 3 ] ) . note that , though friction is not fully mobilized when @xmath159 , the fit of @xmath40 with eq . ( [ 3 ] ) is good ( fig . [ fig13 ] ) ; we obtain a parameter @xmath160 which should not be related to @xmath155 in that case . as a check of consistency , we perform the following dynamical experiment in the stick - slip regime . first , the granular column is pushed upwards in order to mobilize the friction forces downwards and to reach the steady state compacity ; @xmath138 ( referred to as ( a ) on fig . [ fig14 ] ) is then measured for constant stick time @xmath146 . starting from this situation , the friction forces are reversed at the walls by moving the piston downwards . we then measure the force @xmath139 of the first slip ( referred to as ( b ) on fig . [ fig14 ] ) after the same time @xmath146 . note that the roles of @xmath139 and @xmath138 are inverted in the pulling experiment : @xmath139 is the minimum force before a slip , and @xmath138 is the maximum force at the end of a slip . we also measure mean minimum force @xmath139 ( referred to as ( c ) on fig . [ fig14 ] ) when a steady stick - slip regime is reached in the pulling experiment . note that this stationary regime is characterized by the same compacity @xmath72 as in the pushing situation . of the packing scaled by the column diameter @xmath161 , for steel beads in the duralumin column , for stick - time @xmath162 s ; the line is a fit with eq . ( [ 3 ] ) . b : measured mass vs. filling mass for a classical janssen experiment obtained in two different ways : we plot the mass of the first slipping event ( squares ) , and the mean minimum mass in the stationary stick - slip for the pulling situation ( triangles ) ; the line is the janssen curve predicted with the parameter extracted from the pushing experiment of fig . [ fig15]a.,width=264 ] following relation ( [ 3 ] ) , this procedure should imply a change of @xmath73 from @xmath74 to @xmath75 , and consequently , the force at the onset of slipping on the piston should decrease from @xmath158 to @xmath163 . as we observed no noticeable variation of compacity between the pushing and the pulling , we expect janssen s parameter @xmath42 value to be unchanged . in fig . [ fig15 ] the pulling force @xmath163 is measured for different packing heights @xmath36 , for the first slipping event and in the steady stick - slip regime . the fit of experimental results with eq . ( [ 3 ] ) gives @xmath164 for stationary stick - slip which is @xmath165 larger than @xmath166 . it can not be due to a slight change in average compacity @xmath32 as from relation @xmath167 , we would expect a @xmath168 variation in compacity , which would have been observed ; let us recall that we actually measured a compacity variation between the pushing and the pulling experiment @xmath82 . + as in the steady - sliding experiment , according to janssen s picture , this would then imply that vertical stress redirection is more efficient in the pulling situation ; furthermore , this effect is enhanced in the stick - slip regime . for the first slip event in the pulling experiment , we find @xmath169 which is now lower than @xmath170 . we believe this is a clear evidence of a granular structuration effects . we interpret the differences in @xmath171 values by saying that the packing has first been structured to resist the pushing ; when we start pulling , structure is not efficient to resist the pulling , and it results in a lower janssen s constant ( therefore lower @xmath171 ) for the first slip event ; when we continue pulling , the packing gets structured to resist pulling and leans more efficiently on the walls , i.e. janssen s constant ( and @xmath171 ) increases . moreover , steady - state structuration is more efficient for the pulling than for the pushing , which means that both structurations may be different and could reflect the symmetry breaking due to gravity . the results of the simulations of an elastic medium in a column presented in the precedent section for a steady - sliding also carry on to the stick - slip situation . if we impose numerically everywhere at the walls , the relations @xmath172 or @xmath173 , it would merely correspond to a change in the name of the friction coefficient . the equilibrium equations are the same for the static and steady dynamic cases . the important common point is that friction forces are fully mobilized everywhere at the walls . therefore , as for the steady - sliding , the data can be fitted by the elastic predictions , and the @xmath94 difference in janssen s parameter between the pulling and the pushing of an elastic medium of constant poisson ratio is again qualitatively ( though not quantitatively ) in agreement with the experimental results . in the following , as we study the steady - state regime , we will try to distinguish between wall - bead interactions ( i.e. @xmath148 ) and the bulk properties ( i.e. @xmath42 ) as they influence the rheological properties when velocity and relative humidity are changed . ( circles ) and @xmath39 ( squares ) vs. driving velocity in the stick - slip regime for a height @xmath174 ( 250 g ) of steel beads in the duralumin cylinder , for relative humidities @xmath16 ( filled symbols ) and @xmath58 ( open symbols ) . b : @xmath175 as a function of stick time @xmath176 for a height @xmath174 ( 250 g ) of steel beads in the duralumin cylinder , for relative humidities @xmath16 ( filled squares ) and @xmath58 ( open circles).,width=264 ] ( circles ) and @xmath39 ( squares ) vs. driving velocity in the stick - slip regime for a height @xmath104 ( 380 g ) of steel beads in the duralumin cylinder , for relative humidities @xmath58 ( open symbols ) and @xmath15 ( filled symbols ) . b : @xmath175 as a function of stick time @xmath176 for a height @xmath104 ( 380 g ) of steel beads in the duralumin cylinder , for relative humidities @xmath58 ( open circles ) and @xmath15 ( filled triangles).,width=264 ] ( circles ) and @xmath39 ( squares ) vs. driving velocity in the stick - slip regime for a height @xmath177 ( 300 g ) of steel beads in the brass cylinder , for relative humidities @xmath16 ( open symbols ) and @xmath15 ( filled symbols ) . b : @xmath175 as a function of stick time @xmath176 for a height @xmath177 ( 300 g ) of steel beads in the brass cylinder , for relative humidities @xmath16 ( filled squares ) and @xmath15 ( filled triangles ) , and for a height @xmath104 ( 380 g ) for relative humidities @xmath178 ( open circles ) and @xmath15 ( open triangles).,width=264 ] the mean maximum force @xmath138 increases strongly when the driving velocity @xmath43 is decreased and the relative humidity @xmath13 is increased ( fig . [ fig16]a , [ fig17]a , [ fig18]a ) . furthermore , @xmath138 apparent divergence when @xmath43 decreases is enhanced when @xmath13 increases . the mean minimum force @xmath139 is almost independent of the driving velocity , and increases when the relative humidity is increased ( fig . [ fig16]a , [ fig17]a , [ fig18]a ) . for each stick - slip event , an aging time ( the stick time ) is measured : @xmath179 . this time is associated to a parameter @xmath180 extracted from the fit of the maximum force @xmath138 of this stick - slip event with equation ( [ 3 ] ) . in the following , the results of this analysis for duralumin and brass columns are compared and discussed . for steel beads in the duralumin column , the parameter @xmath181 ( fig . [ fig16]b , [ fig17]b ) increases roughly logarithmically with aging time , and increases strongly with relative humidity @xmath13 . its value increases by 50% when @xmath13 increases from @xmath94 to @xmath182 , and doubles when @xmath13 increases from @xmath182 to @xmath111 . the parameter @xmath181 does not depend only globally on @xmath13 , but its logarithmic aging rate @xmath183 with stick time @xmath146 depends strongly on @xmath13 . the parameters @xmath183 extracted from a logarithmic fit of @xmath184 with @xmath185 are given in table [ tablevieillissementdural ] for different relative humidities . note that magnitude of @xmath183 , as @xmath42 is of the order 1 , is consistent with many previous reports on solid friction @xcite . an important feature is that the logarithmic aging rate appears to be 4 times higher in humid atmosphere than in dry atmosphere . [ c]|c|c|relative humidity @xmath13 & @xmath183 + @xmath108 & @xmath186 + @xmath182 & @xmath187 + @xmath111 & @xmath188 + the logarithmic evolution is not perfect : for aging time larger than @xmath189s , we observe that @xmath181 increases more sharply with time ; for @xmath104 of steel beads and @xmath58 , @xmath183 varies from @xmath190 for times @xmath191 s to about @xmath192 in the last time decade ( fig . [ fig17]b ) . this phenomenon was observed for all relative humidities @xmath13 . it is similar to what was reported previously by losert et al . @xcite for plane shearing of glass beads . + it is tempting to link this enhanced aging behavior to the granular material slow restructuring effects since one may think of an aging of the granular structure leading to a better stress redirection at the walls . however , at this point , it is not possible to differentiate between wall and structure properties . we tried to measure independently @xmath148 aging properties in a solid on solid experiment , but we did not obtain satisfactory statistics to conclude . therefore , we need to compare the results obtained in the duralumin column to the one obtained in the brass column . for steel beads in the brass column , the parameter @xmath181 extracted from the experimental measurements ( fig . [ fig18]b ) also increases logarithmically with aging time . however , the increase with relative humidity @xmath13 is much less important than in the duralumin column since @xmath181 increase is only @xmath193 when increasing @xmath13 from dry atmosphere to @xmath111 humidity , whereas the increase is more than @xmath194 in the duralumin column for the same variation of @xmath13 . moreover , the @xmath181 increase with stick time seems now perfectly logarithmic over 4 decades of time variation . as in the duralumin column , @xmath181 increase rate with stick time @xmath146 depends on @xmath13 . the parameters @xmath183 extracted from a logarithmic fit of @xmath184 with @xmath185 are given in table [ tablevieillissementlaiton ] . logarithmic aging rate is now 2.5 times greater in humid atmosphere than in dry atmosphere ( it was 4 times higher in the duralumin column ) . [ c]|c|c|relative humidity @xmath13 & @xmath183 + @xmath108 & @xmath186 + @xmath195 & @xmath196 + @xmath111 & @xmath197 + the only differences between both systems ( duralumin and brass column ) being the contacts at the walls , as for the steady - sliding regime , the differences observed in the rheological behavior suggest that the aging of @xmath181 is mainly an aging of friction coefficient @xmath148 at the walls and that humidity affects principally the contacts at the walls . relative humidity @xmath13 seems to have two effects in the duralumin column : first , the aging rate is higher when @xmath13 is higher , second , the friction level for short times is higher when @xmath13 is higher.in the brass column there seems to be an effect of @xmath13 only on the aging rate : the @xmath198 curves for different @xmath13 cross for @xmath199 ms . the logarithmic aging rate coefficient gets comparable values in both columns , but it is higher in the duralumin column ( @xmath200 ) in humid atmosphere than in the brass column ( @xmath201 ) . our observations are consistent with recent aging experiments in granular media @xcite and solid on solid aging experiments @xcite , who found logarithmic aging of static friction enhanced by an increase in relative humidity @xmath13 . these results have been interpreted by the dynamics of capillary condensation at the contacts @xcite , which is a thermally activated process : as time goes on , there are more and more capillary bridges at the contacts , which are responsible for an adhesion force increasing with aging time ; moreover , condensation goes faster when @xmath13 is higher . in these models , there is no aging in dry atmosphere , as no condensation can occur ; this prediction agrees with most observations @xcite . as in our experiment the logarithmic aging rate coefficient is not zero in dry atmosphere ( even though the air flux we impose is probably not perfectly dry ) , we can conclude that capillary condensation is not the only source of aging . another source of aging can be creep of contacts @xcite . we observed that aging is perfectly logarithmic in the brass column . therefore , we are led to analyze again the aging in the duralumin column since the interpretation based on aging of internal friction we proposed for the duralumin column , should also apply to the brass column , which is in contradiction with the observed behavior . if we suppose that @xmath42 properties depend only on the granular material properties , the acceleration of logarithmic aging in the duralumin column may be not linked to aging of bead / bead contacts nor granular slow restructurations . we may then start from the following experimental observations : in solid on solid friction experiments , static friction coefficients @xmath148 are found to depend on the applied shear exerted for a given waiting time @xcite . for a higher shear during a given waiting time , coefficient @xmath148 is higher , and the logarithmic aging rate is higher too @xcite . in our experiments , during a `` stick '' event , the pushing force increases linearly with time , i.e. shear at the walls ( and contact pressure ) increases . as a consequence , aging occurs with a non - constant applied shear and the mean shear is more important when the aging time is higher . consequently for high waiting time or stick time , i.e. for high mean shear at the walls , the aging of the static friction coefficient may be accelerated.this phenomenology may be a priori different for different surfaces , which would explain why aging seems perfectly logarithmic in the brass column , and not in the duralumin column . now , the high enhancement of the blocking resistance force @xmath138 with decreasing driving velocity can be simply understood within the janssen framework . we start from the experimental observation that the minimum force @xmath139 after a slip does not depend on velocity ( at constant relative humidity @xmath13 ) : @xmath202 . therefore , the aging time @xmath146 before a slip reads : @xmath203 we suppose , as well verified experimentally , that the coefficient of friction @xmath148 at the walls follows a logarithmic aging law : @xmath204 this value of @xmath205 for aging time @xmath146 can be injected in formula ( [ 3 ] ) . we then obtain : @xmath206 with @xmath207.the apparent divergence of @xmath138 with decreasing velocity , as observed experimentally , is thus more important when the height of beads and the logarithmic aging rate @xmath208 of friction at the walls ( i.e. relative humidity as @xmath208 increases with @xmath13 ) are higher . the observed strong dependence of @xmath138 with @xmath7 then corresponds to an exponential amplification of friction aging at the walls . elements of interpretation along these lines were already given in ref @xcite . nevertheless from the whole series of experiments we performed on this system we are led to conclude that the dependence of the phenomenology with the relative humidity @xmath13 remains sometimes unclear . ( stars ) and @xmath40 ( squares ) of a height @xmath137 ( 380 g ) of steel beads in the duralumin cylinder vs. time in the stick - slip regime ( @xmath209 ) . b : evolution of relative humidity @xmath13 during the experiment of fig . a. the dotted line separates two phases : 1 ) forces and @xmath13 evolutions are not correlated ; 2 ) forces and @xmath13 evolutions are correlated.,width=264 ] as an illustration of this statement , we report the following result ( fig . [ fig19 ] , [ fig20 ] ) . in an experiment performed in the stick - slip regime and in the ambient atmosphere , for a driving velocity @xmath210 , the forces @xmath139 and @xmath138 were found to be independent of @xmath13 , which was varying between @xmath211 and @xmath212 , during the first 30 hours of the experiment . we then observed that the @xmath139 and @xmath138 variations became suddenly correlated to the @xmath13 variations and would stay correlated for the next 80 hours . for the first 30 hours , @xmath13 increased from @xmath211 to @xmath212 , while @xmath139 and @xmath138 were nearly constant : @xmath213 n , @xmath214 n. then @xmath13 varied from @xmath212 to @xmath215 in one hour : @xmath139 and @xmath138 started to be correlated to @xmath13 from this `` triggering event '' and for the next 80 hours as if the system had suddenly reached some `` reversible branch '' . such a phenomenology was never obtained when humidities @xmath16 and @xmath15 were imposed . when we imposed dry atmosphere , the force would reach its stationary value in a few minutes , and we never noticed any variation in this value for 5 days ; when we imposed humid atmosphere , the force would reach its stationary value in a maximum of 3 hours and we noticed no variation in this value for 5 days . this may however incite to prudence for results obtained in ambient atmosphere , for experiments performed on the granular column as much as for solid on solid friction experiments . as an example , we see on fig . [ fig20 ] that for @xmath216 , @xmath217 n when forces and @xmath13 are not correlated , whereas @xmath218 n when they are correlated . furthermore , this opens the question of the metastability of capillary condensation . if variation of @xmath138 with @xmath13 is interpreted as variation of @xmath148 due to thermally activated capillary condensation at the contacts @xcite , it seems possible that capillary condensation did not occur for hours . if we notice that @xmath138 starts to be correlated with @xmath13 when @xmath138 constant value is equal to @xmath138 on the @xmath13 dependent part of the curve ( i.e. @xmath219 n and @xmath220 ) , another possibility is that there were as much capillary bridges condensed at the contacts as for @xmath221 in the first part of the experiment , i.e. that evaporation does not always happen . a standard phenomenological model accounting for static and dynamic properties of solid friction is the so - called dieterich - ruina model @xcite . in this picture , the coefficient of friction is @xmath222 where @xmath223 obeys @xcite @xmath224 @xmath225 , @xmath226 , @xmath208 , and @xmath161 are constants dependent of the materials in contact . parameter @xmath161 is usually interpreted as a characteristic length for renewal of contacts and is of order of the creep length before sliding @xcite . in a stationary steady - sliding regime @xmath227 is interpreted as a characteristic time for renewal of contacts . this model accounts for the logarithmic aging of static friction , and the logarithmic velocity strengthening or weakening in the steady - sliding regime with : @xmath228 it also accounts for a third usual phenomenon observed in the transitory regime when the velocity is changed suddenly . in this case there is an abrupt change in friction coefficient followed by exponential relaxation to a new stationary value . if we change velocity @xmath229 to velocity @xmath230 , the instantaneous change in friction coefficient from stationary value @xmath231 is @xmath232 ; friction change from this value to new stationary value @xmath233 is then @xmath234 during typical time @xmath235 , i.e. we get @xmath236 . now we test the dieterich - ruina model with the steel / duralumin and steel / brass friction coefficient extracted from our data . as we observed logarithmic aging and rough logarithmic velocity strengthening , we get parameters @xmath226 and @xmath208 . we are therefore able to test this model predictions with the observation of response to an abrupt variation of velocity . experimentally , we never observed any transitory state : the force in the steady sliding regime was always changed from stationary value @xmath237 to stationary value @xmath238 when changing velocity from @xmath229 to @xmath230 . from the measured value of @xmath239 for steel / duralumin friction and @xmath240 for steel / brass in ambient atmosphere , we would however expect the force at the beginning of the transitory regime to be @xmath241n larger ( steel beads height @xmath104 ) than the stationary force at the new velocity . as the natural fluctuations of force are around @xmath242n , we would therefore expect to observe this transitory regime in @xmath31 due to change in friction at the walls . now we give an interpretation for why , within the dietrich - ruina picture , this transitory regime was not observed , which yields an upper bound for the material contact renewal length @xmath243 . when velocity is changed from @xmath244 to @xmath245 for @xmath104 ( 380 g ) of steel beads in a duralumin cylinder , the change in @xmath31 stationary value is of 1 n. but , due to the finite stiffness of our set - up , this change can not be instantaneous : an increase @xmath246 of force at constant velocity takes minimal time @xmath247 . so an increase of 1 n of force at @xmath248 takes time @xmath249s . this time must be compared to time length @xmath250 of transitory regime : @xmath251 . therefore , in the framework of the dieterich - ruina model , we need @xmath250 to be less than the time of force variation , which means @xmath252 m . let us now consider a sudden decrease in velocity from @xmath245 to @xmath253 . now the decrease of force is governed by inertial time : @xmath254 which can be evaluated around @xmath255 ms . this leads to @xmath256 nm ( so that @xmath257 ) consistently with the dieterich - ruina model if no transitory state is observed . we have no precise measure for @xmath258 but this upper limit on the value is coherent with the usual interpretation in terms of length for contact renewal , as the duralumin column mean roughness is 400 nm . thus , we can neither validate nor rule out the dieterich - ruina model for our system . for experiments performed at given height of beads and relative humidity @xmath13 , we observe that the transition from stick - slip to steady - sliding occurs for a finite stick - slip amplitude ( fig . [ fig21 ] ) . this suggests a subcritical transition but such a transition would be hysteretic . in order to verify this feature , we make the following experiment : we drive the system at constant velocity in the steady - sliding regime near the transition until it reaches a stationary state , we then decrease continuously the velocity , without ever stopping the movement , to a velocity for which stick - slip used to occur . note that the results reported in the previous sections and on fig . [ fig21 ] were obtained for independent experiments on systems driven at a unique velocity . we observe that if we impose @xmath259 directly , we obtain a stick - slip motion , whereas when we decrease velocity from @xmath260 ( in the steady - sliding regime ) to @xmath259 the motion is steady - sliding . this proves the hysteretic character of the transition , which was actually hard to observe as the system seems very sensitive to noise . this transition is similar to the one observed by heslot et al . @xcite in solid on solid experiments ; it correspond to what they call the inertial regime . it is a characteristic of systems for which the dynamical coefficient of friction increases with velocity . this transition scenario was explained by brockley et al . however they find that the force @xmath31 in the steady - sliding regime at the transition is of order of @xmath139 . we actually find ( see fig . [ fig21 ] ) that @xmath31 is less than @xmath139 for steel beads in the duralumin cylinder , a feature we do not understand , whereas @xmath261 for in the brass cylinder . again , the transition we observed is consistent with a picture of solid friction sliding instability at the wall . this is also consistent with the model we have developed in the previous section . after the study of the simplest case , with monodisperse low friction beads , we now want to study the effect of disorder ( friction , polydispersity ) on rheology . therefore , we choose to study a miscellany of glass beads using a three diameters mixture ( 1.5 mm , 2 mm and 3 mm ) with equal volume of each kind in an abraded pmma cylinder . in the following , we present the features obtained in the stick - slip regime for this system . ( open circles ) and @xmath39 ( full squares ) vs. driving velocity in the stick - slip regime for a height @xmath262 ( 132 g ) of a miscellany of glass beads in the pmma cylinder at relative humidity @xmath263% . b : @xmath175 as a function of stick time @xmath176 extracted from @xmath39 with equation ( [ 3]).,width=264 ] ( 126 g ) of a miscellany of glass beads in the pmma cylinder vs. displacement @xmath27 of the translation stage in the stick - slip regime ( @xmath264 ) . b : zoom of fig . [ fig23]a.,width=264 ] for low velocities ( @xmath265 ) , a stick - slip motion is obtained with the same properties as observed with monodisperse steel beads in a duralumin or brass cylinder . it is a regular stick - slip motion characterized by a maximum force before slipping that increases strongly when velocity is decreased , and a minimum force after slipping which does not depend on velocity ( fig . [ fig22]a ) . a coefficient @xmath198 can be extracted from @xmath138 with equation ( [ 3 ] ) for low velocities ( @xmath265 ) . again , it shows a logarithmic aging of friction at the walls ( fig . [ fig22]b ) . however , the behavior changes for higher velocities just before the transition to steady - sliding . the stick - slip motion is then less and less regular as the velocity is increased . furthermore we observe structures in the force signal . figure [ fig23 ] evidences cycles of force increase during which the maximum and minimum forces increase for several stick - slip events , as the force amplitude @xmath266 during a stick is higher than the force amplitude @xmath267 during a slip . then , at the end of a cycle , there is a big slip event and a new cycle starts . this phenomenon is similar to the one observed by albert et al . @xcite for the pushing of a stick in a granular material , and to the one observed for aluminium beads by kolb et al . @xcite in the same display as ours but in 2d . it was however much more important in that last study in 2d : the force @xmath138 would then increase from @xmath268 after a big slip event to @xmath269 before the next big slip event . during a stick and @xmath270 during a slip of a height @xmath271 ( 126 g ) of a miscellany of glass beads in the pmma cylinder for three driving velocities : a ) @xmath272 ; b ) @xmath273 ; c ) @xmath264.,width=264 ] we analyse this phenomenon by studying the distribution of force amplitudes defined as @xmath274 for the @xmath275 stick event , and @xmath276 for the following slip event ( fig . [ fig24 ] ) . we observe the evolution of these distributions with increasing velocity . @xmath266 distribution remains centered around a mean value ; however , this distribution gets larger when the velocity is increased . for lower velocities ( @xmath277 on fig . [ fig24]a ) , stick - slip is regular , and @xmath267 distribution is regular around same mean value as @xmath266 ( with nevertheless a larger distribution ) . when velocity is increased , the distribution is still predominant around same mean value as @xmath266 but many smaller values start to merge in the distribution ( @xmath278 on fig . [ fig24]b ) ; these small values become as frequent as the larger ones for higher velocities ( @xmath279 on fig . [ fig24]c ) . the distribution also enlarges to the higher values which corresponds to occurrence of the big slip events . note that in these experiments , the transition to steady - sliding regime occurs for velocity a velocity around @xmath280 . several explanations can be proposed to account for this complex dynamics exhibiting a large distribution of energy release . in our point of view , the principal ingredient for an explanation seems to be the progressive structuration of the packing between two big slip events . in the janssen model picture , the progressive increase of @xmath138 in a cycle corresponds to progressive increase of janssen s constant @xmath42 . the @xmath42 value increases from an initial value @xmath281 at the beginning of a cycle up to a maximum value @xmath282 at the end of the cycle , and is reinitialised by a big slip event to value @xmath281 . let us recall that @xmath42 was found to increase with density in another report @xcite . a consequence of this dependence was shown in section [ iii.a.1 ] . therefore , a way to obtain the phenomenology is to start from a loose packing at the beginning of a cycle ; if each small slip event densifies the packing ( as can be expected from compaction under vibration ) , it leads to a small increase in @xmath42 , so that the minimum force @xmath139 at the end of the slip is higher than the force @xmath139 at the end of the preceding slip ; the maximum force @xmath138 at the end of the stick is also higher than the force @xmath138 at the end of the preceding stick . sometimes , big events occur ; they probably correspond to convective motion at the walls , as observed by kolb et al . in 2d @xcite , which loosens the packing , which then gets the initial density of the cycle . for the interpretation we give , the stick events last a constant time as the aging time of contacts at the walls is still controlled by the velocity . this explains why the distribution of @xmath266 is still centered on mean value contrary to @xmath267 . but still a criterion for big slip events has yet to be found ; it may be the level of stresses , or a dilatance effect at high density causing shear band and convection rolls . the regularity of stick - slip motion at low velocities can be understood if the criterion for big slip events is at an upper stress level . a slow driving velocity leads to long aging of contacts at the walls so that the maximum force @xmath138 is always greater than the threshold to obtain a big slip . all slip events are then big slip events in this case . this transition from a simple steady stick - slip to a complex dynamical regime with a large distribution of elastic energy release is an interesting issue and we plan to pursue further the investigation trying to clarify in particular the role of disorder and polydispersity in the phenomenology . we presented an experimental report on the dynamical behavior of a granular column pushed vertically . we first investigated the case of a monodisperse assembly of steel beads and we observed at low driving velocities ( @xmath283 ) a steady behavior such that the pushing forces increase roughly logarithmically with velocity . on the other hand , at very low driving velocity ( @xmath1 ) , we evidenced a discontinuous and hysteretic transition to a stick - slip regime characterized by a strong divergence of the maximal blockage force when the velocity goes to zero . all this phenomenology is strongly influenced by surrounding humidity : generally , higher humidity level increases strongly the resistance to pushing . finite elements numerical simulations were used to confront experimental results to a modelling of the granular packing as an isotropic elastic medium . then , we showed that a simple janssen s model is a fair base for analysis as it provides the correct physical interpretation for the pushing resistance . this janssen s model was used to extract an important mechanical parameter combining the effects of stress redirection and wall friction but there is an inherent difficulty to isolate clearly the various contributions either coming from bulk reorganization or from the surface friction properties . using different column materials and measuring directly the friction of a grain with the wall , we accumulated several evidences leading us to conclude that the force dependence with driving velocity and humidity is strongly related to the bead - wall friction properties : ( i ) in the steady limit grain / wall tribology measurements show a friction force increasing with humidity and velocity , ( ii ) in the stick - slip regime , the blockage enhancement can be related to humidity induced aging of the bead - wall friction , ( iii ) the hysteretic transition mechanism from stick - slip to steady - sliding is similar to the one observed in solid - on - solid experiments @xcite . in spite of a dominant surface effect , we could also identify contributions of bulk structurations . for example , we related the transitory part of the response to pushing to a dependance of the coefficient of redirection between horizontal and vertical stress with packing fraction . also a clear difference of the mechanical parameters extracted from pushing and pulling experiments shows a contribution from restructuration of about 20% . a second system , made of polydisperse assemblies of glass beads , was investigated . we emphasize the onset of a new complex dynamical behavior , i.e. the large distribution of blockage forces evidenced in the stick - slip regime close to the transition . we acknowledge collaboration with c. fond , c. gauthier and e. kolb . we thank j. lanuza and p. lepert for technical support . we are grateful to professor t. baumberger , professor r.p . behringer , and professor c. caroli for many interesting discussions .

after the study of the three body molecular system h ( _ j . phys . b : at . mol . opt . phys . _ * 45 * 065101 ) , its isotopomer , the deuterium molecular ion d is studied . the three - body schrdinger equation is solved using the lagrange - mesh method in perimetric coordinates . energies and wave functions for four vibrational states and bound or quasibound states for total orbital momenta from 0 to 56 are calculated . the 1986 fundamental constant is used . the obtained energies have an accuracy from about 13 digits for the lowest vibrational state to at least 9 digits for the third vibrational excited state . quadrupole transition probabilities per time unit between those states over the whole rotational bands were calculated . extensive results are presented with six significant figures .

1302.5234

despite of the simplicity of the molecular ion h@xmath0 and its isotopomers d@xmath0 and t@xmath0 , the three body schrdinger equation can not be solved exactly . however , high precision results , for energies and wave functions , have been obtained using different approaches . for the particular case of the deuterium molecular ion , the ground state as well as some vibrational states are well known . in this sense , non - adiabatic dissociation energies were calculated for several vibrational - rotational levels for the ground electronic state @xcite and also in the first excited electronic state . in particular , the value of the ground state energy has been improved in several papers @xcite . energies for the zero vibrational band for different values of the total angular momentum @xmath3 have been also considered @xcite . complete bound vibrational states below the dissociation limit @xmath4 were presented in @xcite for the two lowest rotational quantum numbers @xmath5 . extension to this work until @xmath6 is presented in @xcite but with a different value of the deuteron mass . some expectation values have been also provided like the mean radii @xcite or quadrupole moments @xcite . however , nowadays quadrupole transitions has not been considered for this system . in this paper , the e2 transitions are obtained from the three - body wave functions calculated with the lagrange - mesh method in perimetric coordinates @xcite , with which the calculation is particularly simple and very precise . the lagrange - mesh method is an approximate variational calculation using a basis of lagrange functions and the associated gauss quadrature . it has the high accuracy of a variational approximation and the simplicity of a calculation on a mesh . in section [ sec : lmtp ] , some expressions are presented but detailed explanations are given in @xcite . in section [ sec : res ] , energies are given for the lowest four vibrational states over the full rotational bands and e2 transition probabilities are tabulated . conclusions are presented in section [ sec : conc ] . throughout atomic units quadrupole oscillator strength for an electric transition between an initial state @xmath7 and a final state @xmath8 is given by @xcite f_if^(2 ) = ^2 ( e_f - e_i)^3 , where @xmath9 is the fine - structure constant and s_if^(2 ) = s_fi^(2 ) = _ m_im_f |_i j_i m_i| q^(2)_| _ f j_f m_f |^2 = |_i j_i || q^(2 ) || _ f j_f |^2 , is the reduced matrix element , defined according to @xcite . here , @xmath10 is a total angular momentum , @xmath11 is its projection and @xmath12 represents the other quantum numbers . the transition probability per time unit for @xmath13 is given ( the atomic unit of time is @xmath14 s ) by w_if^(2 ) = ^5(e_i - e_f)^5 . after the elimination of the center - of - mass motion , the system can be described using the euler angles @xmath15 , defining the orientation of the triangle formed by the three particles and three internal coordinates describing the shape of the triangle . the internal degrees of freedom are described using the perimetric coordinates ( @xmath16 ) defined as rcl x & = & r - r_e2+r_e1 , + y & = & r+r_e2-r_e1 , + z & = & -r+r_e2+r_e1 , where @xmath17 and @xmath18 are the distances between the electron and deuterons 1 and 2 , respectively . the domains of variation of these six variables are @xmath19 $ ] for @xmath20 and @xmath21 , @xmath22 $ ] for @xmath23 and @xmath24 for @xmath25 and @xmath26 . in perimetric coordinates , the transition irreducible tensor operator @xmath27 is given by q^(2)_= [ r^2 - ( 2 ^ 2 - ^2 ) ] d^2_0 ( , , ) - - ^2 [ d^2_2 ( , , ) + d^2_-2 ( , , ) ] . where = 1 - - and the expression of @xmath28 , @xmath29 and @xmath30 in terms of perimetric coordinates are given for example in @xcite . in order to evaluate ( [ 2.3 ] ) it is necessary to solve three - body hamiltonian that involves coulomb forces between the particles but no spin - dependent forces . the total orbital momentum @xmath3 and parity @xmath31 are constants of motion . the wave functions with orbital momentum @xmath3 and parity @xmath31 are expanded as @xcite ^(l^)_m = _ k=0^l d_mk^l ( , , ) _ k^(l^)(x , y , z ) . in practice , the sum can be truncated at some value @xmath32 . the normalized angular functions @xmath33 are defined for @xmath34 by _ mk^l ( , , ) & = & ( 1+_k0)^-1/2 , where @xmath35 represents a wigner matrix element . they have parity @xmath31 and change as @xmath36 under permutation of the deuterons . the @xmath37 functions with symmetry @xmath38 are symmetric for @xmath39 and antisymmetric for @xmath40 , when @xmath41 and @xmath42 are exchanged and are expanded in the lagrange basis as _ k^(l^)(x , y , z ) & = & _ i=1^n _ j=1^i-_k _ k=1^n_z c_kijk^(l^ ) ^-1/2 + & & , where @xmath43 , @xmath44 are the number of points in the @xmath41 and @xmath26 . scale factors @xmath45 and @xmath46 are introduced in order to fit the meshes to the size of the physical problem ( the same number @xmath43 of mesh points and the same scale factor @xmath45 are used for the two perimetric coordinates @xmath41 and @xmath42 ) . because of the symmetrization the sum over @xmath47 is limited by the value @xmath48 , where @xmath49 is equal to 0 when @xmath50 and to 1 when @xmath51 . the three - body hamiltonian in perimetric coordinates for each good quantum number @xmath3 and its discretization on a lagrange mesh are given in @xcite . for given @xmath52 , the eigenvalues in increasing order are labeled by a quantum number @xmath53 related to the vibrational excitation in the born - oppenheimer picture . the corresponding eigenvectors provide the coefficients appearing in expansion . because the @xmath54-value appears in the eigenfunction ( [ 4.1 ] ) , the matrix element ( [ 2.3 ] ) contains terms where @xmath55 , @xmath56 and @xmath57 . these terms are labeled as @xmath58 , respectively ( see @xcite ) . using the lagrange - mesh method , energies of the @xmath59 lowest vibrational bound states for some values of the total angular momentum from @xmath60 to 51 have been calculated in @xcite . at the present study , those calculations are extended for all bound and quasibound states up to @xmath61 and four vibrational states @xmath1 . in order to calculate transition matrix elements involving two different wave functions , a single three - dimensional mesh is used for all the vibro - rotational states in consideration : @xmath62 and @xmath63 . a single set of scaling factors are also employed : @xmath64 , @xmath65 . for a given @xmath54 value , the total number of basis states is 20700 or 19800 depending on the parity of @xmath54 . for @xmath66 , calculations are performed with @xmath67 . they correspond to a size of 61200 . here , the 1986 fundamental constant @xmath2 is adopted and the dissociation energy is @xmath68 . table [ tab:1 ] presents the energies obtained for the three lowest vibrational states @xmath1 as a function of the rotational quantum number . the accuracy is estimated from the stability of the digits with respect to calculations with @xmath69 mesh points . the error is expected to be at most of a few units on the last displayed digit . comparisons when are available are displayed below at the same line for each @xmath3-value . rllll + @xmath3 & @xmath70 & @xmath71 & @xmath72 & @xmath73 + + @xmath3 & @xmath70 & @xmath71 & @xmath72 & @xmath73 + + 0 & -0.5987887843307 & -0.591603121903 & -0.58471220701 & -0.5781085914 + & -0.59878878433068@xmath74&-0.59160312190321@xmath74 & -0.58471220700999@xmath74&-0.57810859143687@xmath74 + & -0.5987887843307@xmath75&-0.5916031219032[4]@xmath75&& + & -0.59878878432@xmath76 & -0.59160312192@xmath76 & -0.58471220699@xmath76 & -0.57810859141@xmath76 + ' '' '' 1 & -0.5986548732205 & -0.591474211529 & -0.58458816962 & -0.5779893120 + & -0.59865487322@xmath74 & -0.59147421153@xmath74 & -0.58458816962@xmath74&-0.57798931196@xmath74 + & -0.5986548732205@xmath75 & -0.5914742115286[7]@xmath75&& + & -0.59865487321@xmath76 & -0.59147421153@xmath76 & -0.58458816960@xmath76 & -0.57798931198@xmath76 + ' '' '' 2 & -0.5983875858100 & -0.591216909624 & -0.58434059838 & -0.5777512419 + & -0.59838758580998@xmath77 & & & + & -0.59838758578@xmath76 & -0.59121690963@xmath76 & -0.58434059839@xmath76 & -0.57775124191@xmath76 + ' '' '' 3 & -0.5979879847467 & -0.590832247071 & -0.58397049374 & -0.5773953525 + & -0.59798798474672@xmath77 & & & + & -0.59798798475@xmath76 & -0.59083224709@xmath76 & -0.58397049375@xmath76 & -0.57739535253@xmath76 + ' '' '' 4 & -0.5974576467838 & -0.590321753376 & -0.58347933994 & -0.5769230848 + & -0.59745764678378@xmath77 & & & + & -0.59745764678@xmath76 & -0.59032175337@xmath76 & -0.58347933994@xmath76 & -0.57692308478@xmath76 + ' '' '' 5 & -0.5967986427008 & -0.589687437079 & -0.58286908594 & -0.5763363306 + & -0.59679864270081@xmath77 & & & + & -0.59679864268@xmath76 & -0.58968743709@xmath76 & -0.58286908595@xmath76 & -0.57633633045@xmath76 + ' '' '' 6 & -0.5960135114695 & -0.588931760548 & -0.58214212082 & -0.5756374089 + & -0.59601351146953@xmath77 & & & + 7 & -0.5951052294020 & -0.588057609916 & -0.58130124447 & -0.5748290372 + & -0.59510522940195@xmath77 & & & + 8 & -0.5940771751293 & -0.587068260974 & -0.58034963445 & -0.5739142993 + & -0.59407717512931@xmath77 & & & + 9 & -0.5929330913285 & -0.585967341945 & -0.57929080974 & -0.5728966100 + & -0.59293309132853@xmath77 & & & + 10 & -0.5916770441414 & -0.584758794049 & -0.57812859247 & -0.5717796780 + & -0.59167704414133@xmath77 & & & + 11 & -0.5903133812204 & -0.583446830801 & -0.57686706841 & -0.5705674674 + & -0.59031338122032@xmath77 & & & + 12 & -0.5888466892923 & -0.582035896884 & -0.57551054715 & -0.5692641596 + & -0.58884668929225@xmath77 & & & + 13 & -0.5872817520563 & -0.580530627415 & -0.57406352270 & -0.5678741147 + 14 & -0.5856235091419 & -0.578935808297 & -0.57253063517 & -0.5664018348 + 15 & -0.5838770167449 & -0.577256338258 & -0.57091663423 & -0.5648519288 + 16 & -0.5820474104456 & -0.575497193053 & -0.56922634458 & -0.5632290787 + 17 & -0.5801398706008 & -0.573663392212 & -0.56746463407 & -0.5615380100 + 18 & -0.5781595905894 & -0.571759968586 & -0.56563638450 & -0.5597834625 + 19 & -0.5761117480916 & -0.569791940871 & -0.56374646533 & -0.5579701664 + 20 & -0.5740014794896 & -0.567764289167 & -0.56179971040 & -0.5561028190 + 21 & -0.5718338574015 & -0.565681933603 & -0.55980089757 & -0.5541860664 + 22 & -0.5696138712928 & -0.563549715941 & -0.55775473137 & -0.5522244863 + 23 & -0.5673464110598 & -0.561372384075 & -0.55566582834 & -0.5502225750 + 24 & -0.5650362534388 & -0.559154579261 & -0.55353870516 & -0.5481847356 + 25 & -0.5626880510684 & -0.556900825932 & -0.55137776916 & -0.5461152700 + 26 & -0.5603063240115 & -0.554615523892 & -0.54918731126 & -0.5440183722 + 27 & -0.5578954535382 & -0.552302942710 & -0.54697150101 & -0.5418981242 + 28 & -0.5554596779645 & -0.549967218120 & -0.54473438364 & -0.5397584939 + 29 & -0.5530030903498 & -0.547612350239 & -0.54247987885 & -0.5376033342 + 30 & -0.5505296378624 & -0.545242203435 & -0.54021178130 & -0.5354363850 + 31 & -0.5480431226351 & -0.542860507670 & -0.53793376262 & -0.5332612750 + 32 & -0.5455472039496 & -0.540470861200 & -0.53564937477 & -0.5310815273 + 33 & -0.5430454016063 & -0.538076734487 & -0.53336205469 & -0.5289005638 + 34 & -0.5405411003522 & -0.535681475235 & -0.53107513031 & -0.5267217137 + 35 & -0.5380375552674 & -0.533288314477 & -0.52879182757 & -0.5245482212 + 36 & -0.5355378980259 & -0.530900373673 & -0.52651527889 & -0.5223832563 + 37 & -0.5330451439779 & -0.528520672798 & -0.52424853276 & -0.5202299270 + 38 & -0.5305622000232 & -0.526152139463 & -0.52199456482 & -0.5180912929 + 39 & -0.5280918732775 & -0.523797619127 & -0.51975629058 & -0.5159703825 + 40 & -0.5256368805684 & -0.521459886554 & -0.51753658000 & -0.5138702126 + 41 & -0.5231998588400 & -0.519141658708 & -0.51533827450 & -0.5117938124 + 42 & -0.5207833765951 & -0.516845609396 & -0.51316420699 & -0.5097442527 + 43 & -0.5183899465717 & -0.514574386104 & -0.51101722587 & -0.5077246832 + 44 & -0.5160220399376 & -0.512330629670 & -0.50890022443 & -0.5057383795 + 45 & -0.5136821024045 & -0.510116997691 & -0.50681617771 & -0.5037888067 + 46 & -0.5113725728331 & -0.507936193030 & -0.50476818997 & -0.5018797063 + 47 & -0.5090959051435 & -0.505790999399 & -0.50275955797 & -0.5000152210 + 48 & -0.5068545947077 & -0.503684327110 & -0.50079385828 & -0.4982000829 + 49 & -0.5046512109696 & -0.501619273844 & -0.49887507334 & -0.4964399169 + 50 & -0.5024884389368 & -0.499599208500 & -0.49700778309 & -0.49474177 + 51 & -0.5003691337052 & -0.497627892183 & -0.49519747673 & -0.4931151 + 52 & -0.4982963948257 & -0.495709662497 & -0.493451107 & + 53 & -0.4962736722626 & -0.4938497343 & -0.4917782 & + 54 & -0.4943049255767 & -0.492054739 & & + 55 & -0.492394879740 & -0.49033383 & & + 56 & -0.490549476 & & & + the ground state @xmath78 is well known and its value has been improved in several papers @xcite . in particular hijikata @xmath79 @xcite reach an accuracy of 27 digits . in this sense we have an accuracy about @xmath80 . energies of the @xmath70 vibrational band for @xmath81 were presented by zong @xmath79 @xcite . moss @xcite presents not only non - adiabatic dissociation - energies but also relativistic and radiative corrections for bound vibrational states up to @xmath82 . seven rovibrational states of the first electronic state are also presented there . more precise results for all the vibrational levels below the dissociation limit @xmath4 are given by hilico @xmath79 @xcite for the two lowest rotational quantum numbers @xmath5 . beyond that , karr @xmath79 @xcite extended this results until @xmath6 but with a different value of the deuteron mass @xmath83 . in order to make some comparison with those results , lagrange - mesh calculations considering this value of the deuteron mass are given in table [ tab:1b ] . although less digits were obtained when comparing with @xcite we can consider that the agreement is in all the digits presented . the accuracies for the first , second and third rotational bands are @xmath84 , @xmath85 and @xmath86 , respectively . the major comparison is possible with the results given by moss @xcite . the obtained spectrum is presented in fig . [ fig:1 ] . ( 140,70 ) ( -20,10 ) ( 0,0 ) + two cases of convergence tests are displayed in table [ tab:2 ] . for the values @xmath87 and @xmath65 , initial and final states are shown as well as the transition probability per second for different values of @xmath43 and @xmath44 . the transition probability @xmath88 is obtained by restricting the matrix element ( [ 2.3 ] ) to the case where @xmath89 while @xmath90 corresponds to @xmath91 . the @xmath92 contributions have an importance smaller than 0.05 % and that the @xmath93 contributions are smaller than 0.005 % . a good convergence with respect to @xmath44 is already obtained for @xmath63 . the convergence with respect to @xmath43 is slower . since the convergence is exponential , one can estimate that the relative accuracy on @xmath94 is about @xmath86 for @xmath95 and still better than @xmath96 for @xmath97 . similar tests have been performed for other transitions . the convergence of the transition probabilities with respect to @xmath32 can be studied by comparison with results from wave functions truncated at @xmath98 and @xmath99 . the relative error when @xmath98 is smaller than 0.3 for all considered transitions while the error for @xmath100 is smaller than @xmath101 . this is rather similar to truncations with respect to @xmath102 . by extrapolation , we estimate that the relative error on the present transition probabilities obtained with @xmath103 should be smaller than @xmath104 . the number of digits presented in tables [ tab:3 ] and [ tab:4 ] are the same if we consider the deuteron mass as @xmath105 . table [ tab:3 ] presents the transitions probabilities per second for transitions within a same rotational band , @xmath106 and @xmath107 . they include some transition probabilities involving quasibound states . six significant digits are given . these probabilities increase slowly with @xmath3 with a maximum around @xmath108 , @xmath109 , @xmath110 and @xmath111 for @xmath112 , @xmath113 , @xmath114 and @xmath115 , respectively , which is due to a maximum of the energy differences around @xmath116 . rllll + + @xmath117&@xmath118&@xmath119&@xmath120&@xmath121 + + @xmath117&@xmath118&@xmath119&@xmath120&@xmath121 + + 2 & 3.06649 - 13 & 3.06947 - 13 & 3.04109 - 13 & 2.98389 - 13 + 3 & 5.02689 - 12 & 5.02874 - 12 & 4.97940 - 12 & 4.88307 - 12 + 4 & 2.96566 - 11 & 2.96408 - 11 & 2.93250 - 11 & 2.87343 - 11 + 5 & 1.08642 - 10 & 1.08455 - 10 & 1.07179 - 10 & 1.04907 - 10 + 6 & 3.01258 - 10 & 3.00301 - 10 & 2.96357 - 10 & 2.89689 - 10 + 7 & 6.95282 - 10 & 6.91874 - 10 & 6.81671 - 10 & 6.65288 - 10 + 8 & 1.40815 - 09 & 1.39848 - 09 & 1.37527 - 09 & 1.33980 - 09 + 9 & 2.58558 - 09 & 2.56215 - 09 & 2.51433 - 09 & 2.44451 - 09 + 10 & 4.39774 - 09 & 4.34729 - 09 & 4.25629 - 09 & 4.12887 - 09 + 11 & 7.03338 - 09 & 6.93441 - 09 & 6.77221 - 09 & 6.55353 - 09 + 12 & 1.06924 - 08 & 1.05123 - 08 & 1.02387 - 08 & 9.88224 - 09 + 13 & 1.55773 - 08 & 1.52692 - 08 & 1.48293 - 08 & 1.42732 - 08 + 14 & 2.18842 - 08 & 2.13842 - 08 & 2.07056 - 08 & 1.98705 - 08 + 15 & 2.97943 - 08 & 2.90183 - 08 & 2.80088 - 08 & 2.67962 - 08 + 16 & 3.94654 - 08 & 3.83072 - 08 & 3.68531 - 08 & 3.51439 - 08 + 17 & 5.10250 - 08 & 4.93541 - 08 & 4.73191 - 08 & 4.49732 - 08 + 18 & 6.45645 - 08 & 6.22254 - 08 & 5.94502 - 08 & 5.63065 - 08 + 19 & 8.01350 - 08 & 7.69471 - 08 & 7.32496 - 08 & 6.91269 - 08 + 20 & 9.77447 - 08 & 9.35026 - 08 & 8.86795 - 08 & 8.33783 - 08 + 21 & 1.17358 - 07 & 1.11833 - 07 & 1.05662 - 07 & 9.89663 - 08 + 22 & 1.38895 - 07 & 1.31839 - 07 & 1.24079 - 07 & 1.15761 - 07 + 23 & 1.62235 - 07 & 1.53382 - 07 & 1.43781 - 07 & 1.33600 - 07 + 24 & 1.87220 - 07 & 1.76288 - 07 & 1.64582 - 07 & 1.52294 - 07 + 25 & 2.13653 - 07 & 2.00354 - 07 & 1.86273 - 07 & 1.71629 - 07 + 26 & 2.41312 - 07 & 2.25347 - 07 & 2.08620 - 07 & 1.91374 - 07 + 27 & 2.69945 - 07 & 2.51018 - 07 & 2.31375 - 07 & 2.11285 - 07 + 28 & 2.99283 - 07 & 2.77098 - 07 & 2.54275 - 07 & 2.31109 - 07 + 29 & 3.29040 - 07 & 3.03308 - 07 & 2.77052 - 07 & 2.50589 - 07 + 30 & 3.58921 - 07 & 3.29366 - 07 & 2.99436 - 07 & 2.69471 - 07 + 31 & 3.88627 - 07 & 3.54986 - 07 & 3.21156 - 07 & 2.87502 - 07 + 32 & 4.17858 - 07 & 3.79885 - 07 & 3.41951 - 07 & 3.04440 - 07 + 33 & 4.46317 - 07 & 4.03786 - 07 & 3.61564 - 07 & 3.20052 - 07 + 34 & 4.73716 - 07 & 4.26425 - 07 & 3.79753 - 07 & 3.34119 - 07 + 35 & 4.99778 - 07 & 4.47548 - 07 & 3.96289 - 07 & 3.46437 - 07 + 36 & 5.24238 - 07 & 4.66915 - 07 & 4.10958 - 07 & 3.56817 - 07 + 37 & 5.46849 - 07 & 4.84305 - 07 & 4.23564 - 07 & 3.65090 - 07 + 38 & 5.67381 - 07 & 4.99514 - 07 & 4.33929 - 07 & 3.71101 - 07 + 39 & 5.85621 - 07 & 5.12355 - 07 & 4.41890 - 07 & 3.74715 - 07 + 40 & 6.01378 - 07 & 5.22661 - 07 & 4.47306 - 07 & 3.75814 - 07 + 41 & 6.14480 - 07 & 5.30284 - 07 & 4.50052 - 07 & 3.74296 - 07 + 42 & 6.24775 - 07 & 5.35096 - 07 & 4.50019 - 07 & 3.70073 - 07 + 43 & 6.32132 - 07 & 5.36982 - 07 & 4.47113 - 07 & 3.63070 - 07 + 44 & 6.36438 - 07 & 5.35847 - 07 & 4.41253 - 07 & 3.53223 - 07 + 45 & 6.37597 - 07 & 5.31608 - 07 & 4.32369 - 07 & 3.40473 - 07 + 46 & 6.35531 - 07 & 5.24194 - 07 & 4.20397 - 07 & 3.24762 - 07 + 47 & 6.30175 - 07 & 5.13543 - 07 & 4.05272 - 07 & 3.06022 - 07 + 48 & 6.21475 - 07 & 4.99594 - 07 & 3.86924 - 07 & 2.84164 - 07 + 49 & 6.09388 - 07 & 4.82284 - 07 & 3.65261 - 07 & 2.59053 - 07 + 50 & 5.93869 - 07 & 4.61536 - 07 & 3.40153 - 07 & 2.306 - 07 + 51 & 5.74875 - 07 & 4.37243 - 07 & 3.11389 - 07 & 1.98 - 07 + 52 & 5.52347 - 07 & 4.09241 - 07 & 2.7861 -07 & + 53 & 5.26195 - 07 & 3.77261 - 07 & 2.411 - 07 & + 54 & 4.96276 - 07 & 3.40820 - 07 & & + 55 & 4.62336 - 07 & 2.990 -07 & & + 56 & 4.23906 - 07 & & & + other transitions probabilities per second are given in table [ tab:4 ] . the columns correspond to transitions between different vibrational states . for each @xmath117 value , the successive lines correspond to increasing values of @xmath122 , i.e. , to @xmath106 for @xmath123 , @xmath124 for @xmath125 , and @xmath126 , respectively . comparison with previous results was not possible . the strongest transition from each state occurs in general towards the nearest vibrational state ( @xmath127 ) . for @xmath127 , in the vicinity of @xmath128 and beyond , the @xmath129 transitions are replaced by @xmath130 transitions because the sign of the energy difference changes ( see the arrows in fig . [ fig:1 ] for the @xmath131 transitions ) . these numbers are indicated in italics in table [ tab:4 ] . for example , the first number in the last line for @xmath128 corresponds to the @xmath132 transition . lllllll + + @xmath117&@xmath133&@xmath134&@xmath135 & @xmath136&@xmath137&@xmath138 + + @xmath117&@xmath133&@xmath134&@xmath135 & @xmath136&@xmath137&@xmath138 + + 0&7.46127 - 08&7.38489 - 09&1.30115 - 07&5.35445 - 10&2.01270 - 08&1.69382 - 07 + ' '' '' 1&3.52148 - 08&4.09047 - 09&6.12020 - 08&3.73819 - 10&1.10072 - 08&7.94026 - 08 + & 3.94786 - 08&3.45944 - 09&6.89688 - 08&2.03661 - 10&9.52149 - 09&8.99401 - 08 + ' '' '' 2&2.02805 - 08&2.70362 - 09&3.51024 - 08&2.92983 - 10&7.19574 - 09&4.53536 - 08 + & 2.50641 - 08&2.91823 - 09&4.35533 - 08&2.67317 - 10&7.85142 - 09&5.64951 - 08 + & 2.94905 - 08&2.25075 - 09&5.15938 - 08&9.99391 - 11&6.26593 - 09&6.73753 - 08 + ' '' '' 3&2.82157 - 08&4.08955 - 09&4.86792 - 08&4.86245 - 10&1.08110 - 08&6.26890 - 08 + & 2.32686 - 08&2.71868 - 09&4.04232 - 08&2.49909 - 10&7.31265 - 09&5.24205 - 08 + & 2.35159 - 08&1.53043 - 09&4.11862 - 08&4.49867 - 11&4.31935 - 09&5.38394 - 08 + ' '' '' 4&3.35316 - 08&5.25537 - 09&5.76414 - 08&6.77275 - 10&1.38038 - 08&7.39567 - 08 + & 2.25034 - 08&2.64149 - 09&3.90807 - 08&2.43932 - 10&7.10260 - 09&5.06609 - 08 + & 1.91012 - 08&1.03072 - 09&3.34791 - 08&1.52208 - 11&2.95897 - 09&4.37935 - 08 + ' '' '' 5&3.75968 - 08&6.34415 - 09&6.43706 - 08&8.77843 - 10&1.65607 - 08&8.22508 - 08 + & 2.20176 - 08&2.59938 - 09&3.82212 - 08&2.41405 - 10&6.98641 - 09&4.95240 - 08 + & 1.55644 - 08&6.70208 - 10&2.72911 - 08&2.05248 - 12&1.96706 - 09&3.57098 - 08 + ' '' '' 6&4.07956 - 08&7.38600 - 09&6.95387 - 08&1.08920 - 09&1.91643 - 08&8.84493 - 08 + & 2.16244 - 08&2.57047 - 09&3.75199 - 08&2.40318 - 10&6.90525 - 09&4.85884 - 08 + & 1.26383 - 08&4.11204 - 10&2.21618 - 08&5.74085 - 13&1.24414 - 09&2.89963 - 08 + ' '' '' 7&4.32541 - 08&8.37990 - 09&7.33719 - 08&1.30922 - 09&2.16146 - 08&9.28551 - 08 + & 2.12586 - 08&2.54700 - 09&3.68635 - 08&2.39943 - 10&6.83822 - 09&4.77075 - 08 + & 1.01914 - 08&2.30667 - 10&1.78658 - 08&7.28493 - 12&7.30041 - 10&2.33653 - 08 + ' '' '' 8&4.50122 - 08&9.31339 - 09&7.59490 - 08&1.53421 - 09&2.38816 - 08&9.55841 - 08 + & 2.08927 - 08&2.52557 - 09&3.62048 - 08&2.39954 - 10&6.77621 - 09&4.68202 - 08 + & 8.14451 - 09&1.11779 - 10&1.42682 - 08&1.94482 - 11&3.81120 - 10&1.86451 - 08 + ' '' '' 9&4.60815 - 08&1.01696 - 08&7.73022 - 08&1.75957 - 09&2.59247 - 08&9.66947 - 08 + & 2.05142 - 08&2.50447 - 09&3.55220 - 08&2.40178 - 10&6.71469 - 09&4.58988 - 08 + & 6.44116 - 09&4.10309 - 11&1.12727 - 08&3.48689 - 11&1.62301 - 10&1.47128 - 08 + ' '' '' 10&4.64690 - 08&1.09308 - 08&7.74583 - 08&1.98023 - 09&2.77014 - 08&9.62414 - 08 + & 2.01171 - 08&2.48275 - 09&3.48050 - 08&2.40507 - 10&6.65112 - 09&4.49303 - 08 + & 5.03563 - 09&7.10949 - 12&8.80046 - 09&5.17980 - 11&4.41020 - 11&1.14675 - 08 + ' '' '' 11&4.61886 - 08&1.15801 - 08&7.64575 - 08&2.19101 - 09&2.91728 - 08&9.42978 - 08 + & 1.96986 - 08&2.45985 - 09&3.40493 - 08&2.40868 - 10&6.58400 - 09&4.39092 - 08 + & 3.88787 - 09&4.40059 - 13&6.78231 - 09&6.88770 - 11&1.43585 - 12&8.81954 - 09 + ' '' '' 12&4.52664 - 08&1.21032 - 08&7.43618 - 08&2.38687 - 09&3.03067 - 08&9.09667 - 08 + & 1.92580 - 08&2.43538 - 09&3.32537 - 08&2.41209 - 10&6.51237 - 09&4.28344 - 08 + & 2.96163 - 09&1.29761 - 11&5.15490 - 09&8.50909 - 11&1.30778 - 11&6.68630 - 09 + ' '' '' 13&4.37424 - 08&1.24889 - 08&7.12579 - 08&2.56316 - 09&3.10795 - 08&8.63831 - 08 + & 1.87954 - 08&2.40911 - 09&3.24187 - 08&2.41487 - 10&6.43561 - 09&4.17070 - 08 + & 2.22372 - 09&3.80667 - 11&3.86000 - 09&9.97233 - 11&6.13535 - 11&4.99140 - 09 + ' '' '' 14&4.16709 - 08&1.27298 - 08&6.72564 - 08&2.71582 - 09&3.14781 - 08&8.07116 - 08 + & 1.83117 - 08&2.38088 - 09&3.15462 - 08&2.41668 - 10&6.35331 - 09&4.05297 - 08 + & 1.64392 - 09&7.03414 - 11&2.84430 - 09&1.12311 - 10&1.31872 - 10&3.66463 - 09 + ' '' '' 15&3.91191 - 08&1.28226 - 08&6.24890 - 08&2.84151 - 09&3.14995 - 08&7.41424 - 08 + & 1.78082 - 08&2.35057 - 09&3.06387 - 08&2.41723 - 10&6.26523 - 09&3.93061 - 08 + & 1.19504 - 09&1.05589 - 10&2.05972 - 09&1.22599 - 10&2.13239 - 10&2.64237 - 09 + ' '' '' 16&3.61652 - 08&1.27680 - 08&5.71046 - 08&2.93776 - 09&3.11512 - 08&6.68850 - 08 + & 1.72866 - 08&2.31813 - 09&2.96992 - 08&2.41629 - 10&6.17122 - 09&3.80407 - 08 + & 8.52971 - 10&1.40625 - 10&1.46351 - 09&1.30497 - 10&2.96738 - 10&1.86803 - 09 + ' '' '' 17&3.28957 - 08&1.25707 - 08&5.12639 - 08&3.00294 - 09&3.04502 - 08&5.91610 - 08 + & 1.67486 - 08&2.28353 - 09&2.87313 - 08&2.41368 - 10&6.07125 - 09&3.67382 - 08 + & 5.96701 - 10&1.73153 - 10&1.01838 - 09&1.36042 - 10&3.75986 - 10&1.29215 - 09 + ' '' '' 18&2.94024 - 08&1.22387 - 08&4.51344 - 08&3.03636 - 09&2.94221 - 08&5.11969 - 08 + & 1.61963 - 08&2.24677 - 09&2.77386 - 08&2.40926 - 10&5.96534 - 09&3.54038 - 08 + & 4.08225 - 10&2.01628 - 10&6.92372 - 10&1.39360 - 10&4.46581 - 10&8.72354 - 10 + ' '' '' 19&2.57798 - 08&1.17835 - 08&3.88851 - 08&3.03814 - 09&2.80999 - 08&4.32172 - 08 + & 1.56317 - 08&2.20788 - 09&2.67249 - 08&2.40291 - 10&5.85359 - 09&3.40428 - 08 + & 2.72393 - 10&2.25117 - 10&4.58594 - 10&1.40641 - 10&5.05760 - 10&5.73023 - 10 + ' '' '' 20&2.21219 - 08&1.12186 - 08&3.26814 - 08&3.00922 - 09&2.65222 - 08&3.54375 - 08 + & 1.50569 - 08&2.16692 - 09&2.56941 - 08&2.39457 - 10&5.73616 - 09&3.26604 - 08 + & 1.76680 - 10&2.43175 - 10&2.94854 - 10&1.40110 - 10&5.52071 - 10&3.64794 - 10 + ' '' '' 21&1.85199 - 08&1.05598 - 08&2.66808 - 08&2.95123 - 09&2.47318 - 08&2.80597 - 08 + & 1.44741 - 08&2.12394 - 09&2.46501 - 08&2.38419 - 10&5.61324 - 09&3.12622 - 08 + & 1.10925 - 10&2.55734 - 10&1.83181 - 10&1.38014 - 10&5.85080 - 10&2.23943 - 10 + ' '' '' 22&1.50599 - 08&9.82403 - 09&2.10290 - 08&2.86637 - 09&2.27740 - 08&2.12669 - 08 + & 1.38854 - 08&2.07902 - 09&2.35968 - 08&2.37177 - 10&5.48506 - 09&2.98535 - 08 + & 6.70446 - 11&2.63001 - 10&1.09312 - 10&1.34599 - 10&6.05113 - 10&1.31706 - 10 + ' '' '' 23&1.18208 - 08&9.02900 - 09&1.58574 - 08&2.75731 - 09&2.06956 - 08&1.52197 - 08 + & 1.32929 - 08&2.03228 - 09&2.25380 - 08&2.35732 - 10&5.35190 - 09&2.84394 - 08 + & 3.87338 - 11&2.65376 - 10&6.21681 - 11&1.30105 - 10&6.13033 - 10&7.35658 - 11 + ' '' '' 24&8.87318 - 09&8.19267 - 09&1.12803 - 08&2.62707 - 09&1.85428 - 08&1.00546 - 08 + & 1.26986 - 08&1.98380 - 09&2.14775 - 08&2.34088 - 10&5.21404 - 09&2.70252 - 08 + & 2.11868 - 11&2.63384 - 10&3.33410 - 11&1.24760 - 10&6.10059 - 10&3.85646 - 11 + ' '' '' 25&6.27795 - 09&7.33270 - 09&7.39377 - 09&2.47890 - 09&1.63610 - 08&5.88158 - 09 + & 1.21045 - 08&1.93372 - 09&2.04188 - 08&2.32250 - 10&5.07181 - 09&2.56157 - 08 + & 1.08286 - 11&2.57617 - 10&1.66145 - 11&1.18769 - 10&5.97615 - 10&1.86576 - 11 + ' '' '' 26&4.08590 - 09&6.46613 - 09&4.27475 - 09&2.31615 - 09&1.41930 - 08&2.78423 - 09 + & 1.15126 - 08&1.88213 - 09&1.93656 - 08&2.30227 - 10&4.92553 - 09&2.42157 - 08 + & 5.07513 - 12&2.48696 - 10&7.52990 - 12&1.12316 - 10&5.77221 - 10&8.12681 - 12 + ' '' '' 27&2.33734 - 09&5.60902 - 09&1.98091 - 09&2.14222 - 09&1.20789 - 08&8.19667 - 10 + & 1.09246 - 08&1.82918 - 09&1.83209 - 08&2.28027 - 10&4.77554 - 09&2.28297 - 08 + & 2.12059 - 12&2.37234 - 10&3.00389 - 12&1.05562 - 10&5.50399 - 10&3.06614 - 12 + ' '' '' 28&1.06212 - 09&4.77618 - 09&5.51016 - 10&1.96044 - 09&1.00552 - 08&1.94502 - 11 + & 1.03424 - 08&1.77498 - 09&1.72881 - 08&2.25659 - 10&4.62218 - 09&2.14618 - 08 + & 7.55275 - 13&2.23816 - 10&9.99844 - 13&9.86467 - 11&5.18615 - 10&9.38687 - 13 + ' '' '' 29&2.79901 - 10&3.98107 - 09&5.67929 - 12&1.77402 - 09&8.15464 - 09&3.90994 - 10 + & 9.76744 - 09&1.71966 - 09&1.62701 - 08&2.23134 - 10&4.46580 - 09&2.01162 - 08 + & 2.12072 - 13&2.08985 - 10&2.52089 - 13&9.16858 - 11&4.83231 - 10&2.06072 - 13 + ' '' '' 30&7.03229 - 13&3.23564 - 09&3.48502 - 10&1.58600 - 09&6.40590 - 09&1.91930 - 09 + & 9.20136 - 09&1.66336 - 09&1.52695 - 08&2.20462 - 10&4.30674 - 09&1.87965 - 08 + & 4.02125 - 14&1.93227 - 10&3.92723 - 14&8.47762 - 11&4.45482 - 10&2.44720 - 14 + ' '' '' 31&2.25530 - 10&2.55031 - 09&1.56742 - 09&1.39923 - 09&4.83362 - 09&4.56897 - 09 + & 8.64555 - 09&1.60619 - 09&1.42889 - 08&2.17653 - 10&4.14534 - 09&1.75061 - 08 + & 3.50986 - 15&1.76973 - 10&2.14434 - 15&7.79962 - 11&4.06460 - 10&6.19847 - 16 + ' '' '' 32&9.47179 - 10&1.93395 - 09&3.63624 - 09&1.21631 - 09&3.45843 - 09&8.28636 - 09 + & 8.10132 - 09&1.54827 - 09&1.33308 - 08&2.14717 - 10&3.98193 - 09&1.62483 - 08 + & 2.96228 - 17&1.60594 - 10&1.30619 - 18&7.14081 - 11&3.67106 - 10&_6.43053 - 22 _ + ' '' '' 33&2.15110 - 09&1.39400 - 09&6.51622 - 09&1.03962 - 09&2.29718 - 09&1.30018 - 08 + & 7.56986 - 09&1.48972 - 09&1.23972 - 08&2.11664 - 10&3.81682 - 09&1.50259 - 08 + & _ 1.61120 - 18_&1.44399 - 10&_8.42415 - 17_&6.50600 - 11&3.28216 - 10&_9.77929 - 16 _ + ' '' '' 34&3.81628 - 09&9.36463 - 10&1.01577 - 08&8.71285 - 10&1.36318 - 09&1.86316 - 08 + & 7.05225 - 09&1.43066 - 09&1.14900 - 08&2.08502 - 10&3.65033 - 09&1.38415 - 08 + & _ 1.13931 - 15_&1.28646 - 10&_6.74528 - 15_&5.89876 - 11&2.90443 - 10&_2.58717 - 14 _ + ' '' '' 35&5.91619 - 09&5.66105 - 10&1.45018 - 08&7.13237 - 10&6.66525 - 10&2.50806 - 08 + & 6.54948 - 09&1.37119 - 09&1.06111 - 08&2.05239 - 10&3.48274 - 09&1.26975 - 08 + & _ 1.69991 - 14_&1.13540 - 10&_6.50338 - 14_&5.32166 - 11&2.54311 - 10&_1.79155 - 13 _ + ' '' '' 36&8.41963 - 09&2.86540 - 10&1.94818 - 08&5.67193 - 10&2.14455 - 10&3.22434 - 08 + & 6.06241 - 09&1.31141 - 09&9.76201 - 09&2.01880 - 10&3.31435 - 09&1.15959 - 08 + & _ 9.47542 - 14_&9.92381 - 11&_3.00677 - 13_&4.77640 - 11&2.20222 - 10&_7.07043 - 13 _ + ' '' '' 37&1.12916 - 08&1.00402 - 10&2.50246 - 08&4.34706 - 10&1.17403 - 11&4.00066 - 08 + & 5.59182 - 09&1.25141 - 09&8.94405 - 09&1.98429 - 10&3.14540 - 09&1.05386 - 08 + & _ 3.33138 - 13_&8.58586 - 11&_9.53697 - 13_&4.26396 - 11&1.88475 - 10&_2.04574 - 12 _ + ' '' '' 38&1.44942 - 08&9.50400 - 12&3.10520 - 08&3.17201 - 10&6.10978 - 11&4.82499 - 08 + & 5.13837 - 09&1.19130 - 09&8.15844 - 09&1.94889 - 10&2.97616 - 09&9.52714 - 09 + & _ 8.96246 - 13_&7.34819 - 11&_2.40564 - 12_&3.78475 - 11&1.59273 - 10&_4.86340 - 12 _ + ' '' '' 39&1.79870 - 08&1.50127 - 11&3.74819 - 08&2.16020 - 10&3.63602 - 10&5.68475 - 08 + & 4.70265 - 09&1.13115 - 09&7.40620 - 09&1.91259 - 10&2.80687 - 09&8.56288 - 09 + & _ 2.02512 - 12_&6.21576 - 11&_5.20446 - 12_&3.33874 - 11&1.32739 - 10&_1.01023 - 11 _ + ' '' '' 40&2.17282 - 08&1.17626 - 10&4.42292 - 08&1.32480 - 10&9.19103 - 10&6.56687 - 08 + & 4.28513 - 09&1.07103 - 09&6.68820 - 09&1.87535 - 10&2.63774 - 09&7.64695 - 09 + & _ 4.04906 - 12_&5.19088 - 11&_1.00913 - 11_&2.92552 - 11&1.08927 - 10&_1.90285 - 11 _ + ' '' '' 41&2.56748 - 08&3.17761 - 10&5.12063 - 08&6.79419 - 11&1.72663 - 09&7.45785 - 08 + & 3.88623 - 09&1.01103 - 09&6.00513 - 09&1.83710 - 10&2.46899 - 09&6.78027 - 09 + & _ 7.39883 - 12_&4.27364 - 11&_1.80342 - 11_&2.54442 - 11&8.78341 - 11&_3.32982 - 11 _ + ' '' '' 42&2.97834 - 08&6.15751 - 10&5.83239 - 08&2.38956 - 11&2.78481 - 09&8.34371 - 08 + & 3.50627 - 09&9.51202 - 10&5.35756 - 09&1.79772 - 10&2.30079 - 09&5.96354 - 09 + & _ 1.26240 - 11_&3.46231 - 11&_3.02744 - 11_&2.19456 - 11&6.94082 - 11&_5.50531 - 11 _ + ' '' '' 43&3.40101 - 08&1.01205 - 09&6.54909 - 08&2.06235 - 12&4.09221 - 09&9.20992 - 08 + & 3.14550 - 09&8.91609 - 10&4.74592 - 09&1.75705 - 10&2.13334 - 09&5.19730 - 09 + & _ 2.04179 - 11_&2.75365 - 11&_4.83939 - 11_&1.87494 - 11&5.35593 - 11&_8.70631 - 11 _ + ' '' '' 44&3.83115 - 08&1.50746 - 09&7.26145 - 08&4.53133 - 12&5.64771 - 09&1.00413 - 07 + & 2.80410 - 09&8.32302 - 10&4.17050 - 09&1.71483 - 10&1.96678 - 09&4.48193 - 09 + & _ 3.16534 - 11_&2.14326 - 11&_7.44171 - 11_&1.58444 - 11&4.01657 - 11&_1.32946 - 10 _ + ' '' '' 45&4.26440 - 08&2.10340 - 09&7.95994 - 08&3.39363 - 11&7.45075 - 09&1.08215 - 07 + & 2.48219 - 09&7.73328 - 10&3.63149 - 09&1.67074 - 10&1.80124 - 09&3.81761 - 09 + & _ 4.74379 - 11_&1.62582 - 11&_1.10966 - 10_&1.32192 - 11&2.90809 - 11&_1.97515 - 10 _ + ' '' '' 46&4.69648 - 08&2.80213 - 09&8.63465 - 08&9.36947 - 11&9.50141 - 09&1.15327 - 07 + & 2.17982 - 09&7.14722 - 10&3.12897 - 09&1.62431 - 10&1.63682 - 09&3.20443 - 09 + & _ 6.91950 - 11_&1.19528 - 11&_1.61501 - 10_&1.08619 - 11&2.01385 - 11&_2.87344 - 10 _ + ' '' '' 47&5.12308 - 08&3.60718 - 09&9.27508 - 08&1.88336 - 10&1.18002 - 08&1.21546 - 07 + & 1.89699 - 09&6.56512 - 10&2.66292 - 09&1.57487 - 10&1.47356 - 09&2.64231 - 09 + & _ 9.87918 - 11_&8.45064 - 12&_2.30712 - 10_&8.76065 - 12&1.31556 - 11&_4.11729 - 10 _ + ' '' '' 48&5.53992 - 08&4.52364 - 09&9.86973 - 08&3.23954 - 10&1.43467 - 08&1.26630 - 07 + & 1.63367 - 09&5.98707 - 10&2.23322 - 09&1.52147 - 10&1.31142 - 09&2.13107 - 09 + & _ 1.38740 - 10_&5.68154 - 12&_3.25159 - 10_&6.90390 - 12&7.93492 - 12&_5.84403 - 10 _ + ' '' '' 49&5.94257 - 08&5.55860 - 09&1.04054 - 07&5.08820 - 10&1.71362 - 08&1.30265 - 07 + & 1.38976 - 09&5.41295 - 10&1.83969 - 09&1.46266 - 10&1.15021 - 09&1.67042 - 09 + & _ 1.92522 - 10_&3.57182 - 12&_4.54396 - 10_&5.28029 - 12&4.26414 - 12&_8.26844 - 10 _ + ' '' '' 50&6.32640 - 08&6.72146 - 09&1.08661 - 07&7.5410 - 10 & 2.0150 - 08&1.3201 - 07 + & 1.16513 - 09&4.84228 - 10&1.48207 - 09&1.3960 - 10 & 9.8944 - 10&1.2600 - 09 + & _ 2.65149 - 10_&2.04426 - 12&_6.33046 - 10_&3.8788 - 12 & 1.913 - 12&_1.1753 - 09 _ + ' '' '' 51&6.68625 - 08&8.02374 - 09&1.12302 - 07 & 1.074 - 09&2.333 - 08&1.312 - 07 + & 9.59611 - 10&4.27385 - 10&1.16003 - 09 & 1.317 - 10&8.28 - 10 & 8.99 - 10 + & _ 3.64153 - 10_&1.01729 - 12&_8.84945 - 10 _ & 2.689 - 12&6.24 - 13 & _ 1.69 - 09 _ + ' '' '' 52&7.01589 - 08&9.4768 - 09&1.1466 - 07 & & & + & 7.73011 - 10&3.7051 - 10&8.7314 - 10 & & & + & _ 5.01492 - 10_&4.0302 - 13&_1.2524 - 09 _ & & & + ' '' '' 53&7.30688 - 08&1.108 - 08&1.1516 - 07 & & & + & 6.05095 - 10&3.130 - 10&6.2 - 10 & & & + & _ 6.97509 - 10_&1.039 - 13&_1.82 - 09 _ & & & + ' '' '' 54&7.54571 - 08 & & & & & + & 4.55587 - 10 & & & & & + & _ 9.903 - 10 _ & & & & & + ( 140,50 ) ( -20,10 ) ( 0,0 ) + figs . [ fig:2 ] , [ fig:3 ] and [ fig:4 ] summarize the oscillator strengths for most of the considered transitions . for @xmath139 ( fig . [ fig:2 ] ) the oscillator strength has a strong variation along the band varying also with @xmath140 . the strong minima for @xmath141 around @xmath128 are due to the change of sign of the energy difference ( see fig . [ fig:1 ] ) . beyond that value , the initial state is lower than the final @xmath142 state and the strengths are negative . otherwise the strengths slowly increase with the vibrational excitation . the @xmath143 strengths are smaller by more than an order of magnitude . the minimum occurring around @xmath144 is here due to a change of sign of the matrix element appearing in . the @xmath145 strengths are smaller by more than an order of magnitude than the @xmath143 strengths . ( 140,55 ) ( -20,10 ) ( 0,0 ) + the @xmath146 oscillator strengths presented in fig . [ fig:3 ] do not vary much along the bands . they slowly decrease for @xmath141 and slowly increase for @xmath145 . they are remarkably flat for @xmath143 . ( 140,55 ) ( -20,10 ) ( 0,0 ) + the @xmath147 strengths presented in fig . [ fig:4 ] behave similarly to the @xmath148 strengths . minima take place around @xmath149 , @xmath150 and @xmath109 for @xmath141 , @xmath114 and @xmath115 , respectively . these minima thus occur now at increasing @xmath117 values with increasing @xmath151 and are all due to a change of sign of the matrix element . the lifetimes can be calculated as = ( _ e_f < e_i w_i f ) ^-1 . using the transitions probabilities of table [ tab:4 ] these are , for states with @xmath60 @xmath152 , @xmath153 and @xmath154 for @xmath71 , @xmath72 and @xmath73 , respectively . lifetimes for the calculated states are displayed in fig . [ fig:5 ] . ( 140,60 ) ( -20,7 ) ( 0,0 ) + applying the langrange - mesh method , the three - body schrdinger equation for the deuterium molecular ion was solved with high precision in perimetric coordinates . four vibrational states @xmath1 and @xmath3 form @xmath155 to @xmath156 were considered . with the same number of mesh points in the @xmath41 and @xmath42 coordinates @xmath157 and @xmath158 for all the sates considered , energies values are presented with 13 significant digits for the zero vibrational band until 9 significant digits for the third vibrational band . with the wave functions provided by the lagrange - mesh method , a simple calculation gives the quadrupole strengths and transition probabilities per time unit . extensive tables and some figures are presented with six significant digits . for high values of the total angular momentum @xmath3 and @xmath159 , the initial and final states are interchange : @xmath160 , @xmath161 and @xmath162 for @xmath163 , @xmath164 and @xmath165 , respectively . in comparison with the hydrogen molecular ion @xcite this values are @xmath166 , @xmath167 and @xmath168 for the same change in the vibrational bands . also in a consistent way , quadrupole transitions for the deuteron molecular ion are less favorable that those for the hydrogen molecular ion , except for some values of the total angular momentum . as a consequence , lifetimes for d@xmath0 are greater than those for h@xmath0 . the difference is around one order of magnitude except for the states in the ground - state rotational band , roughly @xmath169 and @xmath170 , respectively . i indebted to d. baye for initiating me into the subject and a. turbiner for suggesting the problem and for their interest . i would like to thanks the f.r.s .- fnrs ( belgium ) for a postdoctoral grant .

the strong spectral features near 2.2 m in early - type galaxies remain relatively unexplored . yet , they open a tightly focused window on the coolest giant stars in these galaxies a window that can be used to explore both age and metallicity effects . here , new measurements of k - band spectral features are presented for eleven early - type galaxies in the nearby fornax galaxy cluster . based on these measurements , the following conclusions have been reached : ( 1 ) in galaxies with no signatures of a young stellar component , the k - band index is highly correlated with both the optical metallicity indicator and the central velocity dispersion ; ( 2 ) in the same galaxies , the k - band fe features saturate in galaxies with while ( and ) continues to increase ; ( 3 ) [ si / fe ] ( and possibly [ na / fe ] ) is larger in all observed fornax galaxies than in galactic open clusters with near - solar metallicity ; ( 4 ) in various near - ir diagnostic diagrams , galaxies with signatures of a young stellar component ( strong , weak ) are clearly separated from galaxies with purely old stellar populations ; furthermore , this separation is consistent with the presence of an increased number of m - giant stars ( most likely to be thermally pulsating agb stars ) ; ( 5 ) the near - ir vs. or vs. diagrams discussed here seem as efficient for detecting putatively young stellar components in early - type galaxies as the more commonly used age / metallicity diagnostic plots using optical indices ( e.g h vs. ) . the combination of these spectral indices near 2.2 m with high _ spatial _ resolution spectroscopy from ground - based or space - based observatories promises to provide new insights into the nature of stellar populations in the central regions of distant early - type galaxies .

0710.5529

understanding the stellar content of early - type galaxies is fundamental to understanding their star formation and chemical evolution history . most early - type galaxies are too distant to resolve their individual stars with current technology , rendering the direct study of their stellar populations impossible . thus , their stellar populations must be studied using indirect methods . in recent decades , significant effort has gone into trying to better constrain the stellar contents for early - type galaxies using optical spectroscopic data . the most commonly studied features have been ca i h and k 0.38 @xmath0 m , h@xmath4 , mgb 0.52 @xmath0 m , fe @xmath0 m 0.53 , na 0.82 @xmath0 m , and cat 0.86 @xmath0 m . interpretation of all such spectral features is intrinsically complicated by their blended nature each feature is really the super - position of many spectral lines , usually from several different elements , blurred together by the line - of - sight velocity dispersion within each galaxy . there is no way to overcome this problem it must simply be taken into account during analysis . as population synthesis models have become more sophisticated and digital stellar libraries more complete , this problem has become more tractable over time . another challenge arises from the composite nature of galaxies : each observed feature is the luminosity - weighted integrated sum of that feature from all stars in the observed line - of - sight . naturally , luminosity - weighted does not imply mass - weighted . a relatively small fraction of the mass can dominate the observed luminosity and mask the underlying stellar population ( e.g. as happens during a starburst event within a pre - existing galaxy ) . even in relatively quiescent galaxies , light from stars at several important evolutionary stages contribute roughly equally to the observed spectral features between 0.4 1 @xmath0 m range . hence , a feature depth change could be due to ( e.g. ) a change near the ( mostly ) age - driven main - sequence turnoff or the ( mostly ) metallicity - driven red giant branch . the details can become quite complicated , as illustrated by the long standing controversy about whether observed changes in balmer line strength arise from the presence of younger main sequence stars , more metal - poor main sequence stars , or an extended horizontal giant branch ( for recent discussions of this debate , see maraston & thomas 2000 and trager et al . a similar controversy surrounds na 0.82 @xmath0 m feature : is it driven by metallicity - driven red giant branch changes , initial mass function related differences in the relative number of cool dwarf and giant stars or both ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? ? however , the properties of the rgb component can be isolated by observing in the k - band ( centered near 2.2 @xmath0 m ) . at those wavelengths , cool giants near the tip of the first - ascent red giant branch ( rgb ) dominate the integrated light in old ( @xmath5 3 gyr ) stellar populations . in combination with optical observations , k - band observations should facilitate the separation of msto and rgb light contributions . there are two possible complications to this scenario . first , a very young stellar population containing red supergiants will contribute a significant fraction k - band light . fortunately , such a population is obvious from the presence of h ii region emission lines at shorter wavelengths . second , a somewhat older population ( 1 2 gyr , i.e. an intermediate - age population ) may contain bolometrically bright carbon stars that can contribute a detectable amount of k - band light ( see discussions in silva & bothun 1998a , b ) . such a population may or may not be connected to increased h@xmath4 strength . initial development of these ideas can be found in silva et al . ( 1994 ) , mobasher & james ( 1996 ) , james & mobasher ( 1999 ) , mobasher & james ( 2000 ) , all of whom focused on the co 2.36 @xmath0 m feature . origlia et al . ( 1997 ) observed a si dominated feature at 1.59 @xmath0 m as well as co dominated features . these observational studies were limited by small - format detectors to relatively low resolving powers and/or small wavelength ranges per observation . in the cases of silva et al . and origlia et al . , only small , heterogeneous samples of galaxies were observed . a general conclusion of the james & mobasher studies was that changes in co strength between early - type galaxies in high - density and low - density regions were statistically consistent with different fraction contributions of intermediate - age agb light and hence galaxies in low - density regions had younger luminosity - weighted mean ages . origilia et al . argued that [ si / fe ] was super - solar in the four elliptical galaxies they observed . to further develop these ideas and investigate the usefulness of other k - band spectral indices in the study of early - type galaxies , new data have been obtained for eleven e / s0 galaxies in the nearby fornax cluster . only measurements in the central regions of these galaxies are discussed here . in section [ sec : data ] , the galaxy sample and its observations are discussed , while in section [ sec : proc ] the data processing methodology is described . the measurement of spectral feature strength is explained in section [ sec : lines ] while basic observation results are presented in section [ sec : results ] . the broader astrophysical implications of our observational results are discussed in section [ sec : disc ] . a summary is provided at the end . long - slit spectroscopic data obtained with isaac at the eso very large telescope ( vlt ) have been combined with data obtained with sinfoni at the vlt to study a small sample of early - type galaxies in the nearby fornax cluster . details about the sample as well as the instrumental setups for these two instruments are presented in this section . the full sample observed consists of eleven early - type galaxies in the nearby fornax cluster , sub - divided into six elliptical galaxies and five s0 galaxies . the overall sample was selected to cover a significant range in luminosity weighted age , metallicity and mass . basic galaxy properties are summarized in table [ tab : galsample ] . optical long - slit spectroscopy is available for all observed galaxies ( kuntschner 2000 , kuntschner et al . 2002 ) allowing comparison between the k - band line - strength indices derived in this paper with optical indices . when compared to simple stellar population models , three galaxies ( ngc1316 , ngc1344 and ngc1375 ) appear to have a significant contribution from young stars resulting in luminosity weighted ages between 1 and 3gyrs ( see figure [ fig : fornax_pops ] ) . the remaining galaxies are consistent with being dominated by old stellar populations with a range in average metallicity . multi - color hst advanced camera for survey images ( courtesy of a. jorden ) of all the galaxies in our sample were inspected for the presence of dust . three galaxies ( ngc 1316 , ngc 1344 , and ngc 1380 see details below ) have clear central dust features . the other galaxies in our sample have no visible dust feature within the investigated aperture . a few additional object specific remarks seem warranted : * _ ngc 1316 _ fornax a , s0(3)pec . this galaxy has large scale , irregular , central dust features ( based on inspection of hst - acs images ) . * _ ngc 1344 _ this s0 galaxy contains morphological shells in its outer regions @xcite as well as optical spectral signatures of a younger central population ( e.g. kuntschner et al . inspection of hst - acs images reveals patchy dust within the central 5 arcsec . in combination , these characteristics suggest ( 1 2 gyr ) dynamical interaction or merger event that triggered a central star formation episode . * _ ngc 1375 _ kuntschner ( 2000 ) concluded that the mean age of the central population of this s0 galaxy is significantly lower than for the other galaxies in our sample . * _ ngc 1380 _ contains a kinematically decoupled core ( kdc ) @xcite . it also contains an optically thick , well - organized central dust lane , easily visible in available hst images . * _ ngc 1381 _ this s0 galaxy contains x - bulge isophotal pattern , often attributed to a recent dynamical encounter or merger event . * _ ngc 1399 _ cd , e0 * _ ngc 1419 _ this e0 galaxy has anomalously bright k - band surface brightness fluctuations ( sbf ) relative to its i - band surface brightness fluctuations , suggesting the presence of extended giant branches , perhaps due to the presence of an intermediate - age population @xcite . * _ ngc 1427 _ this e4 galaxy also has anomalously bright k - band sbf ( see ngc 1419 sbf references above ) . it also contains a kdc @xcite . llcccccccl ngc 1316 & so(3)pec & 0.0059 & 226 & 132.2 & 90.5 & 50 & & 168 & fornax a ; sinfoni data + ngc 1344 & e5 & 0.0039 & 171 & 26.7 & 15.6 & 165 & 3.3 & 121 & type from rc3 + ngc 1374 & e0 & 0.0043 & 196 & 26.6 & 24.2 & 118 & 3.3 & 124 & kdc + ngc 1375 & s0(cross ) & 0.0025 & 70 & 24.0 & 9.3 & 91 & 3.0 & 48 & central starburst + ngc 1379 & e0 & 0.0044 & 138 & 23.9 & 23.3 & 182 & 2.9 & 78 & + ngc 1380 & s0/a & 0.0063 & 213 & 52.4 & 26.6 & 7 & 3.4 & 213 & + ngc 1381 & s0(9)(boxy ) & 0.0057 & 169 & 23.5 & 8.0 & 139 & 2.9 & 135 & + ngc 1399 & e0 & 0.0048 & 360 & 134.1 & 120.8 & 110 & & 111 & cd ; sinfoni data + ngc 1404 & e2 & 0.0065 & 221 & 26.0 & 23.0 & 157 & 157 & 280 & + ngc 1419 & e0 & 0.0071 & 128 & 9.4 & 9.3 & 188 & 1.2 & 111 & k - band sbf outlier + ngc 1427 & e4 & 0.0046 & 186 & 39.2 & 27.1 & 76 & 4.9 & 104 & k - band sbf outlier , kdc + a small sample of galactic open cluster stars were also observed . their basic properties are summarized in table [ tab : starsample ] . cluster parameters ( age , metallicity ) and star photometry come from houdashelt , frogel , & cohen ( 1992 ) . age has units of gyr . spectral types ( spt ) have been assigned based on their @xmath6 colors following @xcite for k giant stars and @xcite for m giant stars . these stars have several roles : velocity dispersion templates , co 2.36 @xmath0 m index comparisons , and small independent framework for interpreting k - band line strength variations as functions of basic atmospheric parameters . each of these roles will be discussed in turn below . llllccl ngc 2204 & 2.8 & 0.38 & 1146 & 5.71 & 1.12 & m4 + & & & 3304 & 3.10 & 0.79 & k3 + & & & 3325 & 3.84 & 0.94 & m0.5 + & & & 4132 & 5.03 & 1.11 & m3.5 + ngc 2477 & 1.0 & 0.02 & 1069 & 4.07 & 0.95 & m1 + & & & @xmath7 & 2.96 & 0.71 & k3 + & & & 2117 & 4.18 & 0.96 & m2.5 + & & & 6053 & 2.81 & 0.68 & k2 + ngc 2506 & 3.5 & 0.52 & 2401 & 3.65 & 0.91 & k5 + & & & 4228 & 4.07 & 0.99 & m1 + finally , several bright hot stars were observed to measure and correct the telluric absorption features in the k - band . b stars were selected by color from the hipparcos tycho catalog ( esa , 1997 ) . in the spectral range of interest , these stars have no spectral features ; therefore , that any observed features must be telluric absorption lines . in addition to color , proximity to the target objects and brightness were used as secondary selection criteria . the actual stars observed are listed in table [ tab : telluric ] . the _ v _ magnitude and _ b v _ color are from the tycho catalog ( esa , 1997 ) . llcc tycho2 7034 01323 & fornax galaxies & 6.07 & 0.11 + tycho2 5942 02406 & ngc 2204 & 5.29 & 0.19 + tycho2 7645 03250 & ngc 2477 & 6.13 & 0.17 + tycho2 5425 00772 & ngc 2506 & 7.87 & 0.14 + most of the observations discussed in this paper were obtained at the eso very large telescope using the isaac near - ir imaging spectrometer mounted at one of the nasmyth foci of the 8.2 m ut1/antu telescope . galaxy observations were obtained in 2002 january and 2002 november . a summary observations log is given in table [ tab : obslog ] . for each galaxy , the following information is provided : ut date of observation , integration time per frame , number of frames , and total integration time . since ngc 1380 and ngc 1404 were observed multiple times , each observation sequence is assigned a sequence identifier ( a , b , c , ... ) . llccccl 2002 jan 02 & ngc 1404 & 500 & 6 & 3000 & 89.45 & a + 2002 jan 14 & ngc 1404 & 500 & 6 & 3000 & & b + 2002 jan 14 & ngc 1404 & 500 & 6 & 3000 & & c + 2002 nov 12 & ngc 1404 & 500 & 6 & 3000 & & d + 2002 nov 13 & ngc 1404 & 500 & 6 & 3000 & & e + & & & & & + 2002 nov 13 & ngc 1380 & 400 & 8 & 3200 & 89.45 & a + 2002 nov 13 & ngc 1380 & 400 & 8 & 3200 & & b + 2002 nov 12 & ngc 1380 & 400 & 8 & 3200 & & c + & & & & & + 2002 nov 12 & ngc 1344 & 400 & 8 & 3200 & 164.45 & + 2002 nov 13 & ngc 1374 & 400 & 8 & 3200 & 89.45 & + 2002 nov 12 & ngc 1375 & 400 & 8 & 3200 & 89.45 & + 2002 nov 13 & ngc 1379 & 400 & 8 & 3200 & 89.45 & + 2002 nov 13 & ngc 1381 & 400 & 8 & 3200 & 138.45 & + 2002 nov 12 & ngc 1419 & 400 & 8 & 3200 & 89.45 & + 2002 nov 12 & ngc 1427 & 400 & 8 & 3200 & 77.45 & + cccccc 2004 nov 25 & ngc1399 & 300 & 4@xmath84 & 1200 & b + 2004 nov 26 & ngc1399 & 300 & 4@xmath84 & 1200 & c + 2004 nov 26 & ngc1399 & 300 & 4@xmath84 & 1200 & d + 2004 nov 26 & ngc1399 & 300 & 4@xmath84 & 1200 & e + & & & & & + 2005 oct 13 & ngc1316 & 600 & 3@xmath81 & 1800 & a + 2005 oct 13 & ngc1316 & 600 & 4@xmath82 & 2400 & b + isaac was used in short - wavelength spectroscopic mode at medium - resolution ( sws - mr ) . in this mode , the array used is a 1024@xmath91024 hawaii rockwell hgcdte array with a spatial plate scale of 0148 pix @xmath10 . a 120@xmath91 slit was used . the dispersion and spectral resolution measured from our arc lamp data were 1.21 pix@xmath10 and 7.7 ( fwhm , @xmath11 ) , respectively . the approximate wavelength range was 2.12 2.37 @xmath0 m but the central wavelength was always corrected for object redshift to ensure that all relevant spectral features were observed . galaxy observations were obtained using the standard nod - on - slit mode . at the start of each observational sequence , the galaxy was centered on the slit near one end and a individual integration was executed . the galaxy was then moved approximately 60 towards the other end of the slit where two more integrations were executed . the galaxy was then returned to the original slit position where another integration was obtained . this abba pattern was repeated a number of times , resulting in multiple individual two - dimensional spectroscopic images . including time for telescope offsetting and array readout , each sequence took roughly one hour to execute . after each set of galaxy sequences was completed , telluric standard tycho2 7034 01323 was observed ( see table [ tab : telluric ] ) at five different positions along the slit . at each position , a short ( @xmath12 10 secs ) integration was obtained . at the end of the telluric integrations , a xear lamp - on / lamp - off sequence was obtained followed by a quartz lamp - on / lamp - off sequence . the spectrograph setup was not changed during the entire galaxy - standard - lamps sequence , ensuring that the grating did not move . thus , each star - lamp sequence had the same central wavelength as the corresponding galaxy sequence . this is critical for proper correction of telluric features , dispersion , and array pixel - to - pixel response . on various dates , stars from the following open clusters were observed : ngc 2204 ( 2002 november 12 ) , ngc 2477 ( 2001 december 23 ) , and ngc 2506 ( 2002 november 13 ) . the actual stars observed are listed in table [ tab : starsample ] . each star was observed at five ( 5 ) positions along the slit . at each position , a short ( @xmath12 10 secs ) integration was obtained . after each set of cluster star observations , a telluric standard was observed , followed by arc and flat lamp observations ( as described above ) . the exact telluric standards observed are listed in table [ tab : telluric ] . as part of the vlt calibration program , calibration data was also obtained during the day following each set of observations . these data include observations of quartz and arc lamps ( for flat - field and wavelength dispersion corrections ) as well as dark frames with the same detector integration times ( dits ) as the night - time science observations . for this project , the daytime quartz and arc lamps were not useful due to grating angle positioning inaccuracy . observations of ngc1316 ( fornax a ) and ngc1399 ( the central cd galaxy ) were obtained with sinfoni , a near - ir integral field unit ( ifu ) spectrograph mounted at the cassegrain focus of vlt ut4 ( yepun ) . a spatial scale of the 0.1@xmath13 scale was used , corresponding to a total field of view of approximately @xmath14 . the spectra cover the wavelengths between 1.95 2.4 @xmath0 m at 2.45 / pix . the spectral resolution is 6.1 fwhm ( @xmath15 ) . ngc1399 was observed 2004 november 25 and 29 during the sinfoni science verification program ( eisenhauer et al . 2003 ; bonnet et al . the ngc1316 data come from the eso science archive ( program 076.b-0457a , pi ralf bender ) . the original observations were made on 2005 october 13 . details of these observations are given in table [ tab : obssinfoni ] . ngc1399 was observed in a sequence ossoosso ( o = object integration , s = sky / background integration ) to enable accurate background correction . ngc1316 was observed with a slightly different pattern osooso . as with the isaac observations discussed above , telluric and velocity template standard stars were observed periodically . the former were hot stars while the latter covered the range k5iii to m1iii . data processing was done using a combination of shell scripts , eclipse @xcite , and iraf . as described above , each isaac observing sequence consists of a set of galaxy or cluster star observations followed immediately by associated telluric star and lamp exposures . the lamps exposures were processed first . the resultant flat - field and xear arc lamp frames were used in turn to process their associated telluric star and galaxy or cluster star sequences . a raw flat - field frame was constructed by subtracting the lamp - off exposure from the lamp - on exposure to remove thermal background and dark current . the difference frame was used to define the illuminated region of the isaac sw detector ( the entire detector is not illuminated in sws mode ) . the difference frame was trimmed to remove the unilluminated part . the trimmed frame was then collapsed in the spatial dimension into a one - dimensional vector . a one - dimensional polynomial was fit to this vector . this polynomial was then divided into every row of the original background - corrected flat - field frame frame . the resultant frame is suitable for removing pixel - to - pixel sensitivity variations . these flat - field frames obtained immediately after all telluric sequences are preferred to evening or morning flats since tests demonstrated that such so - called daytime calibrations did not accurately remove fringing . this is presumably due to grating re - positioning imprecision . the xear exposures were processed by first subtracting the arc - off frame from the arc - on frame , again to remove any background illumination not associated with the xear lamps . a two - dimensional wavelength dispersion correction was derived , using low - order polynomials in both dimensions . for each telluric absorption correction star sequence , the five individual frames were first trimmed to correspond to the illuminated part of the detector ( see above ) and then divided by their associated flat - field . a background@xmath8dark frame was determined by median combining the individual flattened exposures . this background frame was then subtracted from those exposures . the five flat - field and background corrected frames were summed into a single frame . a spectral tilt correction was determined from the resultant frame . spectral tilt and wavelength dispersion corrections were then applied to the summed , flat - field corrected frame . the five individual spectra were extracted and averaged into a single spectrum using a sigma - clipping algorithm for bad pixel rejection . this combined spectrum was used to produce two things : a normalized telluric absorption line correction spectrum and a instrument sensitivity correction . building the normalized telluric correction spectrum started by fitting a one - dimensional bi - cubic spline to the telluric star spectrum . care was taken to reject high outliers ( e.g. cosmic rays , bad pixels ) but not too many low outliers ( likely to be the telluric absorption features of interest ) . the derived vector was divided back into the original spectrum to normalize it and remove the continuum shape . this bi - cubic spline fit was then re - used to determine an instrument sensitivity curve . since the color of the telluric stars is known from the hipparcos catalog , a reasonable estimate of their effective temperature can be made . furthermore , since these are hot stars , we know that their continuum in the k - band is well approximated by the rayleigh - jeans part of the blackbody spectrum associated with their effective temperature . by computing the appropriate blackbody curve in the relevant wavelength range and then dividing it by the bi - cubic spline fit ( i.e. a representation of the observed stellar continuum ) , the instrument sensitivity function can be estimated . of course , this is not an absolute correction but an absolute correction is not necessary for the spectral line strength measurements described below . cluster star sequences were processed into single combined spectra in a similar manner as telluric star sequences . at that point , the combined spectrum of each star was divided by its associated normalized telluric absorption line correction spectrum . to remove the instrument response signature , the telluric corrected cluster star spectrum is then multiplied by the instrument sensitivity function . to achieve the optimal telluric absorption line correction , it was typically necessary both to shift the telluric star spectrum in the dispersion direction ( usually by less than 1 pixel ) and scale it ( usually by a few percent ) before the division . the physical cause of the physical shift in the dispersion direction is unclear . however , the need to scale is simply due to the slight changes in airmass and time between object and telluric correction standard observations . it is also clear that the strength of every telluric absorption feature in this wavelength window is not changing in concert . thus , the telluric correction is not perfect at all wavelengths . in principle , a more sophisticated correction process is possible . in practice , the current level of correction is good enough for this application . the final processed spectra for all observed cluster stars are shown in figure [ fig : clusterstars ] . the main technical goal of this paper is to make accurate line strength measurements in the central regions of our target galaxies . however , in the future , the radial line strength gradients will also be studied . since each ab galaxy observation is only separated by @xmath16 100 , a simple ab subtraction ( i.e. the typical procedure in the near - ir ) could cause inaccurate dark and background correction in the lower surface brightness parts of the galaxy luminosity profiles . therefore , a procedure derived from the processing of _ optical _ long - slit spectra was applied . the basic data processing challenge is shown in figure [ fig : galprocexample ] where a central 3 diameter extraction of ngc1404 ( sequence d ; see table [ tab : galsample ] ) is shown at three stages : before background correction , after background correction but before telluric absorption line correction , and after all processing . one of the major advantages of isaac over other near - ir spectrometers is the ability to use a high - enough spectral resolution to resolve telluric features so that they can be accurately removed . as shown in table [ tab : galsample ] , each galaxy observation consists of a series of 6 or 8 individual exposures taken in abba sequences . to begin , each galaxy frame was trimmed and the appropriate median dark frame was subtracted . following the advice of the _ isaac data reduction guide _ ( available on - line from the eso isaac web page ) , electronic ghosting was removed using the eclipse recipe _ isaacp ghost_. recall that each galaxy sequence had an associated telluric@xmath8lamp sequence . as described in the telluric star processing section above , each telluric@xmath8lamp sequence was processed into a series of calibrations : flat - field , two - dimensional wavelength dispersion solution , spectral tilt correction , normalized telluric absorption correction vector , and instrument sensitivity function . after electronic ghost correction , each individual galaxy frame was divided by this flat - field and then geometrically rectified using the derived dispersion and spectral tilt solutions . in short , the galaxy sequence and its associated telluric star are calibrated using the same frames and vectors . a slit illumination correction was applied next . since twilight sky exposures were not available , the correction function had to be reconstructed from the actual observations . in each individual galaxy observation , the spatial dimension can be divided into a galaxy sub - section and a sky sub - section . for each ab pair , the sky sub - section is on opposite sides of the slit center . the a and b sky sub - sections were extracted and merged into a single a@xmath8b sky frame , after applying a small additive offset to the b sky sub - section to align it better with the a sky sub - section . this offset is caused by temporal changes in the background between the a and b exposures . this process was repeated for all a and b galaxy exposures on a given night . the resultant frames were then scaled to a common mean value and averaged using a sigma clipping bad pixel rejection scheme . that average sky frame was collapsed in the dispersion direction into a one - dimensional vector . a nightly illumination correction function was derived by fitting a second - order bi - cubic spline to that vector and normalizing the result . this illumination correction was then divided into all galaxy frames from that night . the next step was to remove the background . the standard procedure would be to select a radial position dominated by the background , determine the mean background at this position , and subtract that value at every pixel in the spatial dimension . as applying this procedure to the current dataset did not produce satisfactory results , an alternative two - step approach was developed . first , for each ab pair , an illumination corrected a@xmath8b sky frame was produced as described above and then subtracted from the parent a and b frames . second , on each frame , the standard procedure was applied . this two - step procedure produced a much better background correction than the standard procedure alone . each background corrected frame was then corrected for telluric absorption lines , using the associated normalized telluric absorption correction vector . as described in the cluster star processing section above , small shifts in the dispersion direction as well as small multiplicative scaling were necessary to achieve the best possible correction . finally , the six ( or eight ) individual frames combined into a single frame before this combination , the individual frames were shifted to a common center ( defined by the maximum intensities in the galaxy spectra ) with sub - pixel accuracy using linear interpolation . given the pixel scale ( 0148 pix@xmath10 ) vs. the natural seeing ( estimated to be @xmath16 08 from the profiles of the telluric star observations ) , alignment accuracy is not critical . the frames were then scaled to a common median value at a fixed point in the galaxy spectra and then averaged using a sigma - clipping rejection scheme . the final step was to multiple each combined frame by the sensitivity function derived from its associated telluric standard ( see above ) . basic post - processing data quality can be assessed from figure [ fig : galcenters ] shows the central 3 diameter extraction for all the galaxies observed in 2002 november . the spectra are ordered by velocity dispersion ( @xmath1 ) which corresponds essentially to decreasing signal - to - noise per pixel due to decreasing central intensity . the sharper spectral features seen in the galaxies with lower @xmath1 are intrinsic to the galaxies they are not caused by less accurate telluric line correction in these fainter galaxies . to first order , measurement repeatability can be assessed from figure [ fig : repeatcenters ] where the central 3 diameter extraction of ngc1380 and ngc1404 are shown from two and three different nights , respectively . in the case of ngc1404 , these observations were separated by months . clearly , the gross properties of the extracted spectra are the same . repeatability of line strength measurements is discussed in the next section . the central spectra used for the measurement of spectroscopic features were extracted within a radius corresponding to 1/8 r@xmath17 at the observed position angle for each galaxy ( see also tables [ tab : galsample ] and [ tab : obslog ] ) . for each extracted spectra , an empirical signal - to - noise ratio ( snr ) was derived following stoehr et al . ( 2007 ) and listed in table [ tab : galsample ] . from these extracted spectra , recession velocities and velocity dispersions were measured using the penalized pixel - fitting method ( ppxf ) developed by cappellari & emsellem ( 2004 ) . for simplicity , purely gaussian line shapes were assumed . the resulting measurements are needed as input for the line - strength determinations ( see section [ sec : lines ] ) . the velocity dispersions are tabulated in table [ tab : galsample ] and are used throughput this paper . errors on the kinematics parameters are calculated within the code based on error frames which are produced by our data reduction scripts . see also section [ sec : proc - optical ] for further analysis of the kinematic errors . sinfoni observations of ngc 1316 and ngc 1399 were processed and calibrated using the sinfoni pipeline ( v. 1.6 ) . the pipeline transforms raw science data frames into flat - field corrected , wavelength calibrated and background subtracted three - dimensional data cubes for the galaxies and the calibration stars , including the extraction of one - dimensional spectrum from the central 10 pixels of the observed galaxies and stars . for ngc1399 , the telluric stars were solar - type . in order to obtain a telluric correction spectrum , the star spectra were divided by a scaled and gaussian broadened solar spectrum . this step removes intrinsic stellar features and continuum shape . a correction for telluric features was then applied as described above for the isaac observations . the ngc1316 telluric stars were hotter b stars that have no prominent spectral features in the wavelength range of interest . the telluric correction spectrum was obtained by dividing the stellar spectrum by a scaled black body spectrum with the same temperature as the star . a correction for telluric features was then applied as described above for the isaac observations . the individual data cubes for the two galaxies were then combined to achieve better signal - to - noise ratios . the four ngc1399 data cubes were averaged using a sigma - clipping pixel reject algorithm . since there were only two ngc1316 observations , pixel reject during combination was not possible . since the field - of - view of the vlt - sinfoni observations was limited to approximately 3@xmath93 , an extraction radius of 1/8 r@xmath18 is not possible . therefore , a slit of 2@xmath93 was extracted from the datacube to match the slitwidth of the optical data for these galaxies ( see section [ sec : proc - optical ] ) . as for the isaac data , the recession velocities and velocity dispersions were measured with ppxf ( see also table [ tab : galsample ] ) and errors determined from noise statistics . the long - slit spectroscopy from kuntschner ( 2000 ) and kuntschner et al . ( 2002 , ngc1344 ) was used to extract a central spectrum for each galaxy . the extraction aperture was set to match the effective area covered on the sky by the central isaac data . the position angle of the isaac observations was set to agree with the optical data . due to the rather different slit widths involved ( isaac 1 , optical fornax data 23 and 2 ) the extraction ranges vary . for the two galaxies observed with vlt - sinfoni ( ngc1316 and ngc1399 ) the extraction aperture is limited by the field - of - view of sinfoni . in order to best match the optical and near - ir data we extracted a central 23 @xmath9 3slit from the optical observations . again ppxf was used to derive recession velocities and velocity dispersions . good fits were achieved for all galaxies but ngc1375 where the formal result was @xmath19@xmath3 . this is well below the resolution limit of the optical fornax data ( kuntschner 2000 ; 4.1 fwhm , r @xmath20 ) and is therefore unreliable . in this paper , the velocity dispersion derived from the higher spectral resolution isaac data was adopted and assigned a relatively large error of @xmath2120@xmath3 to reflect the uncertainty of this procedure . a comparison between optical and near - ir measurements of the central velocity dispersions is shown in figure [ fig : sigmas ] . the determinations are in good agreement with each other for all galaxies , including the two galaxies with significant contributions of young stellar populations ( ngc1344 and ngc1316 ) . taking the difference between the optical and near - ir measurements as a guideline for our true errors ( including differences in seeing and remaining aperture differences ) we conclude that our internal velocity dispersion error needs to be scaled by a factor of 2.2 . in the following analysis and figures , these conservative scaled errors are used . the spectral features investigated are listed in table [ tab : features ] . for all but one feature , absorption - line strength is measured as equivalent width via indices , where a central feature bandpass is flanked to the blue and red by pseudo - continuum band - passes . this method is identical to what is used to measure optical lick indices ( worthey et al . the index definitions for and were taken from frogel et al . revised , , and index definitions were used since earlier definitions ( e.g. frster - schreiber 2000 ) did not incorporate continuum band definitions optimized for galaxy spectra with significant velocity broadening . in order to measure the strength of the @xmath22co(2,0 ) absorption we use the definitions of frogel et al . 2001 which provide four pseudo - continuum bandpasses to the blue of the co feature and none to the red . therefore the continuum used to derive the equivalent width is an extrapolation from the blue wavelength region in contrast to the lick indices where the continuum is based on an interpolation between two flanking bandpasses . the feature bandpasses of all measured indices are illustrated in figure [ fig : features ] which shows how the absorption features change as a function of temperature in red giant stars . the spectra shown ( taken from wallace & hinkle 1997 ) are similar in spectral resolution to our spectra ( see figure [ fig : clusterstars ] above ) . it is important to remember that all of these features are really blends of distinct atomic and molecular features , especially at cooler temperatures . this point is richly illustrated by the r @xmath16 40 000 spectra presented in wallace & hinkle ( 1996 ) ( see , in particular , their figures 2 and 3 for the and features , respectively ) . lccccl nai & 21910 21966 & 22040 22107 & 22125 22170 & & frogel et al . 2001 + feia & 22133 22176 & 22250 22299 & 22437 22497 & & this paper@xmath23 + feib & 22133 22176 & 22368 22414 & 22437 22497 & & this paper@xmath23 + cai & 22450 22560 & 22577 22692 & 22700 22720 & & frogel et al . 2001 + mgi & 22700 22720 & 22795 22845 & 22850 22874 & & this paper@xmath23 + index & & feature bandpass & units & source + @xmath22co(2,0 ) & & 22910 23020 & & frogel et al . 2001 + in order to allow a meaningful comparison between observations of stars and galaxies ( with different internal dynamics ) to themselves and to population models , index measurements need to be carefully calibrated to a common system . there are generally three effects to account for : ( a ) the differences in spectral resolution of the instruments used to obtain the index measurements ; ( b ) the internal velocity broadening of galaxies and ( c ) possibly small systematic offsets caused by continuum shape differences ( see also kuntschner 2000 ) . the resolution of our isaac observations ( 6.9 fwhm , or r @xmath24 ) was adopted as the nominal resolution . the vlt - sinfoni spectra were broadened to match this resolution before the measurement of the line - strengths indices . the observed spectrum of a galaxy is the convolution of the integrated spectrum of its stellar population(s ) by the instrumental broadening and line - of - sight velocity distribution ( losvd ) of the stars . these effects typically broaden the spectral features , in general reducing the observed line - strength compared to the intrinsic values . in order to compare the index measurements for galaxies we calibrate the indices to zero velocity dispersion and our nominal resolution . although non - gaussian deviations from the losvd can have significant effects on the losvd correction of indices ( see e.g. , kuntschner 2004 ) , in this study , only the first moments , @xmath25 and @xmath1 , are considered . this is a reasonable approximation for central spectra studied in this study since expected non - gaussian deviations in this region are negligible . observations of k and m giants are used to determine the corrections for each index . these spectra are a reasonable match to the observed galaxy spectra ; however , once stellar population models become available the corrections should be re - derived . by broadening the stellar spectra to velocity dispersions of up to 400@xmath3 in steps of 20@xmath3 a median correction factor is computed for each index , i.e. : @xmath26 and @xmath27 is the median index measured from the stars convolved with the losvd given in brackets . a losvd corrected index is then @xmath28 . the actual correction functions computed here are shown in figure [ fig : veldispcorr ] . for the optical data the velocity dispersion corrections from kuntschner ( 2004 ) were used . the errors on the line - strength indices are evaluated by using error estimates from the spectra and taking into account the errors from the losvd correction procedure . the optical data was calibrated to the lick / ids system . for a more detailed description of the procedure see kuntschner ( 2000 ) . multiple isaac observations of ngc1380 and ngc1404 allow us to constrain possible systematic errors in our line strength measurements . figure [ fig : repeatability ] is plotted with the same y - axis range as figure [ fig : gallinesigma ] but with an smaller @xmath1 range . for the , , and indices our error estimate are comparable to the observed repeatability uncertainty . however , in the case of the index the repeatability uncertainty is significantly larger and we adopt the internal dispersion of the ngc 1380 and ngc 1404 measurements ( 0.3 ) as the formal uncertainty for all subsequent analysis . line - strength indices can be sensitive to continuum shape calibration errors . in particular index definitions spanning a wide wavelength range are prone to suffer from this . in our study the wide band - pass of the co index ( see table [ tab : features ] ) falls in this regime and we consider this to be the explanation for the larger spread between repeat observations . the origin and behavior of the , , and features as a function of effective temperature in k and m giant stars has been extensively reviewed and discussed by ramrez et al . ( 1997 ) and frster schreiber ( 2000 ) ( who also studied the fe and mg features ) . most of the stars used in those studies were field stars with bright apparent magnitudes and hence drawn from a narrow metallicity range biased towards solar metallicity . the behavior of the , , and features as a function of metallicity for galactic globular cluster k and m giants has been investigated by frogel et al . the metallicity range of that study covers both the metallicity range of the clusters investigated here , as well as the luminosity - weighted mean metallicities of the fornax early - type galaxies . in figure [ fig : starlinecolor ] , the index strength ( in ) for all six measured spectral features is plotted against @xmath6 color as tabulated by houdashelt , frogel , & cohen ( 1992 ) ( hereafter hfc92 ) for the open cluster stars listed in table [ tab : starsample ] . these stars are all k or m giant stars . in this context , redder @xmath6 implies cooler effective temperature and lower surface gravity . within a single cluster , it is reasonable to assert that this lower effective temperature is not metallicity - driven . as previously shown in the studies discussed above , all features increase in depth with redder color ( and hence decreasing @xmath29 ) except , which appears to have a more or less constant value for each cluster regardless of effective temperature ( see also frster schreiber 2000 ) . the , , and features are highly correlated with @xmath6 and show no dependency on cluster metallicity . for the co feature , this behavior is consistent with the results of frogel et al . ( 2001 ) over a similar cluster metallicity range . no analogous study exists for the fe features . however , as shown in figure [ fig : starfe ] , the and index strengths are highly correlated with each other . hence , it is convenient to define a mean index : = ( fei a + fei b ) / 2 that has a smaller formal uncertainty than either individual fe index alone . this mean index is used throughout the rest of this paper . as anticipated by the results of frogel et al . ( 2001 ) , at a given @xmath6 ( i.e. @xmath29 ) , giant stars in more metal - rich clusters have stronger and features . the mg feature ( not included in the frogel et al . study ) appears to have similar behavior , although we note again that within a given cluster ( and hence for a given metallicity ) , the mg feature strength is approximately constant with @xmath6 . in figure [ fig : starlineline ] , various combinations of k - band line indices are shown . , and are highly correlated with each other ( linear correlation coefficient , @xmath30 0.9 ) . there is a hint that is stronger in the solar - metallicity stars at equivalent and than in the more metal - poor stars . it is much clearer that at equivalent and , is stronger in the solar - metallicity stars than in the more metal - poor stars . in figure [ fig : gallinesigma ] , central k - band spectral feature measurements are plotted against central velocity dispersion measured from our data . has not been shown here ( and is ignored for the rest of the paper ) because this line is relatively weak and a reliable velocity dispersion correction could not be derived . relative to the stellar spectral line strengths shown in figure [ fig : starlinecolor ] , the galaxy measurements ( except ) fall in the middle of the ranges of the stellar measurements . taken at face value , this is consistent with saying that the luminosity - weighted spectral type for all these galaxies is similar to a solar - metallicity m0 iii star . for most of our galaxies , the feature is stronger than measured in any of our observed stars or any of the field disk giants observed by ramirez et al . only the supergiants in the fster schreiber have such strong na features . the unlikely possibility of a supergiant - dominated population can be ruled out by index values that are completely compatible with first - ascent giant stars . the feature is too strong for these galaxies to be dwarf dominated . at equivalent @xmath1 , the galaxies containing younger populations are separated from purely old galaxies . in the galaxies with purely old populations , is highly correlated with @xmath1 , while and @xmath1 are less well - correlated . in contrast , neither nor are well - correlated with @xmath1 reaches a maximum value for @xmath31 while has significant scatter at any given @xmath31 . and in purely old galaxies if ngc 1399 is removed from consideration . however , there is no sound reason for such data filtering at this time . obviously , a larger sample is needed to investigate this issue . ] figure [ fig : gallinemgfeprime ] shows the relationship between the optical index and our new k - band index measurements . is an abundance ratio insensitive optical index combination in the lick system , designed to track total metallicity @xcite . it is known to be well correlated with @xmath1 ; hence , it is not surprising that figure [ fig : gallinesigma ] and figure [ fig : gallinemgfeprime ] look very similar for the purely old galaxies . again , the galaxies that contain a young population are separated from the purely old galaxies at equivalent and @xmath1 . figure [ fig : gallineline ] shows the relationships between various pairs of k - band spectral indices ( compare to figure [ fig : starlineline ] ) . this figure captures several important observational conclusions already foreshadowed above : * relative to the observed cluster stars , the measured values of are significantly stronger in the galaxies . * in all cases , the galaxies that contain signatures of younger populations ( i.e. stronger and weaker ) have stronger and fe features than comparable purely old galaxies . * saturates circa @xmath32 150 @xmath3 ( @xmath16 3.2 ) in the same galaxies . no self - consistent theoretical spectral synthesis models for the interpretation of integrated k - band spectra of early - type galaxies are widely available . development of such models is on - going ( e.g. marmol queralt , 2007 , in preparation ) . until such models exist , we must rely on basic ( and perhaps imperfect ) astrophysical intuition and available tools , such as the high - resolution near - ir spectral atlas of wallace & hinkle ( 1996 ) ( hereafter wk96 ) . three galaxies in our sample ( ngc 1316 , ngc 1344 , and ngc 1375 ) have stronger and weaker features than the rest of our sample ( see figure [ fig : fornax_pops ] ) . this suggests a simple , two component stellar population model . one stellar component is cold in the luminosity - weighted mean , it is presumably old ( @xmath33 8 gyr ) and metal - rich ( [ fe / h ] @xmath33 0.3 ) with spectral properties consistent with the observed central velocity dispersion . in other words , it has properties similar to the purely old galaxies in our sample ( see figure [ fig : fornax_pops ] ) . the other component is warm it contains a significant number of a / f dwarf or sub - giant stars . this component could have several origins but only two are considered here . each has different consequences for k - band spectral index behavior . first , the warm component could be associated with a young population with a warm ( @xmath162 ) main sequence turnoff ( msto ) . the msto stars are tied to thermally pulsating asymptotic giant branch ( tp - agb ) stars with bolometric magnitudes that place them above the tip of the first - ascent red giant branch ( trgb ) . the tp - agb stars are cooler than stars on the first - ascent rgb and hence have an m spectral type . in the underlying luminosity function ( spectral type vs. number of stars ) , the m - star bins are relatively more populated than in a purely old galaxy . the net effect is that integrated 2.2 @xmath0 m spectra should become more m - like . second , the warm component could be associated with a metal - poor population with a warm msto . these msto stars are tied to rgb stars with similar luminosity as the corresponding metal - rich rgb stars but with warmer . in the underlying luminosity function , the k - star bins will be relatively more populated than in a purely old population . the net effect is to make the integrated 2.2 @xmath0 m spectra more k - like . can the observations presented here distinguish between these two scenarios ? consider figure [ fig : galmodel ] the galaxies with relatively strong and weak have relatively stronger ( i.e. more m - like ) near - ir features . qualitatively , this is consistent with the first scenario . can this conclusion be better quantified ? what is the relative ratio of young to old stellar mass ? are the changes in optical and near - ir spectral features _ quantitatively _ consistent ? to begin to answer these questions , the thomas et al . ( 2003 ) models were used to create very simple two - component models that produced optical index values that matched the observed values in the three galaxies with strong . one component was old ( 11 gyr ) with metallicity set to [ z / h ] @xmath34 0.35 for ngc 1316 and ngc 1344 and 0.0 for ngc 1375 . the other component had the same metallicity but a young ( 1 gyr ) age . the relative mass fractions of these components were varied until the model strength matched the observed strength . the resultant young mass fractions ( where total mass @xmath34 1 ) were 0.135 , 0.075 , and 0.140 for ngc 1316 , ngc 1344 , and ngc 1375 , respectively . the implied differential correction between the purely old model and the two - component model was then applied to the observed data ( see figure [ fig : galmodel ] ) . next , observed strength was adjusted manually until all three galaxies lay within the locus of purely old galaxies with similar @xmath1 in the vs. panel . as part of this adjustment , @xmath35()/@xmath35 ( ) was forced to agree with the value ( 0.62 ) determined for the observed galactic cluster giant stars ( see figure [ fig : starlineline ] ) . this is a trend in effective temperature that is relatively insensitive to metallicity . cooler giant stars ( more late k and early m like ) have strong k - band spectral features . in addition to the @xmath35()/@xmath35 ( ) measured from the galactic open cluster stars , two other key numerical trends can be computed : @xmath36$^\prime$ } } ) & = & 0.70 ( 1.10 ) \\ \delta({\mbox{$<$fe\,{\small i}$>$}})/\delta({\mbox{[mgfe]$^\prime$ } } ) & = & 0.28 ( 0.44)\end{aligned}\ ] ] ngc 1316 and ngc 1344 could be forced to have consistent trajectories . however , ngc 1375 ( values shown in parenthesis ) appears to follow somewhat different trajectories . obviously , this kind of cartoon model is illustrative only and surely hides a plethora of details . indeed , the real trajectories are unlikely to be linear . nevertheless , a clear astrophysical conclusion emerges : for the -strong galaxies , the observed optical and near - ir features all change in concert , consistent with a warm msto tied to an extended rgb , which is in turn consistent with the presence of a young stellar component , not a metal - poor stellar component . liu , graham , & charlot ( 2002 ) found that relative i - band and k - band surface brightness fluctuation strength ( expressed as an sbf color , @xmath37 ) was essentially constant for the majority of the early - type galaxies they studied in fornax . however , two galaxies in common with our sample ( ngc 1419 and 1427 ) were found to have brighter ( larger ) k - band sbf relative to the mean relationship ( see also mei et al . a larger @xmath38 is thought theoretically to arise from the presence of an extended giant branch , i.e. more cool , bright , m giants . in the spectroscopic data discussed here , there are no indications of such extended giant branches in these galaxies their central line indices are consistent with an integrated stellar populations dominated by an old , metal - rich population . of course , the sbf measurements were made in the outer regions of these galaxies , as opposed to the optical and near - ir observations of the central regions discussed here . it may be that the integrated luminosity of such a component ( if present at all ) is not high enough in the central region for detection by our method . in purely old galaxies , we have seen that : * is stronger than in galactic open clusters and is highly correlated with @xmath1 and . * saturates for @xmath31 @xmath3 . * is somewhat correlated with @xmath1 and , while is not ( unless ngc 1399 is not included in the regression fits , see footnote 2 above ) . the observed k - band spectral features are named for the dominant elemental species in the _ solar _ spectrum . however , at the effective temperatures of interest here , the on - band index definition and their companion off - band continuum bands contain lines from other elemental species as well ( as ramirez et al . 1997 discuss comprehensively ) . by referring to the high resolution spectra of wk96 , the contribution from these other absorbers can be investigated and used to explain the overall spectral feature behavior . in turn , underlying astrophysical parameters related to galaxy formation and evolution are revealed . in the current galaxy sample , is significantly stronger than observed in the galactic open cluster stars . the off - bands defined for the feature are relatively line - free , so we can focus on the on - band spectral region ( see the @xmath7 dra , m0 iii spectrum in wk96 , p. 352 ) . in addition to sodium , sc , si , and ( to a lesser extend ) v are significant absorbers in this region . origlia et al . ( 1997 ) argued that [ si / fe ] is super - solar in a small sample of early - type galaxies based in measurements of the si i 1.59 @xmath0 m feature . trager et al . ( 2000 ) have argued that both [ na / fe ] and [ si / fe ] are super - solar in early - type galaxies based on models of optical spectral feature behavior . the k - band measurements presented here are consistent with enhanced silicon in the observed fornax galaxies and hence indirectly confirm the conclusions of origlia et al . and trager et al . the feature contains approximately equal absorption contributions from fe , sc and ti while is dominated by fe absorption features with some additional sc absorption ( in wk96 , see the @xmath7 dra , m0 iii spectra on p. 346 and 348 ) . hence , the index ( a combination of and ) is dominated by absorption lines from fe - peak elements ) . for purely old galaxies with @xmath39 150 @xmath3 , is observed to remain constant while ( an index dominated by absoprtion lines from @xmath40-elements ) and ( a total metallicity [ z / h ] indicator ) become stronger as @xmath1 increases . it is tempting to conclude that above @xmath39 150 @xmath3 luminosity - weighted mean total metallicity [ z / h ] continues to increase as a function of central velocity dispersion , driven by a relative increase of [ @xmath40/h ] , while [ fe / h ] remains constant . however , within the narrow [ fe / h ] range of observed galactic open cluster stars , is more strongly correlated with effective temperature than cluster [ fe / h ] ( see figure [ fig : starlinecolor ] ) . increasing mean total metallicity [ z / h ] in the galaxies should correspond to cooler mean rgb effective temperature and hence increased . recall that the relationship between and mean rgb effective temperature has already been exploited above to explain the observed behavior of ( and ) in galaxies with young stellar components the dilemma is clear : how can a temperature - sensitive index like remain constant while mean rgb effective temperature decreases due to increasing [ z / h ] ? without appropriate population synthesis models or stellar spectra , only a few speculative thoughts can be offered . in their detailed comparison of non - solar aboundance population synthesis models with measurements of spectral indices in the optical lick system , trager et al . ( 2000 ) concluded that their best - fit models included enhancements in c , n , na , and si ( among others ) . as [ @xmath40/h ] increases with @xmath1 ( as suggested by increasing ) , stronger cn bands in the and on - band and off - band windows ( see the wk96 spectra referenced above and the band definitions in table [ tab : features ] ) may have the effect of depressing the local continuum and hence decreasing the value of @xmath34 ( @xmath8 ) /2 . for example , trager et al . noted a similar effect from c@xmath41 bands in the optical mg _ b _ index . therefore , it may be that above some critical mean [ @xmath40/h ] and below some critical mean effective temperature , stays within a narrow range and provides no useful information about mean [ z / h ] , [ fe / h ] , or effective temperature . a conclusive astrophysical explanation of in early - type galaxies with @xmath39 150 @xmath3 awaits larger galaxy samples , observations of appropriate stars ( e.g. in the galactic bulge ) , and detailed population synthesis models . the feature is dominated by the @xmath42co(2,0 ) bandhead and has no significant absorption components from other elements . like above , is highly correlated with @xmath29 in the observed galactic open cluster stars and reaches larger values than observed in the galaxies . as a function of central velocity dispersion @xmath1 in the observed galaxies , does not appear to saturate ( or at least not as definitively as ) . this adds credence to the suspicion that is not tracing mean rgb effective temperature in galaxies with @xmath39 150 @xmath3 . based on the current measurements , no conclusions can be reached about relative carbon or oxygen abundance in these galaxies . no obvious explanation presents itself for the behavior of in this galaxy sample . the feature is intrinsically complex , consisting of contributions from many absorbers : ca , s , si , and ti ( all @xmath40-elements ) as well as sc and fe ( see @xmath7 dra , mo iii spectrum in wk96 , p. 342 ) . as a function of in the range of interest , some of the contributing absorption lines get stronger ( sc and ti , the former faster than the latter ) , some features get weaker ( si ) , and some remain roughly constant ( ca , s , and fe ) ( cf . ramirez et al . relative to solar abundance ratios , trager et al . ( 2000 ) have argued that si and s are over - abundant and ca is under - abundant in early - type galaxies while [ ti / fe ] and [ sc / fe ] have their solar values . in the galactic open cluster stars observed here , the feature is stronger at a given @xmath6 ( @xmath43 ) in the solar - metallicity stars than in the more metal - poor stars . within the galaxies observed here , appears to have significant scatter at any given @xmath1 or . no obvious explanation for this scatter ( or the global behavior of with @xmath1 and ) presents itself at this time . using new , moderate resolution ( @xmath44 ) k - band spectra , spectral indices have been measured in the central regions of eleven early - type galaxies in the nearby fornax cluster . based on these measurements , the following conclusions were reached : 1 . the feature is much stronger in these fornax early - type galaxies than observed in solar - metallicity galactic open cluster stars . this is attributed to relative [ si / fe ] ( and possible [ na / fe ] ) differences between the open cluster stars and the fornax galaxies , i.e. both are larger in the fornax galaxies than in the cluster stars . 2 . in various near - ir diagnostic diagrams , galaxies with optical indices indicative of a warm stellar component are clearly separated from galaxies dominated by colder , presumably old ( @xmath5 8 gyr ) stellar populations . changes in the near - ir spectra features are consistent with the presence of an cool component dominated by late k and/or early m giants stars . in combination , the optical and near - ir observations are consistent with the presence of a young stellar component with a warm msto and a significant extended giant branch consisting of tp - agb stars . 3 . for detecting a young stellar component , the vs. @xmath1 or vs. @xmath1 diagnostic diagram seems as efficient as using h@xmath4 vs. ( or other similar combinations of optical indices ) . the near - ir features have the additional advantage that no emission - line correction is needed ( as it is necessary for the stronger optical balmer lines ) . the index saturates in galaxies with central velocity dispersion @xmath1 @xmath33 150 @xmath3 dominated by old ( @xmath5 8 gyr ) stellar populations . for @xmath39 150 @xmath3 , these fe features are unlikely to be useful for investigating stellar population differences between early - type galaxies . above @xmath39 150 @xmath3 , the continued increase in ( and ) strength with @xmath1 coupled with constant presents an astrophysical challenge . although it is tempting to conclude that [ fe / h ] reaches a maximum value , while [ @xmath40/h ] ( and hence [ z / h ] ) continues to increase , more observational and population synthesis work is needed to understand conclusively the behavior of in these high - mass early - type galaxies . adding these near - ir indices to the standard diagnostic toolkit for analyzing the integrated light of early - type galaxies clearly has great potential to develop this potential , three obvious steps are needed : observe more galaxies over a larger range of central velocity dispersion ( including field galaxies with existing optical data ) and extend current population synthesis models to the near - ir . we know that various groups are working on both steps . it would also be useful to study the radial behavior of these indices within individual galaxies to compare and contrast index behavior between galaxies . for the foreseeable future , the study of the central populations in early - type galaxies will remain the study of integrated light . as distance increases , our ability to study the central regions of early - type galaxies within metric apertures equivalent to nearby galaxies relies on the high _ spatial _ resolution spectroscopy achievable from space or with adaptive optics on the ground . in the former case , the james webb space telescope represents the frontier yet , no optical spectrograph that works below 0.8 @xmath0 m is currently planned . in the latter case , the frontier will be shaped by further development of adaptive optics systems and these systems will achieve their best performance beyond 1 @xmath0 m . hence , understanding how to interpret near - ir spectral indices in nearby galaxies is key to facilitating the kind of investigation and characterization of more distant early - type galaxies that will not be possible at optical wavelengths . we thank the anonymous referee for a quick , enthusiastic , and helpful review that drove us to clarify several key points . 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 research has also made use of the simbad database , operated at cds , strasbourg , france . thanks to valentin ivanov and olivier marco for observing assistance during the acquisition of the isaac data . drs warmly thanks eso santiago for hospitality during july 2003 science leave , especially danielle alloin , during which much of these data were processed . we acknowledge various helpful discussions with andrew stephans , daniel thomas , claudia maraston , scott trager and guy worthey on topics related to stellar populations in early - type galaxies . thanks to chris lidman and rachael johnson for various discussions about reducing isaac spectroscopic data . many of the science and technology concepts in this paper grew out of late night conversations between drs and richard elston ( 1961 2004 ) friend , colleague , fellow observer while we were both postdocs at kitt peak . richard thanks .

the discovery of the nearest young supernova remnant rx j0852.0 - 4622 / gro j0852 - 4642 in the galaxy by rosat and comptel has been reported recently . age and distance are determined to years and pc by the x - ray diameter and the-ray line flux of radioactiveti . here we discuss the implications of the x - ray spectra and of the fact that 1.8 mev-ray line emission from the decay ofal has been measured from the vela region with a certain fraction possibly associated with the new snr . we estimate an uncertainty of the age of 100 yrs for a fixed yield ofti . the highest values ofti yield provided by current supernova explosion models give worst case upper limits of 1100 yrs for the age and of 500 pc for the distance . also the unknown ionization stage ofti adds to the uncertainty of age and distance which is at most another 35% on top . both the energy balance compiled for the remnant and yield predictions forti andal by supernova models favour a core - collapse event . two point sources have been found in the vicinity of the explosion center , either one of these might be the neutron star left by the supernova . if there is a neutron star the x - ray count rates of the two point sources provide an upper limit of the blackbody surface temperature , which is very unlikely to exceed 3 k. the supernova might have been observed some 700 150 yrs ago , but based on the data of sn 1181 , e.g. , there is a realistic chance that it has been missed if the supernova was sub - luminous .

astro-ph9909415

recently , we have published our discoveries of a previously unknown galactic supernova remnant ( snr ) ( aschenbach 1998 , iyudin et al . the x - ray image obtained in the rosat all - sky survey shows a disk - like , partially limb brightened emission region of 2@xmath7 in diameter , which is the typical appearance of a shell - like snr ( cf . [ picture ] ) . the pspc x - ray spectra reveal rather high temperatures of @xmath8 3@xmath510@xmath9 k , which indicate that rx j0852.0 - 4622 is a young object . combining the low age and the 2@xmath7 angular extent it is concluded that rx j0852.0 - 4622 is relatively close - by . comparison with historical snrs limits the age to about @xmath0 1500 yrs and the distance to @xmath101 kpc . the case of rx j0852.0 - 4622 being an snr was clinched by the detection of @xmath1-ray line emission from @xmath2ti , which is a titanium isotope exclusively produced in supernovae . the centre of the @xmath2ti source , called gro j0852 - 4642 , is off - set from the center of rx j0852.0 - 4622 by 0.4@xmath7 , but this is significantly less than the angular resolution of the comptel instrument , so that rx j0852.0 - 4622 and gro j0852 - 4642 are considered to be the same object . using a weighted mean lifetime of @xmath2ti of 90.4 yrs , the angular diameter and adopting a mean expansion velocity of 5000 km / s as well as a @xmath2ti yield of 5@xmath510@xmath11 m@xmath12 age and distance are uniquely determined to @xmath0680 yrs and @xmath0200 pc , respectively . therefore , rx j0852.0 - 4622/gro j0852 - 4642 could be the nearest supernova to earth to have occured during recent human history . the discovery of rx j0852.0 - 4622 and the interpretation as an snr was made by one of us ( ba ) in early 1996 . during the time which followed it was attempted to associate some fraction of the @xmath3al @xmath1-ray line emission from the vela snr region measured by comptel ( oberlack et al . 1994 , diehl et al . 1995 ) with rx j0852.0 - 4622 . the results have not been conclusive basically because of the unknown distance of rx j0852.0 - 4622 ( oberlack 1997 ) . the discovery of @xmath2ti @xmath1-ray line emission , however , made it clear that rx j0852.0 - 4622 is indeed a nearby object , so that we could take up again the discussion of the association of @xmath3al emission with rx j0852.0 - 4622 . for example the combination of just the @xmath3al and @xmath2ti data allow to derive a distance independent estimate of the age of gro j0852 - 4642 . furthermore , if a major fraction of the vela @xmath3al mass would be associated with the snr , a type ia supernova is excluded within the framework of current explosion models . under the assumption of adiabatic expansion ( sedov - like ) of the snr we give an estimate of the supernova explosion energy e@xmath13 related to the progenitor star and the ambient matter density n@xmath13 . the uncertainties in the determination of age and distance by exploiting the x - ray spectra are discussed to come up with a time span in which to search for the historical supernova event . in a recent paper chen & gehrels ( 1999 ) conclude that rx j0852.0 - 4622 was created by a core - collapse supernova of a massive star . their analysis is based on the x - ray data and @xmath1-ray data published published earlier by us ( aschenbach 1998 , iyudin et al . 1998 ) . we discuss their approach and conclusions in the relevant section . basis for the determination of the age and distance of the new snr is the law of radioactive chain - decay @xmath2ti @xmath14sc @xmath14ca , for which we have @xmath15 with the definitions : _ f _ photon flux density , _ d _ distance , _ y@xmath16 _ mass yield of the element a , _ m@xmath16 _ its atomic mass , @xmath17 its mean life time and _ t _ the age . for the @xmath2ti decay chain @xmath18 , so that we can neglect @xmath19 and the second exponential term of eq . 1 . the data available suggest @xmath20 90 yrs , which we adopt for the following . this value is also close to the result of ( 87.7 @xmath41.7 ) yrs recently published by ahmad et al . _ f _ is the flux of the 1.157 mev line which has been measured by iyudin et al . ( 1998 ) to ( [email protected])@xmath2110@xmath11 photons @xmath22 s@xmath23 . apart from the statistical error a systematic error should be added , which is estimated to @xmath4 20% . the range of @xmath24 , @xmath25 and _ y@xmath16 _ of eq . 1 can be constrained by introducing the x - results . the angular radius @xmath26 = 1@xmath7 is related to @xmath24 and @xmath25 by @xmath26 = @xmath27 / _ d _ , with @xmath28 the mean expansion velocity of the snr . by substituting @xmath24 or @xmath25 of eq . 1 by @xmath28 a quantitative relation between @xmath29 and @xmath25 or @xmath29 and @xmath24 can be derived with @xmath28 as a parameter . an estimate of @xmath28 can be obtained from the x - ray spectra . the analysis of the rosat x - ray spectra is affected by the presence of the low energy emission of the vela snr , which is aggravated by the large size of rx j0852.0 - 4622 . but an archival rosat pspc pointing observation centred on the southeastern limb of the vela snr with an exposure of 11000 s happens to contain the northern limb section of rx j0852.0 - 4622 . the number of counts is sufficient to extract a small section of the limb as well as a small section of the vela snr offset by just 10 arcmin to create both uniform source and background spectra . limiting the analysis to these small regions , each 10 arcmin @xmath5 10 arcmin in size , reduces the impact of any spectral and spatial non - uniformity across the source and the background . fits to the residual northern limb spectrum were performed with a two component , optically thin thermal emission equilibrium model ( raymond - smith model ) with kt@xmath30 = 0.21@xmath31 kev , kt@xmath32 = 4.7@xmath33 kev and an absorption column density of n@xmath34 = @xmath352.3@xmath3610@xmath37 @xmath22 . we note that the data can be fit equally well ( @xmath38 1 ) with a straight power law with index @xmath39 = 2.6@xmath40 and absorption column density n@xmath41 = @xmath35 1.2@xmath4210@xmath43 @xmath22 . the spectrum of the rest of the snr can be obtained only from the rosat all - sky survey data and because of the relatively low exposure the spatial region selected for analysis needs to be large which increases the uncertainty in assessing the background level from the vela snr . for the full remnant excluding the bright northern limb an acceptable fit with @xmath38 1 is obtained with a two temperature model with kt@xmath44 = 0.14@xmath45 kev , kt@xmath46 = 2.5@xmath33 kev and n@xmath47 = @xmath354.0@xmath4810@xmath37 @xmath22 . the matter density of the shock - wave heated snr plasma can be derived from the observed x - ray flux @xmath49 via the relation @xmath49 = @xmath50 . @xmath51 is the cooling function for the best - fit values of @xmath52 ; @xmath53 is the electron number density and @xmath54 is the number density of the un - shocked matter , intitially uniformly distributed in a sphere . furthermore , a factor of four has been used for the density jump at the shock . the low and high temperature components are associated with densities of @xmath55 = 0.6@xmath5d@xmath56 @xmath57 and @xmath58 = 0.06@xmath5d@xmath56 @xmath57 , respectively , with d@xmath59 measured in units of 200 pc . despite the acceptable spectral fit the value of @xmath55 and the column density are quite uncertain , as the low temperature component could be significantly affected by the vela snr radiation , which is , however , not the case for @xmath58 . as usual for thermal snrs two components with different temperatures are needed for equilibrium models to fit the observed spectrum . if the plasma is far from ionization equilibrium the low temperature component appears as an artifact because of the under - ionization . the time - scale to reach ionization equilibrium is about 10@xmath60 s @xmath57/n@xmath61 , with n@xmath61 the electron density of the radiating plasma in units of @xmath57 . with @xmath25 = 680 yrs and the densities given above 1.5@xmath510@xmath62 s @xmath63 n@xmath6410@xmath65 s @xmath57 , which demonstrates that rx j0852.0 - 4622 departs significantly from ionization equilibrium . clearly the high temperatures observed are closer to the real electron temperature . but even the high temperatures may underestimate the average temperature of the electrons and ions if the electrons are heated mainly by coulomb collisons with the ions , which occurs on a timescale similar to that of reaching ionization equlibrium . the x - ray temperatures which have been produced by shock wave heating can be used to estimate the velocity @xmath66 of the shock wave : @xmath67@xmath68 ; @xmath69 is the proton mass and @xmath70 is the mean molecular weight , which is 0.6 for a fully ionized plasma of cosmic abundances . again , this relation is for a density jump of a factor of four at the shock . discarding the low temperature components as argued above the x - ray temperatures stretch from 1.8 kev @xmath71 kt@xmath59 @xmath71 9.2 kev including @xmath41-@xmath72 errors . the mean kt@xmath59 , which is consistent at the same significance level with both the radiation from the bulk of the snr and its northern limb section , taking the thermal option , is kt@xmath59 = 4.4 kev . the corresponding best estimates of the minimal and maximal shock velocities using the relation above are 1940 km / s , 1240 km / s and 2800 km / s , respectively . the current shock velocity @xmath66 is related to the mean expansion velocity @xmath28 by the past temporal evolution of the snr . with the limited observations available we are forced to rely on what is known about historical remnants . the compilation of strom ( 1994 ) provides both maximal internal shock velocity and mean expansion velocity and the ratio @xmath73 is 1.5 for the crab nebula and cas a , 2.5 for sn 1006 and 3.5 for the kepler and tycho snrs . for a purely adiabatic expansion in a uniform medium of constant matter density ( the sedov description ) @xmath73 = 2.5 . as strom has pointed out the observed maximal internal velocities may not be strictly related to @xmath66 but they provide a reasonable estimate . more recently measurements of the expansion rate in the x - ray band have become available by comparing images obtained with the einstein and rosat observatories or even just the rosat images taken at different epochs . both koralesky et al . ( 1998 ) and vink et al . ( 1998 ) have found an expansion rate of cas a of 0.002% yr@xmath23 which corresponds to a factor of @xmath0 1.55 for the ratio of mean expansion rate over current expansion rate . hughes ( 1996 ) has found a similar value for the tycho snr . in each of these cases , however , the current expansion velocities , using reasonable distance estimates , are significantly larger than the x - ray spectra and temperatures indicate . therefore a factor of 1.5 for @xmath73 is a very conservative lower limit to estimate @xmath28 from x - ray spectra . for a worst case estimate we define a velocity range for rx j0852.0 - 4622 applying factors of 1.5 , 2.5 , 3.5 to the minimal , best - estimate and maximal @xmath66 , respectively , which leads to a best - estimate expansion velocity @xmath74 = 5000 km / s bracketed by a minimal expansion velocity of @xmath75 = 2000 km / s and a maximal expansion velocity @xmath76 = 10000 km / s ; the values have been rounded off slightly . since we do nt know whether the x - ray temperatures are associated with either the blast wave heated ambient medium or the progenitor ejecta heated by reverse shocks , the expansion velocities derived may even be lower limits . similarly , the values are too low if the electrons have not reached thermal equilibrium with the ions . for the discussion of the impact of the expansion velocity on age and distance we note that the @xmath2ti line appears to be broadened ( iyudin et al . 1998 ) , the origin of which is unknown . in the most extreme case that the width is exclusively attributed to doppler broadening the associated velocity is ( 15300@xmath43700 ) km / s with an upper limit of @xmath77 = 19000 km / s . we include @xmath77 for the sake of completeness , but we stress that the use of @xmath77 for constraining the type of progenitor is overinterpretating the @xmath1-ray data and it is essentially misleading . nevertheless , we add that nagataki ( 1999 ) quotes @xmath78 expansion velocities @xmath7912000 km / s for sub - chandrasekhar mass models of sne ia . [ age ] shows the relation between @xmath29 and the age @xmath25 of rx j0852.0 - 4622 parametrized by @xmath28 . despite the large uncertainty of @xmath28 , @xmath25 is determined to within @xmath4 100 yrs for fixed @xmath29 using @xmath28 of the x - ray data . for @xmath29 = 5@xmath510@xmath11 m@xmath12 and @xmath28 = @xmath80 , @xmath25 = 680 yrs and @xmath24 = 200 pc ( c.f . [ distance ] ) . age and distance are rather insensitive to the exact value of the @xmath2ti @xmath1-ray line flux . a total error of the flux of @xmath440% , which is the sum of the statistical error and the systematic error , broadens the range of @xmath25 by @xmath440 yrs and that of @xmath24 by @xmath4 10 pc . if rx j0852.0 - 4622 is expanding as fast as @xmath77 indicates the age would be as low as @xmath0500 yrs . model calculations provide a range for @xmath29 , which runs for symmetric core - collapse supernovae from 1.4@xmath510@xmath11 m@xmath12 to 2.3@xmath510@xmath81 m@xmath12 , depending on progenitor mass ( woosley & weaver 1995 , thielemann et al . 1996 ) . within this range ( @xmath82 ) is within ( 500 yrs , 400 pc ) and ( 950 yrs , 80 pc ) . nagataki et al . ( 1998 ) have pointed out that @xmath29 could be much higher in an axisymmetric collapse - driven supernova . for example , in their model with the highest degree of asymmetry they obtain @xmath29 = 5.1@xmath510@xmath81 for @xmath83 = 0.07 m@xmath12 . this value of @xmath29 would allow an age of up to 1000 yrs . for type ia supernovae we find 7.9@xmath510@xmath84 m@xmath12 @xmath85 @xmath71 4.7@xmath510@xmath11 m@xmath12 for carbon deflagration models ( nomoto et al . 1984 , iwamoto et al . 1999 ) , which do not significantly differ from the core - collapse sne , and accordingly the range for ( @xmath82 ) is not affected . values of @xmath29 as high as 2@xmath510@xmath86 m@xmath12 are obtained in he - detonation models ( woosley & weaver 1994 ) . this would allow ( @xmath82 ) to lie between ( 850 yrs , 500 pc ) and ( 1100 yrs , 100 pc ) . in summary , for any @xmath29 given by current models the upper limit of the distance of rx j0852.0 - 4622 is 500 pc and 1100 years for the age . chen & gehrels ( 1999 ) have also used the x - ray temperature to derive age and distance , although in a slightly different manner . they use the temperature derived from the rosat data for the central region ( aschenbach 1998 ) which might not be representative for the current expansion velocity at the rim , why we prefer a somewhat higher velocity consistent with the apparently higher temperature observed at the limb . for the mean expansion velocity , which should be higher than the current expansion velocity by some factor , chen & gehrels derive a range of 2000 5000 km / s , whereas we propose a range of 2000 km / s to 10000 km / s by comparison with observational data obtained for the historical remnants . for given velocity and @xmath29 our results agree with those obtained by chen & gehrels , but our estimates allow a wider range of age and distance . @xmath2ti decays by electron capture , which means that lifetime depends on ionization stage , in particular to what extent the k shell is populated . the @xmath2ti lifetime of @xmath87 90 yrs is the mean lifetime for two electrons in the k - shell irrespective of the number of electrons in the higher shells . for just one electron in the k - shell , the hydrogen - like state @xmath2ti@xmath88 , the lifetime is expected to be about twice as long , and for the fully ionized atom @xmath2ti@xmath89 the lifetime is @xmath90 . 1 gives the decay rate for the ionic fraction @xmath91ti@xmath92 = 0 ( full ionization ) , @xmath91ti@xmath93 = 0 ( one electron in the k - shell ) and @xmath91ti@xmath94 = 1 . for @xmath91ti@xmath94@xmath10 1 eq . 1 is modified by introducing the ionic fraction @xmath91ti@xmath94 with @xmath95 = 90 yrs and @xmath91ti@xmath93 with @xmath95 = 2 @xmath5 90 yrs ; the impact of @xmath91ti@xmath92 has been neglected because of its comparatively low contribution to @xmath96 . the solution for @xmath25 of eq . 1 for either the ionization or the no - ionization case is done with the same @xmath96 . as before also @xmath24 and @xmath25 are not independent of each other but constrained by @xmath26 and @xmath28 , which means that not only @xmath25 but also @xmath24 is to change for the ionization case compared to the no - ionization case . so the comparison is done with the same @xmath28 but not with the same @xmath24 . furthermore @xmath28 is constrained by the x - ray spectra and the impact of the uncertainty of @xmath28 on @xmath24 and @xmath25 has been given in the previous section . if @xmath25 = @xmath97 for @xmath98 = 1 and @xmath25 = @xmath99 for @xmath1001 eq . 2 describes the change of the age in terms of @xmath101 . @xmath102\ ] ] @xmath103\ ] ] the ionization stage of @xmath2ti of gro j0852 - 4642 is not yet known , but a case study is useful to demonstrate quantitatively the impact of the ionization on the estimate of @xmath25 and @xmath24 . if @xmath2ti would have been heated to around k@xmath104 = 4.4 kev like the x - ray emitting plasma , e.g. by a reverse shock propagating in the ejecta and if @xmath2ti is in ionization equilibrium the ionic fractions can be extracted from literature . titanium has not been tabulated so far but the distributions of the ionic fraction of calcium and iron are available , which are taken as case representative examples . arnaud & rothenflug ( 1985 ) , for instance , computed @xmath105 = 0.086 , @xmath106 = 0.339 and 57.5% of ca completely ionized for log t = 7.8 . using eq . 2 @xmath97 = 680 yrs increases to @xmath99 = 930 yrs as does @xmath24 by the same factor of @xmath107 . for iron , which has @xmath108 = 0.686 and @xmath109 = 0.269 at log t = 7.8 , @xmath99 = 900 yrs , which is very close to the result obtained for ca , despite a significantly different distribution of the ionic fractions . for higher temperatures , e.g. log t = 8.5 , @xmath105 = 0 , @xmath106 = 0.0465 and 95.4% of ca is completely ionized . for this ionic fraction distribution @xmath107 = 1 and @xmath25 and @xmath24 are unchanged although only 4.65% of the total @xmath110 decays radioactively . for even lower values of @xmath111 , i.e. a larger fraction of totally ionized @xmath2ti , @xmath112 or the age becomes even lower . using the distribution of the ionic fractions of iron at log t = 8.5 , @xmath107 = 1.28 . clearly , the ionization of ti has an impact but of moderate size . values of @xmath25 and @xmath24 may be underestimated by some 30% when the ionization starts to affect the k - shell population and they may be even unchanged if only some 10% or less of the ti has just one electron in the k - shell but is otherwise completely ionized . quite recently mochizuki et al . ( 1999 ) have modelled the heating and ionization of @xmath2ti by the reverse shock in cas - a , for which they report the possibility of a currently increased @xmath2ti activity . with respect to rx j0852.0 - 4622 / gro j0852 - 4642 they find that the reverse shock does not heat the ejecta to sufficiently high temperatures to ionize @xmath2ti because of the low ambient matter density . future x - ray spectroscopy measurements may answer the question of ionization . but independent of the outcome this section shows that even if ionization were significant it does not change the conclusion that rx j0852.0 - 4622 / gro j0852 - 4642 is a young nearby snr . the sedov relation @xmath114 , which has been adopted for describing the adiabatic expansion of an snr of radius r in a homogenous medium of matter density @xmath115 , has been used quite often in the past to estimate the explosion energy e@xmath113 associated with the supernova ( winkler & clark 1974 , pfeffermann et al . the limitations of this approach are well known . the x - ray spectra provide @xmath116 , from which @xmath28 is derived , the x - ray flux is proportional to @xmath117 via @xmath118 ( cf . 2.2 ) and the x - ray image shows the angular extent @xmath26 . with the sedov relation @xmath119 is not yet uniquely determined but can then be expressed as a function of a single variable , for instance @xmath25 . since for rx j0852.0 - 4622 @xmath25 can be related to @xmath29 via eq . ( 1 ) , @xmath119 is a function of @xmath29 . in contrast to the relation @xmath120 , @xmath121 is constrained because of the limited range of @xmath29 , at least towards the higher end . fig . [ energy ] shows @xmath121 for various @xmath28 , because @xmath28 is not uniquely determined by the x - ray spectra . for the reference values of @xmath29 = 5@xmath510@xmath11 m@xmath12 and @xmath28 = @xmath80 = 5000 km / s , @xmath119 = 2.6@xmath510@xmath122 , which is a factor of about 40 less than the canonical @xmath119 = 10@xmath123 erg . with @xmath28 = 5000 km / s this value can not be reached with a realistic @xmath29 ; even a value of @xmath119 = 10@xmath124 erg is hardly consistent with reasonable @xmath29 values at @xmath28 = 5000 km / s . it is interesting to note that a similarly low value of @xmath119 , i.e. @xmath119 @xmath8 4.4@xmath510@xmath122 erg s@xmath23 have been derived by willingale et al . ( 1996 ) for the snr of sn 1006 , with which rx j0852.0 - 4622 shares a number of other similarities like the x - ray appearance and the ratio of radio to x - ray surface brightness ( aschenbach 1998 ) . for the reference value of @xmath24 = 200 pc the total swept - up mass of rx j0852.0 - 4622 is less than one solar mass , which means that the slow - down of the remnant expansion may not be dominated by @xmath115 , so that the applicability of the sedov relation may be questioned . the radial evolution depends then on the details of the explosion rather than just on @xmath119 . for instance , most of the kinetic energy of the sn may be in matter which does not radiate in x - rays . @xmath119 could be raised by increasing @xmath115 . the slow down could have occurred at times @xmath125 when regions of higher density might have been passed by the shock wave , e.g. if the progenitor star had produced a strong stellar wind . for a mass loss rate of 10@xmath11 m@xmath126yr@xmath23 and a wind velocity of 1000 km / s the wind number density would exceed 10@xmath127@xmath57 within a radius of about 6@xmath510@xmath128 cm and x - rays would have been emitted from this region . lower wind velocities like those typical of red supergiants would increase the size accordingly . the emission region would expand to a measurable size over the 700 yrs but both radiative cooling and adiabatic expansion are likely to have reduced the flux below the detection limit . nevertheless , we point out that the rosat image shows weak but enhanced emission from the central 15@xmath129 diameter region ( c.f . fig . [ picture ] ) . for higher values of @xmath28 , e.g. for @xmath28 = @xmath76 = 10000 km / s and @xmath29 = 5@xmath510@xmath11m@xmath12 , @xmath119 = 2.9@xmath510@xmath124 erg is relatively close to the canonical e@xmath113 , but the swept - up , x - ray radiating mass is still just 1 m@xmath12 . basically , because of the low @xmath49 and the maximal value of @xmath24 consistent with @xmath29 the swept - up , x - ray radiating mass never exceeds a few solar masses . in summary , e@xmath113 is not very sensitive to @xmath29 ( c.f . fig . [ energy ] ) but instead to the mean expansion velocity @xmath28 . taking the full range of @xmath28 indicated by the x - ray spectra it follows that 10@xmath122 erg @xmath10 e@xmath113 @xmath10 3@xmath510@xmath124 erg for a sedov - type expansion . the energy budget made of e@xmath113 and the kinetic and thermal energy observed can be used to constrain the mass of the progenitor . the total energy e@xmath130 of the x - ray radiating mass , i.e. the sum of the kinetic energy and the thermal energy , amounts to e@xmath130 = 4@xmath510@xmath131 erg / s@xmath132 with @xmath133 in units of 1000 km / s . since the maximum velocity of the ejecta should not exceed the uniform expansion velocity _ v _ , which for the adiabatic case is 2.5@xmath134 , a lower limit of the ejecta mass m@xmath135 is @xmath136 with m@xmath135 in m@xmath12 and e@xmath137 in 10@xmath123 erg . for @xmath28 = @xmath80 = 5000 km / s @xmath138 4 m@xmath12 , i.e. a massive progenitor is required for e@xmath137 = 1 , whereas a low mass progenitor with m@xmath139 0.9 m@xmath12 is consistent with the data for @xmath28 = @xmath76 = 10000 km / s . a more massive progenitor is required if the bulk of the ejecta mass is moving at significantly lower velocities . another approach to constrain the progenitor and the supervova type has been taken by chen & gehrels ( 1999 ) . they have used the shock wave velocity indicated by the x - ray temperature observed in the central region of the snr and used this as the current expansion velocity . by comparison of this velocity with that predicted by sn explosion models and their subsequent evolution into an ambient medium of constant matter density , they conclude that the likely progenitor of the snr was a massive star of 15 m@xmath12 with a type ii explosion , solely based on the relatively low value of the current expansion velocity v@xmath140 inferred from the x - ray temperature . lower mass progenitors like those leading to a sn of type ia are supposed to have significantly higher ejecta velocities and according to chen & gehrels an ambient matter density @xmath79 500 @xmath57 is required to decelerate the explosion wave from initially 11000 km / s to the current value of 1300 km / s , using the relation v@xmath141 t@xmath142 ( chen & gehrels , 1999 ) . if we use the standard sedov - taylor relation of v@xmath141 t@xmath143 instead , i.e. the asymptotic limit of the evolution into a uniform medium of constant matter density , a much lower ambient density of 1.4 @xmath57 is sufficient to reduce the ejecta speed from 11000 km / s to v@xmath140 = 3900 km / s ( the upper limit of v@xmath140 estimated by chen & gehrels ) in 1000 years for an ejecta mass of one solar mass and e@xmath113 = 10@xmath123 erg . although this ambient density still exceeds the observed value by a factor of @xmath030 it is not unreasonable in comparison with other snrs and ism densities . different ejecta mass , explosion energy and non constant density distributions , in particular , might reduce the required matter density further . in contrast to chen & gehrels we are therefore very reluctant to rule out a snia for rx j0852.0 - 4622 based on just the x - ray temperature . after the discovery of its x - ray emission in early 1996 it was attempted to identify rx j0852.0 - 4622 as a source contributing to the 1.8 mev @xmath3al @xmath1-ray line emission from the vela region , which had been mapped with the comptel instrument ( oberlack et al . 1994 , diehl et al . because of its identifaction as an snr and because of its apparently low distance rx j0852.0 - 4622 was considered a good candidate to provide a measurable amount of the @xmath3al @xmath1-ray line emission . 1.8 mev @xmath1-ray lines are emitted in the radioactive decay of @xmath3al , which is processed and released in supernovae but in other sources as well . the 1.8 mev vela source appears to be extended with a significant peak at about lii = 267.4@xmath7 , bii = -0.7@xmath7 . oberlack ( 1997 ) has used the rosat x - ray map of the vela region to model the 1.8 mev @xmath1-ray map , taking into account the full size of the vela snr , the vela snr explosion fragments ( aschenbach et al . 1995 ) , rx j0852.0 - 4622 and other potential sources . he found two `` comptel point - like '' sources which could contribute significantly to the @xmath1-ray peak , which are the vela snr fragment d / d@xmath129 and rx j0852.0 - 4622 . the peak position and the center position of rx j0852.0 - 4622 agree within the 2-@xmath72 localization accuracy of comptel , and the 1.8 mev point source flux is f@xmath144 = ( [email protected])@xmath2110@xmath11 photons @xmath22 s@xmath23 out of the total vela flux of f@xmath145 = ( [email protected])@xmath2110@xmath11 photons @xmath22 s@xmath23 . recently , diehl et al . ( 1999 ) reported a 2-@xmath72 upper limit of 2@xmath2110@xmath11 photons @xmath22 s@xmath23 for a contribution of rx j0852.0 - 4622 to the overall vela emisson . this result is not really in conflict with the result of oberlack , which we are going to use in the present paper . as we show below most of the our conclusions do not depend on the precise value of f@xmath146 anyway . if f@xmath144 is to be attributed to a single snr with a representative yield of @xmath147 = 5@xmath510@xmath11 m@xmath12 it follows from eq . ( 1 ) that the distance of the source would be ( 160 @xmath4 20 ) pc using a mean lifetime of @xmath148 = 1.07@xmath510@xmath149 yrs . because this excitingly low distance for an snr was not supported by any other measurements at that time and because of other competing @xmath3al sources like the vela snr fragment d / d@xmath129 the results were not published . but after the discovery of the @xmath150 emission , which immediately implies a low age because of its short lifetime and a correspondingly low distance because of the x - ray angular diameter , the situation has changed and both the @xmath2ti and the @xmath3al flux may indeed come from a single supernova now visible as the rx j0852.0 - 4622 snr . because of the uncertainty of the amount of f@xmath146 actually to be attributed to rx j0852.0 - 4622 we discuss two cases in the following chapters : a. ) f@xmath144 is entirely from rx j0852.0 - 4622 ; b. ) f@xmath144 is not entirely associated with rx j0852.0 - 4622 but then the comptel data provide a firm upper limit of f@xmath151 = 3.5@xmath510@xmath11 photons @xmath22 s@xmath23 , which is the total flux observed for the entire vela region . eq . 1 can be used to compute the age @xmath152 of rx j0852.0 - 4622 by using the fluxes of just the two radionuclides , making use of @xmath153 : @xmath154 interestingly , the age determination does not require knowledge of the distance , and it depends only on the ratio of the mass yields of the two elements considered , which might be useful for further searches for young snrs . for f@xmath146 = f@xmath144 and @xmath155 = 1 , @xmath152 = ( 750 @xmath4 25 ) yrs . this age agrees remarkably well with the age @xmath25 = @xmath97 = 680 yrs which has been derived from the @xmath2ti data and the x - ray measurements , and it appears to support the identification of rx j0852.0 - 4622 being the source of both the @xmath2ti and the @xmath3al emission . furthermore the value of @xmath152 is not very sensitive to the precise value of @xmath156 ; even if only one fifth of @xmath156 , e.g. , is actually associated with rx j0852.0 - 4622 , @xmath152 is reduced by just 145 yrs . some interesting conclusions can be drawn about the type of the supernova by making use of model produced values of ( @xmath157 . the core - collapse models of woosley & weaver ( 1995 ) give 0.1 @xmath71 @xmath155 @xmath71 4.1 for progenitor masses between 11 m@xmath12 and 40 m@xmath12 for initial solar metallicity , excluding their models with @xmath158 m@xmath12 . this leads to @xmath152 = ( 540 880 ) yrs @xmath430 yrs . [ limit ] shows the ( @xmath159 ) plane of the core - collapse model data ( s - sequence of solar metallicity ) of woosley & weaver ( 1995 ) ; pairs of ( @xmath159 ) with @xmath147 greater than the values cut by the line of fixed @xmath28 are not consistent with f@xmath151 . [ limit ] demonstrates that the models of woosley & weaver ( 1995 ) are consistent with relatively low expansion velocities , most of them with @xmath160 5000 km / s , which fits nicely the expansion velocity estimated from the x - ray temperature . for increasingly lower metallicity , @xmath147 of the woosley & weaver computations decreases and eventually the data of all the z = 0 models are above the @xmath28 = 5000 km / s cut , except the models for which @xmath16110@xmath162 m@xmath12 . we note that the data of the models s18a , s19a and s25a describing the explosion of the progenitor with a mass of 18 m@xmath12 , 19 m@xmath12 and 25 m@xmath12 , respectively , are closest to the @xmath28 = 5000 km / s line . this appears to be in rather good agreement with the conclusion which has been derived from the energy balance described in sect . the core - collapse models of thielemann et al . ( 1996 ) for masses between 13 m@xmath12 and 25 m@xmath12 show similar values of @xmath29 but significantly lower values of @xmath147 because only the yields of the explosively produced elements are given ( thielemann , private communication , 1999 ) . therefore @xmath147 is to be treated as lower limit , and the applicability to rx j0852.0 - 4622 remains unanswered at this stage . interestingly , woosley & weaver ( 1995 ) have also calculated the yields of models with very little output of @xmath163 , from which the supernova power is being drawn after a possible plateau phase . models with a small yield of @xmath163 , which may explain the sub - luminous supernovae after the early phase , also have low @xmath29 but relatively high values of @xmath147 , which is produced predominantly in the upper envelope by ordinary burning . the yields predicted by woosley & weaver for solar metallicity are shown in fig . [ limit ] as well . clearly , the ratio of @xmath164 is very high and it is too high to be consistent with the observations . given the low value of @xmath29 , @xmath147 is simply too large . such large values have to be checked against f@xmath151 . f@xmath151 requires a minimal @xmath24 for a given @xmath147 , which in turn requires a maximal @xmath25 to be consistent with @xmath29 . minimal @xmath24 and maximal @xmath25 define a minimal expansion velocity for the angular diameter not to exceed @xmath26 . [ limit ] shows that the woosley & weaver models require very large values of @xmath28 such high mean expansion velocities after some 700 yrs are unlikely and it is evident that these models can not explain the rx j0852.0 - 4622 measurements primarily because they are inconsistent with the upper limit of the @xmath3al flux . furthermore , the models of woosley & weaver show a signficant gap for @xmath29 , which covers the range 3@xmath510@xmath162 m@xmath165 10@xmath11 m@xmath12 . this gap might be artificial and further model calculations are required to check , whether the fall - back of matter towards the center of the explosion chokes the production of the high - z elements to the extent shown by the current explosion models . models with somewhat lower values of both @xmath29 and @xmath147 would be consistent with the observations , and a lower value of @xmath29 might mean a low value of @xmath163 as well , which allows for a sub - luminous supernova although the connection between low y@xmath166 and low kinetic energy and luminosity is not yet well established . models for type ia supernovae predict a much higher ratio of ( @xmath155 ) ; iwamoto et al . ( 1999 ) predict 16@xmath167 470 and the sub - chandrasekhar models of woosley & weaver have 280@xmath167 930 . these values correspond to a relatively large @xmath25 ( eq . 3 ) and a low @xmath24 with @xmath168 = @xmath156 , which results in a relatively low value of the mean expansion velocity , i.e. 50 km / s @xmath169 275 km / s for the models of iwamoto et al . and 180 km / s @xmath170 1060 km / s for the models of woosley & weaver . these values are well below the lower limit velocity of @xmath66 = 1240 km / s and are therefore inconsistent with the x - ray temperature measurements . it appears that the type ia model predictions are in serious conflict with the measurements , thus excluding type ia models from explaining rx j0852.0 - 4622 . but this conclusion hinges on the assumption that @xmath168 = @xmath156 is actually associated with rx j0852.0 - 4622 . if only a minor fraction of @xmath10 1 % of @xmath156 is due to rx j0852.0 - 4622 , also type ia models may be reconsidered . for this case @xmath24 and @xmath25 are given in sect . in contrast to type ia supernovae core - collapse supernovae are expected to leave a neutron star or a black hole initially close to the explosion center . here we restrict the discussion to a neutron star . if borne with a significant kick - velocity the neutron star will travel a distance from the centre given by the kick - velocity . kick - velocities as large as 1000 km / s have been reported for pulsars , and for rx j0852.0 - 4622 with a mean expansion velocity of @xmath28 = 5000 km / s any putative neutron star should be within a radius of about 12@xmath129 around the center . the rosat all - sky survey data of this area have been searched for point sources and two candidate sources have been found . excess emission has been detected at ra(2000 ) = 8@xmath171 52@xmath129 3 `` , dec(2000 ) = 46@xmath7 18@xmath129 36 '' , which is off - set from the explosion center by 3.4@xmath129 . with the nominal value of @xmath28 and @xmath25 derived above the separation corresponds to 283 km / s for the transverse component of the kick - velocity or a proper motion of 0.3 `` yr@xmath23 . the center of the explosion has been determined by the circle matching best the snr outer boundary . the uncertainty in the center position is estimated to be about @xmath41.5@xmath129 . this point - like source and the implications concerning a compact remnant have already been reported and discussed ( aschenbach , 1998 ) . here we report excess emission from a second point - like source inside the suspected area at ra(2000 ) = 8@xmath171 51@xmath129 58 '' , dec(2000 ) = 46@xmath7 21@xmath129 33 `` . this source has been detected in the low energy rosat image of the vela snr created from the counts which have been recorded in the central 40@xmath129 diameter field of the pspc . compared to the full field of 2@xmath7 this procedure improves the spatial resolution considerably and thereby the sensitivity of detecting point sources above the diffuse background . the 17 source counts per 40''@xmath540 `` pixel exceed the mean background level of 4.4 counts ( 40''@xmath540")@xmath23 by 6-@xmath72 . the source count rate is 0.12 counts / s . no spectrum and no information about interstellar absorption is available . but the flux can be used to estimate the size of the x - ray emitting area as a function of temperature @xmath172 for a black - body with the interstellar absorption as parameter , which is shown in fig.[blackbody ] . if the source is a black - body radiating neutron star of 10 km radius with emission from the entire surface , there is an upper limit of @xmath172 = 1.2@xmath510@xmath173k imposed by the 5-@xmath72 upper limit of n@xmath174 = 10@xmath175 @xmath22 . but more realistical is a much lower value of n@xmath176 10@xmath43 @xmath22 , which is typical for the southeastern section of the vela snr . then , fig.[blackbody ] implies that @xmath177 3@xmath510@xmath6k . if just a fraction of the full surface is radiating the temperature may be slightly higher by about a factor of two , for instance , if the area of the radiating spot is about 1% of the full neutron star surface area . for a full size area radiating neutron star , which is just 700 yrs old , @xmath172 would be surprisingly low . furthermore , if the supernova left a neutron star somewhere else the black - body surface temperature would be even lower because of a lower x - ray count rate , unless the column density to the putative neutron star should be even higher . a supernova going off at a distance of 200 pc should have been a spectacular sight for the contemporaries in the 13th or 14th century . ( if we take into account the full band of our age determination also the 12th and the 15th century should not be excluded . ) just how spectacular this event was depends on the absolute visual magnitude @xmath178 . for a bright supernova of type ia with @xmath179 19.0 the burst of light would have been as bright as the full moon . for a type ii , ib or ic which are intrinsically fainter the light output can be orders of magnitude less . for the sub - luminous or ultra - dim sne recently discussed ( schaefer 1996 , hatano et al . 1997 , woltjer 1997 ) @xmath178 = 13.0 @xmath4 2 . examples quoted are sn 1181 with @xmath178 = 12.68 @xmath4 1.41 and cas a with @xmath180 13.03 @xmath4 2.69 . taking a somwewhat extreme position with @xmath178 = 11.5 the sn associated with rx j0852.0 - 4622 would still have been brighter than venus . taking the other extreme position of a type ia the sn is a candidate to be recognized even in daylight . in any case it should have been seen . records of astronomical events including the epoch of the supernova proposed by us were taken by the far - east astronomers of china , japan and corea ( clark & stevenson 1977 , ho peng yoke 1962 ) . their observatories were typically located at a geographical latitude of @xmath035@xmath7 north ( clark & stevenson 1977 ) , within @xmath45@xmath7 north or south , so that the sn of rx j0852.0 - 4622 would have risen above the horizon after sun - set by up to 11@xmath7 from middle of december to end of march . if the sn of rx j0852.0 - 4622 would have exploded say in late march it would have re - appeared in the northern hemisphere during night after more than 250 days with significantly lower brightness . the light curve of the sub - luminous sn 1997d ( turatto et al . 1998 ) provides an estimate of @xmath181 5 at @xmath0 250 days after outburst , which appears rather little compared with other sne which show @xmath181 8 like sn 1994w ( sollerman et al . 1998 ) . with @xmath178 = 11.5 , @xmath182 = 5 and a distance modulus of 6.5 the sn of rx j0852.0 - 4622 then had @xmath183 = 0 and it is not unlikely to have escaped the attention of the medieval far - east observers . even with the @xmath178 of sn 1181 the chances to miss it would not have been low . furthermore , within this scenario , that only the tail of the light curve had been caught , the sn may have not been noticed as a `` guest star '' because the change was only against a pattern observed more than 200 days before . the light curve of sn 1997d also demonstrates that even if the sn went off in the december march time frame the detection might have been prevented by a rather short peak / plateau period , and sky visibility conditions become important . for sn 1997d this period probably lasted for @xmath71 60 days ( turatto et al . ) , over which the brightness decreased by @xmath181 3 . the above exercise demonstrates with realistic data that it is indeed possible that the sn of rx j0852.0 - 4622 was not bright enough , despite its proximity . this sort of physical explanation requires a sub - luminous sn . the peak luminosity and the early lightcurve are determined by the ejecta mass , e@xmath113 , pre - sn radius , the structure of the outer layers , @xmath83 and its distribution . as shown by chugai & utrobin ( 1999 ) in their model for sn 1997d it appears that @xmath83 is rather low for this class of sub - luminous sne , and it remains to be seen whether these sne can actually produce enough @xmath2ti ( but c.f . [ age ] , [ distance ] for the minimum amount ) . we stress that we can not exclude that the sn was indeed much brighter and even a daylight object . observers located much further south , like the people of the incas , the aztecs or in middle- and south - africa , should have had better visibility and their traditions are recommended to be searched for an event pointing to a sn . in this context it is interesting to note that the records of the far - east observers as published by ho peng yoke ( 1962 ) appear to be incomplete . the compilation shows three gaps , which are suspiciously long and statistically inconsistent with the mean rate of entries . these periods include the years of 773 814 , 1245 1264 and 1277 1293 , the latter two of which are relevant for rx j0852.0 - 4622 . finally , we point out that there is a chance that the progenitor star of rx j0852.0 - 4622 is shown in ancient star charts if it happened to be a massive star . up to now just one progenitor star of a supernova has been identified , which is the progenitor of sn 1987a , the b3 ia blue supergiant sanduleak 69 202 with m@xmath184 = 6.8 ( west et al . 1987 ) . at a distance of 200 pc the apparent unreddened visual magnitude would have been @xmath183 = 0.3 , which would have made the star the brightest star in the vela constellation located between @xmath1 vel and @xmath185 vel . a wolf - rayet type progenitor star would have been less bright with @xmath186 + 2.5 , but still comparable with the other bright stars in vela . this opens up an interesting explanation for the apparent absence of a historical record , which admittedly is a speculation . if the progenitor star had been so bright the supernova might not have been noted down as a `` guest star '' . the existing star would just have become brighter . and if the short peak of the outburst had been missed of whatsoever reason and only the tail of the supernova lightcurve has been observed the change of brightness might not have been spectacular , and the star would have disappeared slowly over a couple of years . an estimate of the age @xmath25 and distance @xmath24 of the supernova remnant rx j0852.0 - 4622 / gro j0852 - 4642 can be obtained by combining the rosat x - ray and comptel @xmath1-ray data . assuming a @xmath187ti yield of the supenova of 5 @xmath5 10@xmath188 m@xmath189 and an expansion velocity of 5000 km / sec @xmath25 = 680 yrs and @xmath24 = 200 pc are obtained . actually , the expansion velocity is constrained by the x - ray data to lie in the range of 2000 km / sec @xmath10 v @xmath10 10 000 km / sec yielding an uncertainty of the age of @xmath4 100 years for a fixed @xmath187ti yield . for the highest @xmath187ti yield given by current supernova models , a firm upper limit of the distance is 500 pc and 1100 years for the age . the determination of the age depends to some extent on the ionization state of @xmath187ti because @xmath187ti decays by electron capture . the values quoted above have been obtained under the assumption that the k - shell is fully populated . if the k - shell contains only one electron , the @xmath187ti mean lifetime is estimated to increase by a factor of two . but the age of the snr will not increase by the same factor because of the angular diameter , and therefore distance constraint . adopting the same mean expansion velocity @xmath25 and @xmath24 can change by about 35% at most and for a very strong ionization @xmath25 and @xmath24 may be even lower than the `` nominal '' estimate . future x - ray spectroscopy measurements are needed to search for ti x - ray emission lines to determine the ionization state of @xmath187ti and further constrain @xmath25 and @xmath24 . the x - ray surface brightness of rx j0852.0 - 4622 is rather low and implies a rather low matter density of the shock wave heated plasma if the radiation is thermal . a formal analysis of the x - ray data in terms of a sedov - type evolution of the snr using the standard conversion of x - ray temperature in shock velocity turns out a rather low value of a few times 10@xmath122 erg for the explosion energy @xmath119 , which can be raised only significantly if the mean expansion velocity @xmath28 would exceed 10000 km / s . but if @xmath28 is closer to 5000 km / s as the x - ray data indicate then the bulk of @xmath119 resides still in kinetic energy of the ejecta , not radiating in x - rays , which means that any reverse shock has not yet penetrated deep into the ejecta , and that the titanium is not highly ionized . in this case a lower limit for the mass of the progenitor star of 25 m@xmath12 is estimated from the energy balance . there is evidence for @xmath190al 1.809 mev line emission from rx j0852.0 - 4622 , which has been measured by comptel towards the vela region . admittedly this has still to be confirmed . but if a non - neglible part of this 1.809 mev line flux is coming from rx j0852.0 - 4622 a similar age of 600 - 750 yrs for the snr is obtained for similar yields of @xmath190al and @xmath187ti . since @xmath25 depends only on the logarithm of the yields and the fluxes @xmath25 will not change significantly even for large changes of the @xmath190al flux . it is more a matter of whether or not there is @xmath190al emission . if the @xmath190al line flux is about what is indicated by the comptel data the existing type ia supernova models can be ruled out for the progenitor explosion because they predict a ratio of the @xmath190al and @xmath187ti yield which is by far too large . they could be reconsidered only if less than 1% of the @xmath190al line flux from vela in total is associated with rx j0852.0 - 4622 . explosion models of core - collapse supernovae ( woosley & weaver , 1995 ) are in general in agreement with the observations , i.e. the measurements of @xmath28 , @xmath187ti line flux and @xmath190al line flux , judging from their prediction of the yields of @xmath187ti and @xmath190al . their models with @xmath191 10@xmath162 m@xmath12 and @xmath192 10@xmath81 m@xmath12 can definitely be excluded because the yield of @xmath190al predicted is inconsistent with even the upper limit of the @xmath190al line flux measured for the vela region in total . in summary , both the energy balance and the yield predictions of the currently available explosion models point towards a core - collapse event . within the vicinity of the explosion center two point - like x - ray emission regions have been found , either of which could be the manifestation of a neutron star . spectra are not available , but if the radiation is assumed to be black - body emisson , the x - ray flux indicates a surface temperature of @xmath873@xmath510@xmath6 k , which is surprisingly low for a 700 years old neutron star . if neither one of these two sources is a neutron star and the neutron star hides somewhere else with an even lower x - ray count rate or if only a fraction of the radiation observed from the point - like objects is due to thermal radiation the surface temperature of the neutron star could be even lower . in principle , the supernova could have been seen from the far - east astronomers of china , corea or japan or from geographical latitudes further south . the supernova could have been very bright and then there are records expected to exist , which should be searched for . if the supernova would have been of the sub - luminous class with a brightness as low as that of sn 1181 and a short peak - plateau duration it could have been missed . there is some chance that the progenitor star was sufficiently bright and could have been seen by the naked eye . then the stellar pattern of vela was different in ancient times and the astronomers who monitored the sky might not have noted the supernova as a `` guest star '' or a `` new '' star , because the star was existing and just brightening and eventually fading away . ahmad i. , bonino g. , cini castagnoli g. , et al . , 1998 , phys . 80 , 2550 arnaud m. , rothenflug r. , 1985 , a&as 60 , 425 aschenbach b. , 1998 , nature 396 , 141 aschenbach b. , egger r. , trmper j. , 1995 , nature 373 , 587 chen w. , gehrels n. , 1999 , apj 514 , l103 chugai n.n . , utrobin v.p . , 1999 , astro - ph/9906190 10 jun 1999 clark d.h . , stephenson f.r . , 1977 , `` the historical supernovae '' , pergamon , oxford diehl r. , bennet k. , bloemen h. , et al . , 1995 , a&a 298 , l25 diehl r. , oberlack u. , plschke s. , et al . , 1999 , proc . 3rd integral workshop , taormina 1998 , eds . g. palumbo , h. bazzano , c. winkler , astrophys . lett . & comm . , 1999 , in press hatano k. , fisher a. , branch d. , 1997 , mnras 290 , 360 ho peng yoke , 1962 , vistas in astron . 5 , 127 hughes j. , 1996 , baas 189 , 4608 iwamoto k. , brachwitz f. , nomoto k. , et al . , 1999 , iyudin a.f . , schnfelder v. , bennett k. , et al . , 1998 , nature 396 , 142 koralesky b. , rudnick l. , gotthelf e.v . , keohane j.w . , 1998 , apj 505 , l27 mochizuki y. , takahashi k. , janka h .- th . , hillebrandt w. , diehl r. , 1999 , a&a 346 , 831 nagataki s. , 1999 , apj 511 , 341 nagataki s. , hashimoto m .- a . , sato k. , yamada s. , mochizuki y. , 1998 , apj 492 , l45 nomoto k. , thielemann f .- k . , yokoi k. , 1984 , apj 286 , 644 oberlack u. , 1997 , phd thesis , ber die natur der galaktischen @xmath3al - quellen untersuchung des 1.8 mev himmels mit comptel , technische universitt mnchen oberlack u. , diehl r. , montmerle t. , et al . , 1994 , apjs 92 , 433 pfeffermann e. , aschenbach b. , predehl p. , 1991 , a&a 246 , l28 schaefer b.e , 1996 , apj 464 , 404 sollerman j. , cumming r.j . , lundqvist p. , 1998 , apj 493 , 933 strom r.g . , 1994 , a&a 288 , l1 thielemann f .- k . , nomoto k. , hashimoto m .- a . , 1996 , apj 460 , 408 turatto m. , mazzali p.a.,young t.r . , et al . , 1998 , apj 498 , l129 vink j. , bloemen h. , kaastra j.s . , bleeker j.a.m . , 1998 , a&a 339 , 201 west r.m . , lauberts a. , jrgensen h.e . , schuster h .- e . , 1987 , a&a 177 , l1 willingale r. , west r.g . , pye j.p . , stewart g.c . , 1996 , mnras 278 , 749 winkler jr . , clark g.w . , 1974 , apj 191 , l67 woltjer l. , 1997 , a&a 328 , l29 woosley s.e . , weaver t.a . , 1994 , apj 423 , 371 woosley s.e . , weaver t.a . , 1995 , apjs 101 , 181

this study is aimed at making a calculation about the impact of the two most commonly used solar euv flux models solar2000 ( s2k ) of and euvac model of on photoelectron fluxes , volume emission rates , ion densities and co cameron and co uv doublet band dayglow emissions on mars in three solar activity conditions : minimum , moderate , and maximum . calculated limb intensities profiles are compared with spicam / mars express and mariner observations . analytical yield spectrum ( ays ) approach has been used to calculate photoelectron fluxes in martian upper atmosphere . densities of prominent ions and co molecule in excited triplet a state are calculated using major ion - neutral reactions . volume emission rates of co cameron and co uv doublet bands have been calculated for different observations ( viking condition , mariner and mars express spicam observations ) on mars . for the low solar activity condition , dayglow intensities calculated using the s2k model are% higher than those calculated using the euvac model . during high solar activity , due to the higher euv fluxes at wavelengths below 250 in the euvac model , intensities calculated using euvac model are slightly higher (% ) than those calculated using s2k model . irrespective of the solar activity condition , production of cameron band due to photodissociative excitation of co is around 50% higher when s2k model is used . altitude of peak limb brightness of co cameron and co uv doublet band is found to be independent of solar euv flux models . calculated limb intensities of co cameron and co uv doublet bands are on an average a factor of and.5 , respectively , higher than the spicam mars express observation , while they are consistent with the mariner observations . 2

1110.3870

first observation of co cameron and @xmath4 uv doublet emissions in the martian dayglow were made by the mariner 6 and 7 flybys in 19691970 @xcite . these observations provided an opportunity to study the martian upper atmosphere in a greater detail . the co cameron band ( @xmath5 ; 180 260 nm ) system arises due to the transition from the excited triplet a@xmath1 state , which is the lowest of triplet states , to the ground state x@xmath6 of co. doublet transition ( @xmath7 ) from excited @xmath4 ( b@xmath8 ) to the ground state @xmath4 ( x@xmath9 ) gives emission in ultraviolet wavelengths at 288.3 and 288.6 nm . apart from these emissions , fox - duffenback - berger band of @xmath4 ( a@xmath10 x@xmath11 ) , fourth positive band of co , first negative band of co@xmath12 ( b x ) , and several atomic line emissions of carbon and oxygen atoms were also recorded by mariner 6 , 7 , and 9 @xcite . with the help of theoretical calculations and laboratory measurements , @xcite proposed possible mechanisms for the dayglow emission observed in the martian atmosphere . maximum intensity of co cameron band and uv doublet observed by mariner 6 and 7 were 600 kr at @xmath13131 km and 35 kr at 148 km , respectively . emissions from cameron band and @xmath4 uv doublet bands were also observed in 19711972 by mariner 9 , the first spacecraft to orbit mars . @xcite observed a reduction in the intensity of cameron band by a factor of 2.5 compared to mariner 6 and 7 observations . they attributed this difference to the reduction in the solar activity and better calibration of mariner 9 instrument . the observed maximum slant intensities of co cameron band were between 200 and 300 kr and averaged topside scale height for the same band was 17.5 km . @xcite also observed a good correlation between co cameron band intensity and solar f10.7 flux , which suggest that these emissions are controlled by the incident solar photon flux . since the mariner 6 , 7 , and 9 uv observations , spicam ( spectroscopy for the investigation of the characteristics of the atmosphere of mars ) on - board mars express ( mex ) is the first instrument dedicated for the aeronomical studies of mars . spicam has broaden our understanding about the martian dayglow . emissions observed by spicam in uv dayglow are h lyman-@xmath14 emission at 121.6 nm , the atomic oxygen emissions at 130.4 and 297.2 nm , the cameron band ( @xmath15 ) and fourth positive band ( @xmath16 ) of co , and ultraviolet doublet band ( @xmath17 ) emissions of @xmath4 ( cf . * ; * ? ? ? these emission features are similar to those observed by mariner 6 , 7 , and 9 but with better sensitivity , and spatial and temporal coverage . spicam has observed these dayglow emissions on mars throughout the martian year and showed the effect of solar zenith angle ( sza ) , seasonal variation , and martian dust storms on the dayglow emissions @xcite . spicam also provided first observation of @xmath18 uv emissions in martian dayglow @xcite . @xmath18 vegard - kaplan vk ( 0 , 5 ) and ( 0 , 6 ) band emissions at 260.4 nm and 276.2 nm , respectively , have been observed ; @xmath18 vk ( 0 , 7 ) emission at 293.7 nm has also been reported , but it has large uncertainty @xcite . the detailed model of @xmath18 dayglow emissions on mars is presented elsewhere @xcite . several theoretical studies have been made for the dayglow emissions on mars @xcite . first detailed calculation of dayglow emission on mars was carried out by @xcite . calculated overhead intensities of co cameron and @xmath4 uv doublet bands were 49 kr and 12 kr , respectively , for the low solar activity condition similar to viking landing @xcite . @xcite used trans - mars model to calculate limb intensities of cameron and @xmath4 uv doublet emissions for low solar activity condition ( similar to viking landing ) and compared them with spicam - observations . their calculated intensities are higher by @xmath225% than the observation . @xcite had to reduce the viking @xmath19 density in the model atmosphere by a factor of 3 to bring the altitude of peak emission in agreement with the observation . seasonal effects on intensities of cameron and uv doublet bands have been observed by spicam @xcite . @xcite have presented a statistical analysis of cameron band and uv doublet emissions , peak altitude of emissions , and ratios between uv doublet and cameron band . averaged peak emission brightness ( altitude of peak emission ) observed by @xcite for co cameron and uv doublet bands are 118 @xmath20 33 kr ( 121 @xmath20 6.5 km ) and 21.6 @xmath21 7.2 kr ( 119.1 @xmath20 7.0 km ) , respectively . they also presented observations for one particular season , solar longitude ( ls ) = 90 to 180@xmath22 , and compared observational data with model calculations based on monte carlo code , which has been used also by @xcite for the martian dayglow studies . modelling of co cameron and @xmath4 uv doublet dayglow emissions requires a sophisticated input solar euv ( @xmath1350 to 1000 ) flux , which is a fundamental parameter to model physics , chemistry and dynamics of the upper atmosphere of planets . since observations of solar euv irradiance are not frequent and generally not available simultaneously with the observation for the upper atmospheric studies , the solar euv flux model become important for modelling of aeronomical quantities in planetary atmospheres . generally , solar euv flux models are bin - averaged into numbers of wavelength bands and important solar emission lines appropriate for the calculation of photoionization and photoelectron impact production rates . characterisation of the solar euv flux for use in aeronomical and ionospheric studies were developed during the seventies based on the atmospheric explorer - e ( ae - e ) data @xcite . two ae - e reference spectra sc#21ref and f79050n have been published by @xcite at 37 wavelength bins for solar minimum and maximum conditions , respectively . later , based on the measured photoelectron flux , the short wavelength fluxes were increased by @xcite , and they incorporated modified f74113 solar euv flux in their euvac model . detailed discussion on the development of solar euv flux models is beyond the scope of this study ; @xcite , @xcite , and @xcite have provided reviews on solar euv flux models . for a given solar activity there are significant differences between the euv fluxes reproduced by different solar flux models . these models are based on the different input parameters and proxies , e.g. , solar 10.7 cm radio flux ( henceforth referred to as solar f10.7 ) is used as the measure of solar activity , and used for parametrization of solar euv flux models @xcite . euvac model of @xcite is based on solar f10.7 and its 81-day average and also on the f74113 solar flux ( in the euvac model , the f74113 flux below 250 , and below 150 , are doubled and tripled , respectively ) . solar2000 model of @xcite is based on measured solar flux irradiance and various proxies and provides solar flux from x - rays to infrared wavelengths , i.e. , 11000000 . different solar euv flux models have been used to study the solar radiation interaction with upper atmosphere of mars . presently , solar2000 ( s2k ) model of @xcite and euvac model of @xcite are commonly used solar euv flux models in aeronomical studies of mars ; _ e.g. _ , @xcite and @xcite have used euvac model , while @xcite and @xcite have used s2k model for the dayglow calculations on mars . @xcite have studied the effect of different solar euv flux models on calculated electron densities for low and high solar activity conditions . they have used 85315 and 79050 solar fluxes of @xcite and sc#21ref and f79050n ae - e reference solar spectra of hinteregger @xcite for low and high solar activity conditions , respectively . @xcite found that due to smaller fluxes at short wavelength range ( 18200 ) in hinteregger spectra , lower peak in electron density profile is significantly reduced ( 3035% ) compared to that calculated using solar fluxes of @xcite . @xcite calculated ionospheric electron density using euvac model of @xcite and euv94x solar flux model of @xcite . they found that photoionization rate in f2 region calculated by using euv94x model is larger than that calculated using euvac model due to the large euv fluxes in 3001050 wavelength range in euv94x solar flux model . recently , @xcite have studied the effect of solar euv flux models on @xmath18 vk band intensities in martian dayglow and showed that euv flux models does affect the @xmath18 dayglow emissions . similar conclusion have been drawn for @xmath18 dayglow emission on venus @xcite the aim of the present study is to calculate the impact of solar euv flux models on co cameron band and @xmath4 uv doublet band intensities in martian dayglow . we have used 37 bin euvac model of @xcite and s2k version 2.36 of @xcite as the solar euv fluxes . in these models , bins consist of band of 50 width each and few prominent solar euv lines , and are sufficient for the modelling of photoionization and photoelectron flux calculations @xcite . photoelectron flux , volume excitation rates and overhead intensities are calculated using both the solar euv flux models for low , moderate , and high solar activity conditions . line of sight intensities for cameron band and uv doublet emissions are calculated and compared with the latest observations by spicam instrument . model atmosphere consists of five gases ( @xmath19 , n@xmath3 , co , o , and o@xmath3 ) . model atmosphere for solar minimum condition is taken from the mars thermospheric general circulation model ( mtgcm ) @xcite . model atmosphere for high solar activity is taken from @xcite . photoabsorption and photoionization cross sections for gases considered in the present study are taken from @xcite , and branching ratios for ionization in different states are taken from @xcite . production mechanisms for co(a@xmath1 ) are photon and electron impact dissociative excitation of @xmath19 , electron dissociative recombination of @xmath4 , and electron impact excitation of co. since @xmath23 is a forbidden transition , resonance fluorescence of co is not an effective excitation mechanism . co(a@xmath1 ) is a metastable state ; @xcite has measured its lifetime as 7.5 @xmath24 1 ms , consistent with the measurement of @xcite . individual band lifetime can vary , e.g. , lifetime of co(a@xmath1 ) @xmath25 level is around 3 ms @xcite . due to its long lifetime , cross section for the production of cameron bands due to electron impact dissociation of @xmath19is difficult to measure in the laboratory . overall , cameron band cross section can have an uncertainty of a factor of 2 @xcite . @xcite reported relative magnitudes of the cross section for the ( 0 , 1 ) band at 215.8 nm . @xcite estimated the total cameron band emission cross section of @xmath26 @xmath27 at 80 ev , whereas ajello estimated a value of @xmath28 @xmath27 at 80 ev . @xcite have fitted the e-@xmath19 cross sections producing co cameron band based on the estimated value of @xcite . in our study cross section for cameron band production due to electron impact on @xmath19is taken from @xcite . cross section for photodissociation of @xmath19 producing co(a@xmath1 ) is taken from @xcite . to calculate the production rate of co(a@xmath1 ) due to dissociative electron recombination process we have calculated the density of electron and major ions by including ion - neutral chemistry in the model . the coupled chemistry model is similar to that adopted in our other studies @xcite . rate coefficients for ion - neutral reactions are taken from @xcite . recently , @xcite have measured recombination rates for the e-@xmath4 collision , with yield of 0.87 for the channel producing co of which co(a@xmath1 ) production yield is 0.29 @xcite . ion and electron temperatures are taken from @xcite . ion and electron densities are calculated under steady state photochemical equilibrium . to calculate the density of co(a@xmath1 ) we have also included various sources of loss of co(a@xmath1 ) in our coupled chemistry model , which are given in table [ tab : rate - cameron ] . major production sources of @xmath4 in @xmath29 state are photon and electron impact ionization of @xmath19 . cross section for the formation of co@xmath30 state due to electron impact ionization of @xmath19 is taken from @xcite , and cross section due to photoionization of @xmath19 is based on the branching ratio taken from @xcite . for other gases electron impact cross sections have been taken from @xcite . solar euv flux has been taken at 1 au ( based on solar f10.7 at 1 au as seen from the mars , taking the sun - earth - mars angle into consideration ) and then scaled to the sun - mars distance . fig . [ fig : solar - flx ] shows the output of euvac and s2k solar euv flux models for both solar minimum ( 20 july 1976 ) and solar maximum ( 2 august 1969 ) conditions at 1 au . there are substantial differences in the solar euv fluxes of euvac and s2k models ; moreover , these differences are not similar in solar minimum and maximum conditions . in both , solar minimum and maximum conditions , solar flux in bands is higher in s2k than in euvac over the entire range of wavelengths , except for bins below 250 ( 150 for solar minimum condition ) , whereas solar flux at lines is higher in euvac model for entire wavelength range . overall higher solar fluxes above 250 in s2k cause more photoionization . higher photon fluxes below 250 ( during solar maximum condition ) in euvac produce more higher energy electrons in the martian atmosphere causing secondary ionizations ( cf . [ fig : pef ] ) that can compensate for the higher photoionization in s2k model . one big difference between solar euv fluxes of s2k and euvac models is the solar flux at bin containing 1026 h ly-@xmath31 line , which in both solar minimum and maximum conditions is around an order of magnitude higher in s2k compare to euvac solar flux model ( cf . [ fig : solar - flx ] ) . flux at these wavelengths does not contribute to the photoionization , but are very important for dissociative excitation processes . cross section for the photodissociation of @xmath19 producing cameron band lies in longer ( 7001080 ) wavelength regime @xcite . hence , yield of photodissociation excitation of @xmath19 producing co(a@xmath1 ) state would be larger when s2k solar euv flux model is used . to calculate the photon degradation and generation of photoelectrons in the atmosphere of mars , we have used lambert - beer law . solar zenith angle ( sza ) is 45@xmath22 unless mentioned otherwise in the text . photoelectron energy degradation and production rates of excitation of co cameron band and @xmath4uv doublet band in the martian atmosphere are calculated using analytical yield spectrum ( ays ) technique , which is based on the monte carlo model ( cf . * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? details of calculation of photoelectron production rates and photoelectron flux have given in our earlier papers @xcite . below 70 ev , photoelectron flux calculated using s2k is higher compared to that calculated using euvac model for low solar activity condition ( fig . [ fig : pef ] ) . above 70 ev , photoelectron flux calculated using euvac model is higher than that calculated using s2k model due to the larger solar euv fluxes at shorter wavelength ( @xmath32 200 ) in euvac model ( cf . fig . [ fig : solar - flx ] ) . during solar maximum condition photoelectron fluxes calculated by using euvac model is higher than that calculated using s2k model ( cf . [ fig : pef ] ) due to higher solar euv flux at wavelength below 250 in euvac model . during solar minimum , except at peaks and energies above 70 ev , photoelectron flux calculated using s2k model is around 1.4 times higher than that calculated using euvac model . during solar maximum condition photoelectron flux calculated using euvac model are higher than that calculated using s2k model . photoelectron fluxes calculated using both solar euv flux models are similar in shape but peaks around 2030 ev are more prominent in electron flux calculated using euvac due to the higher solar euv flux at lines ( e.g. , he ii lyman-@xmath14 line at 303.78 ) in euvac model ( fig . [ fig : solar - flx ] ) . the peaks in the 2030 ev region of the photoelectron flux arise due to the ionization of @xmath19 in different ionization states by solar euv flux at 303.78 line @xcite . our calculated photoelectron fluxes are in good agreement with other model calculations ( cf . * ? ? ? volume excitation rate is calculated for important processes producing co(a@xmath1)and co@xmath30 states using photoelectron flux as @xmath33 where @xmath34 is the density at altitude @xmath35 , @xmath36 is the electron impact cross section for @xmath37 process , for which threshold is @xmath38 , and @xmath39 is the photoelectron flux . we run the model for low solar activity condition ( similar to viking landing ) , and calculated results are compared with those of @xcite by taking the similar model atmosphere . model atmosphere is based on density data derived from viking 1 @xcite . the sun - mars distance d@xmath40 = 1.64 au and solar zenith angle is taken as 45@xmath22 . [ fig : ionden ] shows the calculated densities of @xmath4 and o@xmath41 in the martian upper atmosphere . the density of @xmath4 around peak and above calculated using s2k model is @xmath230% higher than that calculated using euvac , which is due to higher production rate of @xmath4 when s2k model is used . below 120 km , ion densities calculated by using euvac model are higher due to the higher photoelectron fluxes above 70 ev ( cf . [ fig : pef ] ) . there is a small discontinuity in the density of o@xmath41 ion around 180 km , which is due to the sudden change in the electron temperature at 180 km ( see fig . 2 of * ? our calculated ion densities are consistent with calculations of @xcite . fig . [ fig : ver - viking ] ( upper panel ) shows the profiles of production mechanisms of co(a@xmath1 ) calculated using euvac and s2k solar euv flux models . around the peak of co(a@xmath1 ) production , the major source is photoelectron impact dissociation of @xmath19 , while at higher altitudes photodissociation excitation of @xmath19 takes over . dissociative recombination is about 13% , while photoelectron excitation of co contribute about 3% to the total cameron band excitation at the peak . the shape of volume excitation rate ( ver ) profiles and the altitude at the peak remain the same for all processes for the two solar flux models . however , the magnitude of vers calculated using s2k model are about 40% higher than those calculated using euvac model . contribution of electron impact dissociation of @xmath19 producing co(a@xmath1 ) is higher in our studies than that in @xcite . this is due to the higher value of co(a@xmath1 ) production cross section in e-@xmath19 collision used in the present study ( having the value of @xmath42 @xmath43 at 27 ev ) ; @xcite used the cross section derived from @xcite ( having the value of @xmath44 @xmath43 at 27 ev ) . due to larger photon flux at longer ( 7001050 ) wavelengths ( region where photodissociation of @xmath19 becomes important ) in s2k model compared to euvac model ( cf . [ fig : solar - flx ] ) , co(a@xmath1 ) production due to photodissociative excitation is higher by @xmath250% when s2k model is used . production rates of co cameron band for different processes calculated at the peak along with the peak altitude are given in table [ tab : ver ] . it is also clear from upper panel of fig . [ fig : ver - viking ] that the altitude where the photodissociation of @xmath19 takes over electron impact dissociation of @xmath19 in co(a@xmath1 ) formation is slightly higher when euvac model is used ( 167 km for euvac and 160 km for s2k solar flux model ) . in our model calculations , as well as in the work of @xcite , photodissociation process becomes the major source at higher altitudes ( @xmath45 160 km ) and is a factor of 2 higher than the electron impact dissociation of @xmath19 . for the @xmath4 uv doublet band , we have considered only photoionization and electron impact ionization of @xmath19 producing @xmath4 in the b@xmath46 state . contribution of solar fluorescent scattering is very small , less than 10% ( cf . * ? ? ? * ) , and hence it is not taken into account in the present study . while calculating the emission from @xmath4 uv doublet , we have assumed 50% crossover from @xmath29 to @xmath47 state @xcite . [ fig : ver - viking ] ( bottom panel ) shows the production rates for @xmath4 in @xmath29 state . production of co@xmath30 due to photoionization of @xmath19is about a factor of 34 higher than due to photoelectron impact ionization . here also we find that production rates calculated using s2k are higher than those calculated using euvac flux by about 50% , but peak altitude remains the same in both cases . production rates calculated at peak along with peak altitude for @xmath4 uv doublet band are given in table [ tab : ver ] . volume emission rates are height - integrated to calculate overhead intensities , which are presented in table [ tab : oi - cmp ] for co cameron and uv doublet bands . for cameron band , contribution of e-@xmath19 process is maximum ( 64% ) followed by photodissociation of @xmath19 ( 21% ) . contributions of dissociative recombination and e - co process are around 10% and 3% , respectively . for uv doublet band , the major contribution is coming from photoionization of @xmath19 ( 80% ) , the rest is due to electron impact ionization of @xmath19 . to compare the model output with observed emissions we have integrated volume emission rates along the line of sight . limb intensity at each tangent point is calculated as @xmath48 where @xmath49 is abscissa along the horizontal line of sight , and v(r ) is the volume emission rate ( in @xmath50 s@xmath51 ) at a particular emission point @xmath52 . the factor of 2 multiplication comes due to symmetry along the line of sight with respect to the tangent point . while calculating limb intensity we assumed that the emission rate is constant along local longitude / latitude . for emissions considered in the present study the effect of absorption in the atmosphere is found to be negligible . [ fig : int - viking ] shows the calculated limb profiles of the co cameron and @xmath4 uv doublet bands along with the spicam - observed limb intensity taken from @xcite . observed values are averaged over for the orbits close to viking 1 condition ( ls@xmath2100140@xmath22 ) , low solar activity , and for sza=45@xmath22 . below 100 km there is a sudden increase in the observed intensity of @xmath4 uv doublet band , which , according to @xcite , is due to the very significant solar contamination below 100 km . limb intensities calculated using s2k flux are @xmath24050% higher than those calculated using euvac : clearly showing the effect of input solar euv flux on the calculated intensities . magnitude of the calculated limb intensity of uv doublet band is in agreement with the observation , but for co cameron band calculated intensity is higher than the observed profile . for both bands , the calculated intensity profile peaks at higher ( @xmath25 km ) altitude in comparison with the observation indicating a denser neutral atmosphere in our model . the dashed curves in fig . [ fig : int - viking ] show intensities calculated after reducing the @xmath19 density by a factor of 1.5 ; a good agreement in the altitude of peak emission is seen between calculated and observed limb profiles . though the reduction in @xmath19density shifted the altitude of peak emission downwards , the magnitude of calculated cameron band intensity is still larger than the intensity measured by spicam . as pointed out in section [ sec : model ] , the e-@xmath19 cross sections producing cameron band are uncertain by a factor of @xmath22 . the calculated limb intensity profile for reduced e-@xmath19 cross section by a factor of 2 is also shown in fig . [ fig : int - viking ] . cameron band intensities obtained after reducing the density and cross section are in relatively close agreement with the observed values . in the model calculations of @xcite also the cameron band intensity and its peak emission altitude were higher than the spicam observed values . they have to reduced the density of @xmath19 by a factor of 3 and e-@xmath19 cameron production cross section by a factor of 2 to bring their calculated intensity profile in agreement with the spicam observation . @xcite have presented detailed analysis of spicam data during the period october 2004 to march 2005 , spanning the solar longitude ( ls ) from 101@xmath22 to 171@xmath22 . they divided the total data set in two periods of solar longitude : first , ls = 101@xmath22 to 130@xmath22 , and second , ls = 139@xmath22 to 171@xmath22(cf . table 2 of * ? ? ? @xcite found that the altitude of peak emission for @xmath4 uv doublet and co cameron bands is around 10 km higher for ls @xmath45 138@xmath22 ( 122.5 km and 132.5 km for uv doublet and cameron bands , respectively ) compared to ls @xmath32 130@xmath22 ( 112.5 km and 117.5 km , for the same emissions ) . @xcite could not provide the reason for the higher altitude of peak emission for ls @xmath45 130@xmath22 observations . later , @xcite derived neutral densities in martian upper atmosphere using the spicam instrument in stellar occultation mode for the same observation period . @xcite found that their is a sudden increase in the @xmath19 density in the martian upper atmosphere for ls @xmath2 130@xmath22140@xmath22 , which they attributed to a dust storm . dust storm can heat the lower atmosphere and thus increase the densities at higher altitudes , which could explain the higher altitude for peak emission observed by the spicam for ls @xmath53 130@xmath22 observations . increase in the altitude of peak intensity of dayglow emissions clearly shows the effect of dust storms on martian dayglow emissions . comparisons of spicam observations with model calculated dayglow emissions have been performed by @xcite , @xcite , and @xcite . @xcite have used one dimensional trans - mars model with euvac solar flux model , whereas @xcite and @xcite have used monte carlo model with s2k solar flux . in the present study , we have taken both euvac and s2k models and have calculated the cameron band and uv doublet band emissions using the analytical yield spectrum method ; the results are compared with the spicam observations . to model the spicam observations for ls @xmath54 130@xmath22 the model atmosphere is based on mtgcm of @xcite ( taken from * ? ? ? calculations are made for mex orbit no . 983 ( 24 oct . 2004 ) when d@xmath40 = 1.64 au , f10.7 = 87.7 ( f10.7a = 107.3 ) . [ fig : ver - cam - shem ] ( upper panel ) shows the volume excitation rate of co(a@xmath1 ) . the total ver calculated using s2k flux is only slightly higher than that obtained using euvac flux . however , the production rate due to photodissociative excitation of @xmath19 is around 50% higher when s2k model is used . another interesting feature is the dissociative recombination ( dr ) process , whose contribution is @xmath1318% in the total intensity , roughly equal to the photodissociative excitation process ( dr contribution is even higher than photodissociative excitation around production peak when euvac model is used ) , and it is significantly higher than compared to the dr process in viking condition case ( see table [ tab : oi - cmp ] ) . this is due to the higher density of @xmath4 ion compared to viking condition ( see fig . [ fig : ionden ] ) . @xcite mentioned that higher values of @xmath4 can contribute up to 30% to the cameron band production depending on the solar zenith angle . to account for dr in cameron band production , @xcite and @xcite have taken @xmath4 and electron densities from @xcite for low solar activity condition . since spicam observations are made during moderate solar activity condition , the contribution of dr in cameron band production would be lower in their calculations . [ fig : ver - cam - shem ] ( bottom panel ) shows the production rates of @xmath4 uv doublet band . total rate calculated using both solar flux models is peaking at same altitude ( @xmath2125 km ) , but total production rate calculated using s2k model is higher ( around 10% ) than that calculated using euvac model . table [ tab : oi - cmp ] shows the overhead intensities of co cameron and @xmath4 uv doublet bands calculated using both euvac and s2k solar flux models . contribution of different processes in cameron band production is slightly different than that in the low solar activity condition . contribution from @xmath19 photodissociation is slightly reduced ( 17% , 21% when s2k model is used ) , while dissociative recombination contribution is increased ( @xmath218% ) . contribution of e-@xmath19 and e - co processes remains almost same ; 61 ( 64% ) and 4 ( 4% ) , respectively , when calculated using euvac ( s2k ) model . for uv doublet band photoionization of @xmath19 remains the dominant process contributing around 80% to the total overhead intensity . we have also calculated overhead intensities of major vibrational bands of cameron system , which have been observed in martian dayglow , using frank - condon factors from @xcite and branching ratio from @xcite . table [ tab : vib - oi ] shows the calculated overhead intensities of major vibrational bands of co cameron band system . contribution of major vibrational bands to the total overhead cameron band intensity is around 10 , 10 , 16 , and 8% for ( 0 , 0 ) , ( 0 , 1 ) , ( 1 , 0 ) , and ( 2 , 0 ) bands , respectively . [ fig : int - min ] shows the limb intensity profiles of cameron and @xmath4 uv doublet bands . spicam - observed intensities of cameron and uv doublet bands averaged over ls = 100130@xmath22 observations @xcite are also shown in the fig . [ fig : int - min ] . limb intensities of @xmath4 uv doublet and cameron bands calculated by using s2k model are @xmath26% and @xmath215% , respectively , higher compared to those obtained using euvac model . calculated intensities for both solar flux models are higher than the spicam - observed values . in analogy to the viking case , we reduced the e-@xmath19 cross sections producing cameron band . the resulting intensity profile ( also shown in fig . [ fig : int - min ] ) is still higher than the observation around emission peak . calculated intensity of @xmath4 uv doublet band is also higher near the peak emission than the observed intensity . altitude of the calculated intensity for both co cameron and uv doublet bands peaks @xmath132 to 3 km higher than the observations , which is well within the observational uncertainties . line of sight intensity of different vibrational transitions of cameron band at the altitude of peak emission , which is @xmath13120 km for first case , are shown in table [ tab : vib - oi ] . [ fig : ratio - min ] shows the calculated intensity ratio of uv doublet to cameron band along with the observed ratio derived from spicam observations @xcite . at lower altitudes calculated ratio is in agreement with observation ( @xmath130.18 ) . the ratio remains almost constant up to @xmath2120 km ( where cameron band and uv doublet emission peaks ) , starts gradually decreasing with altitude and becomes almost constant after 150 km . the observed ratio decreases almost monotonically from 100 km all the way to 180 km . @xcite have not found any dependence of sza on the uv doublet to cameron band intensity ratios , though they have observed a weak dependence of this intensity ratio on the solar activity . from the observed intensity ratio profile it is clear that in upper atmosphere cameron band intensity is increasing steadily compare to uv doublet band , which indicates a difference in the source of production of cameron band and uv doublet band . that source could be the dissociative recombination process which is sensitive to the density of @xmath4 ion ( as shown in the fig . [ fig : ionden ] ) . loss of @xmath4 ions at higher altitudes ( @xmath45 200 ) can reduce the intensity of uv doublet and hence decreases the intensity ratio value . as discussed earlier , due to dust storm during spicam observations for ls greater than 130@xmath22 , atmospheric densities were higher resulting in altitude of peak emission shifting to higher altitudes ( @xmath2132.5 km for cameron band emission ) . for mariner 6 and 7 observations the intensity of co cameron band peaked at altitude of @xmath2133 km . mariner observations were carried out during solar maximum condition ( f10.7 @xmath2 180 ) , whereas spicam observations are made during moderate solar activity condition . to model dayglow emissions for ls @xmath53 130@xmath22 , we have made calculation for mex orbit 1426 ( 26 feb . 2005 ) , taking model atmosphere from @xcite for high solar activity condition . sun - mars distance is 1.5 au , f10.7 = 98 ( f10.7a = 97 ) . the euv flux at 1 au calculated using euvac model remains the same for first ( ls @xmath32 130@xmath22 ) and second ( ls @xmath45130@xmath22 ) cases . this is because in the euvac model average of f10.7 and f10.7a ( 81-day average ) is used to scale the solar flux , and on both days average of f10.7 and f10.7a does not change ( it is 97.5 on both days ) , although the f10.7 flux increased by 10 unit . s2k model does not depend on the f10.7 alone , but on other proxies also @xcite , hence flux calculated using s2k model is different on the two days . [ fig : ver - cam-138 ] ( upper panel ) shows the ver of co(a@xmath1)calculated using s2k and euvac models . total ver calculated using s2k is about 17% higher than that calculated using euvac model . major differences are in the co(a@xmath1)production due to dissociative excitation of @xmath19 by photon and electron impact , which are more than 50% and 10% higher , respectively , when s2k model is used . total production rate of cameron band ( calculated using euvac model ) maximises at an altitude of 134 km with a value of about 3328 @xmath50 s@xmath51 , which is around 10 km higher than that in the first case ( ls @xmath32 130@xmath22 ) . although production rate ( 3528 @xmath50 s@xmath51 ) at peak altitude is higher in the first case , but at higher altitudes rate increases faster in the second ( ls @xmath45 130@xmath22 ) case , e.g. , at 200 km , cameron band production rate is 61 @xmath50 s@xmath51 in second case , whereas in first case it is only 3 @xmath50 s@xmath51 . in both , first and second cases , for euvac model , the altitude where photodissociation of @xmath19 takes over electron impact dissociation of @xmath19 is around 30% higher than that for s2k model . bottom panel of fig . [ fig : ver - cam-138 ] shows production rates of @xmath4 uv doublet band . total excitation rate calculated using the s2k model is about 12% higher than that calculated using euvac model . similar to the cameron band , @xmath4 uv doublet production rate at peak is lower than that in the first case , but at higher altitudes uv doublet production rate becomes higher in the second case . table [ tab : oi - cmp ] shows the overhead intensities of cameron and @xmath4 uv doublet bands . contribution of photodissociation of @xmath19 , e-@xmath19 , dr , and e - co processes to the total cameron band production is 16 ( 22% ) , 62 ( 57% ) , 11 ( 11% ) , and 9 ( 8% ) , respectively , when euvac ( s2k ) model is used . [ fig : int - max ] shows the calculated limb intensity of uv doublet and cameron band along with spicam - observed intensities of cameron band and @xmath4 doublet band for mex orbit 1426 on 26 feb . 2005 @xcite . intensities calculated using s2k model are higher by @xmath131218% than those calculated using euvac model . altitude of peak emission of calculated and observed intensity profiles is in good agreement with each other ( e.g. , @xmath2128 km for cameron band ) within the uncertainties of observations and model calculations . however , intensities calculated using both solar flux models are higher than the observations . calculated intensities of co cameron band after reducing the e-@xmath19 cross section by a factor of 2 are also shown in fig . [ fig : int - max ] , which is in good agreement with the observed values . however , the calculated intensity of uv doublet band is 2030% higher than the spicam - observed values . uncertainties in photoionization cross sections for the production of co@xmath30 can also be one of reasons for the differences between observed and calculated intensity of co@xmath30 uv doublet emission . in our calculation ( using euvac model and reduced e-@xmath19 cross section for cameron band production ) , peak brightness of cameron band is 249 ( 255 ) kr with peak located at 120 ( 128 ) km in first ( second ) case . for @xmath4 uv doublet band , peak brightness is 31 ( 33 ) kr with peak located at 119 ( 127 ) km in first ( second ) case . altitude of peak brightness remains the same when s2k model is used but intensities of cameron and @xmath4 uv doublet bands for first ( second ) case are about a factor of 1.1 ( 1.18 ) and 1.07 ( 1.11 ) , respectively , higher than that calculated using euvac model . first dayglow measurements on mars were carried out by mariner series of spacecraft . mariner 6 and 7 observations were taken during the solar maximum conditions ( july - august , 1969 ; f10.7 = 186 at 1 au ) , and are the only martian dayglow observations so far taken during the solar maximum condition . mars was also at perihelion ( distance between sun and mars was around 1.42 au ) during mariner observations . we run our model for the condition similar to the mariner observation . model atmosphere for solar maximum condition is taken from @xcite . [ fig : solar - flx ] ( bottom panel ) shows the solar euv flux during mariner observations ( 2 aug . 1969 , f10.7 = 180 ) from euvac and s2k models . as discussed in section [ sec : model ] , overall the solar euv flux is higher in s2k model , except at wavelengths below 250 where euvac flux is larger . [ fig : pef ] ( bottom panel ) shows photoelectron fluxes calculated at different heights in solar maximum condition using both solar flux models . unlike in the solar minimum condition , the ratio of photoelectron flux calculated using s2k to that of euvac is less than 1 at most of the energies . at electron energies below 30 ev , fluxes calculated using s2k and euvac models are almost equal at altitudes of 130 and 160 km . above 30 ev , photoelectron fluxes calculated using euvac model are higher than those calculated using s2k model . the higher photon fluxes in euvac model at shorter wavelengths below 250 [ cf . [ fig : solar - flx ] ] produce higher energy photoelectrons that can produce more secondary electron through ionization , and hence compensates for the higher solar euv flux above 250 in the s2k model . table [ tab : oi - cmp ] shows the calculated overhead intensities of cameron band and uv doublet band using euvac and s2k models . cameron band production due to electron impact dissociation of @xmath19 is higher when euvac model is used , which is due to the higher photoelectron fluxes . contribution of photodissociation excitation calculated using s2k model is higher , due to higher euv flux at longer wavelengths ( specially flux in the 10001050 bin ) . [ fig : ver - cam - max ] ( upper panel ) shows the ver of co(a@xmath1 ) for higher solar activity condition , calculated using euvac and s2k solar flux models . due to the higher photoelectron flux , co(a@xmath1 ) production due to e-@xmath19 , e - co , and dissociative recombination are higher when euvac model is used . photodissociative excitation of @xmath19 producing cameron band for s2k model is still higher by @xmath250% than for euvac model , which is due to the higher euv fluxes at longer wavelengths in the s2k model . similar to that in the previous cases , the cross over point between photodissociation and electron impact dissociation of @xmath19 forming co(a@xmath1 ) occurs at higher altitude when euvac model is used . bottom panel of fig . [ fig : ver - cam - max ] shows the production rates of @xmath4 uv doublet band . here also calculated values using euvac model is slightly higher than that calculated using s2k model . during solar minimum condition total volume production rate of cameron and uv doublet bands calculated using s2k model is higher than that calculated using euvac model , whereas in solar maximum it is vice - versa . except photodissociation excitation process producing cameron band , production rates due to other processes calculated by using euvac model are higher than that calculated by using s2k model . in both , solar minimum and maximum conditions , cameron band production due to photodissociative excitation is about 50% higher , when s2k model is used . [ fig : int - mariner ] shows model intensities of cameron band and @xmath4(b x ) ultraviolet doublet band calculated using both euvac and s2k models at sza = 45@xmath22 along with intensities observed by mariner 6 and 7 . limb intensities measured by mariner 6 and 7 on mars are at sza = 27@xmath22and 0@xmath22 , and at sza = 44@xmath22 and 0@xmath22 , respectively . calculated limb intensities using euvac model at sza = 0@xmath22 are also shown in the fig . [ fig : int - mariner ] . limb intensities calculated using euvac model are only slightly higher than those calculated using s2k model . there were no observations at emission peak for both cameron band and @xmath4(b x ) band . below the emission peak , there are few observations for cameron band , but ultraviolet doublet observations were not available . solar zenith angle effect is clearly visible at altitudes below 150 km , where intensity is larger and emission peak shift deeper in the atmosphere for lower sza . calculated intensities of cameron and uv doublet bands are lower than the observed values . unlike previous cases , calculated intensities of cameron and uv doublet band emissions at the altitude of peak emission are slightly lower than the observation . @xcite has pointed out that due to calibration problem in mariner 6 and 7 instrument the observed values can be higher . as in the previous cases a reduction in e-@xmath19 cross section is required to get an agreement between observed and calculated intensity . calculated intensity of cameron band after reducing the e-@xmath19 cross section by a factor of 2 is also shown in figure [ fig : int - mariner ] for sza=0@xmath22 . density of co(a@xmath1 ) is calculated under photochemical equilibrium condition . radiative decay is the dominant loss process of co(a@xmath1 ) , the contribution from other processes are negligible ( cf . * ? ? ? [ fig : ionden ] shows the density of co(a@xmath1 ) calculated using euvac and s2k euv flux models for low solar activity condition . the calculated column density of co(a@xmath1)molecule is @xmath55 ( @xmath56 ) @xmath57 for the solar minimum condition using euvac ( s2k ) solar euv flux model . except in the solar maximum condition , density of co(a@xmath1 ) molecule calculated using s2k model is higher than that calculated using euvac model . during solar maximum condition , co(a@xmath1 ) density calculated using euvac model is slightly higher at peak ( around 5% ) , but at altitudes above 140 km , density calculated using s2k model becomes higher ( @xmath210% at 200 km ) . the shape of the density of co(a@xmath1 ) is similar to that of its production rate ( cf . [ fig : ver - viking ] ) since the main loss mechanism of co(a@xmath1 ) is radiative decay whose value is independent of altitude . hence , the density of co(a@xmath1 ) in the martian atmosphere can be represented by @xmath58 = \frac{\mathrm{[co_2]\,(k_1 + k_2 ) + [ co_2^+]\,[n_e]\ , k_3}}{\mathrm{k_4 } } \label{eq : coa - den}\ ] ] where @xmath59 , @xmath60 , @xmath61 , and @xmath62 are as described in table [ tab : rate - cameron ] . @xmath59 and @xmath60 are photodissociation rate and electron impact dissociation rate of @xmath19 , respectively , @xmath61 is dissociative recombination , @xmath62 is radiative decay loss , and @xmath63 is the electron density . the values of @xmath59 ( photodissociation rate ) in units of s@xmath51 at the top of atmosphere in case of euvac ( s2k ) model are @xmath64 ( @xmath65 ) , @xmath66 ( @xmath67 ) , @xmath68 ( @xmath69 ) , and @xmath70 ( @xmath71 ) in the solar minimum , first case , second case , and solar maximum , respectively . present study deals with the model calculations of co cameron band and @xmath4doublet ultraviolet emissions in martian dayglow and the impact of solar euv flux on the calculated intensities . photoelectrons generated due to photoionization in the martian atmosphere have been degraded in the atmosphere using monte carlo model based analytical yield spectrum technique . emission rates of cameron and @xmath4 uv doublet bands due to photon and electron impact on @xmath19 have been calculated using euvac and s2k solar euv flux models . densities of prominent ions and co(a@xmath1 ) in martian upper atmosphere are calculated under steady state photochemical equilibrium condition . production rates of cameron and @xmath4 uv doublet bands are height - integrated to calculate overhead intensity and along the line of sight to obtain limb intensities . limb intensities are compared with the spicam / mars express and uv spectrometer / mariner observed intensities . due to higher euv fluxes at longer ( 7001050 ) wavelengths in the s2k model , the contribution of photodissociation of @xmath19 in producing cameron band is about 50% higher in low as well as in high solar activity conditions . variations in euv fluxes at longer wavelengths from solar minimum to solar maximum are less prominent in the euvac solar euv flux model compared to the s2k model . for low solar activity condition , limb intensities of cameron and @xmath4 uv doublet bands around peak brightness calculated using s2k model are @xmath23040% higher than those calculated using euvac model . comparison of calculated intensities has been made with the spicam - observed values for condition similar to the viking . intensities calculated using both s2k and euvac models are higher than the observed values . calculated altitude of emission peak of co cameron and @xmath4 uv doublet bands is also higher by @xmath25 km than the observed value . a reduction in the e-@xmath19 cross section forming cameron band by a factor of 2 and the density of @xmath19 in model atmosphere by a factor of 1.5 brings the calculated intensity ( using euvac model ) of cameron band in close agreement with the spicam observation . while modelling the recent observations made by spicam on - board mars express , we have taken two set of conditions with different model atmospheres and solar longitudes . in the first case , ls @xmath54 130@xmath22 ( ls = 100 - 130@xmath22 ) , atmosphere is taken from mars thermospheric general circulation model @xcite and calculations are made for the day 24 october 2004 with moderated solar activity flux ( f10.7 = 88 ) . total intensities of co cameron and @xmath4 uv doublet bands calculated using s2k model are around @xmath2615% higher than those calculated using euvac model . contribution of @xmath19 photodissociative excitation in cameron band production is 50% higher when s2k model is used . dissociative recombination of @xmath4 is an important source of cameron band in this case ( cf . [ fig : ver - cam - shem ] ) due to higher densities of @xmath4 ion ( fig . [ fig : ionden ] ) compared to those calculated for low solar activity condition ( viking condition ) . calculated intensities of cameron and uv doublet bands have been compared with the spicam - observed limb intensities . intensities calculated using s2k and euvac solar flux models are higher than the observed values by a factor of 1.7 to 2 for cameron and a factor of 1.4 for uv doublet bands ( see fig . [ fig : int - min ] ) . we found that altitude of peak emission of both cameron and uv doublet bands are 2 to 3 km higher than observed profiles . this discrepancy in observed and calculated intensities and altitudes could be due to the fact that observed values are averaged over several days of observations while calculation are carried out for a particular day . due to the dust storm during ls @xmath45 130@xmath22 , observed emission peak is around 10 km higher for both cameron and uv doublet compared with the spicam - observed values for mex orbit 1426 on 26 feb . 2005 @xcite . to model the emission during ls @xmath53 130@xmath22 , we have taken atmospheric model for solar maximum condition . intensities calculated using the s2k solar flux model are @xmath2818% higher than those calculated using the euvac model . these calculated intensities are higher than the observed - averaged values by a factor of @xmath132 for cameron band and @xmath1350% for uv doublet band . the calculated intensity of cameron band ( after reducing the e-@xmath19 cross sections by a factor of 2 ) is in agreement with the observed values ( fig . [ fig : int - max ] ) . in all three conditions discussed above , i.e. , low solar activity ( viking ) , and first ( ls @xmath32 130@xmath22 ) and second ( ls @xmath45 130@xmath22 ) cases , calculated intensities of both cameron and uv doublet bands are higher than observations . on an average , model values of @xcite for cameron and @xmath4 uv doublet bands are around a factor of 1.74 and 1.41 , respectively , higher than the spicam observations . @xcite also found that their calculated intensities of cameron and uv doublet bands are around 25% higher than the spicam - observed values . this shows that these discrepancies in the model and observed values are due the uncertainties in the input physical parameters in the model . uncertainties in cross sections , namely , e-@xmath19 cross section producing cameron band and photoionization of @xmath19 forming uv doublet band can be one of the causes of discrepancies in the model and observations . calculations are also made for the high solar ( mariner 6 and 7 observations ) activity condition . though the contribution of @xmath19 photodissociative excitation in cameron band production is higher when s2k model is used , but the total intensity of co cameron band calculated using the euvac model is slightly higher than that calculated using the s2k model . this is because of the higher photoelectron flux when euvac model is used ( fig . [ fig : pef ] ) . the calculated intensities of both cameron and uv doublet band are lower than the observed values ( fig . [ fig : int - mariner ] ) . 1 . generally , solar euv fluxes in bands are higher in s2k model except at few bands at shorter wavelength range ( @xmath32 250 ) . solar euv fluxes at longer wavelengths are higher in s2k model , specially in the 1000 - 1050 bin , where flux is around an order of magnitude higher than the corresponding flux in euvac model . solar euv flux at lines is smaller in the s2k model compared to that in the euvac model . 2 . due to higher euv flux at lines in the euvac model , peaks at 2030 ev range in the photoelectron flux are more prominent when euvac model is used . 3 . during high solar activity condition , calculated photoelectron fluxes are higher for euvac model due to higher euv fluxes below 250 in the euvac model . hence , intensities calculated using euvac model are higher by 510% than those calculated using s2k model . 4 . during solar minimum condition , the cameron and @xmath4 uv doublet intensities calculated using s2k solar flux model are @xmath23040% higher than those calculated using the euvac model . 5 . during both , solar minimum as well as maximum conditions , cameron band production due to photodissociative excitation of @xmath19 is about 50% higher when s2k solar euv flux model is used . altitude of peak production rate of cameron and @xmath4 uv doublet bands is independent of solar euv flux model used in the calculations . however , for the cameron band the altitude where photodissociation of @xmath19 takes over electron impact dissociation is higher in the euvac model compared to that in the s2k model . 7 . reduction in the e-@xmath19 cross section producing cameron band and photoionization cross section producing @xmath4 uv doublet band is required to bring the model calculations in agreement with the observations . 8 . for a given set of observation , and accounting for the uncertainties in the cross sections , intensities calculated using the euvac model are in better agreement with the observation than those calculated using the s2k model . simultaneous observation of solar euv flux with dayglow measurements would be very helpful in improving our understanding of the processes that governs the dayglow emissions on mars . more accurate measurements of cross sections for electron impact dissociation of @xmath19 producing cameron band and photoionization of @xmath19 in @xmath29 state are required for the better modelling of co cameron band and @xmath4 uv doublet band in martian atmosphere , as well as in other @xmath19containing atmospheres , like venus and comets . avakyan , s. v. , iiin , r. n. , lavrov , v. m. , ogurtsov , g. n. , 1998 . in : avakyan , s. v. ( ed . ) , collision processes and excitation of uv emission from planetary atmospheric gases : a handbook of cross sections . gordon and breach science publishers . barth , c. a. , hord , c. w. , pearce , j. b. , kelly , k. k. , anderson , g. p. , stewart , a. i. , 1971 . mariner 6 and 7 ultraviolet spectrometer experiment : upper atmosphere data . j. geophys . 76 , 2213 2227 . . bhardwaj , a. , 1999 . on the role of solar euv , photoelectrons , and auroral electrons in the chemistry of c(@xmath72d ) and the production of ci 1931 in the inner cometary coma : a case for comet p / halley . j. geophys . 104 , 1929 1942 . . bhardwaj , a. , singhal , r. p. , 1993 . optically thin h lyman alpha production on outer planets : low - energy proton acceleration in parallel electric fields and neutral h atom precipitation from ring current . j. geophys . 98 ( a6 ) , 9473 9481 . . bougher , s. w. , engel , s. , hinson , d. p. , murphy , j. r. , 2004 . mgs radio science electron density profiles : interannual variability and implications for the martian neutral atmosphere . j. geophys . res . 109 . . bougher , s. w. , engel , s. , roble , r. g. , foster , b. , 1999 . comparative terrestrial planet thermospheres : 2 . solar cycle variation of global structure and winds at equinox . j. geophys . 104 , 16591 16611 . . bougher , s. w. , engel , s. , roble , r. g. , foster , b. , 2000 . comparative terrestrial planet thermospheres : 3 . solar cycle variation of global structure and winds at solstices . 105 , 17669 17692 . . buonsanto , m. j. , richards , p. g. , tobiska , w. k. , solomon , s. c. , tung , y. .-k . , fennelly , j. a. , 1995 . ionospheric electron densities calculated using different euv flux models and cross sections : comparison with radar data . j. geophys . 100 ( a8 ) , 14569 14580 . . forget , f. , montmessin , f. , bertaux , j. l. , galindo , f. g. , lebonnois , s. , qumerais , e. , reberac , a. , dimarellis , e. , valverde , m. a. l. , 2009 . density and temperatures of the upper martian atmosphere measured by stellar occultations with mars express spicam . j. geophys . res . 114 . . gilijamse , j. j. , hoekstra , s. , meek , s. a. , metsl , m. , van de meerakker , s. y. t. , t , s. y. , meijer , g. , groenenboom , g. c. , c. , g. , 2007 . the radiative lifetime of metastable co ( a@xmath76=0 ) . j. chem . 127 , 2211024 . . haider , s. a. , bhardwaj , a. , 2005 . radial distribution of production rates , loss rates and densities corresponding to ion masses @xmath7740 amu in the inner coma of comet halley : composition and chemistry . icarus 177 , 196 216 . . halmann , m , laulicht , i , february 1966 . isotope effects on vibrational transition probabilities.iv . electronic transitions of isotopic c@xmath3 , co , cn , h@xmath3 , and ch molecules . astrophysical journal supplement 12 , 307 321 . . lean , j. l. , woods , t. n. , eparvier , f. g. , meier , r. r. , strickland , d. j. , correira , j. t. , evans , j. s. , 2011 . solar extreme ultraviolet irradiance : present , past , and future . j. geophys . res . 116 . . rosati , r. e. , johnsen , r. , golde , m. f. , 2003 . absolute yields of co ( @xmath79 , @xmath80 , @xmath81 ) + o from the dissociative recombination of co@xmath0 ions with electrons . j. chem . 119 , 11630 11635 . . seiersen , k. , al - khalili , a. , heber , o. , jensen , m. j. , nielsen , i. b. , pedersen , h. b. , safvan , c. p. , andersen , l. h. , aug 2003 . dissociative recombination of the cation and dication of co@xmath3 . a 68 ( 2 ) , 022708 . . shematovich , v. i. , bisikalo , d. v. , grard , j .- c . , cox , c. , bougher , s. w. , leblanc , f. , 2008 . monte carlo model of electron transport for the calculation of mars dayglow emissions . j. geophys . res . 113 . . , m. p. , gougousi , t. , johnsen , r. , golde , m. f. , may 1998 . measurement of the absolute yield of co(a@xmath82@xmath83)+o products in the dissociative recombination of co@xmath84 ions with electrons . j. 108 , 8400 8407 . . tobiska , w. k. , woods , t. , eparvier , f. , viereck , r. , floyd , l. , bouwer , d. , rottman , g. , white , o. r. , 2000 . the solar2000 empirical solar irradiance model and forecast tool . 62 , 1233 1250 . . lccccccccc band & band origin & & & + ( @xmath86 ) & & euvac & s2k & & euvac & s2k + 0 - 0 & 2063.0 & 716 & 845 & & 20 & 24 + 0 - 1 & 2158.4 & 713 & 840 & & 20 & 23.2 + 0 - 2 & 2261.0 & 311 & 367 & & 9 & 10 + 0 - 3 & 2374.0 & 79 & 93 & & 2 & 2.6 + 1 - 0 & 1992.5 & 1151 & 1358 & & 32 & 38 + 2 - 0 & 1927.5 & 603 & 711 & & 17 & 20 + 3 - 0 & 1868.1 & 181 & 213 & & 5 & 6 + 4 - 0 & 1813.0 & 38 & 45 & & 1 & 1.3 + . the ratio of the photoelectron flux at 130 km calculated using the two solar flux models is also shown with magnitude on right side y - axis . thin dotted horizontal line depicts the s2k / euvac ratio = 1 . ] and o@xmath41 ions and co(a@xmath1 ) molecule for solar minimum condition calculated using euvac ( solid curve ) and s2k ( dashed curve ) solar euv flux models . density of co(a@xmath1 ) molecule is plotted after multiplying by a factor of 100 . dotted curves show the densities of @xmath4 and o@xmath41 ions for first case ( ls @xmath54 130@xmath22 ) using euvac solar flux . ] uv doublet bands ( left panel ) and co cameron ( right panel ) for low solar activity condition . solid squares with error bars represents the spicam - observed values taken from @xcite . dashed curves show the calculated intensity ( using euvac model ) after reducing the density of @xmath19by a factor of 1.5 . dash - dotted curve shows the calculated intensity ( using euvac ) of cameron band with reduced density ( by a factor of 1.5 ) and reduced ( by a factor of 2 ) e-@xmath19 cameron band production cross section . ] uv doublet band ( left panel ) and co cameron band ( right panel ) for ls @xmath32 130@xmath22 . symbols represent the spicam - observed values taken from @xcite . dash - dotted curve shows the calculated intensity of cameron band with reduced ( by a factor of 2 ) e-@xmath19 cross section . ] uv doublet band ( left panel ) and co cameron band ( right panel ) for ls @xmath45 130@xmath22 . dash dotted curve shows the cameron band intensity for euvac model with e-@xmath19 cross section forming co(a@xmath1 ) reduced by a factor of 2 . open circles with error bars represent the spicam - observed intensity taken from @xcite . ] . solid curve with symbols shows the limb intensity calculated using s2k model at sza = 45@xmath22 . dashed curve shows the calculated intensity ( using euvac model ) at sza = 0@xmath22 . dash dotted curve shows the calculated intensity ( using euvac model ) at sza = 0@xmath22 and after reducing the e-@xmath19 cross sections forming co(a@xmath1 ) by a factor of 2 . symbol represents the observed intensity of cameron band and uv doublet band measured by mariner 6 and 7 @xcite . observed values are shown for 2 orbits each of mariner 6 ( for sza = 27 and 0@xmath22 ; open and solid triangle , respectively ) and mariner 7 ( for sza = 44 and 0@xmath22 ; open and solid circle , respectively ) . ]

the decadal irac botes survey ( dibs ) is a mid - ir variability survey of the sq . deg . of the ndwfs botes field and extends the time baseline of its predecessor , the _ _ deep , wide - field survey ( sdwfs ) , from 4 to 10 years . the _ spitzer space telescope _ visited the field five times between 2004 and 2014 at 3.6 and 4.5 microns . we provide the difference image analysis photometry for a half a million mostly extragalactic sources . in mid - ir color - color plane , sources with quasar colors constitute the largest variability class ( 75% ) , 16% of the variable objects have stellar colors and the remaining 9% have the colors of galaxies . adding the fifth epoch doubles the number of variable agns for the same false positive rates as in sdwfs , or increases the number of sources by 20% while decreasing the false positive rates by factors of 2 - 3 for the same variability amplitude . we quantify the ensemble mid - ir variability of spectroscopically confirmed agns using single power - law structure functions , which we find to be steeper ( index ) than in the optical ( ) , leading to much lower amplitudes at short time - lags . this provides evidence for large emission regions , smoothing out any fast uv / optical variations , as the origin of infrared quasar variability . the mid - ir agn structure function slope seems to be uncorrelated with both the luminosity and rest - frame wavelength , while the amplitude shows an anti - correlation with the luminosity and a correlation with the rest - frame wavelength .

1509.02467

there is increasing evidence that understanding the evolution of active galactic nuclei ( agns ) is crucial for understanding galaxy formation and evolution ( e.g. , ) . this has driven a rapid growth in attempts to comprehensively characterize the properties and demographics of agns . the _ spitzer space telescope _ has played a significant role in this effort through the development of mid - ir agn selection methods ( e.g. , @xcite ) and detailed studies of agn spectral energy distributions ( seds ; @xmath4 ) in the near / far - ir ( e.g. , @xcite ) . these studies have expanded further based on the all - sky _ wide - field infrared survey explorer _ ( _ wise _ ; @xcite ) mid - ir survey data ( e.g. , @xcite ) . there is another dimension in which _ spitzer _ can significantly expand its studies of agn demographics and physics , namely the time domain . agns are variable sources at all wavelengths ( e.g. , ) . their seds have a minimum near a rest wavelength of 1 and rise towards both the `` blue '' and the `` red '' ( e.g. , @xcite ) . the blue , optical side corresponds to emission from the accretion disk surrounding a super massive black hole at the center of an agn , while the red mid - ir side is due to emission by hot dust ( e.g. , ) as shown by near - ir reverberation mapping studies of nearby agns ( e.g , @xcite ) but also by direct observations of the torus ( @xcite ) . depending on redshift , the 3.64.5 emission probes from near the sed minimum ( @xmath5 ) into the dust - dominated regime ( @xmath6 ) . we know a great deal about how quasars vary , but we know relatively little about the origins of that variability . possible origins include accretion disk instabilities , surface temperature fluctuations , or variable heating from coronal x - rays ( e.g. , ) . the introduction of stochastic process models , particularly the damped random walk ( drw ; @xcite ) , has provided a means to quantitatively describe the variability of increasingly improving ( in length and cadence ) quasar light curves from optical surveys ( such as sdss , @xcite , ogle , @xcite , la silla - quest , @xcite , or crts , @xcite ) or reverberation mapping campaigns ( e.g. , @xcite ) . the drw model is characterized by two parameters , the time scale and amplitude , where the time scale has been tied to either disk orbital or thermal time scales ( @xcite ) . while it well - describes quasar light curves with a typical cadence of days ( @xcite ) , there appears to be a break in the power spectrum distribution at short times scales observed for highly sampled _ kepler _ agns ( @xcite ) . sparser light curves for hundreds or thousands of agns are typically analyzed with an ensemble structure function analysis ( e.g. , @xcite ) , commonly described by a single power - law function with the slope index @xmath3 and amplitude @xmath7 at a fixed time scale , although on long time scales there is clearly a break ( @xcite ) . in @xcite , we used a structure function analysis to study the ensemble quasar variability of the largest repeatedly observed mid - ir agn sample from the _ spitzer _ deep , wide - field survey . the survey observed the ndwfs botes field ( ( ra , decl.@xmath814:30 , @xmath934:05 ) ) four times between 2004 and 2008 ( @xcite ) . it is a unique 9 deg@xmath10 area of the sky , far from the galactic plane ( @xmath067 deg ) , that has been observed at many different wavelengths to provide a good census of extragalactic objects . the agn and galaxy evolution survey ( ages ; @xcite ) provides redshifts for over 23,500 objects . the noao deep wide - field survey ( ndwfs ; @xcite ) provides deep optical images and photometry in the @xmath11 bands , while the _ spitzer _ deep , wide - field survey ( sdwfs ; @xcite ) provides magnitudes in four mid - ir bands at 3.6 , 4.5 , 5.8 , and 8.0 microns . it was also observed at x - ray ( xbotes ; @xcite ) , at uv by _ galex _ ( @xcite ) at 24 ( mages survey ) , at far - ir by _ herschel space observatory _ ( ) , and radio ( first ; @xcite and wsrt ; @xcite ) wavelengths . in this paper , we extend our previous analysis using the decadal irac botes survey ( dibs ) , which stretched the time baseline of the survey to 10 years by adding a fifth epoch . this expansion enables two improvements over our earlier study . first , the longer temporal baseline provided by the fifth epoch significantly improves the structure function slope measurement , our ability to characterize the differences between optical and mid - ir variability , and to subdivide the sample based on other criteria ( e.g. , x - ray or radio flux ) . the ensemble quasar variability in mid - ir can be described by a single power - law structure function with the index @xmath12 , hence increasing the time baseline from 4 ( sdwfs ) to 10 years ( dibs ) , typically increases quasar variability by 50% . second , existing mid - ir agn samples are strongly biased against low luminosity agn and variability provides a means of removing this bias to obtain a more complete census the addition of the fifth epoch should improve the completeness and/or purity of the variability selection . please note , however , that the typical @xmath13 of mid - ir quasar variability is @xmath00.1 mag , which is comparable to the uncertainties provided by dibs or sdwfs for typical sources , hence we detect only significantly variable or brighter quasars . the paper is organized as follows . in section [ sec : data ] , we present the available data and data analysis . in section [ sec : definitions ] , we introduce the variability measures , while in section [ sec : results ] , we discuss the results of the quasar selection methods . the structure function analysis is presented in section [ sec : sfunction ] , and the paper is summarized in section [ sec : summary ] . lc|ccccc|ccccc 217.659222 & 32.452357 & 18.677 & 18.742 & 18.777 & 18.920 & 18.977 & 0.141 & 0.139 & 0.137 & 0.131 & 0.153 + 217.730372 & 32.452616 & 19.048 & 18.793 & 19.001 & 19.217 & 19.029 & 0.158 & 0.159 & 0.155 & 0.147 & 0.173 + 217.787424 & 32.452656 & 19.349 & 19.573 & 20.245 & 19.596 & 99.999 & 0.207 & 0.204 & 0.196 & 0.192 & 99.999 + 217.628739 & 32.452706 & 19.119 & 19.483 & 19.180 & 19.391 & 19.950 & 0.179 & 0.175 & 0.174 & 0.166 & 0.190 + 217.618530 & 32.452833 & 18.907 & 18.853 & 19.024 & 18.676 & 18.973 & 0.147 & 0.146 & 0.143 & 0.139 & 0.160 lc|ccccc|ccccc 217.659222 & 32.452357 & 18.875 & 19.046 & 19.653 & 18.559 & 18.317 & 0.192 & 0.158 & 0.170 & 0.172 & 0.164 + 217.730372 & 32.452616 & 18.844 & 18.433 & 19.255 & 18.652 & 18.526 & 0.170 & 0.143 & 0.152 & 0.150 & 0.143 + 217.787424 & 32.452656 & 99.999 & 99.999 & 99.999 & 19.855 & 19.689 & 99.999 & 99.999 & 99.999 & 0.321 & 0.304 + 217.628739 & 32.452706 & 18.837 & 18.673 & 19.173 & 19.397 & 19.468 & 0.206 & 0.171 & 0.184 & 0.177 & 0.167 + 217.618530 & 32.452833 & 19.314 & 19.029 & 18.570 & 18.273 & 18.591 & 0.173 & 0.143 & 0.161 & 0.158 & 0.147 + + a detailed description of sdwfs is presented in @xcite . the first epoch was taken on 2004 january 10 - 14 as the irac shallow survey ( @xcite ) . then , we added three more epochs on 2007 august 8 - 13 , 2008 february 2 - 6 , and 2008 march 6 - 10 as the _ spitzer _ deep , wide - field survey ( @xcite ; d. stern pi , pid 40839 ) . the new , fifth epoch was taken on 2014 april 23 - 29 as the decadal irac botes survey ( dibs ; m.l.n . ashby pi , pid 10088 ) . the intervals between any two epochs span the range from 1 month to 10 years . each epoch consists of @xmath020,000 individual @xmath14 pixel irac ( @xcite ) frames that were combined into a single mosaic . we have re - reduced the original sdwfs images and created new mosaics , each approximately @xmath15 pixels , with a pixel scale of 0.60 arcsec . these mosaics cover the same area of the sky , but at a different pixel resolution from sdwfs ( @xmath16 pixels and 0.84 arcsec ) . we used the same photometric catalogs as provided in @xcite . due to the slightly different mosaicing of the fifth epoch and a slightly modified image subtraction procedure ( described below ) , we report the photometric measurements for a total of 514975 objects . the photometry of our sources was done using difference imaging methods ( e.g. , @xcite ) , where we use the implementation of @xcite . first , all ten mosaics ( five for each of the @xmath17 m and @xmath18 m bands ) were aligned to the same reference frame . we then cut them into a slightly overlapping @xmath19 grid of @xmath20 pixel subimages ( with 20 pixel overlaps ) . because the _ spitzer _ point spread function ( psf ) is `` triangular '' with diffraction spikes , we pre - filtered these images to obtain gaussian - like psfs ( see @xcite for details ) , which greatly helps the difference image analysis ( dia ) software obtain clean subtractions and photometry . having the corrected images , we chopped each of them into even smaller @xmath21 grids of @xmath22 pixel chunks ( again with 20 pixels of overlap ) to speed up the dia analysis . we stacked all five epochs in order to create template images and then use dia to match and subtract the reference images from each individual epoch . we measured the fluxes of the objects in the sdwfs catalogs on the template image to photometrically align them to the sdwfs catalog . they were measured using the psf fitting routines from the dia package of @xcite . dia typically returns underestimated uncertainties , hence they were corrected with @xmath23 , where @xmath24 and @xmath25 were set so the new error bars @xmath26 match the dispersions for non - variable field objects . after subtraction of the template , the difference fluxes at the positions of all sources were measured and added to the fluxes measured on the template , thereby creating the light curves ( tables [ tab : sdwfslc1 ] and [ tab : sdwfslc2 ] ) . our catalogs are limited to the objects detected on the template image . a transient which does not produce a detectable source in the ( averaged ) template image and is not present in the sdwfs catalog will not be identified . the brightest objects create multiple variable artifacts in their vicinity . we masked the regions around these objects using the two micron all sky survey ( 2mass ) catalog ( @xcite ) , where the masking region radius was @xmath27 arcsec for @xmath28 mag stars . the absolute magnitudes were derived using a standard @xmath29cdm cosmological model with @xmath30 and using k - corrections derived from @xcite . -[8.0]$ ] colors ( `` photometry group 3 '' sources ; see [ sec : definitions ] ) with 4 or 5 epochs of data for variability level 2 , 3 and 4 ( @xmath31 , @xmath32 and @xmath33 ) sources . the true variables are the sources to the right of the vertical dashed line , in the high correlation region ( @xmath34 ) . the poorly correlated ( @xmath35 ) sources were used to estimate our false positive rates inside the high correlation region ( tables [ tab : resultsdibs ] , [ tab : resultswave ] , and [ tab : resultsages ] ) . [ fig : covariance],width=309 ] magnitude . red , blue , and black lines are for high correlation sources with @xmath34 , for @xmath36 , and all agns , respectively . _ bottom panel : _ ratios of histograms and cumulative distributions for highly correlated agns to all agns . [ fig : corr_hist],width=309 ] we were interested in sources with at least three epochs ( @xmath37 ) measured in both the 3.6 and 4.5 bands . from the fluxes we compute the standard deviations @xmath38)_i$ ] as a function of apparent magnitude @xmath39)_i$ ] ( figure [ fig : var3645 ] , panel ( a ) ) . we also calculated the variability covariance @xmath40 and pearson s correlation @xmath41 coefficients between the two bands , where @xmath42)^i & = & { \frac{1}{n_{\rm ep}-1}\sum_{j=1}^{n_{\rm ep}}\left(m([x])^i_j-\langle m([x])^i \rangle\right)^2 } , \\ c_{12}^i & = & \frac{1}{n_{\rm ep}-1}\sum_{j=1}^{n_{\rm ep}}\left(m([3.6])^i_j-\langle m([3.6])^i \rangle\right ) \nonumber \\ & \times & \left(m([4.5])^i_j-\langle m([4.5])^i \rangle\right ) , \\ r^i & = & \frac{c_{12}^i}{v([3.6])^i v([4.5])^i},\end{aligned}\ ] ] and @xmath43)^i \rangle$ ] is the average magnitude for the @xmath44th light curve in band @xmath45 . a correlation coefficient of @xmath46 ( @xmath47 , @xmath48 ) means that the 3.6 and 4.5 light curves are perfectly correlated ( anti - correlated , uncorrelated ) . we expect the majority of sources in the field to be non - variable galaxies , hence the distribution of @xmath38)_i$ ] is dominated by contributions from noise and not true variability . in order to select truly variable objects , we defined a variability significance threshold . let @xmath49)$ ] be the median dispersion as a function of magnitude @xmath50 and @xmath51)$ ] be the dispersion of the @xmath38)_i$ ] values around the median . the more distant the source is from the median , @xmath52)_i - v_m([x]))/\sigma([x])>0 $ ] , and the more its 3.6 and 4.5 light curves are correlated , the more likely it is to be the true variable . the typical @xmath13 of mid - ir quasar variability ( marked by dots in fig . [ fig : var3645 ] ) is @xmath00.1 mag , which is comparable to the uncertainties provided by dibs or sdwfs , hence we can detect only bright and/or significantly variable quasars . we define a combined significance of the variances in the light curves as @xmath53)_i - v_m([3.6])(m_i)}{\sigma([3.6])(m_i)}\right)^2 + \nonumber \\ & + & \left(\frac{v([4.5])_i - v_m([4.5])(m_i)}{\sigma([4.5])(m_i)}\right)^2 , \end{aligned}\ ] ] which quantifies the degree to which the object deviates from the median variances in both bands . the best candidates for true variables are objects with positive excess variances ( large @xmath54 and @xmath55 ) and strong correlations @xmath56 between the two bands . we define three variability levels 2 , 3 , and 4 ( in short vl2 , vl3 , and vl4 ) for sources exceeding the joint variability significance @xmath31 , 3 , and 4 , respectively , that will be frequently used to select variable objects and discuss selection yields ( see figure [ fig : var3645 ] , panels ( b - d ) ) . by definition half of the sources have negative @xmath52)_i - v_m([x]))/\sigma([x])$ ] and anomalously non - variable sources could also give rise to high values of @xmath54 . the distributions of @xmath52)_i - v_m([x]))$ ] are , however , very non - symmetric and while for the 3.6 or 4.5 bands there are 7443 or 7538 objects with @xmath52)_i - v_m([x]))/\sigma([x])>1 $ ] and @xmath57 , there are only 0 or 17 sources with @xmath52)_i - v_m([x]))/\sigma([x])<-1 $ ] and @xmath57 . this means that the high variability significance sources are those with positive ( high ) variability . to 3 ( to 6 ) for galaxies ( agns ) that start at @xmath58 bulls eye . the solid black trapezoid is the modified agn wedge , the dashed wedge is the old agn wedge , and the black solid rectangle is the stellar box . [ fig : col - col_explain],width=309 ] the distribution of correlation strengths @xmath59 for sources with @xmath54 exceeding a range of variability significance thresholds is shown in figure [ fig : covariance ] . it is obvious that the true variables tend to be objects with high correlations @xmath34 ( dashed vertical line ) . we can estimate the contamination as follows . we assume that uncorrelated sources ( @xmath35 ) in figure [ fig : covariance ] represent the density of false positives . if the number of identified variables with @xmath60 ( or 3 , 4 ) and @xmath34 is @xmath61 , and the number of sources with @xmath60 ( or 3 , 4 ) and @xmath35 is @xmath62 , then we estimate that the number of false positives is @xmath63 , where @xmath64 is the ratio of the intervals @xmath34 and @xmath65 . the fraction of the identified variables that are false positives is then @xmath66 . in figure [ fig : corr_hist ] , we show both the histograms and cumulative distributions for spectroscopically confirmed quasars from the ages survey with the variability exceeding vl2 as a function of the observed [ 3.6 ] magnitude . bright sources with highly correlated light curves ( @xmath34 ) constitute @xmath0100% of the agn sample , but this proportion decreases to @xmath080% with decreasing source brightness . the variability properties of dibs sources are presented in table [ tab : varcat ] . knowledge of the mid - ir colors is key to assigning objects to various classes such as agns , galaxies or stars ( figure [ fig : col - col_explain ] ) . we use the same definitions as in @xcite , where the stellar box is @xmath67-[4.5]<0.15 $ ] and @xmath68-[8.0]<0.15 $ ] , the old agn wedge ( @xcite ) is defined as @xmath69-[8.0]>0.6 $ ] mag ( left side ) , + @xmath70-[4.5]>2.5\times([5.8]-[8.0])-3.5 $ ] mag ( right side ) , + @xmath70-[4.5]>0.2\times([5.8]-[8.0])+0.18 $ ] mag ( bottom ) , + and the modified agn wedge is defined as + @xmath70-[4.5]<2.5\times([5.8]-[8.0])-0.6 $ ] mag ( left side ) , + @xmath70-[4.5]>2.5\times([5.8]-[8.0])-3.5 $ ] mag ( right side ) , + @xmath70-[4.5]>0.2\times([5.8]-[8.0])+0.08 $ ] mag ( bottom ) . + the mid - ir colors depend vitally on the brightness of objects , therefore we divide our sources into three classes that increase with our ability to characterize sources in the mid - ir color - color plane : * photometry group 1 ( pg1 ) consists of all objects that after masking have 4 or 5 epochs of data , with no constraints on either their magnitudes or magnitude uncertainties . there are 438270 photometry group 1 objects after masking , of which 9811 , 4229 and 2472 meet the variability level 2 , 3 and 4 criteria . * photometry group 2 ( pg2 ) is the subset of the photometry group 1 sources for which we can measure the [ 3.6]@xmath71[4.5 ] color . these sources must satisfy @xmath70 < 19.8 $ ] mag and @xmath72 < 19.3 $ ] mag with uncertainties in these bands smaller than 0.1 mag . there are 214433 photometry group 2 objects , of which 5325 , 2878 and 1851 meet the variability level 2 , 3 and 4 criteria . * photometry group 3 ( pg3 ) is the subset of photometry group 2 objects for which we can also measure the [ 5.8]@xmath71[8.0 ] color . these sources are required to have uncertainties smaller than @xmath73 mag for the 5.8 and 8.0 bands . there are 39591 photometry group 3 objects , of which 1615 , 840 and 504 meet the variability level 2 , 3 and 4 criteria . in table [ tab : resultsdibs ] , we present the significance statistics for the dibs sources as a function of photometry group , variability level , and the mid - ir color - color class . the most variable , highly correlated ( @xmath34 ) sources in the field are agns and stars , peaking at @xmath74 . lcccccccrrrrrrc j143038.21 + 322708.5 & 217.6592 & 32.4523 & 18.56 & 18.56 & 99.99 & 17.03 & 5 & 0.13 & 0.51 & @xmath710.04 & @xmath710.65 & 0.85 & 1 + j143055.29 + 322709.4 & 217.7303 & 32.4526 & 18.89 & 18.51 & 17.44 & 16.26 & 5 & 0.15 & 0.33 & 0.01 & 0.22 & 0.26 & 1 + j143108.98 + 322709.6 & 217.7874 & 32.4526 & 19.52 & 20.45 & 99.99 & 99.99 & 1 & 99.99 & 99.99 & 99.99 & 99.99 & 1.39 & 1 + j143030.90 + 322709.7 & 217.6287 & 32.4527 & 19.31 & 18.51 & 99.99 & 99.99 & 5 & 0.33 & 0.35 & 0.06 & 0.48 & 0.28 & 1 + j143028.45 + 322710.2 & 217.6185 & 32.4528 & 18.81 & 18.50 & 99.99 & 99.99 & 5 & 0.13 & 0.41 & 0.02 & 0.29 & 0.59 & 1 ) sources , while the open blue squares mark less correlated sources ( @xmath36 ) . the @xcite agn selection region is shown by the dashed wedge , the modified wedge from @xcite is shown as the solid line trapezoid , and the stellar box is marked with a rectangle . the contours are based on 39591 photometry group 3 objects , that are counted in 0.05 mag bins in both axes . the contours are at levels of 2 , 10 , 20 , 50 , 100 , and 150 ( counting from the outside ) . _ right panel : _ histogram of the fraction of the photometry group 2 , variability level 2 objects relative to all photometry group 2 objects as a function of the @xmath70-[4.5]$ ] color . [ fig : colcol_hist],width=309 ] we will now methodically study the variability of mid - ir sources based on their classifications . we will start with a general description of all variable objects . there are 438270 photometry group 1 objects ( with measurements in 4 or 5 epochs in both bands ) . there are 9811 ( 2.2% ) sources that show variability @xmath31 ( variability level 2 ) and 1569 ( 0.4% ) objects that also show a highly significant correlation ( @xmath34 ) . for the sources with well - defined @xmath70-[4.5]$ ] colors ( 214433 objects ) , we find 5325 variability level 2 objects ( 2.5% ) and 1206 ( 0.6% ) with a high correlation . photometry group 3 , the sources with well measured @xmath70-[4.5]$ ] and @xmath69-[8.0]$ ] colors , contains 39591 sources , of which 1615 ( 4.1% ) are variability level 2 objects and 870 ( 2.2% ) are also highly correlated . increasing the photometry group , means focusing on bright sources , where it is easier to measure variability , so it is not surprising that the variable fraction increases . for example , for photometry group 1 , variability level 2 and highly correlated ( @xmath34 ) sources with @xmath70<20 $ ] , 19 , 18 , 17 , and 16 mag , we find @xmath75 , @xmath76 , @xmath77 , @xmath78 , and @xmath79 variable sources , respectively . ccccccc + + total & & 42689 & 74751 & 40343 & 323176 & 438270 + 2 & @xmath80 & 684 ( 1.6%/14% ) & 803 ( 1.1%/21% ) & 178 ( 0.4%/37% ) & 588 ( 0.2%/100% ) & 1569 ( 0.4%/57% ) + 3 & @xmath32 & 377 ( 0.9%/11% ) & 438 ( 0.6%/18% ) & 113 ( 0.3%/21% ) & 256 ( 0.1%/100% ) & 807 ( 0.2%/47% ) + 4 & @xmath33 & 213 ( 0.5%/13% ) & 242 ( 0.3%/20% ) & 83 ( 0.2%/17% ) & 152 ( 0.0%/100% ) & 477 ( 0.1%/48% ) + + + + total & & 29339 & 55055 & 18108 & 141270 & 214433 + 2 & @xmath80 & 651 ( 2.2%/9% ) & 755 ( 1.4%/15% ) & 164 ( 0.9%/19% ) & 287 ( 0.2%/100% ) & 1206 ( 0.6%/36% ) + 3 & @xmath32 & 371 ( 1.3%/8% ) & 427 ( 0.8%/14% ) & 108 ( 0.6%/11% ) & 167 ( 0.1%/100% ) & 702 ( 0.3%/34% ) + 4 & @xmath33 & 208 ( 0.7%/11% ) & 235 ( 0.4%/17% ) & 80 ( 0.4%/10% ) & 111 ( 0.1%/100% ) & 426 ( 0.2%/38% ) + + + + total & & 6692 & 12801 & 7477 & 19313 & 39591 + 2 & @xmath80 & 593 ( 8.9%/2% ) & 652 ( 5.1%/3% ) & 143 ( 1.9%/11% ) & 75 ( 0.4%/38% ) & 870 ( 2.2%/7% ) + 3 & @xmath32 & 341 ( 5.1%/1% ) & 376 ( 2.9%/3% ) & 95 ( 1.3%/4% ) & 43 ( 0.2%/34% ) & 514 ( 1.3%/5% ) + 4 & @xmath33 & 192 ( 2.9%/2% ) & 209 ( 1.6%/3% ) & 68 ( 0.9%/3% ) & 27 ( 0.1%/40% ) & 304 ( 0.8%/6% ) in figure [ fig : colcol_hist ] , we examine the distribution of variable photometry group 3 sources in the mid - ir color - color plane . we only consider this group because both colors are needed to assign classes to objects based on their location on the color - color plane . out of 39591 objects , 870 ( 2.2% ) are variable ( vl2 ) with as little as 7% contamination ( table [ tab : resultsdibs ] ) . most of them ( 652 ; 75% ) are agn and are located in the modified agn wedge . there are also 143 vl2 stars ( 16% ) and 75 ( 9% ) objects outside these regions , hence likely agns with bright host galaxies . increasing the significance threshold to vl3 , we are left with 514 highly correlated variable objects , of which 376 ( 73% ) are agns , 95 ( 18% ) are stars , and 43 ( 8% ) are outside these regions . the final and the highest threshold , vl4 , selects 304 sources , of which 209 ( 69% ) are agns , 68 ( 22% ) are stars , and 27 ( 9% ) are other sources . it is easy to see that with the increasing significance threshold , the fractional number of agn decreases while that for stars increases . this is likely due to differences in the typical variability amplitudes : agns typical vary by only tenths of a magnitude , while stellar variability can be very spectacular , especially in the infrared ( e.g. , miras or long period variables ; see @xcite for the mid - ir variability and @xcite for the optical variability at 0.8 ) . we inspected the mid - ir color - magnitude diagram ( cmd ; @xmath70-[5.8]$ ] color vs. [ 3.6 ] magnitude ) and found that ( 1 ) all the pg3 vl2 variable objects in the stellar box ( but one ) are brighter than @xmath70<15.5 $ ] mag and ( 2 ) the stellar locus at the bright end is bi - modal , with a color difference of 0.15 mag , and that the variable stars occupy the redder branch . we suspect that our typical variable star is an rr lyrae variable . these show mid - ir variability amplitudes of @xmath00.3 mag ( @xcite ) that are similar to our sources , and with @xmath70_{\rm abs}\approx-1 $ ] mag they would have distances ( 220 kpc ) consistent with being in the galactic halo . the primary alternative , variable red giants , tend to have higher amplitudes ( 0.51.0 mag ; @xcite ) and would have implausibly large distances ( 100 kpc ) because they typically have @xmath70_{\rm abs}<-9 $ ] mag ( e.g. , @xcite ) . since a fair fraction of variable stars are bright , we also note that some of their variability may not be real but rather be due to residuals from gaussian filtering the images . by increasing the significance threshold , we lose less variable quasars more quickly than the more variable stars . . _ right panel : _ histograms of the fraction of the photometry group 2 x - ray , 24 and radio sources identified as variability level 2 variables as a function of the @xmath70-[4.5]$ ] color . [ fig : col - col_wave],width=309 ] in figure [ fig : z_abs ] , we present the dependence of variability of the ages confirmed quasars on the redshift , absolute and observed magnitudes for both bands . the most variable objects seem to be those with redshifts @xmath81 and absolute magnitudes higher than @xmath82 . we have also matched our dibs variability catalog to the xbotes x - ray catalog of the field ( @xcite ) , mips detections at 24 ( from mages ) and radio ( first ; @xcite and wsrt ; @xcite ) catalogs . we summarize the results in table [ tab : resultswave ] and in figure [ fig : col - col_wave ] . agns are known to have x - ray emission , so it is not surprising to have a significant number of matches and that they mostly lie in the agn color region ( figure [ fig : col - col_wave ] ; see also @xcite ) . out of 39591 pg3 sources , we found 2307 matches to x - ray sources , of which 448 ( 19% ) were at least as variable as variability level 2 . by increasing the variability threshold to vl4 , we are left with 160 ( 6.9% ) x - ray - selected , mid - ir variable agns . quasars are also known to have strong emission at mid- and far - ir . we matched our variability catalog to mips sources detected at 24 to find 26280 matches . of them , 713 ( 2.7% ) were vl2 objects and 238 ( 0.9% ) were vl4 sources . we also found nearly 1500 matches to radio catalogs . however , only 51 ( 3.4% ) objects were vl2 sources and there were only 17 ( 1.1% ) vl4 objects . @xcite classified the botes sources based on up to 17-band sed fits using the @xcite templates and provided photometric redshifts for the sources lacking spectroscopy . they used the @xmath83-test to compare fits with and without an agn component , where increasing values of @xmath83 make the presence of an agn more probable . because non - agns greatly outnumber agns , there will be a high false positive rate for samples of objects at @xmath83-test values that would strongly indicate the presence of an agn if only considering a single object . @xcite sources with @xmath84 almost all lie in the agn wedge , while those with @xmath85 track both the wedge and some of the more extended color region where x - ray sources are found . we have matched 3794 pg3 sources with @xmath86 , of which 440 ( 11.6% ) turned out to exceed vl2 and 137 ( 3.6% ) to be vl4 sources . these highly correlated variable sources also have very low estimated false positive rates of @xmath01% . of the 3794 pg3 sources , 2929 ( 77% ) lie inside the stern wedge , 3282 ( 87% ) lie inside the modified wedge , and only a single object ( 0% ) in the stellar box . the ages survey provides spectroscopic redshifts and classifications for roughly 23,500 quasars , galaxies and stars ( @xcite ) . we matched the dibs variability catalog to the ages spectroscopic catalogs and obtained 2179 matches for agns , 490 for stars , and 11199 for galaxies in photometry group 3 ( table [ tab : resultsages ] ) . at variability level 2 , we identified 402 ( 18.4% ) agns , 18 ( 3.7% ) stars , and 164 ( 1.5% ) galaxies ( figure [ fig : col - col_ages ] ) . by increasing the variability significance threshold to vl4 , we were left with 145 ( 6.7% ) agns , 8 ( 1.6% ) stars , and 54 ( 0.5% ) galaxies . what are the implications of adding the fifth epoch ( this study ) to the sdwfs variability catalog ? consider the stern wedge sources first . the number of sources in the wedge is nearly identical in both surveys with 6677 for sdwfs and 6692 for dibs . there are 496 ( 7.4% ) sdwfs vl2 pg3 variable objects and 593 ( 8.9% ) dibs sources . dibs increases the number of variable agn candidates by nearly a hundred ( 20% ) and reduces the estimated false positive rate ( from 6% to 2% ) . combining the two effects , the number of agn with robustly measured variabilty is increased from roughly 470 to 581 or about 25% . increasing the variability significance threshold to vl4 results in 204 ( 3.1% ) agn in sdwfs and only 192 ( 2.9% ) in dibs , and the false positive rate stays at 2% . on the other hand , if we keep the false positive rates from sdwfs ( i.e. , use a lower significance levels for dibs ) , we obtain 496 ( sdwfs ) and 936 ( dibs ) pg3 sources for a 6% false positive rate , or 204 ( sdwfs ) and 678 ( dibs ) pg3 objects for a 2% false positive rate . the number of variable agn detections is roughly doubled in dibs as compared to sdwfs at fixed false positive rates . we observe a similar situation for the modified agn wedge . there are 584 ( 4.6% ) sdwfs vl2 sources and 652 ( 5.1% ) for dibs ( 12% more ) out of 12741 and 12801 pg3 sources , respectively . along with the increased total number of these variable agns in dibs , we observe a decreased false positive rate from 7% to 3% . when increasing the variability significance threshold to vl4 , we have 224 ( 1.8% ) and 209 ( 1.6% ) sources , respectively . for the fixed false positive rates from sdwfs we have 584/997 ( sdwfs / dibs ) sources for a 7% false positive rate or 224/719 for a 3% false positive rate . we also study the content of the stellar box , where there are 7567 ( sdwfs ) and 7477 ( dibs ) objects . while in sdwfs we have 86 ( 1.1% ) vl2 stars , there are 143 ( 1.9% ) in dibs . increasing the variability significance threshold to vl4 results in a collection of 21 ( 0.3% ) stars in sdwfs and 68 ( 0.9% ) in dibs . . _ right panel : _ the fraction of objects with stellar , galaxy or qso template matches that are the photometry group 2 , variability level 2 variables as a function of the @xmath70-[4.5]$ ] color . [ fig : col - col_ages],width=309 ] cccccc & & + + total & & 3667 & 376 & 2300 & 9258 + 2 & @xmath80 & 469 ( 12.8%/1% ) & 11 ( 2.9%/3% ) & 42 ( 1.8%/6% ) & 454 ( 4.9%/3% ) + 3 & @xmath32 & 286 ( 7.8%/1% ) & 5 ( 1.3%/2% ) & 27 ( 1.2%/6% ) & 254 ( 2.7%/2% ) + 4 & @xmath33 & 163 ( 4.4%/1% ) & 4 ( 1.1%/3% ) & 14 ( 0.6%/9% ) & 139 ( 1.5%/2% ) + + & & + + total & & 3258 & 342 & 2082 & 5550 + 2 & @xmath80 & 468 ( 14.4%/1% ) & 11 ( 3.2%/3% ) & 42 ( 2.0%/6% ) & 451 ( 8.1%/2% ) + 3 & @xmath32 & 286 ( 8.8%/1% ) & 5 ( 1.5%/2% ) & 27 ( 1.3%/5% ) & 254 ( 4.6%/1% ) + 4 & @xmath33 & 163 ( 5.0%/1% ) & 4 ( 1.2%/3% ) & 14 ( 0.7%/8% ) & 139 ( 2.5%/1% ) + + & & + + total & & 2307 & 206 & 1278 & 3794 + 2 & @xmath80 & 448 ( 19.4%/1% ) & 11 ( 5.3%/2% ) & 40 ( 3.1%/1% ) & 440 ( 11.6%/1% ) + 3 & @xmath32 & 277 ( 12.0%/0% ) & 5 ( 2.4%/2% ) & 25 ( 2.0%/2% ) & 250 ( 6.6%/1% ) + 4 & @xmath33 & 160 ( 6.9%/0% ) & 4 ( 1.9%/3% ) & 13 ( 1.0%/4% ) & 137 ( 3.6%/1% ) ccccc & & + + total & & 2911 & 1102 & 18039 + 2 & @xmath80 & 415 ( 14.3%/1% ) & 22 ( 2.0%/16% ) & 181 ( 1.0%/16% ) + 3 & @xmath32 & 245 ( 8.4%/1% ) & 13 ( 1.2%/16% ) & 110 ( 0.6%/15% ) + 4 & @xmath33 & 146 ( 5.0%/1% ) & 9 ( 0.8%/10% ) & 62 ( 0.3%/21% ) + + & & + + total & & 2793 & 960 & 17368 + 2 & @xmath80 & 415 ( 14.9%/1% ) & 22 ( 2.3%/11% ) & 178 ( 1.0%/13% ) + 3 & @xmath32 & 245 ( 8.8%/1% ) & 13 ( 1.4%/10% ) & 109 ( 0.6%/14% ) + 4 & @xmath33 & 146 ( 5.2%/1% ) & 9 ( 0.9%/8% ) & 62 ( 0.4%/20% ) + + & & + + total & & 2179 & 490 & 11199 + 2 & @xmath80 & 402 ( 18.4%/1% ) & 18 ( 3.7%/5% ) & 164 ( 1.5%/7% ) + 3 & @xmath32 & 240 ( 11.0%/1% ) & 11 ( 2.2%/4% ) & 100 ( 0.9%/6% ) + 4 & @xmath33 & 145 ( 6.7%/1% ) & 8 ( 1.6%/3% ) & 54 ( 0.5%/9% ) quasars are well known as aperiodically variable sources ( ) . such brightness changes are now routinely and successfully modeled using drw model ( e.g , @xcite ) , however with some departures on very short time scales and low amplitudes that are presently only probed by _ kepler _ ( @xcite ) . because our light curves are very sparsely sampled , we are restricted to comparing average structure functions ( sf ; e.g. , @xcite ) , which quantify the mean magnitude difference at the observed ( @xmath87 ) or rest ( @xmath88 ) frame time difference ( or `` lag '' ) between two epochs . we define it as in @xcite , @xmath89 where @xmath90 and @xmath91 are the number of quasars and galaxies ( providing the variance / noise estimate ) used , and @xmath92 is the magnitude difference for a given time - lag @xmath93 . for each qso , we match four galaxies having nearly identical brightness that serve as a noise estimate . at the relevant magnitudes , galaxies are essentially unresolved in the dibs images . we calculated 1000 bootstrap resamplings for both the quasar and galaxy lists , and we report here the median values along with the 1@xmath94 ( 68.3% ) uncertainties ( tables [ tab : sf_fits_qso][tab : sf_fits_chung ] ) . we do not expect a strong agn variability signal at short time lags , which makes these estimates very sensitive to any problems with the estimated noise . having five epochs , we measure @xmath95 different time - lags , blurred by the @xmath96 factor in the rest frame . we model the sf by a power - law of the form @xmath97 adopting @xmath98 ( 4 ) years for the rest - frame ( observed - frame ) estimates . + + the ages survey provides us with over 23500 sources in the field that have measured redshifts and/or type classification ( quasar , galaxy , or star ) . because the number of galaxies in this magnitude limited survey falls off very quickly with increasing redshift , all sources beyond the redshift @xmath99 are quasars ( see figure 6 in @xcite ) . we will frequently use this information , in the selection of `` true '' quasars and in the calculation of the observed and rest - frame structure functions . for the structure function , we must analyze both the detected and undetected variable agns in dibs . we do this by computing the rest - frame structure function of all sources in a given sample with @xmath70<$ ] 18 mag . we consider several cases independent of their variability level . we calculated the rest - frame structure functions for spectroscopically confirmed quasars from ages ( table [ tab : sf_fits_agesqso ] ) . the sf slopes are steeper than in the optical , with @xmath100 , and the amplitudes are @xmath101 mag at @xmath98 years ( figure [ fig : sf_bin ] ) , in agreement with our earlier findings in @xcite . the observed frame structure functions show slightly shallower slopes with @xmath102 and amplitude @xmath103 mag at @xmath104 years . let us now consider the dependence of the structure function on the redshift ( or the emission wavelength ) and the absolute luminosity of ages quasars . we select three redshift ranges of @xmath105 , @xmath106 , and @xmath107 , each containing 248336 agns . the sample of quasars in each redshift bin was then divided into two equal halves after sorting them by absolute magnitude , hereafter called the faint and bright subsamples . for the bright subsample , we observe constant sf slope across the three redshift bins ( rest frame wavelengths @xmath1082.1@xmath109 m ) in both bands with mean @xmath110 . we observe a mild decrease of the amplitude @xmath7 as a function of increasing redshift ( correlation with wavelength ) . for the faint subsample , the sf slope is most likely constant at @xmath111 , but we observe a decline to @xmath112 in both bands in the high-@xmath113 bin . from figure [ fig : z_abs ] , we see that with increasing redshift we are losing intrinsically faint agns ( that seem to be more variable ) due to the observational magnitude limit . the observed decline of sf slope in the faint sample may simply reflect this fact , because we do not observe such a decline in the more complete bright subsamples . for both the bright and faint subsamples , the amplitudes @xmath7 decline to higher redshifts and decreasing emission wavelengths . also fainter subsamples show larger amplitudes than for the bright samples . we repeated the sf calculation in equal magnitude redshift bins by dividing the ages quasars into absolute magnitude bins ( spaced every 1 mag ) and redshift bins ( spaced every @xmath114 ) , and considered only bins with a minimum of 10 agns . the amplitude at a fixed time scale increases with increasing rest frame wavelength ( or decreasing redshift ) . the slope is either constant with the rest frame wavelength ( or redshift ) or it is very weakly correlated with the rest frame wavelength ( weakly anti - correlated with redshift ) . to increase the statistical sample , we also define another larger sample of 1000 agns with @xmath81 . we again sort it by absolute brightness and split it into two brightness bins ( figure [ fig : sf_bightness ] ) . we do not find the slopes of the fainter agns to differ from the brighter ones ( figure [ fig : sf_bightness ] ) , but the amplitudes are larger for fainter agns . this sample consists of 40% highly correlated objects and 60% with low or no correlation . for the highly correlated objects , we again observe larger amplitudes for fainter agns , but also steeper sf slopes than for the bright agns . we can also study the structure function of all agn candidates lying inside the modified agn color wedge . these sources must meet the pg3 requirement but again with no restriction on their variability . there are 999 sources inside the wedge that are brighter than @xmath70<18 $ ] mag and at redshifts @xmath99 . the latter requirement guarantees that we are dealing with true agns , because galaxies are too faint to be observed at such redshifts in ages . the rest frame structure function slopes are @xmath1150.48 and amplitudes of @xmath101 mag ( table [ tab : sf_fits_qso ] ) , in agreement with the true ages quasars . we are also interested in the observed structure functions for @xmath70<18 $ ] mag pg3 sources inside the agn color wedge with either @xmath99 ( 999 sources ) or ignoring redshift information ( 11302 sources ) . the observed slopes for the first sample are @xmath116 and @xmath117 for the 3.6 and 4.5 bands . the amplitudes @xmath7 at the time scale @xmath118 years are @xmath119 mag and @xmath120 mag , respectively . the second sample is characterized by sfs with @xmath121 and @xmath122 for the 3.6 and 4.5 bands , and the amplitudes of @xmath123 mag and @xmath124 mag , respectively . we analyzed the x - ray - matched sources with known redshifts from ages that lie inside the modified agn wedge . we calculated the rest - frame structure functions ( table [ tab : sf_fits_xray ] ) and obtained the following slopes @xmath111 and amplitudes @xmath125 mag for agns at @xmath99 , and obtained slopes @xmath110 and amplitudes @xmath126 mag for all agns . these values again stay in general agreement with the measurements for the confirmed ages quasars . the observed structure functions for all @xmath70<18 $ ] mag pg3 x - ray sources inside the agn color wedge , with or without known redshifts ( 1720 sources ) have slopes @xmath127 and @xmath128 for the 3.6 and 4.5 bands , and the amplitudes @xmath7 at the time scale @xmath118 years are @xmath129 mag and @xmath130 mag , respectively . we also inspected the variability of 183 radio sources from inside the modified agn wedge with known redshifts . we measure the rest - frame slopes @xmath131 for 3.6 and @xmath132 for 4.5 bands . this selection returns slightly shallower slopes than for the confirmed agns . the amplitudes at 2 years are @xmath133 and @xmath134 mag , respectively . the observed structure functions for 555 sources from inside the modified agn wedge with either known or unknown redshifts have slopes @xmath135 and @xmath136 and amplitudes at the time scale @xmath118 years @xmath137 mag and @xmath138 mag , in the two bands , respectively . @xcite provide a catalog of agn candidates based on up to 17-band sed fits using the @xcite templates . we have kept only the sources with @xmath70<18 $ ] mag , @xmath83-ratio @xmath86 or @xmath139 for which we adopted the ages spectroscopic redshifts and if unavailable then we used the estimated photometric redshifts . we find the rest - frame slopes @xmath112 and amplitudes @xmath140 mag for the @xmath86 sources , and @xmath141 and amplitudes @xmath142 mag for the @xmath139 sources . these values are slightly lower than the ones for the confirmed agns . this is most likely an indication of the level of contamination in the @xcite selection method . in fact , we can use the amplitude of the structure function to estimate the purity of the @xcite quasar candidate sample as a function of the @xmath83-test values . we assume that the spectroscopically confirmed ages agn are representative of agn variability with @xmath143 mag ( 0.13 ) mag at @xmath144 years and @xmath145 ( 0.45 ) in the 3.6 ( 4.5@xmath109 m ) band . if we measure amplitude @xmath146 ( at fixed slope ) for a sample where fraction @xmath147 is composed of sources varying with an average amplitude of @xmath148 and fraction @xmath149 are non - variable , then the fraction of real sources is @xmath150 . the scaling of @xmath151 with @xmath83 is robust , but the quantitative estimate depends strongly on using the correct value of @xmath148 , particularly as @xmath152 . the formal uncertainty is @xmath153 so even our roughly @xmath154 measurement uncertainties represent a @xmath155 uncertainty in @xmath151 as @xmath156 . if we consider the @xcite sources in `` logarithmic '' bins of @xmath157 , @xmath158 , @xmath159 and @xmath160 , we find @xmath161 , @xmath162 , @xmath163 and @xmath164 , respectively , at 3.6@xmath109 m . this then implies @xmath165 , @xmath166 , @xmath167 and @xmath168 . the results for 4.5 are similar . we clearly see that the purity of the sample increases with @xmath83 . the samples with @xmath169 are dominated by contaminants , as expected . while these values of @xmath83 would imply a significant agn component when considering a single object , they are not as significant when used for a sample of objects where only @xmath170 of the sources are agn ( see the discussion in @xcite ) . the low fraction found for @xmath139 sources suggests that the uncertainties in @xmath148 are dominated by systematic effects ( e.g. , are the ages agn representative ? ) rather than the measurement errors . the fraction @xmath151 is , of course , a crude statistical estimate and likely subject to a number of systematic effects . from studies of optical quasar variability , we know that the structure function slopes are shallower at @xmath2 ( @xcite ) than the mid - ir ones reported here . in figure [ fig : sf_bin ] , we compare the optical and mid - ir structure functions . it is clear that at short time scales the variability amplitude in the mid - ir is significantly lower than that in optical , and the two match each other only at time - lags of several years . this is consistent with the wavelength dependent smoothing seen in ngc 2617 ( @xcite ) . lccr + + + 3.6 & @xmath171 & @xmath172 & 1518 + 4.5 & @xmath173 & @xmath174 & 1518 + 3.6 & fixed 0.5 & @xmath175 & 1518 + 4.5 & fixed 0.5 & @xmath176 & 1518 + + + + + 3.6 & @xmath177 & @xmath178 & 1518 + 4.5 & @xmath179 & @xmath180 & 1518 lccr + + + 3.6 & @xmath181 & @xmath182 & 999 + 4.5 & @xmath183 & @xmath184 & 999 + 3.6 & fixed 0.5 & @xmath185 & 999 + 4.5 & fixed 0.5 & @xmath186 & 999 + + + + 3.6 & @xmath187 & @xmath188 & 999 + 4.5 & @xmath189 & @xmath190 & 999 lcccr + + + 3.6 & y & @xmath191 & @xmath192 & 625 + 4.5 & y & @xmath193 & @xmath194 & 625 + 3.6 & y & fixed 0.5 & @xmath195 & 625 + 4.5 & y & fixed 0.5 & @xmath196 & 625 + 3.6 & n & @xmath197 & @xmath198 & 1079 + 4.5 & n & @xmath199 & @xmath200 & 1079 + 3.6 & n & fixed 0.5 & @xmath201 & 1079 + 4.5 & n & fixed 0.5 & @xmath202 & 1079 + + + + + 3.6 & y & @xmath203 & @xmath204 & 625 + 4.5 & y & @xmath205 & @xmath206 & 625 + 3.6 & n & @xmath207 & @xmath208 & 1720 + 4.5 & n & @xmath209 & @xmath210 & 1720 lccr + + + 3.6 & @xmath211 & @xmath212 & 1518 + 4.5 & @xmath213 & @xmath214 & 1518 + 3.6 & fixed 0.5 & @xmath215 & 1518 + 4.5 & fixed 0.5 & @xmath216 & 1518 + + + + + 3.6 & @xmath217 & @xmath218 & 1518 + 4.5 & @xmath219 & @xmath220 & 1518 + + + + 3.6 & @xmath221 & @xmath222 & 4160 + 4.5 & @xmath223 & @xmath224 & 4160 + 3.6 & fixed 0.5 & @xmath225 & 4160 + 4.5 & fixed 0.5 & @xmath226 & 4160 + + + + + 3.6 & @xmath227 & @xmath228 & 4160 + 4.5 & @xmath229 & @xmath230 & 4160 in this paper , we have combined the images from sdwfs with a fifth epoch from dibs , extending the time baseline of the survey to 10 years . we performed dia photometry on the five epochs of dibs data , providing the light curves and variability measures for half a million mostly extragalactic sources . we cross - correlated the dibs variability data with the ages redshift survey , x - ray , mips , radio data , and photometrically selected agn candidates from the same area of the sky . in comparison to sdwfs , the variability yields are by 1525% larger for x - ray and 24 quasars , and nearly identical for radio sources exceeding the variability significance @xmath31 . the yield of variable agns selected from the ages survey seem to be also increased by @xmath020% as compared to sdwfs , while the number of variable stars and galaxies stay the same . the mid - ir variability selection detects 18% of the known agns from the ages survey . keeping the false positive rates as in sdwfs ( i.e. , lowering the variability significance level used for the dibs sample ) doubles the number of selected variable agns . the addition of the fifth epoch significantly lowered our estimated false positive rates . the primary reason for the low overall yields is that quasar variability amplitudes are comparable ( for @xmath70<18 $ ] mag ) to the photometric quality of the dibs data . the main motivation of this paper was to study the variability parameters ( the sf slope and amplitude ) as a function of physical quantities ( luminosity or wavelength ) or selection method ( true agns , or mid - ir , x - ray , or radio selected , or from sed fitting ) . the variability analysis of @xmath01500 spectroscopically confirmed sources from the ages survey returned the rest - frame structure function slopes in both bands of @xmath1 and amplitudes @xmath103 mag ( at @xmath98 years ) . except for the small sample of radio - selected agns the mid - ir structure functions of agns are significantly steeper than in optical ( @xmath2 ) . the optical and mid - ir variability amplitudes match each other for time - lags of several years , but for the shorter lags the mid - ir variability amplitudes are significantly lower than the ones in optical . the lack of short time scale variability can be interpreted as the accretion disk or dust torus smoothing out short - time scale variations . this is easily understood , as the accretion disk radius is @xmath231 light days for a typical black hole mass of @xmath232 solar masses , radiative efficiency of 0.1 , and eddington ratio of @xmath233 , hence the optical variations originate in a region of 1 light - day in size and the mid - ir ones from a region 25 times bigger . we also explored the dependence of the structure function slope and amplitude as a function of the luminosity and rest - frame wavelength . we find significant anti - correlation ( lack of correlation ) of the agn amplitude ( slope ) with luminosity . in accordance with optical studies , the brighter the agn the lower its variability amplitude . we also find no clear evidence for the correlation of the variability slope with the rest - frame wavelength , but the range of rest - frame wavelengths explored is rather narrow ( @xmath1082.1@xmath109 m ) . on the other hand , there is a correlation of the sf amplitude with the rest - frame wavelength . we thank prof . igor soszyski for discussions on variable stars and the anonymous referee for helpful suggestions that improved the manuscript . this work is based on observations made with the _ spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of the technology under contract with the national aeronautics and space administration ( nasa ) . sk acknowledges the financial support of the polish national science center through the grant number 2014/15/b / st9/00093 . rja was supported by gemini - conicyt grant number 32120009 and fondecyt grant number 1151408 .

an optimal control strategy is developed to construct nanostructures of desired geometry along line segments by means of directed self - assembly of charged particles . such a control strategy determines the electric potentials of a set of electrodes located at fixed points in the line segment . the particles move under the electric forces generated by these electrodes and by the interactions between the particles themselves to form a desired pattern eventually . due to technology limitations , the particle positions can not be measured during the course of control , so that the control is open - loop in nature . such an open - loop control optimally changes the electrode potentials in time in order to create a desired pattern with the highest probability , despite the inherent uncertainty in the initial positions and the dynamical behaviors of the particles . two models are proposed to describe the uncertain dynamics of the particles : a continuous model relying on a set of nonlinear stochastic differential equations , and a discrete ising model consisting of a large dimensional continuous - time markov chain . while the first model is more mathematically tractable , the second one more precisely describes particles at the nanometer scale . the control design procedure begins with the continuous model and identifies the structure of its stable equilibria , which is used later to propose a piecewise constant structure for the control and to demonstrate that the optimal value of each piece is independently obtained from a certain static optimization problem . it is shown next that the design procedure can be applied to the discrete model with only minor modifications . a numerical example of control design is presented . directed self - assembly , fokker - planck equation , ising model , nanostructure , optimal control .

1603.00113

self - assembly is the process of forming an ordered structure from initially disordered components that only interact locally , without external direction . at the molecular level , this process is a common technique for fabrication of nanostructures with periodic patterns @xcite . due to the important role of this fabrication technique in nanotechnology , several researchers have studied self - assembly phenomena at a theoretical level based on abstract models @xcite . self - assembled nanostructures usually demonstrate periodic patterns that only depend on the nature of their components and the environmental conditions under which the patterns are formed . however , several applications require fabrication of nanostructures with certain non - periodic geometries @xcite . given the major role of molecular self - assembly in fabrication of periodic nanostructures , it is reasonable to ask if this process can be externally directed to fabricate nanostructures of desired geometry which are not necessarily periodic . such a _ directed self - assembly _ process is the focus of this paper . in directed self - assembly , a number of charged nanoparticles ( e.g. , dna tiles ) are manipulated by external electrical fields to form a nanostructure of desired geometry . the directing electrical fields are generated and controlled by relatively small number of electrodes ( compared to the number of particles ) located at fixed locations on the substrate containing the particles . the dynamics of the particles are primarily governed by the interactions between them ( self - assembly ) , and is modified to some extent by manipulation of the electrical potentials of these electrodes ( external direction ) . the particles are initially distributed randomly on the substrate and are perturbed by random disturbances during the assembly process . since the particle positions can not be measured during the course of control , a feedback loop can not be established and the electrodes are actuated only by open - loop controls . the control objective is to direct the particles towards formation of a desired pattern despite the uncertainty in their dynamics and initial positions . under an optimal design , this control must maximize the probability of forming the desired pattern by the end of the assembly process , and maintain the formed structure under a static control afterward . such a constant static control creates a stable equilibrium representing the desired pattern . in addition to this intended stable equilibrium , the static control inherently creates multiple undesired stable equilibria , and a major challenge of an optimal control is to prevent the system falling into such _ kinetic traps_. given the large number of these kinetic traps and the inherent uncertainty in the initial distribution and dynamics of the particles , the system will most likely be trapped by an undesired stable equilibrium ( formation of a wrong pattern ) , unless a phase of dynamic control ( time - dependent ) is applied prior to the static control . this paper intends to develop an analytic framework for study of directed self - assembly , including a systematic method for control design . rather than focusing on a detailed model of the physical process , the main emphasis is on providing a clear understanding of the fundamental concepts such as static control , kinetic traps , and dynamic control . to be consistent with this approach , the models and control problem considered in this paper are abstractions of real - world directed self - assembly : they capture the essence of this phenomenon but do not reflect all its details . in particular , the paper focuses on directed self - assembly of linear structures ( one - dimensional patterns along straight lines ) , a simplified model also adopted in @xcite to study `` undirected '' self - assembly . this special case of the more general planar patterns demonstrates certain properties that facilitate exact characterization of the stable equilibria of the system , which in turn , allows for a rigorous analysis and control design methodology . the concepts and methods developed here for linear patterns are equally valid in two dimensions , while generalization of some computational procedures might not be immediate . such a generalization , at least approximately , is the subject of our future work . this paper adopts two different but closely related models to describe directed self - assembly . a continuous model is presented in section [ lineargeometry.modelsection ] which allows the particles to position continuously at any arbitrary point in a line segment . this model is precise for larger particles of micrometer diameter and its continuous nature facilitates our analysis in section [ lineargeometry.controldesignsection ] . later in section [ lineargeometry.discretemodelsection ] , a discrete model is presented for nano - scale particles , and it is shown how the results of section [ lineargeometry.controldesignsection ] , originally developed for the continuous model , can be tailored to this discrete model only with minor modifications . our main results on the structure of the kinetic traps , control design , and optimization of the electrodes are presented in section [ lineargeometry.controldesignsection ] . the system of particles considered in this paper is described at the nano - scale ( @xmath0 ) by a discrete ising model and a master equation @xcite . in this model , the particles can occupy only a finite set of positions along a line segment , in contrast to a continuous model used for larger particles ( @xmath1 ) in which the particles can position continuously at any arbitrary point along the line segment . the latter model is directly derived from the classical newton s second law of motion , and coulomb s law that governs the interactions between the particles and the forces applied to the particles by the electrodes . such a continuous model is more intuitive and mathematically more tractable , thus it is a convenient point of departure to explain the concepts and control design methodology developed in this paper . we begin with this continuous model and construct our control design method on this basis . later in section [ lineargeometry.discretemodelsection ] , we present the discrete model and show that for the purpose of control design using our proposed method , the two models are mathematically equivalent and can be interchanged with minor modifications . in particular , our control design relies on the steady - state behaviors of these models which match closely despite their different dynamical behaviors . it is emphasized that the two models describe different physical phenomena and they are not necessarily interchangeable for other purposes such as simulations . throughout this paper , the time - dependent state and control vectors are shown by the boldface letters @xmath2 and @xmath3 , so that @xmath2 and @xmath3 are mappings from time into the state space and the control set , respectively . the values of the state and the control vectors at time @xmath4 are denoted by @xmath5 and @xmath6 or simply by @xmath7 and @xmath8 as a shorthand . all constant vectors and other functions of time or other variables are shown in plain letters . referring to fig . [ lineargeometry.controlsparticles ] , consider a line segment and assume that @xmath9 electrodes are located at the fixed points @xmath10 in this line segment . suppose that @xmath11 identical charged particles are located between @xmath12 and @xmath13 at the points @xmath14 . the particle positions specify the state vector @xmath15 in @xmath16 . the control vector @xmath17 is defined in such a manner that its @xmath18th component @xmath19 represents the electric charge of the electrode @xmath20 normalized to the charge of a single particle . it is assumed that at any time , the value of the control vector @xmath8 can be arbitrarily chosen within the control set @xmath21 . + the total energy associated with the state @xmath7 of the particles and the control value @xmath8 is given by @xmath22 , where the normalized energy function @xmath23 is given by @xmath24 here , @xmath25 is a constant defined as @xmath26 where @xmath27 denotes the charge of a single particle , @xmath28 stands for the permittivity of free space , and the dimensionless constant @xmath29 is the relative permittivity of the environment containing the particles . the negative gradient @xmath30 of the total energy ( @xmath31 denotes the gradient operator with respect to the first argument ) is a vector in @xmath16 whose @xmath18th component is the total force applied to the @xmath18th particle by the remaining @xmath32 particles ( first term on the right - hand side of ) and by the @xmath9 electrodes ( second term on the right - hand side of ) . assume that the particles start from the initial state @xmath33 at time @xmath34 and their state @xmath35 evolves in time under a time - varying control @xmath36 . the dynamics of the particles is determined by three factors : the coulomb forces caused by the interactions between the particles and the electrodes and the interactions between the particles themselves , the friction between the particles and their surrounding fluid ( drag ) , and the brownian motion . in the absence of the brownian motion , the particles accelerate under the coulomb forces and the opposing resistance of the surrounding fluid . by stokes drag law @xcite , such resistive forces are negatively proportional to the velocity of the particles with a proportionality constant @xmath37 . in response to a sudden change in the control coulomb forces , the particles accelerate for a short period of time before the opposing drag forces balance this change in the control forces . in a large friction regime , this acceleration period is short and negligible @xcite , so that it is a reasonable approximation to take the drag and the coulomb forces as equal . then the velocity of each particle will be proportional to its applied coulomb force ( smoluchowski approximation ) . by normalizing time to @xmath38 , the proportionality constant is unit and the equation of motion of the particles can be simply written as @xmath39 the contribution of the brownian motion is incorporated into the equation of motion using a @xmath11-dimensional standard wiener process @xmath40 as described by the it stochastic differential equation @xcite @xmath41 here , @xmath42 is a constant depending on the boltzmann constant @xmath43 , the temperature @xmath44 in kelvin , and the normalizing factor @xmath45 of the energy function . it is assumed that the initial state @xmath46 is a random vector with the known probability density function @xmath47 satisfying @xmath48^n.\ ] ] the stochastic differential equation represents the langevin equation for the particle positions @xcite . suppose that the interval @xmath49 $ ] is partitioned into @xmath50 subintervals @xmath51 of the equal length @xmath52 . it is assumed that the number @xmath50 of these subintervals is larger than the number @xmath11 of the particles . further , assume that the distance @xmath53 between the electrodes is an integer multiple of the grid size @xmath52 . a pattern @xmath54 is defined as a binary vector of dimension @xmath50 with exactly @xmath11 ones ( @xmath55 s ) and @xmath56 zeros ( @xmath57 s ) . the total number of patterns is given by the combination @xmath58 . each binary component of a pattern represents one of the subintervals @xmath59 . it is said that a pattern @xmath60 is formed by the particles , if exactly one particle is inside the subintervals associated with the components of value @xmath55 in @xmath60 . notice that every state of the particles does not necessarily define a pattern since it is possible that more than one particle belong to a certain subinterval . [ lineargeometry.figdynamiccontrolexample ] illustrates a nanostructure created by @xmath61 particles in a grid of @xmath62 cells with @xmath63 electrodes . the binary vector @xmath64 represents this nanostructure ( pattern ) in such a manner that each occupied cell corresponds to a @xmath55 in this vector . particles in a grid of @xmath62 cells with @xmath63 electrodes . the pattern is represented by the binary vector @xmath64 , where each filled cell is corresponding to a @xmath55 in this vector . the boxes mark the locations of the electrodes and the disks show the locations that the particles can occupy in the discrete model of section [ lineargeometry.discretemodelsection].,title="fig : " ] + each pattern @xmath60 is uniquely mapped into a subset @xmath65 of the state space @xmath49^{n}$ ] such that the formation of that pattern at time @xmath4 occurs if @xmath66 . let @xmath67 , @xmath68 denote the indices of the @xmath18th @xmath55 in the binary vector ( pattern ) @xmath60 . then the value of the mapping @xmath65 is defined as the union @xmath69 taken over the set of all @xmath70 permutations of @xmath71 . the control goal is to move the particles in such a manner that they form a desired pattern @xmath72 with the highest probability at a final time @xmath73 . this must be achieved despite the inherent uncertainty in the system dynamics and the initial state and by means of an open - loop control since the particle positions ( components of the state vector ) can not be measured during the course of control due to technology limitations . therefore , the objective is to obtain an open - loop control @xmath36 on @xmath74 $ ] to form a desired pattern @xmath72 at the final time @xmath73 with the highest possible probability , and to maintain this maximum probability under the constant control @xmath75 afterward ( @xmath76 ) . these requirements are mathematically expressed by maximizing the payoff function @xmath77 under the inequality constraint @xmath78 for all @xmath79 and some small @xmath80 . the final time @xmath81 is a free parameter that is preferred but not constrained to be reasonably short . this optimization problem can be formulated as an optimal control problem with deterministic but infinite - dimensional dynamics . it is well known that the probability density function @xmath82 of @xmath5 solves the fokker - planck equation @xcite @xmath83 with the initial condition @xmath84 , where @xmath85 denotes the divergence operator with respect to the first argument . then , subject to this infinite - dimensional dynamics , an admissible control @xmath36 is sought on @xmath74 $ ] to maximize the payoff function @xmath86 while maintaining the terminal condition @xmath87 under the constant control @xmath88 for @xmath76 . this new formulation represents a standard optimal control problem although its solution is complicated by the infinite dimensions of the fokker - planck equation . in section [ lineargeometry.controldesignsection ] , we obtain a solution to this optimal control problem within a certain class of piecewise constant controls . this solution exploits a certain structure of the nonlinear system to convert the optimal control problem above to a sequence of static optimization problems with tractable computational complexity . our control design procedure consists of two steps : design of a static control , and design of a dynamic control . the static control @xmath89 is a constant control intended to create a stable equilibrium @xmath90 inside the subset @xmath91 of the state space that represents a desired pattern @xmath72 . in the absence of the brownian motion , a stable equilibrium is a point of the state space with a sustainable balance of forces under which the particles are at rest , i.e. , the state vector @xmath5 settles at this point in the steady - state so that @xmath92 . in the presence of the brownian motion , the state vector moves towards the stable equilibrium and eventually reaches a stationary regime under which it randomly jitters in the vicinity of @xmath93 . the desired pattern is formed in this regime as @xmath93 is inside @xmath91 and the state vector remains close to this point . the optimal design of the static control @xmath94 is discussed in section [ lineargeometry.staticcontrolsection ] . the static control @xmath94 and the corresponding equilibrium @xmath93 must jointly satisfy the conditions indicates that @xmath95 is a positive definite matrix . ] @xmath96 where @xmath97 is the hessian matrix @xmath98 for a fixed @xmath94 , the solution @xmath93 of the algebraic equation is a stationary point of the energy function @xmath99 , and if this stationary point is a strict local minimum of @xmath99 , it is a stable equilibrium of the deterministic dynamical system . as noted in , if the hessian matrix @xmath100 is positive definite , the stationary point is a strict local minimum . for any given static control @xmath94 , the algebraic equation can have multiple stable solutions for @xmath93 , not necessarily inside @xmath91 to form a desired pattern this is caused by the fact that the energy function @xmath99 can have multiple strict local minima ( see fig . [ lineargeometry.multipleenergyminima](a ) ) that allow for the formation of multiple stable patterns . as shown in fig . [ lineargeometry.multipleenergyminima](a ) , the energy function consists of several _ potential wells _ with a single stable equilibrium ( a strict local minimum ) at the bottom of each one . each potential well specifies a _ region of attraction _ ( roa)an open subset of the state space containing exactly one stable equilibrium and marked by the property that if the state @xmath5 of the dynamical system is initially inside a certain roa , it remains inside that roa and moves towards its equilibrium . this property is demonstrated by the inequality @xmath101 where the equality holds if and only if @xmath5 is an equilibrium . this implies that the total energy inside a single roa monotonically decreases before the system settles at the stable equilibrium at the bottom of the potential well . since the energy level inside a roa can never exceed its initial value , the state vector can not escape its initial roa . + for a static control with multiple stable equilibria , the state space is partitioned by the set of roas as illustrated schematically in fig . [ lineargeometry.multipleenergyminima](b ) . however , only a certain equilibrium forms the desired pattern , and only those initial states belonging to the roa of that equilibrium end up with the desired pattern ( see fig . [ lineargeometry.multipleenergyminima](b ) ) . thus , before starting the phase of static control , it is necessary to bring the initial state inside the desired roa . this task is performed by a dynamic control , a time - varying open - loop control that drives the state vector @xmath5 towards the desired roa regardless of the inherent uncertainty in the initial state . notice that the state vector particle locations is not known to the controller during the course of control . it is shown in section [ lineargeometry.dynamiccontrolsection ] that the dynamic control can be decomposed into a sequence of static controls , so that a piecewise constant structure is proposed for this control . before proceeding with the control design in sections [ lineargeometry.staticcontrolsection ] and [ lineargeometry.dynamiccontrolsection ] , the structure of the roas and their stable equilibria is studied in section [ lineargeometry.roastructuresection ] . in the deterministic system under the constant control @xmath102 , the state vector remains in the same roa that it takes its initial value . this property does not generally hold for stochastic systems perturbed by a wiener process . for such stochastic systems , the random disturbance causes the state vector to jitter around one of the equilibria , often inside the same roa . occasionally , the deviations from the equilibrium are large enough to drive the state vector outside that roa , allowing other stable equilibria to attract it . this migration from one potential well to another can be viewed as being caused by a high level of energy absorbed from a disturbance that exceeds the depth of the departure potential well . as illustrated in fig . [ lineargeometry.multipleenergyminima](a ) , the depth of a potential well is defined as the energy difference between its deepest point and the lowest energy on its boundary . in the stochastic system , however , the state vector can not leave its initial roa ( almost surely ) , even though the system is disturbed by a wiener process . this unusual property is a consequence of the infinite depth of the potential wells in this system . to show this property , consider an electrode charged with the same polarity as the particles ( assumed positive ) , and a particle that is pushed towards the electrode by an external force . for example , in fig . [ lineargeometry.controlsparticles ] suppose that the particle located at @xmath103 is pushed left towards the electrode at @xmath104 . the external force required to maintain the particle at the distance @xmath103 from the electrode is proportional to @xmath105 which increases unboundedly as @xmath103 tends to @xmath57 . this implies that the particle can not reach the electrode using a bounded external force , and clearly can not pass through it . with a similar argument , two particles can not hit each other under bounded external forces that squeeze them together . assume that all electrodes have constant positive charges ( same polarity as the particles ) , i.e. , @xmath102 is a vector of positive components . let @xmath106 , @xmath107 be the number of particles in the interval @xmath108 so that @xmath109 . based on the above argument , the integers @xmath106 remain constant over time , i.e. , at any time after applying the constant control @xmath102 , the number of particles in the interval @xmath108 is equal to its value just before application of this control . in addition , the order of the particles is preserved over time as the particles can not jump over each other . since the particles are identical and their order does not change over the course of control , it can be assumed without loss of generality that they are labeled by @xmath110 from left to right , as shown in fig . [ lineargeometry.controlsparticles ] . this requires the initial distribution of the state vector to satisfy @xmath111 where the simplified state space @xmath112 is defined as @xmath113 in addition , the fixed order of the particles allows the mapping @xmath65 to be simplified to @xmath114 defined as @xmath115 where @xmath67 , @xmath68 denote the indices of the @xmath18th @xmath55 in the pattern @xmath60 . let @xmath116 be a vector in @xmath117 whose @xmath18th component is the number of particles in the interval @xmath108 . since the total number of particles is @xmath118 , this vector must satisfy the constraint @xmath119 . the total number of such vectors is the `` _ _ weak compositions of @xmath118 into @xmath120 parts _ _ '' ( * ? ? ? 5.2 ) and is given by the combination @xmath121 each instance of @xmath122 uniquely specifies a convex subset of @xmath112 defined as , the sum @xmath123 is taken equal to @xmath57 . ] @xmath124 the convexity of this set is straightforward to show . these subsets are disjoint and their union is equal to the state space @xmath112 excluding a zero - measure set @xmath125 containing the boundaries of the open sets @xmath126 in @xmath112 , i.e. , @xmath127 theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] in this section states that each subset @xmath126 of the state space @xmath112 contains exactly one stable equilibrium of and that the energy function is convex over @xmath126 , concluding that each @xmath126 is a roa of the dynamical system . [ lineargeometry.existenceuniquenessequilibriumtheorem ] for any constant control @xmath102 with positive components , the energy function @xmath99 is strictly convex over each convex subset @xmath126 ( with @xmath122 satisfying @xmath119 ) , the dynamical system has exactly one equilibrium in @xmath126 , and that equilibrium is stable . see appendix . formation of a pattern @xmath60 at time @xmath4 is confirmed if the state vector @xmath5 is in the subset @xmath114 of the state space @xmath112 . on the other hand , the definition of @xmath114 and the structure of the roas imply that each @xmath114 is entirely inside a single roa . that specific roa is characterized as follows . assume that the pattern @xmath60 has @xmath128 particles in the interval @xmath108 and let @xmath129 denote a vector in @xmath117 containing the integers @xmath128 , @xmath107 . then @xmath114 is a subset of @xmath130 as defined in . thus , to form a pattern @xmath60 , it is necessary to first bring the state vector inside the roa @xmath130 . for simplicity of notation in the rest of the paper , @xmath130 is abbreviated into @xmath131 to represent the roa containing the pattern @xmath60 . suppose that a dynamic control has been applied to the stochastic system during the time interval @xmath132 to bring its state inside the roa @xmath133 that contains the desired pattern @xmath72 . in section [ lineargeometry.dynamiccontrolsection ] , it is shown how to design such a dynamic control to maximize the probability of hitting the target set @xmath133 . at @xmath134 , the constant static control @xmath94 is applied to the system and the system gradually reaches the steady - state as @xmath135 . in the steady - state regime , the probability of forming the desired pattern remains constant , i.e. , the event of @xmath136 has a constant probability . the objective of the static control is to maximize this constant probability assuming that at @xmath134 the state vector is inside the desired roa , i.e. , @xmath137 . this goal is mathematically represented by the optimization problem @xmath138 in practice , the system can get arbitrarily close to the steady - state within a bounded but long enough settling time @xmath139 . since the problem statement in section [ lineargeometry.modelsection ] does not constrain the final time @xmath73 , this quantity can be chosen sufficiently large to ensure that the conditional probability @xmath140 is close enough to its final value in . under this value of the final time , the static control that solves the optimization problem , nearly maximizes this conditional probability . here and in the rest of this paper , the static control @xmath94 is chosen from the control set @xmath141 defined as a subset of @xmath142 with positive control charges at two end points @xmath12 and @xmath13 , and nonnegative charges for the rest of the electrodes , so that @xmath143 under this assumption , theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] is applied to the roas of the system . notice that in the statement of theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] , all control charges are assumed positive , while allows some electrodes to be inactive with zero charges . this provides more flexibility to the control vector without jeopardizing the use of theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] : when some electrodes are inactive , still this theorem is applied , albeit to a system with smaller number of electrodes ( with a control vector of smaller dimension ) . in , only the electrodes at two end points @xmath12 and @xmath13 are constrained to be active to ensure that the particles can not escape the line segment . for @xmath144 , define @xmath145 as the conditional probability density function of @xmath5 given @xmath137 . the evolution of this function for @xmath146 is governed by the fokker - planck equation @xmath147 the definition of @xmath148 implies that at @xmath134 this function is identically @xmath57 for every @xmath149 . since the state vector is almost surely inside @xmath133 at @xmath134 and almost surely can not leave this roa , it stays in @xmath133 for every @xmath144 with probability @xmath55 . this implies @xmath150 for every @xmath149 and every @xmath144 . based on this analysis , the steady - state solution of this fokker - planck equation is given by @xcite @xmath151 for @xmath152 and by @xmath153 for @xmath149 . note that the normalizing factor in the denominator is an integral over @xmath133 rather than the entire state space following the fact that the conditional probability distribution is identically @xmath57 outside this roa . using the conditional density function , the conditional probability @xmath154 is expressed as @xmath155 the condition @xmath156 on probabilities is reflected in this expression by the fact that @xmath157 . using the explicit form , the optimal static control @xmath158 is obtained from the optimization problem @xmath159 in this optimization problem , it is computationally expensive to determine @xmath160 from , since numerical approximation of this expression requires computation of the energy function @xmath161 at a large number of points . the computational complexity can be significantly reduced by _ saddle point approximation _ @xcite of the integrals in . this approximation relies on the fact that the negative exponential integrands in take their significant values in the roa @xmath133 only around the unique minimizer @xmath162 of the energy function . hence , without significant loss of accuracy , @xmath161 can be replaced with a simpler function that approximates it well only around @xmath162 . a reasonable choice for such an approximation is the truncated taylor series @xmath163 in which the dependence of @xmath93 on @xmath94 is not explicitly shown for the sake of simplicity . substituting this approximate expression into and multiplying both its numerator and denominator by an appropriate constant , @xmath160 is approximated by @xmath164 in terms of the mapping @xmath165 defined as @xmath166 here , @xmath167 denotes a multivariate normal distribution with the mean vector @xmath168 and the covariance matrix @xmath169 . using the payoff function - , the optimization problem can be reformulated as the constrained optimization problem [ lineargeometry.gaussianapproximationoptimizationproblem ] @xmath170 it is shown next that the approximate formula can be directly derived from a linearized model around the unique stable equilibrium of the roa @xmath133 . such a linear model is later used to provide an intuitive explanation for the static control design . let @xmath171 be a fixed control and assume that @xmath93 is its associated equilibrium in @xmath133 . if the disturbance strength @xmath172 is small compared to the norm of the hessian matrix @xmath100 , the dynamical system can be linearized around @xmath173 to approximate the state vector as @xmath174 in which the small deviation @xmath175 is the solution of the linear stochastic differential equation @xmath176 the linearity of this equation indicates that @xmath175 is a zero - mean gaussian random vector with the steady - state covariance matrix @xmath177 that solves the algebraic lyapunov equation @xcite @xmath178 it is concluded that @xmath5 is approximately a gaussian random vector with the mean vector @xmath93 and the covariance matrix @xmath179 . for this approximation , @xmath180 defined in represents the conditional probability on the right - hand side of . based on the linear model - , an approximate design method is introduced below to provide further intuition on the optimal static control . to facilitate the discussion , @xmath160 is expressed as the ratio @xmath181 with @xmath182 and @xmath183 defined as @xmath184 to achieve the maximum of @xmath185 , the value @xmath182 must be kept as large as possible compared to @xmath183 . this requires to shift the concentration of probability towards the central point of the desired set @xmath186 and push it away from the forbidden set @xmath187 . let @xmath188 be a vector containing the mid points of the intervals @xmath59 corresponding to @xmath55 s in the desired pattern @xmath72 . then the problem is to keep the state vector @xmath5 at steady - state as close as possible to @xmath189 with respect to some appropriate norm . since the desired set @xmath186 is a hypercube , a reasonable choice of the distance measure is @xmath190 \simeq\lim_{t\to+\infty}\mathrm{e}\left[\,\left\|x_\mathrm{ss}-\xi_d+\mathbf{{\delta}x}\left(t\right)\right\|_\infty\right].\ ] ] however , this measure is not mathematically tractable unless the disturbance power @xmath191 is small enough to justify the approximation @xmath192 . in this case , an approximation for the optimal static control is given by the optimization problem @xmath193 such a suboptimal static control places the stable equilibrium @xmath93 as close as possible to @xmath189 that represents the desired pattern . when @xmath191 is not negligible , one can alternatively adopt the mean squared distance measure @xmath194\simeq\left\|x_\mathrm{ss}-\xi_d\right\|^2 + \mathrm{tr}\left\{\sigma_\mathrm{ss}\right\}\ ] ] which leads to the optimization problem @xmath195 it is reasonable at this point to ask whether the optimization problem has always a bounded solution or it is possible for the optimal static control to be unbounded . an informal treatment of this problem comes below . let @xmath94 be a constant control in @xmath141 with a stable equilibrium @xmath196 . for every @xmath197 , the value @xmath161 of the energy function tends to @xmath198 as @xmath199 since the components of @xmath94 are nonnegative . under this limit , the ratio tends to either @xmath57 or @xmath55 , depending on which of the sets @xmath187 or @xmath186 contain @xmath162 . in particular , the limiting value of @xmath160 is explicitly given by @xmath200 this implies that @xmath158 in can be unbounded if under this control the equilibrium @xmath201 remains inside @xmath186 . assume that the desired pattern @xmath72 at least in one of the intervals @xmath108 has two or more particles . any unbounded control , necessarily squeezes these particles together , and two particles closer than a grid length can not form a valid pattern , i.e. , the probability of forming @xmath72 under an unbounded control is identically @xmath57 . it is concluded that the optimal control @xmath158 can be potentially unbounded only for the sparse patterns with at most one particle in each interval @xmath108 . thus , except for such sparse patterns ( that are not of much practical interest ) , it is not necessary to impose an upper bound on the control set @xmath141 to secure a well defined solution for the optimization problem . in theory , an unbounded control can achieve the probability @xmath202 for a sparse pattern . in this degenerate case , the number of particles is smaller than the number of controls ( @xmath203 ) so that the algebraic equation @xmath204 can have multiple solutions for @xmath94 , including a solution with an infinite magnitude . such an unbounded control leads to a hessian matrix @xmath205 with an unbounded norm , which in turn , results in a zero covariance matrix according to . two numerical techniques are proposed here to compute the stable equilibrium inside each roa . the first method relies on the numerical solution of the ordinary differential equation . starting from any arbitrary initial state inside the desired roa , the solution to asymptotically approaches the unique stable equilibrium of that roa . although the exact equilibrium is approached only as @xmath135 , after a bounded but long enough time , the solution of will be close enough to the equilibrium to provide an acceptable approximation for it . the second method makes use of the proof of theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] ( see appendix ) . this method starts from an initial vector @xmath206 and generates the sequence of vectors @xmath207 from the recursive equation @xmath208 , @xmath209 until the distance between two successive vectors drops below a given threshold . then the last vector in this sequence is taken as an approximation for the equilibrium . at each step @xmath18 , the @xmath210th component of @xmath211 is computed by numerically solving ( e.g. using newton s method ) the algebraic equation @xmath212 for @xmath213 , as explained in theorem [ lineargeometry.existenceuniquenessequilibriumtheorem ] . determining an estimate for the settling time @xmath139 is the last step to complete the design of the static control . this quantity closely depends on the second smallest eigenvalue ( in absolute value ) of the fokker - planck operator @xcite @xmath214 note that the first eigenvalue of this operator ( smallest in the absolute value ) is @xmath57 associated with the steady - state solution of the fokker - planck equation . let @xmath215 be the second smallest eigenvalue of @xmath216 in the absolute value . then a rule of thumb for computation of the settling time is given by @xmath217 direct computation of @xmath218 is generally a difficult task . in the case of this paper , the settling time can be approximated using an alternative method . this approximation is based on the observation that the settling time @xmath139 is the time required for the state vector @xmath5 to reach a stationary regime starting from an initial state inside the roa @xmath133 . the temporal evolution of the state vector is governed by the stochastic differential equation under the constant control @xmath102 . the sample trajectories of @xmath5 almost surely remain inside the same roa @xmath133 and tend towards the equilibrium @xmath93 . the large energy gradient near the boundary ( large repulsive forces between closely placed point charges ) of the roa strongly pushes the trajectories towards the equilibrium so that they rapidly move away from the boundary and spend most of their transition time near the equilibrium . this justifies linearizing the nonlinear dynamics around the equilibrium @xmath93 and computing the settling time from the linear model rather than the original nonlinear model . then the approximate settling time is expressed in terms of the smallest eigenvalue @xmath219 of the positive definite hessian matrix @xmath100 as @xmath220 in the procedure proposed in section [ lineargeometry.staticcontrolsection ] for design of the static control , the positions @xmath221 of the electrodes are assumed fixed and given in advance . however , the specific choice of these positions , that are directly involved in the shape of the energy function , can affect the performance of the static control for better or worse . to maximize the probability of forming a desired pattern , the electrode positions can be optimized simultaneously with the static control , of course within certain physical constraints . as mentioned before , the electrode positions are integer multiples of the grid size @xmath52 . let @xmath222 be an integer quantifying the distance between the electrodes @xmath18 and @xmath223 as @xmath224 then , in terms of @xmath225 , the electrode positions are given by @xmath226 since the length @xmath227 of the line segment is fixed and includes exactly @xmath50 grid cells , the integers @xmath225 must satisfy the equality constraint @xmath228 . moreover , technology limitations do not allow to fabricate the electrodes closer than @xmath229 , imposing the inequality constraints @xmath230 for some integer @xmath231 . substituting into the energy function , the dependence of the probability on the energy function results in an explicit expression @xmath232 for this probability . then the joint optimization of the static control and the electrode positions can be formulated as the mixed integer nonlinear program ( minlp ) @xcite @xmath233 at the initial time @xmath34 , the state vector @xmath234 is randomly distributed in the state space @xmath112 according to some probability density function @xmath235 . the dynamic control is an open - loop control applied to the system of particles during @xmath236 to bring the random initial state inside the desired roa @xmath133 with the highest probability . the dynamic control proposed in this paper consists of a sequence of static controls , each one designed through a procedure similar to section [ lineargeometry.staticcontrolsection ] . the concept of a multistage control for directed self - assembly has been established in @xcite and is illustrated in fig . [ lineargeometry.dynamiccontrolstages ] . + consider a constant control vector @xmath237 whose first and last components are positive and the rest of its components are @xmath57 ( only the electrodes at @xmath12 and @xmath13 are active ) . under this control , all particles are distributed inside a single large interval @xmath238 and the system has a single roa @xmath239 covering the entire state space @xmath112 . application of this constant control at @xmath34 drives the state vector towards the unique equilibrium of @xmath239 regardless of its initial value . after a settling time of @xmath240 , the constant control @xmath237 is switched to another constant control @xmath241 with more than two positive components ( e.g. the electrode in the middle is activated ) . under this new control , the system has more than one roa while a specific one of these multiple roas ( denoted by @xmath242 ) is the target set of the state vector at @xmath243 . the optimal value of @xmath237 at the first stage is determined to maximize the probability of the state vector being inside this target roa at @xmath243 , i.e. , @xmath244 . this procedure is repeated in @xmath245 stages by activating more controls at each stage while getting closer to the final desired equilibrium . the target set @xmath246 of the last stage is the roa containing the desired pattern in the static control problem , i.e. , @xmath247 . at the end of the last stage of the dynamic control , the static control @xmath94 activates all electrodes to create the desired pattern . design of the target roas @xmath248 is explained below by an example using fig . [ lineargeometry.figdynamiccontrolexample ] . in this figure , a system of @xmath61 particles with @xmath63 controls and @xmath62 grid cells is illustrated . the dynamic control is designed with @xmath249 stages and the following sequence of controls : in the first stage , only the electrodes at @xmath12 and @xmath250 are active , while in the second and last stage the electrode at @xmath251 is turned on . the electrodes at @xmath252 and @xmath253 are simultaneously activated in the static control phase . according to this sequence , the controls @xmath237 , @xmath241 , and @xmath94 have the structure @xmath254 where @xmath255 and @xmath256 denote positive and nonnegative components , respectively . the target roa @xmath242 is a subset of the state space with @xmath257 particles on the left side of @xmath251 and @xmath258 particles on its right side , and is explicitly given by @xmath259 similarly , @xmath260 is a subset with @xmath258 , @xmath261 , @xmath55 , and @xmath261 particles in the intervals @xmath262 , @xmath263 , @xmath264 , and @xmath265 , respectively . consider the family of piecewise constant controls @xmath266 where the constants @xmath267 are the switching times of the control and @xmath73 is the final time . for @xmath268 , define the increasing sequence of control sets @xmath269 with @xmath141 given by , @xmath270 defined as @xmath271 and with the property that the controls in @xmath272 can possess at least one more nonzero component than those in @xmath273 . it is assumed that at each stage @xmath268 , the constant control @xmath274 is in the control set @xmath273 , while the static control @xmath94 belongs to @xmath141 . the goal is to obtain within the class of controls , the one that solves the optimal control problem of section [ lineargeometry.modelsection ] . the controls in this class consist of a static part applied for @xmath144 and a dynamic part during @xmath275 . optimization of the static part was discussed in section [ lineargeometry.staticcontrolsection ] and the procedure for optimizing the dynamic part is presented below . consider the sequence @xmath248 of roas associated with the sequence @xmath276 of control sets , and for @xmath268 define the conditional probabilities @xmath277 where @xmath5 is the state of under the constant control @xmath274 applied at @xmath278 . as discussed in section [ lineargeometry.staticcontrolsection ] , these probabilities are explicitly expressed as @xmath279 it is assumed that the time durations @xmath280 between the switching times are long enough for the system to reach the steady - state before application of a new segment of the control ( as discussed in section [ lineargeometry.settlingtimesection ] ) . under this condition , the following approximation holds : @xmath281 at the initial time @xmath34 , the state vector belongs to @xmath282 with probability @xmath55 . thus , the probability of formation of a desired pattern @xmath72 under the control at @xmath283 is given in terms of the conditional probabilities @xmath160 and @xmath284 , @xmath268 by the product @xmath285 note that this approximation tends to exact as @xmath286 and @xmath287 , @xmath268 . to maximize the probability , each multiplicative term must be maximized independently with respect to its argument . for the static control term , the optimization problem was already discussed in section [ lineargeometry.staticcontrolsection ] . for the rest of the terms , the optimal controls @xmath288 , @xmath268 are obtained by solving the optimization problems @xmath289 in terms of these optimal controls , the maximum probability of forming a desired pattern @xmath72 at a large final time @xmath73 achieved by a control in the class of controls is given by @xmath290 the activation sequence of the electrodes is not unique and can be regarded as an additional optimization variable . for any specific activation sequence , the maximum probability to form a desired pattern is obtained from . then in a higher level of the optimization process , the maximum of these optimized probabilities is determined over all possible activation sequences . for the small size problem of fig . [ lineargeometry.figdynamiccontrolexample ] with only @xmath291 possible activation sequences , this level of optimization can be performed by simply enumerating all sequences . for a problem of larger size , more advanced techniques can be developed based on the outer approximation @xcite or branch and bound @xcite methods . we applied the design procedure developed in this section to the example of fig . [ lineargeometry.figdynamiccontrolexample ] . in this figure , self - assembly of @xmath61 particles using @xmath63 electrodes is considered along a line segment . the line segment is partitioned into @xmath62 cells and the distance between the electrodes is assumed to be @xmath55 unit of length . throughout the design procedure and its following simulations the disturbance power is set at @xmath292 . the goal is to generate a desired pattern of @xmath293 for design of both static and dynamic controls , the constrained optimization problem was utilized to approximate the original problems and . this constrained optimization problem , was converted to an unconstrained problem by solving the constraint for @xmath93 using the second numerical procedure proposed in section [ lineargeometry.numericalcomputationequilibriumsection ] . the resulting unconstrained problem was solved using the ` fminsearch ` function of @xmath294 . the designed optimal control consists of a static control and a two - stage dynamic control as illustrated in fig . [ lineargeometry.figdesignedcontrrol ] . the optimal values of the control vector are shown for the two stages of the dynamic control and for the static control in figs . [ lineargeometry.figdesignedcontrrol](a ) , [ lineargeometry.figdesignedcontrrol](b ) , and [ lineargeometry.figdesignedcontrrol](c ) , respectively . further , the most likely patterns formed at the end of each stage are illustrated in these figures . the small boxes in these figures represent the locations of the electrodes , while the disks mark the optimal equilibrium of each stage . using the approximation method of section [ lineargeometry.settlingtimesection ] , the switching times of the control were computed as @xmath295 , @xmath296 , and @xmath297 . also , the highest probability of success at each stage was obtained as @xmath298 , @xmath299 , and @xmath300 , which lead to the total probability of success @xmath301 + under this control , the deterministic dynamical system and its stochastic version were numerically simulated . the results of these simulations are shown in fig . [ lineargeometry.figsiumulationresults ] for both deterministic ( thick line ) and stochastic ( thin line ) models . the trajectory of the particles start at random positions ( small boxes ) and end at the optimal final equilibrium ( small disks ) that represents the desired pattern shown on the top . the vertical axis in this figure shows progress in time while the horizontal axis stands for the line segment on which the particles move . the heavy vertical lines represent the energy barriers created by the electrodes which can not be crossed by the particles . note that the specific sample path illustrated in fig . [ lineargeometry.figsiumulationresults ] succeeds to create the desired pattern ; however , not all sample paths of the stochastic system end up with the desired pattern . for example , the sample path of fig . [ lineargeometry.figsiumulationresults ] could fail if the deviation marked by the dashed ellipse would occur shortly later . with the probability of success estimated as @xmath302 , the sample paths fail to form the desired pattern at an average rate of @xmath303 . of them fail to form this specific pattern . the sample path of fig . [ lineargeometry.figsiumulationresults ] could fail if the deviation marked by the dashed ellipse would occur shortly later.,title="fig : " ] + in the prior work on directed self - assembly , the system of particles has been described by a discrete ising model @xcite rather than the continuous model of this paper @xcite . in such a discrete model , the particles can occupy only a finite set of positions along the line at most one particle in each position . as shown in fig . [ lineargeometry.figdynamiccontrolexample ] , these positions are located at the centers of the intervals @xmath59 . in this model , the discrete state of the particles is a vector @xmath304 in @xmath16 taking its value in the discrete state space @xmath305 at any time @xmath306 . each vector @xmath307 in this discrete state space corresponds to a pattern @xmath308 and its @xmath118 components represent the discrete locations of @xmath118 particles along the line . the number @xmath309 of the elements in the discrete state space is the same as the number of patterns that can be formed by placement of @xmath118 particles in @xmath310 positions and is given by the combination @xmath58 . equivalently , the discrete state of the particles at time @xmath4 can be described by an integer @xmath311 such that @xmath312 . similar to the continuous model , the discrete state @xmath313 is characterized by a stochastic process , namely by a continuous - time markov chain @xcite . for a continuous - time markov chain , the evolution of probability is described by a _ master equation _ , the analogue of the fokker - planck equation in the continuous model . for @xmath314 define the probabilities @xmath315 these probabilities evolve in time according to the set of linear differential equations @xmath316 defined for @xmath314 . the initial state of these equations ( the initial probability distribution ) is assumed known and the nonnegative scalars @xmath317 , @xmath318 describe the transition rates from pattern @xmath319 to pattern @xmath210 . the contribution of the control to the system dynamics is reflected in the model through the dependence of the transition rates on the control vector @xmath6 . by collecting the probabilities @xmath320 in a single vector @xmath321^\mathrm{t}$ ] , the master equation can be written in the compact form @xmath322 where @xmath323 is a @xmath324 matrix with off - diagonal elements @xmath317 and diagonal elements @xmath325 for any fixed @xmath306 , the transition rate @xmath317 from a pattern @xmath319 to another pattern @xmath210 exponentially decreases with the difference between the energy @xmath326 of the pattern @xmath319 and the energy barrier @xmath327 between the two patterns . the mapping @xmath328 that maps the instantaneous value of the control vector into the instantaneous value of the transition rate has the form of @xcite @xmath329 where @xmath330 is a constant increasing with the absolute temperature @xmath44 , and the energy barrier satisfies the conditions @xmath331 for the explicit form of @xmath332 , the reader is referred to @xcite . for the analysis of this paper , it is enough to know that the energy barrier between two patterns is positive , and it is of infinite magnitude if in transition from one pattern to another a particle has to jump over an active electrode . this latter property is caused by the unbounded level of energy at a point charge that represents an active electrode . due to this property , the transition rates between two such patterns are identically @xmath57 which parallels the property observed in the continuous model that a particle can not jump over an active electrode . because of this property , the central role of the roas in the continuous model has an analogue in the discrete model . each roa contains a certain subset of the patterns @xmath333 , and similar to the continuous model , the roas partition the discrete state space @xmath334 into @xmath335 subsets ( @xmath335 is the number of roas given by ) . the roas are marked by unbounded energy barriers encircling them , and such infinite energy barriers block the transition of a pattern inside a roa to any pattern outside it . this implies that for any pair of patterns @xmath210 and @xmath319 inside two different roas , both othe transition rates @xmath317 and @xmath336 are identically @xmath57 . as a result , the state - space equation is decomposed into @xmath335 decoupled smaller state - space equations @xmath337 where @xmath338 is a column vector containing the probabilities of the patterns belonging to the @xmath210th roa and @xmath339 is its corresponding square block of @xmath323 . equivalently , the markov chain @xmath313 is reducible into @xmath335 smaller markov chains @xmath340 which are statistically independent conditioned on the initial value @xmath341 . each of the @xmath335 decoupled state - space equations has a steady - state solution which follows a gibbs probability distribution . this property helps to determine a simple expression for the discrete counterpart of the conditional probability . let @xmath72 be a desired pattern inside the roa @xmath133 and assume that it is represented in the discrete state space by the vector @xmath189 . similar to , define the conditional probability @xmath342 under the constant control @xmath102 applied at @xmath134 . using the analysis above , this conditional probability can be explicitly expressed as @xmath343 similarly , the discrete counterpart of is defined as @xmath344 under the constant control @xmath345 for @xmath346 , and is explicitly expressed as @xmath347 finally , the probability of successful formation of a pattern @xmath72 at time @xmath283 under the control is approximately given as @xmath348 evidently , the procedure of control design for the discrete model parallels the one proposed for the continuous model : it is enough to maximize the payoff function instead of . further , each of these payoff functions closely approximates the other one as the sums in are discrete approximations of the integrals in . the only major difference in the control design procedure is in computation of the practical values of the settling times @xmath139 and @xmath349 . for the continuous model , these quantities are determined in terms of the eigenvalues of the fokker - planck operator as explained in section [ lineargeometry.settlingtimesection ] . for the discrete model , the settling time @xmath139 is determined in terms of the smallest nonzero eigenvalue of the matrix @xmath350 ( smallest in the absolute value ) , where the subscript @xmath351 refers to the roa containing the desired pattern . directed self - assembly of charged particles along line segments has been considered . in the assembly process , a number of particles move in one dimension along a line segment under the repulsive forces experienced from interactions with other particles , and the process is directed towards formation of a desired pattern by external forces applied from charged electrodes located at fixed points in the line segment . the potentials of these electrodes are precisely controlled in time so that the formation of a desired pattern is secured with the highest probability despite the inherent uncertainty in the initial position and the dynamical behaviors of the particles . a challenging aspect of such a control is that the actual positions of the particles are not measurable during the assembly process . two models have been proposed to describe the uncertain dynamics of the particles . the first model which is mathematically more tractable consists of a set of nonlinear stochastic differential equations and is suitable for larger particles of micrometer scale . the second model is a discrete ising model consisting of a continuous - time markov chain and is more accurate for nanometer scale particles . a class of piecewise constant controls has been proposed for these models and the optimal values of the constant pieces have been determined as the solutions to certain optimization problems . consider @xmath126 with @xmath119 . the goal is to show that the algebraic equation @xmath352 has exactly one solution @xmath93 in @xmath126 and that the jacobian matrix of the vector field @xmath353 is negative definite over the set @xmath126 . to that end , denote the @xmath18th component of @xmath354 by @xmath355 and let @xmath356 be the @xmath319th component of @xmath94 , where by hypothesis @xmath357 . then using , @xmath355 can be written as @xmath358 where @xmath359 denotes the signum function . further , for @xmath360 , the partial derivatives of @xmath361 with respect to @xmath362 and @xmath363 , @xmath364 exist and are given by given @xmath360 , for each @xmath68 , let @xmath366 denote the smallest integer satisfying @xmath367 and define @xmath368 fix all components of @xmath360 except for @xmath362 which is allowed to vary in the segment @xmath369 . the negative derivative implies that @xmath361 is strictly decreasing in @xmath370 with all other variables fixed . in addition , implies that this function tends to @xmath198 and @xmath371 as @xmath362 tends to @xmath372 and @xmath373 , respectively . thus , it is concluded that @xmath374 has one and only one solution for @xmath362 in the interval @xmath370 , with all other variables fixed . this solution depends on @xmath375 and is denoted by @xmath376 let @xmath377 be a vector - valued function with the components @xmath378 , @xmath68 . then , @xmath93 is a solution to , if and only if it is a fixed point of the mapping @xmath379 , i.e. , if it solves @xmath380 it is proven next that has exactly one solution in @xmath126 by showing that @xmath379 is a contraction map on @xmath126 , and thereby it has exactly one fixed point in this set . the gradient @xmath381 of the scalar function @xmath378 is obtained as follows . based on the definition of @xmath378 , the identity @xmath382 holds true for any @xmath383 , where the components of @xmath384 are similar to @xmath122 except for @xmath385 which is equal to @xmath386 . differentiating this identity with respect to @xmath387 leads to @xmath388 solving this equation for @xmath389 results in @xmath390 since @xmath378 does not depend on @xmath362 , its partial derivative with respect to @xmath362 is identically @xmath57 so that @xmath391 substituting the explicit expressions into and noting that @xmath357 , it is straightforward to verify that @xmath392 let @xmath393 and consider the line segment @xmath394.\ ] ] since @xmath126 is a convex set , all points on this line segment are inside the set . applying the mean value theorem @xcite to the scalar function @xmath395 implies that there exists @xmath396 such that @xmath397 substituting @xmath398 , @xmath399 , and @xmath400 into this equality and taking absolute values of its sides lead to @xmath401 because this inequality holds for every @xmath68 , it is concluded that @xmath402 where @xmath403 this verifies that @xmath379 is a contraction map . finally , it is shown that @xmath99 is strictly convex on @xmath126 , and thereby the solution @xmath93 to is a stable equilibrium . the gergorin circle theorem @xcite applied to the jacobian matrix of the vector field @xmath353 implies that the eigenvalues @xmath404 of this matrix are inside the circles of the form @xmath405 since @xmath406 according to and the right - hand side of the inequality is smaller than @xmath407 , the eigenvalues of the jacobian matrix have negative values . s. ouk kim , h. h. solak , m. p. stoykovich , n. j. ferrier , j. j. de pablo , and p. f. nealey , `` epitaxial self - assembly of block copolymers on lithographically defined nanopatterned substrates , '' _ nature _ , vol . 424 , no . 6947 , pp . 411414 , 2003 . j. c. love , l. a. estroff , j. k. kriebel , r. g. nuzzo , and g. m. whitesides , `` self - assembled monolayers of thiolates on metals as a form of nanotechnology , '' _ chem . _ , vol . 105 , no . 4 , pp . 11031170 , 2005 . a. khaled , s. guo , f. li , and p. guo , `` controllable self - assembly of nanoparticles for specific delivery of multiple therapeutic molecules to cancer cells using rna nanotechnology , '' _ nano letters _ , vol . 5 , no . 9 , pp . 17971808 , 2005 . l. adleman , q. cheng , a. goel , m. huang , and h. wasserman , `` linear self - assemblies : equilibria , entropy , and convergence rates , '' in _ proc . of the 6th international conference on difference equations and applications , icdea 2001 _ , ( augsburg , germany ) , jun . 2001 . n. stephanopoulos , e. o. p. solis , and g. stephanopoulos , `` nanoscale process systems engineering : toward molecular factories , synthetic cells , and adaptive devices , '' _ aiche journal _ , vol . 51 , no . 7 , pp . 18581869 , 2005 . a. winkleman , b. d. gates , l. s. mccarty , and g. m. whitesides , `` directed self - assembly of spherical particles on patterned electrodes by an applied electric field , '' _ advanced materials _ , vol . 17 , no . 12 , pp . 15071511 , 2005 . m. p. stoykovich , h. kang , k. c. daoulas , g. liu , c .- c . liu , j. j. de pablo , m. mller , and p. f. nealey , `` directed self - assembly of block copolymers for nanolithography : fabrication of isolated features and essential integrated circuit geometries , '' _ acs nano _ , vol . 1 , no . 3 , pp . 168175 , 2007 . c. h. lalander , y. zheng , s. dhuey , s. cabrini , and u. bach , `` dna - directed self - assembly of gold nanoparticles onto nanopatterned surfaces : controlled placement of individual nanoparticles into regular arrays , '' _ acs nano _ , vol . 4 , no . 10 , pp . 61536161 , 2010 . e. o. p. solis , p. i. barton , and g. stephanopoulos , `` controlled formation of nanostructures with desired geometries . 1 . robust static structures , '' _ ind . eng . _ , vol . 49 , no . 17 , pp . 77287745 , 2010 . e. o. p. solis , p. i. barton , and g. stephanopoulos , `` controlled formation of nanostructures with desired geometries . 2 . robust dynamic paths , '' _ ind . _ , vol . 49 , no . 17 , pp . 77467757 , 2010 . roux , `` brownian particles at different times scales : a new derivation of the smoluchowski equation , '' _ physica a : statistical mechanics and its applications _ , vol . 188 , no . 4 , pp . 526552 , 1992 . d. liberzon and r. w. brockett , `` spectral analysis of fokker - planck and related operators arising from linear stochastic differential equations , '' _ siam journal on control and optimization _ , vol . 38 , no . 5 , pp . 14531467 , 2000 . p. kesavan , r. j. allgor , e. p. gatzke , and p. i. barton , `` outer approximation algorithms for separable nonconvex mixed - integer nonlinear programs , '' _ mathematical programming _ , 100 , no . 3 , pp . 517535 , 2004 . m. tawarmalani and n. v. sahinidis , _ convexification and global optimization in continuous and mixed - integer nonlinear programming : theory , algorithms , software , and applications_. dordrecht ; boston : kluwer academic publishers , 2002 .

we investigate legendrian graphs in . we extend the classical invariants , thurston - bennequin number and rotation number to legendrian graphs . we prove that a graph can be legendrian realized with all its cycles legendrian unknots with and if and only if it does not contain as a minor . we show that the pair does not characterize a legendrian graph up to legendrian isotopy if the graph contains a cut edge or a cut vertex . for the lollipop graph the pair determines two legendrian classes and for the handcuff graph it determines four legendrian classes .

1108.2281

in this paper we begin the systematic study of legendrian graphs in @xmath6 with the standard contact structure . these are embedded graphs that are everywhere tangent to the contact planes . legendrian graphs have appeared naturally in several important contexts in the study of contact manifolds . they are used in giroux s proof of existence of open book decompositions compatible with a given contact structure , see @xcite . legendrian graphs also appear in the proof of eliashberg s and fraser s result which says that in a tight contact structure the unknot is determined up to legendrian isotopy by the invariants @xmath7 and @xmath8 , see @xcite . yet no study of legendrian graphs , until now , has been undertaken . we remedy this by establishing the foundations for what we expect will be a very rich field . a _ spatial graph _ is an embedding of a graph into @xmath9 an _ abstract graph _ is a set of vertices together with a set of edges between them , without any specified embedding . we will throughout this paper refer to an abstract graph as simply a graph . in section [ graphs ] , we show there is no obstruction to having a legendrian realization of any spatial graph . we extend the classical invariants thurston - bennequin number , @xmath7 , and rotation number , @xmath8 , from legendrian knots to legendrian graphs . in @xcite , mohnke proved that the borromean rings and the whithead link can not be represented by legendrian links of _ trivial unknots _ , that is unknots with @xmath1 and @xmath2 . the trivial unknot is the one unknot among all unknots that attains the maximal thurston - bennequin number of its topological class . as an application of our invariants , we ask which graphs admit legendrian embeddings with all cycles trivial unknots . in section [ maxtb ] , we give a full characterization of these graphs , in the form of the following : a graph @xmath10 admits a legendrian embedding in @xmath0 with all its cycles trivial unknots if and only if @xmath10 does not contain @xmath3 as a minor . the proof of this theorem relies partly on the fact that the trivial unknot has an odd thurston - bennequin number , that is @xmath1 . we prove the reverse implication in more generality , for @xmath11 which represents the set of topological knot classes with odd maximal thurston - bennequin number . let @xmath10 be a graph that contains @xmath3 as a minor . there does not exist a legendrian realization of @xmath10 such that all its cycles are knots in @xmath12 realizing their maximal thurston - bennequin number . it is known that certain legendrian knots and links are determined by the invariants @xmath7 and @xmath8 : the unknot ( see @xcite ) , the torus knots and the figure eight knot ( see @xcite ) , the links consisting of an unknot and a cable of that unknot ( see @xcite ) . in section [ classification ] , we ask what types of spatial graphs are classified up to legendrian isotopy by the pair @xmath4 . we show that the pair @xmath4 does not classify graphs which contain both cycles and either cut vertices or cut edges , independent of the chosen topological class . this means that even uncomplicated graphs carry more information than the set of knots represented by their cycles . in order to have a classification by the pair @xmath4 , we must first narrow to a specific topological class . we investigate _ topological planar graphs _ , that is an embedded graph that is ambient isotopic to a plannar embedding . not to be confused with _ planar graphs _ , which refers to an abstract graph that has a planar embedding . in the case of the handcuff graph , for topological planar graphs , we prove there are exactly four legendrian realizations for each pair @xmath4 . for the lollipop graph , for topological planar graphs , we prove there are exactly two legendrian realizations for each pair @xmath4 . the authors would like to thank tim cochran for his support and interest in the project . they would also like to thank john etnyre for helpful conversations . a _ spatial graph _ is an embedding @xmath13 of a graph @xmath10 into @xmath6 ( or @xmath14 ) , also called a _ spatial embedding _ or _ graph embedding_. we remind the reader that an abstract graph , one without an embedding well be referred to as simply a graph . we will be considering spatial graphs in @xmath6 throughout this paper . two spatial graphs @xmath15 and @xmath16 are _ ambient isotopic _ if there exits an isotopy @xmath17 such that @xmath18 and @xmath19 similar to knots , there is a set of reidemeister moves for spatial graphs , described by kauffman in @xcite . here we remind the reader of some basic graph theoretic terminology . we are considering the most general of graphs , so there can be _ multi - edges _ ( edges that go between the same pair of vertices ) , and _ loops _ ( edges that connect a vertex to itself ) . the _ valence _ of a vertex is the number of endpoints of edges at the given vertex . two vertices are _ adjacent _ if there is an edge between them . the complete graph on @xmath20 vertices , denoted @xmath21 , is the graph with @xmath20 vertices were every vertex is adjacent to every other vertex , with exactly one edge between each pair of vertices . two edges are _ adjacent _ if they share a vertex . a graph @xmath22 is a _ minor _ of @xmath10 if @xmath22 can be obtained from a subgraph of @xmath10 by a finite number of edge contractions . since we are considering topological questions about legendrian graphs throughout this paper , it is important to be aware of intrinsic properties of graphs . a property is called _ intrinsic _ if every embedding of @xmath10 in @xmath6 ( or @xmath14 ) has the property . a graph @xmath10 is _ minor minimal _ with respect to a property if @xmath10 has the property but no minor of @xmath10 has the property . such properties are characterized by their full set of minor minimal graphs . a spatial graph is said to _ contain a knot ( or link ) _ if the knot ( or link ) appears as a subgraph of @xmath10 . a graph , @xmath10 , is _ intrinsically knotted _ if every embedding of @xmath10 into @xmath6 ( or @xmath14 ) contains a nontrivial knot . a link @xmath23 is _ split _ if there is an embedding of a @xmath24sphere @xmath25 in @xmath26 such that each component of @xmath27 contains at least one component of @xmath23 . a graph , @xmath10 , is _ intrinsically linked _ if every embedding of @xmath10 into @xmath6 ( or @xmath14 ) contains a nonsplit link . the combined work of conway and gordon @xcite , sachs @xcite , and robertson , seymour , and thomas @xcite fully characterized intrinsically linked graphs . they showed that the petersen family is the complete set of minor minimal intrinsically linked graphs . that is no minor of the petersen family is intrinsically linked and any intrinsically linked graph contains a graph in the petersen family as a minor . unlike the intrinsically linked graphs , the set of intrinsically knotted graphs has not been characterized . however , there are many graphs that are known to be minor minimal intrinsically knotted . as consequence of how intrinsically linked graphs were characterized , it is known that all intrinsically knotted graphs are intrinsically linked , see @xcite . as we examine legendrian graphs and their properties , it is important to be aware of what topological properties ( without restricting to legendrian embeddings ) are known to be intrinsic to the graph . these issues will be present when considering theorem [ thmg ] and corollary [ unknot ] . let m be an oriented 3-manifold and @xmath28 a 2-plane field on @xmath29 . then @xmath28 is a _ contact structure _ on @xmath29 if @xmath30 for some @xmath31form @xmath32 on @xmath29 satisfying @xmath33 on @xmath34 , the @xmath31form @xmath35 defines a contact structure called the standard contact structure , @xmath36 . there is a diffeomeorphism of @xmath34 taking the standard contact structure , @xmath36 , to the symmetric contact structure , @xmath37 given in cylindrical coordinates by @xmath38 . in this paper , we switch between @xmath36 and @xmath39 when convenient . darboux s theorem says that any contact structure on a manifold @xmath29 is locally diffeomorphic to @xmath36 . a curve @xmath40 is called _ legendrian _ if for all @xmath41 and @xmath42 the contact plane at @xmath43 , @xmath44 . in @xmath0 , legendrian curves , in particular knots and links , are studied via projections , and a common projection is the front projection ( on @xmath45plane ) . the legendrian condition implies that @xmath46 ( i.e. the @xmath47coordinate can be recovered as the slope in the @xmath45plane ) , and so front projections of legendrian knots do not have vertical tangencies . figure [ unknot_trefoil ] shows two front projections of legendrian unknots and that of a legendrian right - handed trefoil . since the positive @xmath48-axis points inside the page , at each crossing the overstrand is always the one with smaller slope . ( 385 , 72 ) ( 0,0 ) apart from the topological knot class , there are two classical invariants of legendrian knots , the thurston - bennequin number , @xmath7 , and the rotation number , @xmath8 . the thurston - bennequin number measures the amount of twisting of the contact planes along the knot and does not depend on the chosen orientation of @xmath49 . to compute the thurston - bennequin number of a legendrian knot @xmath49 consider a non - zero vector field @xmath50 transverse to @xmath28 , take @xmath51 the push - off of @xmath49 in the direction of @xmath50 , and define @xmath52 if @xmath49 is null - homologous , @xmath53 measures the twisting of the contact framing on @xmath49 with respect to the seifert framing . for a legendrian knot @xmath49 , @xmath53 can be computed from its front projection @xmath54 as @xmath55 the rotation number , @xmath56 , is only defined for oriented null - homologous knots , so assume @xmath49 is oriented and @xmath57 , where @xmath58 is an embedded oriented surface . the contact planes when restricted to @xmath59 form a trivial @xmath24dimensional bundle , and the trivialization of @xmath60 induces a trivialization on @xmath61 . let @xmath50 be a non - zero vector field tangent to @xmath49 pointing in the direction of the orientation on @xmath49 . the winding number of @xmath50 about the origin with respect to this trivialization is the rotation number of @xmath49 , @xmath56 . in @xmath6 , the vector fields @xmath62 and @xmath63 define a positively oriented trivialization for @xmath36 . therefore , @xmath56 can be computed by counting with sign ( + for counterclockwise and @xmath64 for clockwise ) how many times the positive tangent vector to @xmath49 crosses @xmath65 as we travel once around @xmath49 . the tangent vector aligns with with one of the vectors @xmath65 or @xmath66 at the points corresponding to cusps in the front projection , @xmath54 , and one can check that @xmath67 given a legendrian knot @xmath49 , legendrian knots in the same topological class as @xmath49 can be obtained by _ stabilizations_. a strand of @xmath49 in the front projection of @xmath49 is replaced by one of the zig - zags as in figure [ stabilizations ] . the stabilization is said to be positive if down cusps are introduced and negative if up cusps are introduced . the legendrian isotopy type of @xmath49 changes through stabilization and so do the thurston - bennequin number and rotation number : @xmath68 and @xmath69 . ( 300 , 115 ) ( 0,0 ) ( 53,67)@xmath49 ( 135,85)@xmath70 ( 135,28)@xmath71 it is known , @xcite , that if @xmath49 is a null - homologous legendrian knot in @xmath0 and @xmath59 is a seifert surface for @xmath49 then @xmath72 in particular , if @xmath49 is the unknot then @xmath73 and if @xmath74 is the right hand trefoil knot then @xmath75 . both inequalities are sharp and are realized for the first and third knots in figure [ unknot_trefoil ] . a _ legendrian graph _ in a contact 3-manifold @xmath76 is a graph embedded in such a way that all its edges are legendrian curves that are non - tangent to each other at the vertices . when the graph @xmath10 must be specified this will also be referred to as a _ legendrian embedding of @xmath10 _ or a _ legendrian realization of @xmath10_. it should be noted for legendrian graphs , if all edges around a vertex are oriented outward , then no two tangent vectors at the vertex coincide in the contact plane . however , two tangent vectors may have the same direction but different orientations resulting in a smooth arc through the vertex . it is a result of this structure that the order of the edges around a vertices is not changed under legendrian isotopy . given any embedded graph @xmath10 in @xmath34 there exists a legendrian realization of @xmath10 in @xmath0 . [ leggraph ] denote the vertices of @xmath10 by @xmath77 , and fix these points . every point @xmath78 has an @xmath79neighborhood @xmath80 contactomorphic to a neighborhood of the origin in @xmath81 . via this diffeomorphism , the contact plane @xmath82 is identified with the plane @xmath83 at the origin . near each vertex @xmath78 , we modify @xmath10 through ambient isotopy , such that the edges incident with @xmath78 are segments which lie in the contact plane @xmath82 and are identified with @xmath84-constant segments in the plane @xmath83 . the edges of @xmath10 are thus legendrian near each vertex . consider @xmath85 an edge between two vertices @xmath78 and @xmath86 and let @xmath85 be identified with the @xmath87ray near @xmath78 and with the @xmath88ray near @xmath86 . denote by @xmath89 and @xmath90 the two points which are identified with @xmath91 and @xmath92 . denote by @xmath93 the legendrian segment between @xmath78 and @xmath94 identified with the segment @xmath95 , and by @xmath96 the legendrian segment between @xmath86 and @xmath97 identified with the segment @xmath98 . denote by @xmath99 the arc of @xmath85 between @xmath94 and @xmath97 . we can @xmath100approximate @xmath99 by a legendrian arc @xmath101 , in such a way that the union of arcs @xmath102 is a @xmath103 curve . we do this by choosing a @xmath100close approximation of the front projection of @xmath99 by a regular curve @xmath104 with no vertical tangencies , isolated cusps , and such that at each point @xmath105 the slope of @xmath104 is close to the @xmath48-coordinate of the point on @xmath99 projecting to @xmath43 . we can do this while keeping the endpoints @xmath94 and @xmath97 fixed and in such a way that the legendrian lift of @xmath99 , @xmath101 , coincides with @xmath85 near @xmath94 and @xmath97 . for more details see @xcite , section 3.3.1 . the curve @xmath102 is a legendrian curve which is @xmath100close to @xmath85 . by making @xmath104 have only transverse intersections , no self - intersection in the approximating curve are introduced , thus @xmath102 is an embedded arc . additionally , we can avoid introducing any non - trivial knotting . we approximate all other edges in the same way , and we obtain a legendrian @xmath100 approximation of @xmath10 ( i.e. a legendrian graph topologically ambient isotopic to @xmath10 ) . legendrian graphs can also be studied via front projections and two generic front projections of a legendrian graph are related by reidemeister moves i , ii and iii together with three moves given by the mutual position of vertices and edges , see @xcite . see figure [ moves ] . ( 360 , 188 ) ( 0,0 ) ( 84 , 153)i ( 82,99)ii ( 80,33)iii ( 256 , 150)iv ( 328 , 151)iv ( 295,93)v ( 291,30)vi through legendrian isotopy , the order of the edges around a vertex does not change . since two edges which share a vertex can never have the same ( oriented ) tangent vector throughout a legendrian isotopy , when seen in the contact plane at the vertex , the order of the edges around the vertex is the same up to a cyclic permutation . we extend the classical invariants @xmath7 and @xmath8 to legendrian graphs . a cycle in a legendrian graph is a _ piecewise smooth legendrian knot _ , that is , a simple closed curve that is everywhere tangent to the contact planes , but has finitely many points , at the vertices , where it may not be smooth . we first define the invariants , @xmath7 and @xmath8 , for piecewise smooth legendrian knots and then we extend the definition to legendrian graphs . given any piecewise smooth legendrian knot @xmath49 , there is a natural way to construct a smooth knot from @xmath49 . we define the _ standard smoothing _ of a cycle @xmath49 of a legendrian graph to be the @xmath103curve obtained by replacing @xmath49 in an @xmath79neighborhood of each vertex by the lift of a minimal front projection of a smooth curve , which coincides with @xmath49 outside of the @xmath79neighborhood . the projection is minimal , in the sense that no extra stabilizations or knotting are introduced with this approximation . see figure [ smoothing ] . we denote the standard smoothing of a piecewise smooth legendrian knot @xmath49 by @xmath106 . + [ isosmoothing ] if @xmath107 and @xmath108 are piecewise smooth legendrian knots which are isotopic as legendrian graphs , then their standard smoothings are isotopic legendrian knots . since @xmath107 and @xmath108 are isotopic as legendrian graphs , their front projects , @xmath109 and @xmath110 , are related by a finite sequence of the standard reidemeister moves i , ii , iii , as well as , the moves iv , v , vi , and legendrian planar isotopy . thus to show that the standard smoothings of @xmath107 and @xmath108 are isotopic , we need only show that the standard smoothings of valence two subgraphs of the reidemeister moves iv , v , and vi are moves that result in isotopic legendrian knots . for reidemeister move iv there are two different possibilities of valence two subgraphs up to reflection and planar isotopy to be considered , see figure [ moveivi](a ) . after smoothing the move either shows no changed , or a minor planar isotopy . for reidemeister move v there are two different possibilities of valence two subgraphs up to reflection and planar isotopy to be considered , see figure [ moveivi](b ) . after smoothing the move either shows a minor planar isotopy , or a difference of a reidemeister ii move . for reidemeister move vi there are three different possibilities of valence two subgraphs up to reflection and planar isotopy to be considered , see figure [ moveivi](c ) . after smoothing the move either shows no change , a minor planar isotopy , or a difference of a reidemeister i move . in all cases the standard smoothing can be moved to the other standard smoothing by a single reidemeister move or planar isotopy . ( 360 , 100 ) ( 0,0)curve near a vertex . the blue edges which are part of the cycle @xmath49 are replaced by the red @xmath103arc near the vertex . , title="fig : " ] ( 410 , 216 ) ( 10,0 ) ( -10,15)(c ) ( -10,100)(b ) ( -10,175)(a ) ( 57,181)iv ( 119,181)iv ( 267,181)iv ( 337,181)iv ( 111 , 101)v ( 300 , 102)v ( 56,22)vi ( 198,22)vi ( 345,22)vi of course , for any piecewise smooth legendrian knot there are isotopic legendrian knots , which can be obtained by a small isotopy near each vertex moving the edges so that they have parallel tangents at the vertex . in light of this , proposition [ isosmoothing ] shows that the isotopy class of a piecewise smooth legendrian knot contains legendrian embeddings from exactly one isotopy class of a legendrian knot , the knot obtained by its standard smoothing . so given a piecewise smooth legendrian knot @xmath49 , and its standard smoothing @xmath111 one could take the definition of the classical invariants to be : @xmath112 and @xmath113 . however , since we will be using the said definitions to define invariants for legendrian graphs , we would like to defined them in such a way that smoothings are not needed . it should be noted that any other definition of @xmath7 and @xmath8 that coincides with the invariants for smooth legendrian knots will be equivalent to the above definition , as a result of proposition [ isosmoothing ] . let @xmath49 represent a piecewise smooth legendrian knot and @xmath50 a vector field along @xmath49 which is transverse to the contact planes . take the push - off of @xmath49 , @xmath51 in the direction of @xmath50 and let @xmath114 . this definition coincides with that for smooth knots . there is no obstruction given by the direction change at the vertices . for a legendrian graph @xmath10 , we fix an order on the cycles of @xmath10 and we define _ the thurston - bennequin number of @xmath10 _ , denoted by @xmath115 , to be the ordered list of the thurston - bennequin numbers of the cycles of @xmath10 . if @xmath10 has no cycles , we define @xmath115 to be the empty list . we define the rotation number of a null - homologous piecewise smooth legendrian knot @xmath116 as follows : consider @xmath59 with @xmath117 and endow @xmath49 with the orientation induced by that on @xmath59 . consider the trivialization of @xmath36 given by the two vectors @xmath62 and @xmath63 . denote by @xmath118 , @xmath119 , ... , @xmath120 , @xmath121 the vertices on @xmath49 , in cyclic order as given by the orientation , denote by @xmath122 , @xmath123 , the smooth edge of @xmath49 between @xmath94 and @xmath124 , and denote by @xmath78 , @xmath123 , the unit vector field tangent to @xmath122 pointing in the direction of the orientation on @xmath49 . we follow the @xmath125 in the trivialization given by @xmath126 and @xmath127 and count the number of times @xmath126 is passed , with sign . at each vertex @xmath94 , @xmath128 if @xmath129 and @xmath78 do not coincide we complete the rotation counterclockwise if @xmath130 is a positively oriented basis for @xmath131 and clockwise if @xmath130 is a negatively oriented basis for @xmath131 . this is equivalent to completing by a rotation from @xmath129 towards @xmath78 in the direction of the shortest angle between the two . note that , since one edge is oriented towards the vertex and one edge is oriented away from the vertex , @xmath129 and @xmath78 can not be opposite to each other . denote by @xmath132 the number of times @xmath133 is passed in the counterclockwise direction and by @xmath134 the number of times @xmath133 is passed in the clockwise direction as @xmath49 is traced once . we define @xmath56 by @xmath135 let @xmath136 denote the legendrian isotopy class of @xmath49 . the above discussion gives a recipe of how to compute the rotation number for a particular embedding of a piecewise smooth legendrian knot . however , when @xmath49 changes through legendrian isotopy , the tangent vector at @xmath49 changes continuously and the edges can not pass over one another at the vertices . thus we get a continuous map @xmath137 . the rotation number is therefore a legendrian isotopy invariant for piecewise smooth legendrian knots . if @xmath49 is a smooth legendrian knot then we recover the rotation number for @xmath49 . for a legendrian graph @xmath10 , we fix an order on the cycles of @xmath10 with orientation and we define _ the rotation number of @xmath10 _ , denoted by @xmath138 , to be the ordered list of the rotation numbers of the cycles of @xmath10 . if @xmath10 has no cycles , we define @xmath138 to be the empty list . in @xcite , mohnke proved that the borromean rings and the whithead link can not be represented by legendrian links of trivial unknots . a _ trivial unknot _ is one with @xmath1 and @xmath2 ( like the first unknot shown in figure [ unknot_trefoil ] ) . the trivial unknot attains the maximal possible thurston - bennequin number for an unknot . the obstruction comes from upper bounds on @xmath7 given by the minimal degree of one of the variables in the kauffman polynomial . in this section we determine which graphs can be legendrian realized in such a way that all cycles are trivial unknots . we will see that there are many graphs ( even planar graphs ) with no such legendrian realization . we give a full characterization of these graphs . we also present a more general result about which graphs can be realized with all their cycles having maximal @xmath7 for a class of knots . the following two lemmas will be useful . ( 144 , 50 ) ( 0,0).,title="fig : " ] ( 15,42)@xmath139 ( 20,6)@xmath140 ( 107,3)@xmath141 ( 111,38)@xmath142 the front projections of any two legendrian realizations of a graph in @xmath0 are related by a finite sequence of legendrian isotopies and changes involving 1 . one edge ( i.e. stabilizations , crossings of the edge with itself ) 2 . two adjacent edges ( i.e. change in the number of crossings between the two edges ) 3 . two nonadjacent edges ( i.e change in the number of crossings between the two edges ) [ two_fronts ] if the two legendrian realizations are legendrian isotopic then the two front projections differ by legendrian isotopy only . if not , the two front projections may differ in number of edge stabilizations or number of crossings . the stabilizations occur on a single edge while the crossings may occur on a single edge , between two adjacent edges or between two nonadjacent edges . for any legendrian embedding @xmath23 of @xmath3 in @xmath143 @xmath144 where @xmath145 is the set of cycles in @xmath23 . [ mod2 ] denote by @xmath139 , @xmath140 , @xmath141 and @xmath142 the vertices of an abstract @xmath3 . there are seven cycles in @xmath3 : four 3-cycles ( @xmath146 and three 4-cycles @xmath147 + @xmath148 each edge appears in four different cycles and each pair of edges appears in two different cycles . consider the embedding @xmath49 of @xmath3 shown in figure [ k4 ] . for this embedding , there are six cycles with @xmath1 and one cycle ( @xmath149 ) with @xmath150 , thus @xmath151 take an arbitrary legendrian embedding of @xmath3 , call it @xmath23 . by lemma [ two_fronts ] , the front projection for this embedding differs from @xmath49 by a finite sequence of legendrian isotopy and changes of the form described above in one edge , two adjacent edges , or two non - adjacent edges . first , since @xmath7 is an invariant , @xmath152 does not change under isotopy . 1 . if a change in a single edge is made , the @xmath7 for all four cycles containing this edge is modified by the same quantity . 2 . if a change in two adjacent edges is made , the @xmath7 for both cycles containing this pair of edges is modified by the same quantity . if a change in two non - adjacent edges is made , the @xmath7 for both cycles containing this pair of edges is modified by the same quantity . thus , the parity of the sum of the @xmath7 s over all cycles remains unchanged throughout the process and @xmath153 we define @xmath12 to be the set of topological knot classes with odd maximal thurston - bennequin number . we have the following theorem : let @xmath10 be a graph that contains @xmath3 as a minor . then there does not exist a legendrian realization of @xmath10 such that all of its cycles are knots in @xmath12 realizing their maximal thurston - bennequin number . [ thmg ] it suffices to prove the theorem for @xmath154 , since any graph that contains @xmath3 as a minor contains a subdivision of @xmath3 ( see definition [ sub ] ) . assume all seven cycles of @xmath3 can be realized with maximal odd thurston - bennequin number , @xmath155 . then , @xmath156 , contradicting the conclusion of lemma [ mod2 ] . since the unknot has odd maximal @xmath1 we obtain the following corollary . recall that the trivial unknot is the unknot with maximal thurston - bennequin number . let @xmath10 be a graph that contains @xmath3 as a minor . then there does not exist a legendrian realization of @xmath10 such that all of its cycles are trivial unknots . [ unknot ] when we consider a graph @xmath10 that contains @xmath3 as a minor and an arbitrary legendrian embedding @xmath157 corollary [ unknot ] guaranties , either @xmath15 contains a nontrivial knot , or that all of the cycles of @xmath15 are unknots , but they do not all have maximal @xmath7 . for any intrinsically knotted graph ( see section [ spatialgrph ] ) corollary [ unknot ] is not surprising , since every embedding of an intrinsically knotted graph contains a nontrivial knot . while there are many graphs that contain @xmath3 as a minor and are not intrinsically knotted , all of the intrinsically knotted graphs contain @xmath3 as a minor . this can be seen , by noting that @xmath3 is a minor of all graphs in the petersen family . then since all intrinsically knotted graphs are also intrinsically linked , we see @xmath3 is a minor of all of the intrinsically knotted graphs . the same should be considered for theorem [ thmg ] . given an arbitrary legendrian embedding @xmath157 of a graph @xmath10 that contains @xmath3 as minor , theorem [ thmg ] implies that either @xmath15 does not only contain knots from @xmath11 or @xmath15 contains only knots from @xmath11 and they do not all attain their maximal @xmath7 . so theorem [ thmg ] gives more information about intrinsically knotted graphs . one example of an intrinsically knotted graph is @xmath158 , ( see @xcite ) . there are embeddings of @xmath158 with unknots as all but one cycle which is a trefoil . both the unknot and the right handed trefoil have odd maximal thurston - bennequin number , so then theorem [ thmg ] implies for such a legendrian embedding not all of the cycles will attain their maximal @xmath7 . there are many other knots whose maximal thurston - bennequin number is odd . a few examples are the right handed trefoil ( figure [ unknot_trefoil ] ) , the figure eight knot , the @xmath159 knot and its mirror image , the @xmath160 knot and its mirror image . there is a more general version of theorem [ thmg ] that is an immediate consequence of the proof , though it is cumbersome to state . let @xmath10 be a graph that contains @xmath3 as a minor . let @xmath152 be the subdivision of @xmath3 that @xmath10 contains . then there does not exist a legendrian realization of @xmath10 such that @xmath152 contains precisely an odd number of cycles that are knots in @xmath11 where all the knots in @xmath152 realize their maximal thurston - bennequin number . if we focus on embeddings with only unknots as in corollary [ unknot ] the converse also holds . before proving this we introduce some needed definitions and observations . 1 . 2 . a _ path _ between two vertices @xmath139 and @xmath140 of a graph @xmath10 is a finite sequence of at least two edges starting at @xmath139 and ending at @xmath140 , with no repetition of vertices . a vertex of the graph @xmath10 is said to be _ a cut vertex _ if by deleting the vertex ( and all incident edges ) the resulting graph has more connected components than @xmath10 . an edge of @xmath10 is said to be a _ cut edge _ if by deleting the edge the resulting graph has more connected components than @xmath10 . 5 . a _ subdivision _ of the graph @xmath10 is a graph obtained by replacing a finite number of edges of @xmath10 with paths ( one can think of this as adding a finite number of vertices along edges of @xmath10 ) . [ sub ] let @xmath10 be a graph which does not contain @xmath3 as a minor . let @xmath139 , @xmath140 , @xmath141 , @xmath142 be four vertices in a cycle of @xmath10 , appearing in this order . for any such formation , there are not two edges or paths , other than the ones already contained in the cycle , connecting @xmath139 and @xmath141 and connecting @xmath140 and @xmath142 . otherwise @xmath139 , @xmath140 , @xmath141 , and @xmath142 represent the vertices of a ( subdivision of ) @xmath3 . see figure [ nok4_remarks](a ) . [ alternates ] let @xmath10 be a graph which does not contain @xmath3 as a minor . let @xmath139 , @xmath140 , @xmath141 , be three vertices in a cycle of @xmath10 . for any such formation , there is not an additional vertex @xmath50 with distinct edges or paths , connecting @xmath50 to the other vertices @xmath139 , @xmath140 , and @xmath141 . otherwise @xmath139 , @xmath140 , @xmath141 , and @xmath50 represent the vertices of a ( subdivision of ) @xmath3 . see figure [ nok4_remarks](b ) . [ triangle ] ( 200 , 70 ) ( 0,0 ) as described in remarks [ alternates ] and [ triangle].,title="fig : " ] ( 28,-12)(a ) ( 169,-12)(b ) ( 63,63)@xmath161 ( 0,63)@xmath162 ( -2,9)@xmath163 ( 65,9)@xmath164 ( 178,38)@xmath165 ( 172,77)@xmath161 ( 129,16)@xmath162 ( 208,16)@xmath163 let @xmath10 be a graph , if @xmath10 does not contain @xmath3 as a minor then it can be legendrian realized in @xmath0 in such a way that all its cycles are trivial unknots . [ thmgrec ] we need only prove the theorem for @xmath10 connected , with no cut edges and no cut vertices . in all other cases such components ( connected components with no cut edges nor cut vertices ) of @xmath10 can be realized in the same way and cut edges can be realized in any fashion ( as they do not appear in any cycle ) . in what follows , all edges of @xmath10 are realized as non - stabilized arcs . let @xmath166 be one of the cycles of @xmath10 and @xmath167 the number of vertices of @xmath166 . realize @xmath166 in such a way that its front projection consists of @xmath168 horizontal edges and one edge on top of these . label the vertices on @xmath166 with @xmath139 , @xmath140 , ... , @xmath169 , in this order , from left to right . realize all the other edges between vertices of @xmath166 on top of the horizontal edges of @xmath166 , in a nested fashion as shown in figure [ nok4_one](a ) . by remark [ alternates ] , we can realize these edges without any crossings . otherwise , the four endpoints of two crossing edges represent the vertices of a @xmath3 . next , for each pair of vertices @xmath170 , @xmath171 , of @xmath166 for which there exists at least one path between @xmath78 and @xmath86 not containing any edges of @xmath166 , realize one of these paths under the horizontal line of @xmath166 . call this path @xmath172 . for each @xmath173 , if there is another path between @xmath78 and @xmath86 not containing edges of @xmath166 or edges of @xmath172 , ... , @xmath174 , realize this path under @xmath174 . see figure [ nok4_one](b ) . denote by @xmath175 this is a finite set . by construction , all paths in @xmath176 are nested , without any crossings , they have the vertices @xmath78 and @xmath86 in common and are disjoint otherwise . by remark [ alternates ] , we can realize all the paths appearing thus far without any crossings . additionally , for any vertices @xmath78 , @xmath86 , @xmath177 and @xmath178 of @xmath166 , with @xmath179 , no element of @xmath180 has any vertices ( distinct from @xmath78 , @xmath86 , @xmath177 and @xmath178 ) or edges in common with any element of @xmath181 . we prove the observation by contradiction . there are four different cases of @xmath179 , that is : ( 1 ) @xmath182 and @xmath183 ( 2 ) @xmath184 and @xmath185 ( 3 ) @xmath186 , and ( 4 ) @xmath187 and @xmath188 are all distinct . suppose a path @xmath189 in @xmath180 and a path @xmath190 in @xmath181 have at least one vertex other than @xmath78 , @xmath86 , @xmath177 and @xmath178 in common . let @xmath191 be the non - empty set of vertices that the paths @xmath189 and @xmath190 have in common distinct from @xmath78 , @xmath86 , @xmath177 and @xmath178 . let @xmath192 be the left most vertex on @xmath189 with @xmath193 , and let @xmath194 be the right most vertex on @xmath189 with @xmath195 . ( we need only show that a contiguous overlap can not occur , for if there are paths with multiple overlaps there are paths with a single overlap . ) for the subcases ( 1 ) , ( 3 ) , and ( 4 ) , let @xmath196 be the vertex in this set closest to @xmath178 on the path @xmath190 . then the part of the path @xmath190 from @xmath197 to @xmath178 is distinct from the path @xmath189 . so this path between @xmath197 and @xmath178 together with the two parts of @xmath189 going from @xmath78 to @xmath197 and from @xmath197 to @xmath86 are all distinct paths joining @xmath197 to the cycle @xmath166 . thus by remark [ triangle ] we have a contradiction . similarly , for the case ( 3 ) let @xmath196 be the vertex in this set closest to @xmath78 on the path @xmath189 . then the part of the path @xmath189 from @xmath197 to @xmath78 together with the two parts of @xmath190 going from @xmath177 to @xmath197 and from @xmath197 to @xmath178 are all distinct paths joining @xmath197 to the cycle @xmath166 . thus by remark [ triangle ] we have a contradiction . further , we treat each of the elements in @xmath180 , @xmath198 , @xmath199 , as if it were the path of horizontal edges in @xmath166 . we realize the edges with vertices on each such path above the path and the paths with endpoints on each such path below the path . we continue to realize each new edge on top of the path its vertices end lie on and each new path below the path its vertices lie on . ( 450 , 85 ) ( 0,0 ) with all cycles unknots of maximal @xmath7 . the cycle @xmath166 is blue.,title="fig : " ] ( 85,0)(a ) ( 305,0)(b ) ( 60,85 ) ( 288,87 ) ( 0,32)@xmath161 ( 28,32)@xmath162 ( 263,49)@xmath200 ( 310,49)@xmath201 ( 160,32)@xmath202 ( 180,32)@xmath203 claim : 1 . this is a legendrian embedding of @xmath10 ( all edges and vertices of @xmath10 are realized ) . all cycles in this embedding have maximal @xmath1 . proof of ( 1 ) : the above construction is a prescription for a front projection thus as long as all vertices and edges of @xmath10 are realized we have a legendrian embedding . the two cases of assuming a vertex was not realized and assuming an edge was not realized will be treated concurrently . if there exists an edge of @xmath10 which has not been realized , then there is an edge or a path in @xmath10 between two realized vertices that has not been realized . if there exists a vertex of @xmath10 which has not been realized , then since @xmath10 was assumed without cut edges and connected , there is a path in @xmath10 between two realized vertices that has not been realized . we show that the existence of such an edge or path leads to a contradiction . let @xmath50 and @xmath204 represent two realized vertices of @xmath10 such that there is a non - realized edge or path ( containing exclusively edges that have not been realized ) between @xmath50 and @xmath204 . the vertices @xmath50 and @xmath204 can not both be in @xmath166 , since all paths and edges between these vertices have been realized . for the same reason , @xmath50 and @xmath204 can not both be vertices on the same element of a @xmath180 , or on the same path realized at a later stage . from the vertex @xmath204 we will form a cycle , @xmath205 . the cycle @xmath205 is made by taking the path to the right of @xmath204 formed by choosing the edge that is to the right and upper most at each vertex ( this path will connect @xmath204 with @xmath169 ) , the path to the left of @xmath204 made by taking the edge that is to the left and upper most at each vertex ( this path will connect @xmath204 with @xmath139 ) , together with the edge connecting @xmath139 and @xmath169 . the vertex @xmath50 is not on the cycle @xmath205 , or we would have realized a path between these vertices . now let @xmath206 be the path formed by starting at @xmath50 choosing the edge that is to the right and upper most at each vertex until the vertex is a vertex of @xmath205 , call this vertex @xmath207 , and let @xmath208 be the path formed by starting at @xmath50 choosing the edge that is to the left and upper most at each vertex until the vertex is a vertex of @xmath205 , call this vertex @xmath48 . the vertices @xmath204 , @xmath207 and @xmath48 are all on the cycle @xmath205 , the ( edges or ) paths @xmath206 , @xmath208 , and the unrealized edge or path from @xmath50 to @xmath204 are all distinct , thus by remark [ triangle ] , this can not occur . therefore there is no such unrealized edge or path . proof of ( 2 ) : since the construction does not contain any crossings , we need only show that no cycle exhibits a stabilization . assume there is a cycle in the embedding which represents a stabilized unknot . since all edges were realized as non - stabilized arcs in the first place , at each cusp of this unknot there is a vertex of @xmath10 . since a stabilization occurs , there are at least two left cusps and two right cusps . we denote the vertices at the left cusps by @xmath192 and @xmath209 and the vertices at the right cusps by @xmath194 and @xmath210 , with the four vertices appearing in order @xmath192 , @xmath194 , @xmath209 , @xmath210 in the cycle . without loss of generality we may assume that either @xmath194 , @xmath209 , and @xmath210 are at consecutive right - left - right cusps and @xmath192 is on the left of @xmath209 ( figure [ nok4_three ] , ( a ) ) or @xmath192 , @xmath210 , and @xmath209 are consecutive left - right - left cusps and @xmath194 is on the right of @xmath210 ( figure [ nok4_three ] , ( b ) ) . we show that a @xmath3 necessarily exists . ( 380 , 108 ) ( 0,0 ) ( -8,52)@xmath192 ( 144,78)@xmath194 ( 66,52)@xmath209 ( 146,22)@xmath210 ( 174,78)@xmath192 ( 326,51)@xmath194 ( 172,22)@xmath209 ( 251,52)@xmath210 ( 78,-10)(a ) ( 248,-10)(b ) we prove the case pictured in figure [ nok4_three](a ) . the other case is similar . we consider a path @xmath211 which starts at @xmath194 and always follows along an edge to the right at each vertex . we also consider a path @xmath212 which starts at @xmath210 and always follows along an edge to the right at each vertex . denote the first vertex where @xmath211 and @xmath212 intersect by @xmath213 . the vertex @xmath213 may be @xmath194 , @xmath210 , @xmath169 , or a point in between . the path which follows @xmath211 from @xmath194 to @xmath213 and then @xmath212 from @xmath213 to @xmath210 is a path between @xmath194 and @xmath210 , which is disjoint from both @xmath192 and @xmath209 , since both @xmath192 and @xmath209 lie on the left of both @xmath194 and @xmath210 . this is a third path between @xmath194 and @xmath210 , in addition to the two included in the stabilized cycle . now , consider a path @xmath214 which starts at @xmath209 and always follows along an edge to the left at each vertex . the path @xmath214 will eventually reach @xmath139 , and will intersect the stabilized cycle in a vertex @xmath215 situated on the arc of the cycle which goes between @xmath194 and @xmath210 and does not contain @xmath209 . the vertices @xmath215 , @xmath194 , @xmath209 , @xmath210 are the vertices of a ( subdivision of ) @xmath3 . recall , a graph @xmath10 is _ minor minimal _ with respect to a property if @xmath10 has the property , but no minor of @xmath10 has the property . corollary [ unknot ] together with theorem [ thmgrec ] show that @xmath3 is minor minimal with respect to the property of not having a legendrian embedding with all cycles trivial unknots . not only that , but theorem [ thmgrec ] shows that @xmath3 is the only graph in this minor minimal set , thus characterizing this property . it is known that certain types of legendrian knots and links are determined by the classical invariants @xmath7 and @xmath8 in @xmath143 . in @xcite , eliashberg and fraser showed that the legendrian unknot is determined by @xmath7 and @xmath8 . in @xcite , etnyre and honda showed the same holds for torus knots and the figure eight knot , and in @xcite , ding and geiges showed that links consisting of an unknot and a cable of that unknot , are classified by their oriented link type and the classical invariants in @xmath143 . in this section we investigate what types of spatial graphs are classified up to legendrian isotopy by the pair @xmath4 . it is useful to recall that within each legendrian isotopy class , at each vertex , the edges appear in a fixed cyclic order . for now we consider graphs that have cut edges or cut vertices . [ cute ] let @xmath10 be a graph containing a vertex of valence at least three which is incident to at least one cut edge . for different legendrian realizations of @xmath10 , the order of edges at this vertex can differ , while the classical invariants for all cycles are the same . this is because the cut edge @xmath85 does not appear in any cycle . see figure [ tbrot](a ) . that is , a legendrian embedding of @xmath10 is not determined by the pair @xmath4 . [ cutv ] let @xmath10 be a graph containing a cut vertex of valence at least four . for different legendrian realizations of @xmath10 , the order of edges at this vertex can differ , while the classical invariants for all cycles are the same . for an example see figure [ tbrot](b ) . note that there are no cycles containing edges from both @xmath49 and @xmath22 . in particular , the modified edge is not in such a cycle , thus @xmath7 and @xmath8 are the same for the two embeddings . that is , a legendrian embedding of @xmath10 is not determined by the pair @xmath4 . suppose that @xmath10 has at least one cut edge or vertex and is determined by the pair ( @xmath7 , @xmath8 ) . suppose @xmath10 has a cut vertex @xmath50 , then by remark [ cutv ] the vertex @xmath50 must have valence 3 or less . if @xmath50 is a cut vertex with valence 3 then it must be adjacent to a cut edge , but by remark [ cute ] this would imply that @xmath10 is not determined by the pair ( @xmath7 , @xmath8 ) . so @xmath50 must have valence 2 or less . however , to be a cut vertex it must have at least valence two . for a valence 2 cut vertex both incident edges are cut edges , so by remark [ cute ] both of the vertices defining these edges must have valence 2 or less . therefore the only such graph with a cut vertex is a path graph . so @xmath10 has no cycles . suppose @xmath10 has a cut edge @xmath85 with no cut vertices . by remark [ cute ] the valences of the two vertices must be 2 or less . since the two vertices can not be cut vertices they are of valence one . thus the graph @xmath10 is a single edge . thus there does not exist a graph containing at least one cycle and at least one cut edge or one cut vertex that is determined by the pair ( @xmath7 , @xmath8 ) . next , we focus on legendrian embeddings of the lollipop graph and the handcuff graph . see figure [ lollipop ] . both these graphs have one cut edge . for any topological class of these graphs the legendrian class can not be determined by the pair @xmath4 , by proposition [ cutedge ] . a _ planar spatial graph _ is a spatial graph which is ambient isotopic to an embedding in the plane . a _ legendrian planar graph _ ( or a _ planar legendrian realization _ ) is a legendrian realization of a planar spatial graph . we show that if we restrict to planar spatial graphs , a pair @xmath4 determines exactly two legendrian isotopy classes of the lollipop graph and a pair @xmath4 determines exactly four legendrian isotopy classes of the handcuff graph . we do this by constructing a legendrian isotopy between an arbitrary embedding and a _ standard form _ embedding . we define a standard form embedding below . 1 . 2 . we say a legendrian unknot is in _ standard form _ if it is the lift of a front projection as in figure [ standard_unknotef](a ) or ( b ) . the front projection in figure [ standard_unknotef](a ) represents two distinct legendrian classes , depending on the chosen orientation . for the front projection shown in figure [ standard_unknotef](b ) both orientation give the same legendrian class , we fix the orientation to be the one which makes the left cusp a down cusp . we say a planar legendrian realization of the lollipop graph is _ in standard form _ if it is the lift of a front front projection consisting of one front projection of an unknot @xmath216 in standard form as in figure [ standard_unknotef ] and a nonstabilized arc at the lower right cusp of the unknot . the arc can sit in one of two ways with respect to the other edge segments coming together at the vertex . we say the planar legendrian realization of the lollipop graph is in _ standard form a _ or _ b _ if the cut edge sits as in figure [ standard_vertex_lollipop](a ) or ( b ) , respectively . we say a planar legendrian realization of the handcuff graph is _ in standard form _ if it is the lift of a front projection consisting of two non - crossing front projections of unknots @xmath217 and @xmath218 each in standard form as in figure [ standard_unknotef ] , one on the left and one on the right , and a nonstabilized arc between the lower right cusp of the unknot on the left and the leftmost cusp of the unknot on the right . the arc can sit in one of two ways with respect to the other edge segments coming together at each vertex . we say the planar legendrian realization of the handcuff graph is in _ standard form aa , ab , ba _ or _ bb _ if the cut edge sits as in figure [ standard_vertex](a ) , ( b ) , ( c ) or ( d ) , respectively . ( 320 , 90 ) ( 0,0 ) , ( b ) standard form @xmath226 , ( c ) standard form @xmath227 , ( d ) standard form @xmath228.,title="fig : " ] ( 80,58)(a ) ( 232,58)(b ) ( 80,-5)(c ) ( 232 , -5)(d ) ( 28 , 18)@xmath217 ( 130 , 18)@xmath218 ( 184,18)@xmath217 ( 286 , 18)@xmath218 ( 28 , 82)@xmath217 ( 130 , 82)@xmath218 ( 184,82)@xmath217 ( 286 , 82)@xmath218 in @xcite , eliashberg and fraser showed that a legendrian unknot @xmath49 is legendrian isotopic to a unique unknot in standard form . the number of cusps and crossings of the unknot in standard form ( see figure [ standard_unknotef ] ) are uniquely determined by @xmath53 and @xmath56 as follows : let @xmath10 be a legendrian graph consisting of a legendrian knot and a cut edge connected to it . through legendrian isotopy , the cut edge can be moved to be connected at any point of the knot . [ slide0 ] we work with a front projection of the graph @xmath10 . away from the cusps the cut edge can be moved by planar isotopy . a cut edge can be passed through a right cusp as in figure [ slide ] ( below or above , depending on how it sits with respect to the cusp ) . passing from the lower strand to the upper strand of a right cusp can be obtained by vertical reflection of the two illustrated cases . diagrams for passing through a left cusp can obtained by horizontal reflection of the diagrams for the right cusp . we construct a legendrian isotopy between @xmath23 and one of the two standard forms . denote by @xmath139 the valence three vertex of @xmath23 , by @xmath140 the valence one vertex of @xmath23 , denote by @xmath216 the loop edge of @xmath23 , and denote by @xmath85 the cut edge of @xmath23 . we work with a front projection of @xmath23 . _ step 1 _ ( remove the crossings of the cut edge with itself and with @xmath216 ) . starting from @xmath140 towards @xmath139 , retract the edge @xmath85 in a @xmath238neighborhood of @xmath139 , and remove all its self crossings in the front projection , as well as the crossings between @xmath85 and @xmath216 . _ step 2 _ ( put @xmath216 in standard form ) . change @xmath23 by legendrian isotopy in a neighborhood of @xmath139 such that the unknot @xmath216 is everywhere smooth . by @xcite , there exists a unique unknot in standard form which is legendrian isotopic to @xmath216 . take @xmath216 to standard form through legendrian isotopy , while keeping @xmath140 and its neighborhood containing @xmath85 away from the isotopy . we can do this by sliding the cut edge when necessary , as in lemma [ slide0 ] . _ step 3 _ ( slide the cut edge to the lower right cusp of @xmath216 ) . using lemma [ slide0 ] , slide the cut edge so that it connects to the rest of the graph at the lower right cusp of @xmath216 . starting from @xmath140 towards @xmath139 , retract the edge @xmath85 in a @xmath238neighborhood of @xmath139 , and remove all self crossings in the front projection , as well as the all crossings between @xmath85 and @xmath216 . now the graph is in one of the two standard forms . since a standard form of the unknot is uniquely determined by @xmath7 and @xmath8 , each of the standard forms of the lollipop graph are also determined by @xmath7 and @xmath8 . thus we have two legendrian isotopy classes for the lollipop graph for each pair @xmath4 . we construct a legendrian isotopy between @xmath23 and one of the four standard forms . denote by @xmath139 and @xmath140 the two vertices of @xmath23 , denote by @xmath217 and @xmath218 the two loop edges of @xmath23 , and denote by @xmath85 the cut edge of @xmath23 . we work with a front projection of @xmath23 . _ step 1 _ ( make @xmath217 and @xmath218 disjoint in the front projection ) . since the graph @xmath23 is topologically equivalent to the embedding in figure [ lollipop](b ) , the two unknots @xmath217 and @xmath218 bound disks @xmath239 and @xmath240 which are disjoint from each other and disjoint from the rest of the graph . shrink the disks @xmath239 and @xmath240 in @xmath238neighborhoods of @xmath139 and @xmath140 , with @xmath241 small enough for there to exist no crossings between @xmath217 and @xmath218 in the front projection . ( 390 , 280 ) ( 0,0 ) undoes the stabilization in the cut edge for the case when the cut edge connects _ outside _ the cusp at @xmath140 . , title="fig : " ] ( 110 , 256)iv ( 244 , 256)iv ( 370 , 254)vi ( 100 , 181)ii ( 235 , 182)ii ( 120 , 107)ii s ( 373 , 182)ii ( 249 , 108)ii ( 117 , 33)i ( 40 , 260)@xmath223 ( 180 , 187)@xmath224 ( 313 , 187)@xmath224 ( 57 , 115)@xmath242 ( 171 , 128)@xmath224 ( 58 , 43)@xmath223 ( 117 , 33)i _ step 2 _ ( remove crossings of the cut edge with itself and with @xmath217 and @xmath218 ) . there exists an embedded @xmath24sphere @xmath243 such that @xmath217 is contained in the @xmath244ball @xmath245 bounded by @xmath243 and @xmath243 intersects the cut edge at one point , @xmath192 . shrink @xmath245 in a small neighborhood of @xmath192 . starting from @xmath192 retract the cut edge while carrying along the neighborhood of @xmath192 and undo its knotting outside of @xmath245 as well as all crossings with @xmath218 . next , there exists an embedded @xmath24sphere @xmath246 disjoint from the sphere @xmath243 such that @xmath218 lies in the @xmath244ball @xmath247 bounded by @xmath246 and @xmath246 intersects the cut edge at one point , @xmath194 . shrink @xmath247 in a small neighborhood of @xmath194 . this move may introduce a crossing between @xmath85 and @xmath218 in the front projection . starting from @xmath194 retract the cut edge while carrying along the neighborhood of @xmath194 and undo its knotting outside of @xmath247 as well as all crossings between the cut edge and @xmath217 in the front projection . _ step 3 _ ( put @xmath217 and @xmath218 in standard form , slide one end of the cut edge to the lower right cusp of @xmath217 , and slide the other end to the left cusp of @xmath218 ) . take @xmath217 into a small neighborhood of @xmath139 . modify @xmath218 through a legendrian isotopy which takes it to the unknot in standard form having the assigned @xmath7 and @xmath8 . by sliding the cut edge repeatedly ( as in lemma [ slide0 ] ) , we can keep @xmath139 and @xmath217 away for this isotopy . once @xmath218 is in standard form , using lemma [ slide0 ] slide the cut edge so that it connects to @xmath218 at the left cusp of @xmath218 . the cut edge can sit in two ways with respect to the other two edge segments at this cusp . leaving @xmath218 in standard form and leaving @xmath85 connected to @xmath218 at the left cusp , move @xmath217 and @xmath85 through legendrian isotopy so that the front projection of @xmath217 lies outside and to the right of the bounded region in the plane determined by the front projection of @xmath218 . modify @xmath217 through a legendrian isotopy which takes it to the unknot in standard form having the assigned @xmath7 and @xmath8 . by sliding the cut edge repeatedly ( as in lemma [ slide0 ] ) without contracting it , we leave @xmath218 unchanged . once @xmath217 is in standard form , using lemma [ slide0 ] slide the cut edge without contracting it so that it connects to @xmath217 at the lower right cusp of @xmath217 . ( 380 , 240 ) ( 0,0 ) undoes the stabilization in the cut edge for the case when the cut edge connects _ inside _ the cusp at @xmath140.,title="fig : " ] ( 107 , 224)iv ( 237 , 224)vi ( 357 , 224)vi ( 107 , 155)ii ( 243 , 156)iv ( 359 , 155)iv ( 108 , 91)ii ( 241 , 91)ii ( 363 , 93)ii s ( 234 , 27)i , vi , iv ( 28 , 230)@xmath223 ( 190,162 ) @xmath224 ( 305 , 99)@xmath242 ( 22 , 47)@xmath223 _ step 4 _ ( undo stabilizations of the cut edge and reach one of the standard forms ) . the cut edge can connect in two ways at @xmath140 relative to the other two edge segments , _ outside _ the cusp , or _ inside _ the cusp . 1 . if the cut edge connects outside the cusp , then the stabilizations of the cut edge can be removed by flipping @xmath218 horizontally , as in figure [ flip1 ] . the other type of stabilization is solved by reflecting the diagrams . after undoing the additional stabilizations , the graph is in one of the standard forms @xmath219 or @xmath227 , depending on how the cut edge sits at @xmath139 relative to the other two edge segments . if the cut edge connects inside the cusp , then the stabilizations of the cut edge can be removed by flipping @xmath218 horizontally , as in figure [ flip2 ] . the other type of stabilization is solved by reflecting the diagrams . after undoing the additional stabilizations , the graph is in one of the standard forms @xmath226 or @xmath228 , depending on how the cut edge sits at @xmath139 relative to the other two edge segments .

recent numerical simulations have shown that the distribution of conductances in three dimensional strongly localized systems differs significantly from the expected log normal distribution . to understand the origin of this difference analytically , we use a generalized dorokhov - mello - pereyra - kumar ( dmpk ) equation for the joint probability distribution of the transmission eigenvalues which includes a phenomenological ( disorder and dimensionality dependent ) matrix containing certain correlations of the transfer matrices . we first of all examine the assumptions made in the derivation of the generalized dmpk equation and find that to a good approximation they remain valid in three dimensions ( 3d ) . we then evaluate the matrix numerically for various strengths of disorder and various system sizes . in the strong disorder limit we find that can be described by a simple model which , for a cubic system , depends on a single parameter . we use this phenomenological model to analytically evaluate the full distribution for anderson insulators in 3d . the analytic results allow us to develop an intuitive understanding of the entire distribution , which differs qualitatively from the log - normal distribution of a q1d wire . we also show that our method could be applicable in the critical regime of the anderson transition .

cond-mat0501101

the full distribution of conductances @xmath0 for non - interacting electrons at zero temperature has recently been studied in detail in quasi one dimension ( q1d ) both analytically @xcite and numerically @xcite . large mesoscopic fluctuations lead to several remarkable features in the distribution , including a highly asymmetric ` one - sided ' log - normal distribution at intermediate disorder between the metallic and insulating limits @xcite , and a singularity in the distribution near the dimensionless conductance @xmath2 in the insulating regime @xcite . while some numerical studies exist for 3d finite size systems @xcite , there is no analytic method currently available to study the full distribution @xmath0 in 3d , especially at strong disorder . theoretical work based on @xmath3 dimensions ( @xmath4 ) , where a weak disorder approximation can be applied , has been used to propose that the critical distribution at the anderson transition point has a gaussian center with power law tails @xcite , but this can not be compared with numerical results in 3d @xcite that show a highly non - trivial asymmetric distribution similar to the one - sided log - normal form of q1d . it has not been possible so far to study analytically even the simpler case of @xmath0 in the deeply insulating regime in 3d , where numerical results point to non - trivial deviations from the expected log - normal form @xcite . the @xmath0 in q1d systems were studied analytically within the transfer matrix approach @xcite . in this paper we use a recently proposed generalization of the q1d approach @xcite to obtain analytically for the first time the full @xmath0 for strongly disordered 3d systems . a brief account of the work has been published earlier @xcite . in sections ii and iii we review briefly the dmpk equation and its generalization , respectively . in sect . iv we analyze in detail the numerical data for 3d disordered systems in all three transport regimes : metallic , insulating and critical . numerical data allow us to determine the free parameters of a matrix @xmath1 which appears in the generalized dmpk equation . in sect v we use them to formulate a simple model for @xmath1 , and solve the generalized dmpk equation analytically . in sect . vi , an analytical formula for the conductance distribution @xmath0 is derived in detail . in our model , the form of @xmath0 is determined by two parameters , @xmath5 , which measures the strength of the disorder , and @xmath6 , which determines the strength of the interaction term in the generalized dmpk equation . @xmath7 in the q1d systems . the fact that @xmath8 in 3d makes the statistics of the conductance in 3d different from that in q1d . although we introduced two new disorder dependent parameters , they turn out to be related to each other and we show that the present model is not in contradiction with the single parameter scaling theory @xcite . in sect . vii we compare the analytical formula for the conductance distribution with the numerical data and analyze how the distribution @xmath9 depends on the parameter @xmath6 . in the limit @xmath10 , we recover the q1d results . sect . viii discusses the possible extension of our solution to the critical point . we show that our results describe the critical regime qualitatively correctly , including the non - analyticity of the critical conductance distribution . finally , summary and conclusions are given in sect . the distribution of conductances for non - interacting electrons at zero temperature can be studied within the transfer matrix approach . in this approach , a conductor of length @xmath11 and cross - section @xmath12 is placed between two perfect leads ; the scattering states at the fermi energy then define @xmath13 channels . the @xmath14 transfer matrix @xmath15 relates the flux amplitudes on the right of the system to those on the left @xcite . flux conservation and time reversal symmetry ( we consider the case of unbroken time reversal symmetry only ) restricts the number of independent parameters of @xmath15 to @xmath16 and @xmath15 can be written in general as @xcite @xmath17 where @xmath18 are @xmath19 unitary matrices , and @xmath20 is a diagonal matrix , with positive elements @xmath21 . microscopic distribution of impurities will lead to a distribution @xmath22 of the transfer matrices where @xmath23 is an invariant measure which we rewrite as @xmath24 if we know the marginal distribution @xmath25 then the distribution of conductances @xmath0 can be written as @xmath26 where @xmath27 is the landauer conductance @xcite . a systematic approach to evaluate the @xmath28-dimensional integral , based on a mapping to a one - dimensional statistical mechanical problem , has been developed @xcite , so the full distribution @xmath0 can be obtained if the marginal distribution @xmath29 is known . note that the distribution of other transport variables which can be written as @xmath30 , e.g. shot noise power @xcite @xmath31 or conductance of n - s ( normal metal - superconductor ) microbridge @xcite @xmath32 , can also be obtained in the same way . the above approach is valid in principle for all strengths of disorder , in all dimensions . if we assume that the distribution @xmath33 is _ independent _ of @xmath18 , then the evolution of the distribution with length @xmath11 can be obtained from a fokker - planck equation first derived by dorokhov and by mello , pereyra and kumar @xcite which has become known as the dmpk equation : @xmath34,\cr j & \equiv & \prod_{a < b}^n|\lambda_a-\lambda_b|^{\beta}.\end{aligned}\ ] ] here @xmath35 is the mean free path and the parameter @xmath36 is equal to @xmath37 or @xmath38 depending on orthogonal , unitary or symplectic symmetry of the transfer matrices . we will consider only the case with time - reversal symmetry , for which @xmath39 . although the parameters @xmath40 are not eigenvalues of @xmath15 , it turns out that they determine the eigenvalues of the matrix @xmath41 ( @xmath42 is the transmission matrix@xcite ) [ tt ] tt^=v^*(1+)^-1v which characterizes the conductance given by eq . ( [ eq - landauer ] ) , and the matrix @xmath43 contains the eigenvectors of @xmath41 . so we will loosely refer to these as the eigenvalues and the eigenvectors in the text . note that the parameter @xmath36 determines the strength of ` level repulsion ' between eigenvalues . the assumption that @xmath33 is independent of @xmath18 restricts the validity of the dmpk equation to quasi one dimension ( q1d ) . quasi 1d means not only that @xmath44 where @xmath11 is the direction of the current and @xmath45 is the cross - sectional dimension , but it also requires that @xmath46 , where @xmath47 is the localization length . in this limit , all channels become ` equivalent ' , the matrices @xmath48 and @xmath43 become isotropic and the distribution becomes independent of @xmath48 or @xmath43 . the distribution of conductances @xmath0 for such q1d systems has been studied recently ; it has many surprising features arising from large mesoscopic fluctuations . these include a highly asymmetric ` one - sided ' log - normal distribution at intermediate disorder between the metallic and insulating limits @xcite , and a singularity in the distribution near @xmath2 in the insulating regime @xcite . it is not clear if these features persist in higher dimensions . to study 3d systems , a phenomenological generalization of the dmpk equation has recently been proposed in which the q1d restriction is lifted in favor of an unknown matrix @xmath49 where the angular bracket represents an ensemble average . in terms of this matrix , the marginal distribution @xmath50 satisfies an evolution equation given by @xcite @xmath51,\cr \bar{j}&\equiv&\prod_{a < b}^n|\lambda_a-\lambda_b|^{\gamma_{ab } } ; \;\;\ ; \gamma_{ab}\equiv\frac{2k_{ab}}{k_{aa}}.\end{aligned}\ ] ] in q1d under the isotropy condition , the matrix @xmath1 reduces to @xmath52 and one recovers the dmpk equation ( with @xmath39 ) . in 3d , @xmath1 is not known analytically , and must be obtained from independent numerical studies . there are two major assumptions made in ref . [ ] in deriving eq . ( [ three ] ) : * the elements @xmath53 can be replaced by their mean values @xmath54 , and * the @xmath11-dependence of @xmath54 is negligible . these assumptions need to be verified before the equation can be used . note that the matrix @xmath1 depends on the choice of representation . since the assumptions are most natural in the position representation , we will study the matrix in this representation . the generalized dmpk equation apparently introduces a large number of new parameters , elements of the matrix @xmath54 . there is no theoretical prediction about how these parameters should depend on the size of the system or on disorder . we only know that in the q1d limit they should follow eq . ( [ dmpk - q1d ] ) . therefore our first goal is to study numerically various 3d and q1d systems systematically in detail in order to answer the following questions : * are the assumptions ( i ) and ( ii ) discussed in section iii valid at all strengths of disorder ? * how do the elements @xmath54 depend on disorder and on the system size ? * how do the elements @xmath54 depend on the indices @xmath55 and @xmath56 ? * given the size , disorder and index dependence of @xmath54 , is it possible to construct a simple model of @xmath1 at all disorder with only a small number of parameters ? we will address all of the above in this section , but let us first briefly discuss the numerical procedure used to evaluate @xmath1 . we consider the tight binding anderson model defined by the hamiltonian @xmath57 } t_{nn'}c^\dag_n c_{n'}.\ ] ] in eq . ( [ andham ] ) , @xmath58 counts sites on the simple cubic lattice of the size @xmath59 , and @xmath60 are random energies , uniformly distributed in the interval @xmath61 $ ] . the parameter @xmath62 measures the strength of disorder . the fermi energy is chosen as @xmath63 . the hopping term @xmath64 between the nearest - neighbor sites @xmath65 is unity for hopping along the @xmath66 direction and @xmath67 for hopping in the @xmath68 and @xmath69 directions . then the dispersion relation is [ ef ] e=2k_z + 2tk_x + 2tk_y for a given cross section of the sample : @xmath70 , @xmath71 and @xmath72 possess values @xmath73 ( we consider hard wall boundary conditions ) . at fixed energy e , given values of @xmath71 and @xmath72 determine the value of @xmath74 , which is either real ( if @xmath75 ) or imaginary . the latter case corresponds to closed channels which do not transmit current in perfect leads . to avoid these closed channels , which are not considered in the dmpk formulation , we use @xmath76 . then the model eq . ( [ andham ] ) exhibits a metal - insulator transition @xcite at @xmath77 . to obtain transport properties , we use the transfer matrix method of [ ] . the main difference from previous works @xcite is that we also calculate eigenvectors of the matrix @xmath41 . using eq . ( [ one ] ) , we calculate numerically the matrix @xmath41 . owing to eq . ( [ tt ] ) , diagonalizing @xmath41 gives us @xmath20 as well as all elements of the matrix @xmath43 . note that the eigenvectors depend on the representation . in the original formulation of the dmpk approach , semi - infinite leads consist of mutually independent and equivalent 1d wires . therefore , the transfer matrix in the leads is diagonal in both channel and space representations . in numerical work , we need to distinguish between these two , since the transfer matrix is diagonal only in the channel representation . we therefore calculate the matrix @xmath41 in the channel representation , find eigenvalues and eigenvectors , and transform the latter back to the space representation to obtain the matrix @xmath43 . elements of @xmath43 are then used for the calculation of the matrix @xmath1 in the space representation . we now go back and address the questions raised at the beginning of this section . for q1d systems . ( @xmath78 : @xmath79 , @xmath80 : @xmath81 , @xmath82 , @xmath83 : @xmath84 ) . full symbols : @xmath85 , @xmath86 , @xmath87 ; open symbols : @xmath85 , @xmath86 , @xmath88 . lines with symbols : @xmath85 , @xmath89 , @xmath90 . data confirm that the distribution becomes narrower when @xmath45 increases . inset shows the same for @xmath91 and @xmath92 , @xmath93 . as expected , distributions are broader . , scaledwidth=35.0% ] for 3d systems with @xmath91 and various system sizes . ( critical disorder @xmath77 . ) open symbols : @xmath79 ; full symbols : @xmath81 , @xmath82 . inset shows the @xmath45-dependence of @xmath94 which decreases when @xmath45 increases . data confirm that distribution is self - averaging ( it becomes narrower and @xmath95 when @xmath45 increases ) although it is broader than in the q1d case.,title="fig:",scaledwidth=40.0% ] . ( a ) @xmath96 ( @xmath97 and 3 ) at the critical point @xmath98 . distribution is @xmath45-independent ( apart from the exponential tail which is broader for larger @xmath45 since mean value @xmath99 ) . note the logarithmic scale on the @xmath69 axis . ( b ) distribution of off - diagonal elements @xmath100 and @xmath101 possess sharp maxima close to zero , and long exponential tails . insets show standard deviations of distributions as a function of mean values for @xmath102 . these data also show the accuracy of our estimate of the critical point since we expect both the mean and the standard deviation to be @xmath45-independent at @xmath103 . the distributions for q1d systems @xmath104 are almost identical to those for cubes ( data not shown ) . , title="fig:",scaledwidth=40.0% ] . ( a ) @xmath96 ( @xmath97 and 3 ) at the critical point @xmath98 . distribution is @xmath45-independent ( apart from the exponential tail which is broader for larger @xmath45 since mean value @xmath99 ) . note the logarithmic scale on the @xmath69 axis . ( b ) distribution of off - diagonal elements @xmath100 and @xmath101 possess sharp maxima close to zero , and long exponential tails . insets show standard deviations of distributions as a function of mean values for @xmath102 . these data also show the accuracy of our estimate of the critical point since we expect both the mean and the standard deviation to be @xmath45-independent at @xmath103 . the distributions for q1d systems @xmath104 are almost identical to those for cubes ( data not shown ) . , title="fig:",scaledwidth=40.0% ] in the insulating regime ( @xmath105 ) . although distribution becomes narrower when system size @xmath45 increases , it is not self - averaging . inset shows the size dependence of the @xmath106 for @xmath105 and @xmath107 . @xmath108 converges to a non - zero constant when @xmath109 . , scaledwidth=40.0% ] in order to check assumption ( i ) , i.e. if the elements @xmath53 can be replaced by their average @xmath54 in 3d , we analyze the probability distribution @xmath110 . we start with the weakly disordered metallic regime . two values of disorder were used : @xmath85 and @xmath91 . more relevant than the actual strength of the disorder is the mean free path @xmath35 which can be estimated from the mean conductance @xcite [ mfp ] g= with @xmath111 . from the @xmath11 dependence of @xmath112 in q1d systems we estimate @xmath113 and @xmath114 , in units of the lattice spacing . , both @xmath115 and @xmath116 converge to @xmath11-independent values when @xmath117 . the limiting value @xmath115 depends on @xmath118 . shown are also data for the insulating regime @xmath107 , @xmath76 . all data for @xmath119 almost coincide so that they are not distinguishable in the figure . , scaledwidth=35.0% ] first , we test how assumption ( i ) is fulfilled in the weakly disordered limit , where the dmpk equation is known to describe the universal features of transport statistics correctly . we show in figure [ 3d_t04_q1d_k ] the distribution @xmath110 in the weakly disordered q1d regime . as expected , the width of the distribution increases when @xmath62 increases , but @xmath110 is self - averaging ( it becomes narrower when @xmath45 increases ) . figure [ 3d_t04_kov_k ] shows the same distribution for 3d systems . the distribution is again self - averaging , although much broader than in the q1d case . in the critical regime , the probability distribution @xmath110 is no longer self - averaging but tends to be @xmath45-independent in the limit @xmath109 ( fig . [ pkaa_cp ] ) . although the distributions possess long exponential tails , they have well defined sharp maxima , which do not depend on the system size . in the insulating regime , the distribution @xmath120 becomes narrower when @xmath45 increases ( fig . [ loc_a ] ) . however , on the basis of our numerical data we conclude that the distribution is not self - averaging . although var @xmath121 decreases when @xmath109 ( data not shown ) , the normalized width @xmath106 ( shown in inset of fig . [ loc_a ] ) slightly increases when @xmath45 increases . as @xmath122 itself is non - zero in the limit @xmath109 ( fig . [ 3d_t04_k11 ] ) , @xmath108 should converge to an @xmath45-independent function for large @xmath45 . the distribution of off - diagonal elements @xmath123 in the insulating regime ( not shown ) is qualitatively the same as that at the critical point . we conclude that both in the critical and localized regimes the distributions converge to @xmath45 independent functions with a well - defined peak , but the standard deviation is of the same order of magnitude as the mean . we note that the most - probable value of @xmath121 is always very close to its mean value ; we therefore expect that replacing @xmath121 by its mean value @xmath122 is a reasonable approximation as long as one is interested in qualitative results only . thus , although to leading order assumption ( i ) remains valid for all disorder , we have to keep in mind that fluctuations of the elements @xmath53 in the strongly disordered regime might become important if the final results are sensitive to the exact values of these elements . we have checked that the final distribution of conductances do not change in any appreciable way if fluctuations of @xmath121 are included by overaging the conductance distribution over @xmath120 . to check assumption ( ii ) , namely if the @xmath11 dependence of @xmath54 is negligible , we studied the @xmath124 dependence of @xmath122 and @xmath119 . figure [ 3d_t_cp ] confirms that for both critical and insulating regimes the parameters @xmath122 and @xmath119 converge to non - zero ( although @xmath118 - dependent ) limits when the length of the system increases . it shows that the properties of the matrix @xmath1 depend only slightly on the ratio @xmath124 and reach @xmath11-independent limiting values when @xmath117 in all transport regimes . the assumption ( ii ) is therefore reasonably well satisfied at all disorder as long as @xmath125 . thus we conclude from our numerical studies that to leading approximation the generalized dmpk equation ( [ three ] ) remains qualitatively valid at all disorder in 3d , but the effect of fluctuations of @xmath53 on the final results has to be evaluated in more detail before a quantitative comparison with numerical results can be made . we start with the weak disorder regime . to distinguish the generic @xmath62 dependence of @xmath122 from finite size effects , we analyzed in fig . [ 3d_t04_k11_metal ] the @xmath45 dependence of the parameter = ( n+1)k_11/2 . as expected , @xmath126 decreases when @xmath45 increases . however , from the analysis of q1d systems ( also shown in fig . [ 3d_t04_k11_metal ] ) we conclude that @xmath126 converges to 1 only for very small values of disorder . as shown in fig . [ 3d_t04_k11_metal ] , @xmath127 for @xmath91 . thus , deviations from eq . ( [ dmpk - q1d ] ) already appear in the metallic limit , probably due to the decrease of the mean free path . dependence of @xmath128 ( open symbols ) and @xmath129 ( full symbols ) in the metallic regime . data confirms that @xmath126 depends on @xmath45 . this agrees with data in fig . [ metal_kaa ] . also , @xmath130 differs from @xmath122 for small @xmath45 . this agrees with data in fig . [ metal_gamma1a ] . to estimate limiting behavior of @xmath122 , we also considered q1d systems @xmath131 . @xmath126 is close to 1 only for very weak disorder @xmath85 . for @xmath91 we obtained @xmath127 . as this value does not depend on @xmath45 for @xmath102 ( @xmath132 ) , we expect that 3d data for @xmath91 will converge to the same value when @xmath109 . , scaledwidth=35.0% ] and @xmath6 for various system sizes . note the common crossing point at @xmath98 . note also that @xmath10 for @xmath133 and @xmath109 , as expected from dmpk , but @xmath6 decreases with the system size for @xmath134.,title="fig:",scaledwidth=35.0% ] and @xmath6 for various system sizes . note the common crossing point at @xmath98 . note also that @xmath10 for @xmath133 and @xmath109 , as expected from dmpk , but @xmath6 decreases with the system size for @xmath134.,title="fig:",scaledwidth=35.0% ] as a function of the system size for various strengths of disorder . in the metallic regime , @xmath6 converges to 1 when @xmath109 . at the critical point , @xmath6 is @xmath45-independent , and in the insulating regime @xmath6 decreases when @xmath45 increases . solid lines are linear fits @xmath135 with @xmath136 for @xmath137 . , scaledwidth=35.0% ] for various values of disorder . solid lines are linear fits . data in legend presents limiting values @xmath138 . shown are also data for the metallic regime ( @xmath91 ) for which @xmath139 ( table i ) . while @xmath140 in metal and at the critical point ( @xmath141 ) , it is non - zero in the insulating regime . thus , it may be used as the order parameter of the anderson transition . , scaledwidth=40.0% ] for various values of disorder in the insulating regime . data confirm that @xmath142 and that @xmath143 . , scaledwidth=40.0% ] and from @xmath115 . both parameters @xmath112 and @xmath115 are @xmath45-independent at @xmath103 . @xmath144 ( @xmath145 in the insulating ( metallic ) regime , respectively . from data we conclude that @xmath146 and use @xmath147 throughout the paper . , title="fig:",scaledwidth=35.0% ] and from @xmath115 . both parameters @xmath112 and @xmath115 are @xmath45-independent at @xmath103 . @xmath144 ( @xmath145 in the insulating ( metallic ) regime , respectively . from data we conclude that @xmath146 and use @xmath147 throughout the paper . , title="fig:",scaledwidth=35.0% ] in the next step we analyze how the matrix @xmath1 changes when the disorder increases . typical results are given in fig . [ w_dependence ] . our data show that the @xmath45 dependence of the parameters @xmath115 and @xmath6 is different for different transport regimes . while in the metallic regime @xmath115 decreases as @xmath148 and @xmath6 converges to unity when @xmath45 increases , qualitatively different behavior is obtained at strong disorder . in the insulating regime @xmath122 converges to a non - zero @xmath45-independent constant when @xmath109 ( fig . [ 3d_t04_k11 ] ) and @xmath119 converges to zero as @xmath142 ( fig . [ 3d_t04_k12 ] ) . this means that @xmath149 in the insulating regime ( fig . [ loc_gamma ] ) . figures 7 a , b also show that there exists a critical disorder @xmath103 where both @xmath115 and @xmath6 are independent of @xmath45 . note that @xmath8 at @xmath103 . we found that the critical value @xmath150 depends on the anisotropy ( fig . [ 3d_t_cp ] ) . for the present case @xmath76 , @xmath151 . the qualitative @xmath45 dependence in different transport regimes is summarized in table [ table : one ] . .typical @xmath45-dependence of the parameters @xmath122 and @xmath6 in the metallic , critical and localized regimes . [ cols="<,^,^,^,^,^,^",options="header " , ] we find that the index dependence of @xmath152 can be ignored in the metallic regime . the @xmath55-dependence is more pronounced at the critical point and in the localized regime . on the other hand , higher channels ( @xmath153 ) do not contribute to the transport either at the critical point or in the localized regime , so the actual values of @xmath54 for large @xmath55 and @xmath56 are not important . therefore , we conclude that the weak index dependence of the matrix elements @xmath54 is less relevant for transport properties compared to the dependence of the matrix elements on disorder in the @xmath154 limit . it is also worth mentioning that since we are interested only in @xmath45-independent quantities at the critical point , the @xmath55-dependence of any parameter is relevant only for @xmath155 . when @xmath55 becomes comparable to @xmath45 , we can not distinguish the true @xmath55-dependence from finite size effects . finally we ask the question if it is possible to construct a simple model of @xmath54 with only a small number of independent parameters . we just concluded in the previous section that the weak index dependence of the matrix at strong disorder is not very important . we expect that the crude approximations @xmath156 capture the major qualitative features of the matrix @xmath1 . approximation ( [ crude ] ) introduces two new parameters , @xmath122 and @xmath6 . we show that both parameters @xmath122 and @xmath6 are unambiguous functions of the localization length . to do so , we calculated the limiting values of @xmath122 ( fig . [ 3d_t04_k11 ] ) and of the parameter @xmath157 ( fig . [ 3d_z1 ] ) and plot @xmath122 _ versus _ @xmath47 ( fig . [ 3d_univ ] ) . we observe that the @xmath47 dependence of @xmath122 depends also on the anisotropy parameter @xmath118 ( data not shown ) . in the same way , we analyzed parameter @xmath6 . as was shown in fig . [ loc_gamma ] , we expect @xmath158 in the localized regime . data in figures [ 3d_t04_k11 ] and [ 3d_t04_k12 ] support this assumption . to estimate the disorder dependence of @xmath6 , we plot in fig . [ 3d_univ_lg ] the quantity [ lg ] _ l l_12=2l which indeed increases linearly with localization length in the strongly localized regime . from the linear @xmath45 dependence @xmath159 . ( @xmath160 ) . values of @xmath47 are given in the legend . @xmath161 . , scaledwidth=35.0% ] depends on @xmath157 for various disorder in the localized regime . data confirm linear relation @xmath162 . figure ( b ) shows how limiting values of @xmath122 , obtained from the @xmath45-dependence of @xmath163 ( fig . [ 3d_t04_k11 ] ) , depend on values of @xmath164 ( fig . [ 3d_z1 ] ) . data confirms that there is an unambiguous @xmath47-dependence of @xmath122 . , title="fig:",scaledwidth=35.0% ] depends on @xmath157 for various disorder in the localized regime . data confirm linear relation @xmath162 . figure ( b ) shows how limiting values of @xmath122 , obtained from the @xmath45-dependence of @xmath163 ( fig . [ 3d_t04_k11 ] ) , depend on values of @xmath164 ( fig . [ 3d_z1 ] ) . data confirms that there is an unambiguous @xmath47-dependence of @xmath122 . , title="fig:",scaledwidth=35.0% ] defined by eq . ( [ lg ] ) as a function of @xmath47 . data confirms that there is an unambiguous @xmath47-dependence of @xmath6 . this is important for the formulation of the single parameter scaling theory . dashed line is the linear dependence @xmath165 , considered in [ ] . , scaledwidth=35.0% ] , given by eq . [ gamma ] for a cubic system , as a function of disorder for 3d systems in the insulating regime @xmath134 . , scaledwidth=35.0% ] while modeling the full @xmath166 at the critical point needs more careful numerical studies , the insulating limit is simpler and provides a test case for the generalized dmpk equation ( 8) . it predicts that the logarithmic interaction between the transmission eigenvalues @xmath40 vanishes as @xmath167 in the insulating limit . in the rest of the paper we will test this prediction by evaluating the full distribution of conductances in the insulating limit for a 3d conductor as described by eq . ( 8) , using the simple approximate model for @xmath1 suggested by our numerical studies , namely @xmath168 it is useful to introduce another parameter , = and the ratio [ gamma ] = = . @xmath169 in the insulating regime , both in 3d ( @xmath170 ) and q1d ( @xmath171 ) systems . it measures the strength of the disorder . the ratio @xmath172 reduces to @xmath5 in the q1d limit when @xmath173 . for 3d ( @xmath174 ) we see in fig . [ 3d_t04_gamma ] that @xmath175 varies smoothly between about @xmath176 and @xmath177 when disorder increases from @xmath178 to infinity . thus , the strength of disorder is characterized predominantly by the parameter @xmath5 . for definiteness , we will use @xmath179 appropriate for a cubic system when needed for comparison with numerical results @xcite . note that fluctuations in @xmath180 , ignored in the model , would lead to fluctuations of @xmath5 and @xmath6 as well , but will not change either the length or the disorder dependence of these parameters . a brief description of the results has appeared in [ ] . here we provide many of the details . we rewrite the generalized dmpk as @xcite @xmath181\ ] ] where @xmath182 we define @xmath183 . then following [ ] , @xmath184 satisfies the imaginary time schrodinger equation @xmath185 with @xmath186}\\ ~&\displaystyle{+ \eta\sum_{a < b}\left[\frac{1}{\sinh^2(x_a - x_b)}+\frac{1}{\sinh^2(x_a+x_b)}\right ] } , \end{array}\ ] ] and @xmath187 constant , where the strength of the interaction is given by @xmath188 the interaction term in q1d vanishes for the unitary case , when @xmath189 . note that the interaction also vanishes in the limit @xmath190 . in the 3d insulating case , @xmath191 , and the interaction can be considered negligible , for all symmetries . we can therefore use the @xmath192 solution of [ ] for our 3d insulators . the solution in the insulating limit is then given by @xcite @xmath193 , with @xmath194}\\ ~&\displaystyle{-\sum_{a=1}^n \left[\frac{1}{2}\ln \sinh 2x_a+\ln x_a-\gamma x^2_a\right]}. \end{array}\ ] ] the replacement of @xmath195 in ref . [ ] by @xmath190 in eq . ( [ ham ] ) has the consequence that while all @xmath196 in the insulating regime , the difference @xmath197 is _ not _ of the same order as @xmath198 . for example , if we keep only the first two levels , the saddle - point solutions for @xmath157 and @xmath199 give @xmath200 and @xmath201 . we therefore do not assume that @xmath202 . however , we do make the simplifying approximation that @xmath203 and @xmath204 for @xmath205 and @xmath206 . ( [ ham ] ) then becomes @xmath207,\ ] ] where @xmath208 @xmath209 @xmath210 we can now use the method developed in refs . [ ] to obtain the full distribution @xmath0 . as in [ ] , we separate out the lowest level @xmath157 and treat the rest as a continuum beginning at a point @xmath199 . then @xmath211\ ] ] where we have used @xmath212 the density satisfies the normalization condition @xmath213 @xmath0 can be obtained from @xmath214 as @xmath215 where the @xmath216-function represents the landauer formula for conductance . it turns out that because of the nonlinear dependence of the hamiltonian on the density , the complex delta function representation used in [ ] is not suitable for the present case . we therefore use a real representation @xmath217 following [ ] , @xmath0 may be expressed as @xmath218\int_0^{\infty}dx \int_{x_1}^{\infty } dx_2 e^{-f(x_1,x_2;\sigma(x))}\ ] ] where the free energy functional @xmath219 is given by @xmath220}\\ ~&\displaystyle{+ \frac{1}{\lambda^2}\left[(g - h_1)-\frac{\kappa}{2}\right]^2 } \end{array}\ ] ] where @xmath221 and @xmath222 the saddle point density is to be obtained by minimizing @xmath219 with respect to @xmath223 , subject to the normalization condition . we therefore define @xmath224\ ] ] and minimize @xmath225 . it is useful to rewrite the free energy using the normalization condition in a way that removes the upper limit from the resulting equation . we use @xmath226 \nonumber \\ & = & f(x ) \int_{x_2}^{x}dx'\sigma ( x')-\int_{x_2}^{x}dx'\sigma ( x')f(x')\end{aligned}\ ] ] to obtain @xmath227\ ] ] where we have defined @xmath228h(x).\ ] ] eq . ( [ yl ] ) evaluated at @xmath229 fixes @xmath230 . taking a derivative of eq . ( [ yl ] ) with respect to @xmath68 ( represented by a prime ) gives @xmath231 evaluated at @xmath229 , this fixes @xmath199 as the beginning of the continuum @xmath232 taking another derivative with respect to @xmath68 , it is now possible to obtain the density @xmath233 we check that plugging @xmath223 in eq . ( [ sigma ] ) back to eqs . ( [ yl ] , [ y ] ) satisfy those equations . the density has the form @xmath234 plugging this form in the definition of @xmath126 , we obtain @xmath235dx } { 1-\frac{2}{\lambda^2}\int_{x_2}^{b}h(x)b(x)dx}\ ] ] which , expanded in powers of @xmath20 , is given by @xmath236 where @xmath237\ ] ] and @xmath238 using the expansion for @xmath126 , given by eq . ( 48 ) , we obtain in the limit @xmath239 from eq . ( 42 ) @xmath240 the free energy can then be written as @xmath241.\ ] ] there are two additional constraints that were not included in the variational scheme and will be enforced directly , @xmath242 the conductance distribution now becomes @xmath243 with equations ( [ v ] ) , ( [ f ] ) , ( [ sigma ] ) and ( [ y ] ) defining @xmath244 , @xmath245 , @xmath223 , and @xmath246 , respectively . let us write @xmath247 and define @xmath248 then @xmath249 and @xmath250.\ ] ] on the other hand , defining @xmath251 and using partial integration , we get @xmath252,\ ] ] where we have neglected an irrelevant term @xmath253 independent of @xmath199 and we have used @xmath254 . using @xmath255 and @xmath256 , we rewrite the above as @xmath257 the two alternate expressions for @xmath258 can now be combined to obtain @xmath259 where @xmath260 is a constant . plugging this back to eq . ( [ f21 ] ) , we obtain @xmath261 we can again use partial integration to rewrite the first term as an integral over @xmath262 , using again the fact that @xmath254 . then the two terms can be combined to obtain @xmath263.\ ] ] we define @xmath264 then @xmath265 ; \;\;\ ; \mu_1=-\frac{v'(x_2)}{h'(x_2)}\ ] ] we already have one constraint @xmath266 . we also demand @xmath267 which requires @xmath268 this defines @xmath269 . also , from eq . ( [ mu ] ) , @xmath270 or @xmath271 on the other hand defining @xmath272 and @xmath273 , it is easy to see that @xmath274 using partial integration and the fact that @xmath266 , we can also rewrite @xmath275.\ ] ] we started with the assumption that while @xmath276 in the insulating regime , the difference @xmath201 . it is therefore important to estimate the difference from the above results . we will use saddle points of the free energy @xmath277 where @xmath258 is given in eq . ( [ f2 ] ) . let us define @xmath278 and @xmath279 . we assume @xmath280 and @xmath281 . the saddle point solutions for @xmath282 and @xmath283 are obtained from @xmath284 and @xmath285 . using chain rule to write the partial derivatives in terms of @xmath286 and @xmath287 we obtain @xmath288 where prime denotes derivatives with respect to the arguments . combining the two gives @xmath289 from the definition of @xmath258 we have @xmath290+\frac{\mu_1}{\gamma_{12 } } \frac{\partial \mu_1}{\partial x_2}j_3.\ ] ] the integrals @xmath291 and @xmath292 depend on @xmath199 only via the lower limits of the integrals . derivatives w.r.t . @xmath199 are simply the negatives of the integrands evaluated at the lower limit . using definition of @xmath293 this gives the first term in eq . ( [ f2 ] ) equal to zero , leaving @xmath294 using @xmath295 and @xmath296 for @xmath297 , we get @xmath298.\ ] ] then @xmath299 where @xmath300 , @xmath301 and @xmath302 . we neglect @xmath303 compared to the other terms in @xmath304 ( eq . ( [ ds ] ) ) and expand @xmath305 in taylor series around @xmath306 . the dominant term is @xmath307 where we have used @xmath308 . using @xmath309 , we finally obtain @xmath310 therefore , the saddle point solutions are given by @xmath311 since @xmath312 , we confirm our expectation that @xmath313 . the results are also consistent with our assumption that both @xmath282 and @xmath283 are much smaller than @xmath157 , so the free energy calculations remain valid . however , numerically we find that while @xmath314 , @xmath315 independent of disorder . thus our result @xmath316 is only qualitatively correct . the fact that actual @xmath282 is much smaller than what we find is related to the inaccuracy in our evaluation of the density . indeed , we can obtain the density directly from eq . ( [ sigma ] ) . in the limit @xmath317 , @xmath318 we find @xmath319 \end{array}\ ] ] where we have used @xmath320 ( eq ( [ f ] ) ) . in the limit @xmath321 but @xmath322 , the density simplifies to @xmath323 the linear @xmath68-dependence as well as the @xmath5 dependence agrees with numerical results . however , the slope turns out to be too large . this is possibly the consequence of our simplification of the hamiltonian eq . ( [ ham ] ) to eq . ( [ happ ] ) , where all the interaction terms were neglected except for the one between the first and the second levels . as shown in [ ] , it should be possible to obtain an integral equation for the saddle point density which can then be solved at least approximately . as we will show , the actual density of the levels play a minor role in the distribution @xmath0 , which is dominated by the first few levels . therefore our results will be qualitatively correct , although there would be quantitative discrepancies due to the difference in the density . note that in the opposite limit @xmath324 , @xmath325 , the density becomes @xmath326 this corresponds to a uniform average spacing @xmath327 of eigenvalues of order unity ( @xmath170 ) , compared to the uniform spacing @xmath328 in q1d . in contrast , 3d metals are similar to q1d metals having uniform @xmath223 extending down to @xmath329 and @xmath330 . the opening of a gap in the spectrum of lyapunov exponents @xmath331 may be considered as the signature of the anderson transition . with the above caveat in mind , the saddle point free energy @xmath332 has the form ( eq . ( [ fsp ] ) ) @xmath333,\ ] ] where primes denote @xmath68-derivatives . eq ( [ pg ] ) can then be rewritten as @xmath334 where the integration over @xmath157 is eliminated by a constraint arising from the minimization of the free energy : @xmath335 ; \\ ~&~\\ ~&g_0 = -\frac{1}{\gamma_{12}}\int_{x_2}^b \frac{dx}{f'(x)}h'(x)[v'(x)+\mu_1h'(x ) ] . \end{array}\ ] ] the lower limit @xmath269 is the larger of the additional constraints imposed by the conditions @xmath267 and @xmath336 , @xmath337 real . it is instructive to consider first a simple approximate solution of eq . ( [ model ] ) , which is dominated by the lower limit of the integral . to a good approximation , @xmath338 is negligible compared to @xmath339 in the insulating limit , and @xmath340 . the condition @xmath341 or equivalently @xmath342 gives @xmath269 from the condition @xmath343 . this gives @xmath344 and hence @xmath345 . this immediately leads to @xmath346 and variance var @xmath347 as a function of @xmath6 in the strongly insulating regime , calculated from analytical distribution ( [ model ] ) . as expected , var @xmath348 for @xmath349 but var @xmath350 for @xmath173 . note that both mean and variance depends only weakly on @xmath6 when @xmath6 is small ( which is always the case when the system size @xmath45 is large ) . , scaledwidth=35.0% ] we do a saddle point analysis of eq . ( [ pa ] ) to obtain @xmath351 and var(@xmath347 ) as a function of @xmath5 . in order to illustrate the difference between q1d and 3d insulators , we will keep the general expressions without using the condition @xmath318 . the free energy can be written as @xmath352 the saddle point solution of the mean @xmath353 is obtained from @xmath354 where the prime denotes derivative with respect to @xmath68 . denoting @xmath355 this gives @xmath356 , leading to @xmath357.\ ] ] the variance @xmath358 can be estimated from @xmath359 , giving @xmath360 in the limit @xmath190 appropriate for our 3d insulators , @xmath361 . on the other hand in the q1d limit @xmath362 , @xmath363 . thus compared to the q1d result @xmath364 , the 3d result is shifted by @xmath365 . similarly , compared to the q1d result var(@xmath366 , the 3d insulators have a much sharper distribution , with half the variance @xmath367 . both results agree with numerical data . although our model is not in general valid for @xmath368 because of our neglect of the interaction terms , in the insulating limit the interaction terms are negligible and it is useful to see how the mean and the variance changes with @xmath6 as it is changed from the 3d limiting value of zero to the q1d limiting value of unity . since @xmath369 , for a given disorder this will correspond to starting from a cubic sample of width @xmath370 ( where @xmath11 is the length ) and decreasing the width to @xmath371 to reach the q1d limit . in the strongly insulating regime . as expected , skewness is zero for @xmath173 ( q1d limit ) . , scaledwidth=35.0% ] obtained from analytical formulas ( eqs . [ res-1 ] , [ model ] and [ res-2 ] ) for @xmath372 and various values of @xmath6.,scaledwidth=35.0% ] figure [ var_gamma ] shows how the mean and the variance changes with @xmath6 according to eqs . ( [ spmean ] ) and ( [ spvar ] ) . it is not possible to obtain a simple formula for the skewness @xmath373^{3/2}$ ] except that in the limit @xmath374 it approaches a number of order unity . direct evaluation of the quantity as a function of @xmath6 is shown in fig . [ skew_gamma ] , which shows that for a given disorder , the skewness starts from zero in the q1d limit , as is well known , but saturates to a finite value ( depending on disorder ) in the 3d limit . it shows that the distribution is never log - normal for 3d insulators . it also shows that the distribution @xmath9 is almost independent of @xmath6 provided that both @xmath6 and @xmath5 are small . this explains why the distribution shown in fig . [ mmwk ] does not depend on the ratio @xmath375 . finally , fig . [ ann_plng ] shows how the entire distribution @xmath9 changes as a function of @xmath6 , from a sharp , skewed form for @xmath318 to a broad gaussian form for @xmath376 . , @xmath377 ( circles ) , and from eq . ( [ model ] ) for @xmath378 and @xmath379 ( solid line ) . both have the same mean value @xmath380 . dashed and dotted lines show eq . ( [ pa ] ) with @xmath381 and @xmath382 , respectively , with @xmath378 . shown are also numerical data for @xmath383 and @xmath86 ( triangles ) . similar agreement is obtained for other values of @xmath384 if @xmath5 is used as a free parameter ; see ref . [ ] for the case @xmath385 fitted with @xmath386 @xcite . , scaledwidth=40.0% ] ( @xmath86 ) in contrast to 3d system ( fig . [ mmwk ] ) , parameter @xmath387 . from numerical data , we have @xmath388 , @xmath389 for @xmath105 , and @xmath390 , @xmath391 for systems with @xmath392 and @xmath393 ( 40 ) , respectively . solid lines show analytical result , eq . ( [ model ] ) with @xmath6 and @xmath384 as given form numerical data . , scaledwidth=40.0% ] when comparing our theoretical prediction , eq . ( [ model ] ) , with numerical data , we distinguish two cases : ( 1 ) for 3d systems , both @xmath5 and @xmath6 decreases @xmath148 , while the ratio @xmath172 does not depend on @xmath45 . ( 2 ) for systems @xmath394 , @xmath395 while @xmath149 does not depend on @xmath11 ( see fig . [ 3d_t_cp ] ) . therefore , contrary to 3d geometry , @xmath396 for @xmath171 . this different behavior of parameters @xmath5 and @xmath6 explains difference between the shape of @xmath397 in 3d and q1d strongly disordered systems , shown in figures [ mmwk ] and [ mmwk - b ] . figure [ mmwk ] shows eq . ( [ pa ] ) compared with the results from direct integration of eq . ( [ model ] ) , both compared with numerical results based on eq . ( [ andham ] ) . for the analytic curves , we chose @xmath378 and @xmath379 to have the same @xmath384 as in the numerical case @xcite . note that using the q1d result @xmath362 gives @xmath398 , leading to a log - normal distribution ( see dotted line in fig . ( [ mmwk ] ) ) . as shown in [ ] , variance and skewness calculated from direct integration of eq . ( [ model ] ) compares well with numerical results , consistent with saddle point results from eq . ( [ pa ] ) . figure [ mmwk - b ] compares theoretical formula , eq . ( [ model ] ) with numerical data for @xmath397 for insulating samples @xmath394 . while @xmath5 is small , decreasing as @xmath399 , @xmath6 does not depend on @xmath11 and is constant for fixed @xmath45 . consequently , ratio @xmath400 increases with increasing @xmath11 . both figures [ mmwk ] and [ mmwk - b ] show qualitative agreement with numerical data and theoretical model . quantitative differences between eq . ( [ model ] ) and numerical results have origin in our simplified model eq . ( [ simple ] ) , which still overestimates the strength of the interaction for higher channels . it is important to note that in both figures [ mmwk ] and [ mmwk - b ] we compared numerical data with theoretical model with the same mean @xmath384 . this is consistent with scaling theory of localization since there is only one parameter - for instance @xmath384 - which determines @xmath397 completely . in order to make sure that the analytical model has the same @xmath384 as the numerical data for a given disorder , we used @xmath5 as a free fitting parameter . of course we could use eq . ( 20 ) to obtain @xmath5 independently for a given disorder . however , in order to do that we will need a good estimate of the mean free path @xmath401 . this is difficult in the strongly disordered regime because the mean free path defined as the decay length of the single particle green s function is actually smaller than the lattice spacing in the strongly disordered regime @xcite , and our numerical model does not allow us to obtain such small lengths with good accuracy . while independent calculations of the mean free path are available for cubic systems below critical disorder @xcite ( e.g. @xmath402 for @xmath403 in the isotropic case ) , there is no data available in the insulating regime . we therefor use @xmath5 as a free parameter . nevertheless , as a consistency check , we estimate @xmath401 from a plot of @xmath404 vs @xmath157 ( fig . 19a ) , where the slope should give the mean free path ( equivalently , we could identify @xmath405 with numerical value of @xmath157 ) . the results are plotted as the inset in fig . [ g - lng ] to show the mean free path as a function of disorder . using this result , we can estimate the value of @xmath5 corresponding to the disorder @xmath105 used in fig . we find that @xmath406 and @xmath407 , which is close to the fitting value @xmath408 . this shows that while we can not obtain @xmath5 accurately enough in our present numerical scheme , the fitted values are consistent with our crude estimates . as a further consistency check , we use the above estimate of the mean free path to plot in fig . 27 the @xmath5 dependence of @xmath384 . it shows that @xmath384 is indeed an unambiguous function of @xmath5 , as required by the theory . finally , we have checked the effects of fluctuations of @xmath121 on @xmath9 by integrating the conductance distribution in fig . 25 over the distribution @xmath120 ( fig . [ loc_a ] ) . we find that the effects are negligible . as a function of @xmath367 . open symbols : analytical result , eq . ( [ model ] ) . data confirm linear dependence @xmath409 for small @xmath5 ( strong disorder ) . for smaller disorder ( larger @xmath5 ) analytical model is less accurate . data confirm that @xmath410 does not depend on @xmath6 provided that both @xmath5 and @xmath6 are small . full symbols : numerical data with @xmath411 . we estimate @xmath5 for a given disorder using the mean free path @xmath401 obtained from data in fig . 19a ( @xmath412 is shown in inset ) and the relation @xmath413 . numerical @xmath384 is an unambiguous function of @xmath5 , as required in the analytical model . small deviations are due to finite size effects which affect actual values of parameters @xmath157 and @xmath122 . , scaledwidth=35.0% ] for @xmath414 where critical regime is expected . the distribution agrees qualitatively with numerical data for the critical regime . figure ( b ) shows detail of the distribution shown for @xmath415 , @xmath379 and for @xmath416 , @xmath173 . it shows that there is indeed a non - analyticity in the distribution close to @xmath417 with position of the non - analyticity at @xmath418 , in agreement with analytical results @xcite . , title="fig:",scaledwidth=35.0% ] for @xmath414 where critical regime is expected . the distribution agrees qualitatively with numerical data for the critical regime . figure ( b ) shows detail of the distribution shown for @xmath415 , @xmath379 and for @xmath416 , @xmath173 . it shows that there is indeed a non - analyticity in the distribution close to @xmath417 with position of the non - analyticity at @xmath418 , in agreement with analytical results @xcite . , title="fig:",scaledwidth=35.0% ] and @xmath173 ) . data confirm that the distribution @xmath0 is gaussian with variance var @xmath419 for @xmath420 and var @xmath421 for @xmath422 which is comparable to the ucf value.,scaledwidth=35.0% ] from numerical simulations @xcite we know that the critical regime in 3d is also dominated by only a few eigenvalues @xmath423 . however , since @xmath6 is neither @xmath424 nor @xmath425 , it seems that we may not be able to use the free fermion ( @xmath426 ) solution of eq . ( [ ch ] ) to obtain the distribution of the transmission levels . however , as shown in [ ] , the solution is independent of the strength of the interaction @xmath427 in the strong disorder regime characterized by @xmath423 . this means that our solutions might be used , albeit only qualitatively , even near the critical regime . we show in fig . ( [ plng_g1 ] ) the distribution @xmath9 for @xmath428 which is expected to be near the critical regime . it agrees qualitatively well with numerical results at the critical point , including a discontinuity in the slope near @xmath429 . it is known from analysis of the q1d systems @xcite that separating out an additional level helps to study the non - analyticity near @xmath429 . we therefore expect to obtain better results near the critical regime by separating out an additional level . finally , we show in fig . [ ann_metal ] the distribution @xmath0 for @xmath430 and for @xmath173 which corresponds to the metallic regime . although we do not expect that our approximate formula works quantitatively for the metallic regime , eq . ( [ model ] ) gives , for this choice of the parameters , a gaussian distribution of the conductance . also the width of the distribution qualitatively agrees with the universal conductance fluctuations @xcite in this regime . this shows that our simple model already captures the essential qualitative features at all strengths of disorder . we systematically analyzed the length and disorder dependence of the matrix @xmath1 to check if the generalized dmpk equation proposed in [ ] is valid in three dimensional systems at all strengths of disorder . we studied the matrix @xmath1 in detail . the goal was to test the assumptions on which the generalized dmpk equation was derived and to construct a simple analytically tractable model for @xmath1 which captures all the important qualitative features . in particular , since q1d systems have been studied in great detail , we looked for any major qualitative differences in the structure of @xmath1 between q1d and 3d systems . in 2d disordered systems compared with gaussian distribution with the same mean value and variance . @xmath431 , var @xmath432 and skewness is 0.251 . figure indicates that the deviation from the gaussian distribution exists in 2d as well . , scaledwidth=35.0% ] we find that to a good approximation , the generalized dmpk equation remains _ qualitatively _ valid for any disorder . we also conclude that to a good approximation , we can use only two parameters , @xmath433 and @xmath434 , to characterize the qualitative changes in transport at different strengths of disorder in different dimensions . we find that although fluctuations in @xmath121 at strong disorder are large ( non self - averaging ) , the effect of these fluctuations on @xmath9 is negligible . more importantly , we do not have an independent way to estimate the mean free path to obtain @xmath435 , but all qualitative features of the entire distribution @xmath9 is obtained correctly once an effective @xmath5 is used as a free parameter . we also find how these parameters depend on disorder and show their unambiguous dependence on the localization length . this is important since it indicates that the introduction of new parameters does not necessarily invalidate the single parameter scaling theory of localization . we have also shown that the matrix @xmath1 contains information about the anderson transition . the scaling of the parameters @xmath115 or @xmath6 clearly identifies the critical point which agrees with numerical results . we then concentrate on the strong disorder limit where our numerical results allowed us to construct a simple _ one - parameter model _ of the matrix @xmath1 , containing @xmath436 and @xmath437 , with @xmath179 . by varying @xmath6 , we show how one can go from a q1d ( @xmath46 ) to a truly 3d ( @xmath438 ) system in the insulating regime , which clearly shows the difference between a q1d and a 3d insulator . we then use the model to obtain the full distribution @xmath0 which agrees qualitatively with numerical results . it is indeed remarkable that even though the generalized dmpk equation ( [ gendmpk ] ) neglects fluctuations in @xmath53 and the model eq . ( 19 ) neglects the index dependence of @xmath54 , the theory still captures all the essential features of length , disorder as well as dimensionality dependence of the entire conductance distribution and provides in particular a simple understanding of the 3d distribution at strong disorder , which is qualitatively different from a log - normal distribution in q1d . at the same time , our numerical studies suggest that having an independent estimate of the mean free path could provide a more quantitative description of the conductance distribution in 3d at all disorder . we emphasize that there are large differences between q1d and higher dimensions . in q1d defined in [ ] , disorder is always _ weak _ enough to assure that the localization length @xmath46 where @xmath45 is the transverse dimension . the q1d insulator corresponds to the _ weakly disordered _ systems of length @xmath439 . it is this length - induced insulating behavior that is described by the dmpk equation . this is different from localization in 3d which occurs at _ strong _ disorder , where @xmath47 is much less than both @xmath11 and @xmath45 . this difference is clearly reflected in the matrix @xmath1 , where table i summarizes how the scale dependence of @xmath122 and @xmath6 depend on disorder in 3d . in contrast , @xmath1 in q1d is independent of disorder . our model recovers all the peculiarities of the 3d localized regime : we found that the distribution @xmath9 is narrower than in q1d and possesses non - zero skewness . although we concentrated on 3d systems , general considerations about the properties of the matrix @xmath1 in the insulating regime should be valid in any dimension @xmath440 . in particular , the assumption that @xmath122 is a nonzero @xmath45-independent constant in the localized regime is correct independent of dimensionality . therefore , we expect that our theory of the insulating regime is valid for any @xmath441 . consequently , the distribution @xmath9 is not gaussian in the localized regime in any @xmath441 , although the nature of the deviation from the gaussian form might be dimensionality dependent . in fact we expect @xmath9 to be different from gaussian distribution even in 2d @xcite . this expectation is supported by fig . [ 2d ] which shows @xmath442 for a 2d square system obtained numerically using eq . [ andham ] . as discussed in the paper , deviations from the gaussian form are due to the changes of the spectrum of parameters @xmath68 . although the contribution of the first channel to the conductance is dominant , higher channels do influence the statistical properties of the smallest parameter , @xmath157 . this effect was probably not considered in previous analytical works which predict gaussian distributions of @xmath347 in the insulating regimes in dimensions @xmath443 @xcite . when applied to the critical regime , our theory recovers typical properties of the conductance distribution , including the non - analyticity of the distribution in the vicinity of @xmath429 . although our present results are only qualitatively correct in the critical regime , we believe that the method developed in the paper represents a good starting point for further development of the theory of anderson transition . kam thanks u. karlsruhe for support and hospitality during his visit . pm thanks apvt , grant n. 51 - 021602 for financial support . pw gratefully acknowledges support of a visit and hospitality at the u. florida as well as partial support by a max - planck research award . note that for @xmath445 , @xmath166 must decrease as @xmath446 . indeed , from ( [ two ] ) we have @xmath447 so that the sum of @xmath448 _ positive _ parameters @xmath449 is @xmath450 in the insulating regime . we ignore this detail in our model . p. marko j. phys . : condens . * 7 * , 8361 ( 1995 ) . in ref . [ ] we used @xmath451 . this difference is not important , since , as we show in fig . [ skew_gamma ] , the actual value of @xmath172 does not influence the value of the skewness when both @xmath5 and @xmath6 are small .

we review basic design principles underpinning the construction of mimetic finite difference and a few finite volume and finite element schemes for mixed formulations of elliptic problems . for a class of low - order mixed - hybrid schemes , we show connections between these principles and prove that the consistency and stability conditions must lead to a member of the mimetic family of schemes regardless of the selected discretization framework . finally , we give two examples of using flexibility of the mimetic framework : derivation of higher - order schemes and convergent schemes for nonlinear problems with small diffusion coefficients .

1610.05850

the mixed formulation allows us to calculate simultaneously the primary solution of a pde and its flux . for this reason , mixed formulations are very useful for numerical solution of multiphysics systems . the focus of this work is on a single diffusive process that is a part of almost any complex multiphysics system . in this paper , we present design principles used in the derivation of mimetic finite difference ( mfd ) schemes on polygonal and polyhedral meshes and establish bridges to design principles used by a few other discretization frameworks ( finite volumes and finite elements ) . the focus on the design principle allows us to avoid technical details and provide a more clear connection between different frameworks in comparison with the work performed in @xcite . we also illustrate the flexibility of the mimetic framework with two challenging examples : derivation of arbitrary - order accurate schemes for linear problems and convergent schemes for nonlinear problems with degenerate diffusion coefficients . many ideas underpinning the mfd method were originally formulated in the sixties for orthogonal meshes using the finite difference framework from which the name of the method was derived . over the years , the mfd method has been extensively developed for the solution of a wide range of scientific and engineering problems in continuum mechanics @xcite , electromagnetics @xcite , fluid flows @xcite , elasticity @xcite , obstacle and control problems @xcite , diffusion @xcite , discretization of differential forms @xcite , and eigenvalue analysis @xcite . an extensive list of people who contributed to the development of the mfd method can be found in the recent book @xcite and review paper @xcite . the paper summarizes almost all known results on cartesian and curvilinear meshes for various pdes including the lagrangian hydrodynamics . the book complements the paper by providing numerous examples and describing basic tools used in the convergence analysis of mimetic schemes for elliptic pdes . the mfd method preserves or mimic essential mathematical and physical properties of underlying partial differential equations ( pdes ) on general polygonal and polyhedral meshes . for the elliptic equation , these properties include the _ local flux balance _ and the _ duality between gradient and divergence operators_. the latter implies symmetry and positive definiteness of the resulting matrix operator and is desirable for robustness and reliability of numerical simulations . the duality of the primary and derived mimetic operators is one of the major design principles . the _ definition of the primary mimetic operators is coordinate invariant _ , which is another design principle that allows us to build discrete schemes for non - cartesian coordinate systems . the discrete operators are also built to satisfy _ exact identities _ , the property that is critical for avoiding spurious numerical solutions , providing accurate modeling of conservation laws , and making the convergence analysis possible . the related discretization frameworks considered in this paper include the finite volume methods @xcite , the mixed finite element ( mfe ) method @xcite , and the virtual element method ( vem ) @xcite . other finite volume frameworks exist that are based on mimetic principles such as the discrete duality finite volume methods ( ddfv ) , see , e.g. , @xcite , but these methods do not fit in the mfd framework and will not be considered here . the fv methods , originally introduced in @xcite for the heat equation and dubbed as the integrated finite difference method , form , perhaps , the largest class of schemes that can handle unstructured polygonal and polyhedral meshes , non - linear problems , and problems with anisotropic coefficients . an introduction to the finite volume nethoddology can be found in the recent review @xcite . almost all fv methods starts with a discrete representation of the flux balance equation . this representation is exact and this property is so important that all the methods that we consider in this paper use the same discrete form of the balance equation and the difference between them is only in the discretization of the constitutive equation . the classical cell - centered fv scheme uses a two - point flux formula that is second - order accurate for special meshes such as the voronoi tessellations . to overcome this limitation , a class of fv methods , consistent by design , is proposed by introducing additional unknowns on mesh faces . examples of such methods are the _ hybrid finite volume _ method @xcite , and the _ mixed finite volume _ method @xcite . these fv methods start with different definitions of the cell - based discrete gradient that are exact for linear solutions . the formula for the numerical flux based on this gradient needs a stabilization term . construction of the stabilized flux uses two principles . first , the stabilization term should be zero on linear solutions . second , the stabilized flux is defined as the solution of a certain equation with a symmetric and positive definite bilinear form . we will show that these design principles imply the duality principle in the mimetic framework . the vem was originally introduced as an evolution of the mfd method . in the classical finite element spirit , the duality principle is incorporated directly in the weak formulation . the exact identities are replaced by the exact sequence of virtual finite element spaces . a new design principle is the unisolvency property where the space of degrees of freedom is isomorphic to a space of finite element functions and includes polynomial as well as non - polynomial functions . the bilinear forms are split explicitly into consistency and stability forms using problem - dependent @xmath0 and @xmath1 projectors . we discuss how the new design principles are connected to the stability and consistency conditions in the mimetic framework . the recent developments of the mfd framework exploit it flexibility for selecting non - standard degrees of freedom , optimization of inner products , and non - standard approximations of primary operators to build schemes with higher order of accuracy and convergence schemes for nonlinear pdes with degenerate coefficients ( see also section [ sec : divk ] ) . extension to higher - order mixed scheme is almost straightforward in the mimetic framework . the key step is the proper selection of degrees of freedom that ( a ) simplify the discretization of the primary divergence operator and ( b ) allows us to formulate a computable consistency condition . a new design principle is introduced in this case , which states that a commuting relation exists between the interpolation operators defining the degrees of freedom of scalar and vector fields , and the divergence operators in the discrete and continuum settings . all of the aforementioned methods discretize effectively the divergence operator `` _ _ div__@xmath2 '' . to solve nonlinear parabolic equations , we employ new mfd schemes where the primary operator discretizes the combined operator `` _ _ div__@xmath3 '' , where @xmath4 is the non - constant scalar diffusion coefficient . the resulting scheme uses both cell - centered and face - centered values of the diffusion coefficient . this model has applications in heat diffusion @xcite and moisture transport in porous media @xcite . the duality property mentioned above guarantees that the schemes can be formulated as algebraic problems with symmetric and positive definite matrices . matrices with these properties lead to better performance of scalable iterative solvers , such as algebraic multigrid solvers and krylov solvers such as the preconditioned conjugate gradient . finally , we mention other discretization methods that work on general meshes . our necessarily incomplete list include the polygonal / polyhedral finite element method ( pfem ) @xcite , hybrid high - order method @xcite , the discontinuous galerkin ( dg ) method @xcite , hybridized discontinuous galerkin ( hdg ) method @xcite , and the weak galerkin ( wg ) method @xcite . the outline of the paper is as follows . in section [ sec : basic : principles : mfd : framework ] , we review the basic discretization principles of the mimetic framework . in section [ sec : mixed : mimetic : formulation ] , we derive the mimetic finite difference method for elliptic problems through the consistency and stability properties . we also prove that any mixed - hydrid method that uses the same degrees of freedom leads to a member of the mimetic family of schemes . in section [ sec : recent : developments ] , we review the recent progress in the development of mimetic methods for mixed formulations of elliptic problems . our final remarks and conclusions are given in section [ sec : conclusions ] . the mfd method mimics important mathematical and physical properties of underlying pdes . we give two examples showing the importance of preserving such properties in physics simulations . consider a polygonal or polyhedral mesh @xmath5 . we denote the sets of mesh nodes , edges , faces , and cells by symbols @xmath6 , @xmath7 , @xmath8 and @xmath9 , respectively , the set of vectors collecting the degrees of freedom associated with those mesh objects by the corresponding symbol with the subscript `` @xmath10 '' , and the restriction to cell `` @xmath11 '' by the subscript `` @xmath12 '' . each set of vectors of degrees of freedom with the ( obvious ) definitions of addition and multiplication by a scalar number is a linear space . for example , @xmath13 is the linear space of vectors formed by the degrees of freedom located on the mesh faces , and @xmath14 is its restriction to cell @xmath11 . its precise definition depends on the scheme . an illustration of particular discrete spaces restricted to a single cell is shown in fig . [ fig : dofs ] . from left to right are @xmath15 , @xmath16 , @xmath14 , @xmath17 . the degrees of freedom are shown only on the visible objects . ] the mimetic finite difference method operates with discrete analogs of the first - order operators . these operators are designed to satisfy exact identities and duality principles . in the mimetic framework , we usually discretize pairs of adjoint operators , such as the primary divergence @xmath18 and the derived gradient @xmath19 . hereafter , we will distinguish the derived operators from the primary operators by using a tilde on the operator s symbol . it is convenient to think about these operators as matrices acting between finite dimensional linear spaces . to discretize a large class of pdes , we need three pairs of primary and derived operators , which are discrete analogs of gradient , curl and divergence operators . each pair of operators satisfies a discrete integration by parts formula , e.g. @xmath20_{\cc_h } = -\big[\bu_h,\ , \gradt\ , q_h\big]_{\cf_h } \qquad \forall \bu_h\in \cf_h,\ \forall q_h \in \cc_h.\ ] ] this formula represents one of the mimetic discretization principles as it mimics the continuum green formula @xmath21 the brackets @xmath22_{\cc_h}$ ] and @xmath22_{\cf_h}$ ] in stand for an approximation of the integrals in and will be referred to as the _ mimetic inner products _ ( or , simply _ inner products _ ) . the inner products are typically constructed from local ( e.g. cell - based or node - based ) inner products , which simplifies their derivation . for example , the two inner products in can be reformulated as @xmath23_{\cf_h } = \sum\limits_{c \in \omega_h } \big[\bu_c,\ , \bv_c\big]_{\cf_h , c } , \qquad \big[p_h,\ , q_h\big]_{\cc_h } = \sum\limits_{c \in \omega_h } \big[p_c,\ , q_c\big]_{\cc_h , c}\ ] ] where @xmath24 , @xmath25 , @xmath26 and @xmath27 denote the restriction to mesh cell @xmath11 of the corresponding global vectors in the left - hand side and @xmath28_{\cf_h , c}$ ] and @xmath28_{\cc_h , c}$ ] are the local contribution from @xmath11 to the global inner products @xmath28_{\cf_h}$ ] and @xmath28_{\cc_h}$ ] , respectively . let @xmath29 and @xmath30 be the symmetric positive definite matrices induced by the inner products @xmath22_{\cc_h}$ ] and @xmath22_{\cf_h}$ ] , respectively . then , the explicit formula for the derived gradient operator is @xmath31 this formula shows that this operator has a nonlocal stencil when matrix @xmath30 is irreducible as is typical for unstructured meshes . note that the same property holds true for many other discretization frameworks . formula implies that in general the discrete operators can not be discretized independently . if we discretize one of the operators , e.g. , the divergence , and select the inner products , the other operator , the gradient , must be derived from formula . the existing freedom is in the selection of the inner products . the selection of the discrete spaces is typically done to simplify the discretization of the primary mimetic operator . for a different pair of discrete spaces , e.g. , @xmath32 and @xmath33 , one can find that the gradient operator can be discretized in a natural way . in such a case , the gradient operator @xmath34 is the primary mimetic operator and the discrete divergence operator @xmath35 is the derived operator . this pair of operators satisfies another discrete integration by parts formula that mimics : @xmath36_{\cn_h } = -\big[\bv_h,\ , \grad\ , p_h\big]_{\ce_h } \qquad \forall p_h\in \cn_h,\ \forall \bv_h \in \ce_h.\ ] ] the explicit formula for the derived divergence operator is @xmath37 where @xmath38 and @xmath39 are symmetric positive definite matrices induced by the inner products . note that in some low - order mimetic schemes , matrix @xmath39 is diagonal , so that the derived operator has a local stencil . the third pair of discrete operators approximates the continuum operators `` _ _ curl _ _ '' . let @xmath40 and @xmath41 satisfy the discrete integration by parts formula @xmath42_{\cf_h } = \big[\bv_h,\ , \curlt\ , \bu_h\big]_{\ce_h } \qquad \forall \bv_h\in \ce_h,\ \forall \bu_h \in \cf_h,\ ] ] which mimics the continuum formula @xmath43 the explicit formula for the derived curl operator is @xmath44 so that this derived operator has typically a non - local stencil . the spaces of discrete functions that we have introduced so far satisfy homogeneous boundary conditions . in @xcite , these spaces were enriched conveniently to approximate the boundary integrals that appear in general green formulas . the resulting derived mimetic operators include an approximation of the boundary conditions . we will not follow this approach here , since the focus of this paper is on mixed - hybrid formulations , which provide another way to incorporate boundary conditions in a numerical scheme . the duality of the discrete operators helps us to build numerical schemes that satisfy discrete conservation laws . for example , consider the euler equations in the lagrangian form : @xmath45 where @xmath46 is the pressure , @xmath47 is the density , @xmath48 is the velocity , and @xmath49 is the internal energy . the system is closed by an equation of state . a mimetic discretization of is given by @xmath50 where @xmath51 , @xmath52 , @xmath53 and @xmath54 are the discrete analogs of the corresponding continuum quantities that appear in and @xmath55 and @xmath56 are the mimetic operators acting , respectively , as divergence and gradient . let us assume that no external work is done on the system , e.g. , @xmath57 of @xmath58 . the integration by parts and the continuity equation from lead to the conservation of the total energy @xmath59 : @xmath60 to mimic this property , we need the discrete gradient and divergence operators @xmath56 and @xmath55 to satisfy a discrete integration by parts formula like . using the same argument that leads to , we obtain the conservation of the total discrete energy @xmath61 : @xmath62_{\cf_h } - \big[p_h,\,\div \bu_h\big]_{\cc_h } = 0.\ ] ] we emphasize that numerical methods that conserve energy usually have other important properties such as correct prediction of a shock position and bounded numerical solution . another discretization principle is to derive primary operators that mimic exact identities . this is typically achieved by using the first principles ( the divergence and stokes theorems ) to define the primary operators , e.g. ( [ primary - div ] ) . as the result , we have @xmath63 another consequence of the duality principle is that similar identities hold for the derived mimetic operators . using the aforementioned explicit formulas for these operators , we immediately obtain that @xmath64 these exact identities allows us to design numerical schemes without non - physical spurious modes . for instance , in the numerical solution of maxwell s equations , such operators guarantee that the magnetic field @xmath65 remains divergence - free for all times . applying the primary divergence operator to a semi - discrete form of faraday s law of induction , i.e. , @xmath66 , we obtain @xmath67 therefore , if @xmath65 is such that @xmath68 at time @xmath69 , this relation will be satisfied at any time @xmath70 . from this section , we limit our discussion to elliptic problems and one pair of the primary and derived operators . for the practical implementation of mimetic schemes , it is convenient to write a local integration by parts formula that implies the global one . to do it , we need an additional space @xmath71 of pressure unknowns defined typically on mesh faces . its restriction to cell @xmath11 is denoted by @xmath72 and consists of the vectors @xmath73 . we recall that the subscript `` @xmath11 '' is added to denote the local mimetic operators and the local discrete spaces corresponding to cell @xmath11 . let @xmath74 be the primary divergence operator . the derived gradient operator @xmath75 satisfies the discrete integration by parts formula @xmath76_{\cc_h , c } - [ \bu_c,\ , { { \bm\lambda}}_c]_{\lambda_{h , c } } = -\big[\bu_c,\ , \gradt_c\ , \left(\!\!\begin{array}{c}q_c \\ { { \bm\lambda}}_c \end{array}\!\!\right)\big]_{\cf_h , c } \nonumber\\[0.5em ] & \hspace{7.5 cm } \forall \bu_c\in \cf_{h , c},\ \forall q_c\in \cc_{h , c},\ \forall { { \bm\lambda}}_c\in\lambda_{h , c } , \label{eq : formula : b}\end{aligned}\ ] ] which mimics the continuum green formula for cell @xmath11 : @xmath77 in order to recover formula for the global discrete gradient operator , we impose the continuity of @xmath73 and @xmath25 on the mesh faces , define the local divergence operator as the restriction of the global one , define the local spaces as restrictions of global ones , require that the interface terms cancel each other , @xmath78_{\lambda_{h , c } } = 0,\ ] ] and that the local inner products are summed up into global inner products as in ( [ additivity ] ) . the derivation of the mimetic method follows three generic steps . first , we select the degrees of freedom such that the local primary operator , e.g. , @xmath79 , has a simple form . second , we define the inner products in the discrete spaces that satisfy the consistency and stability conditions . third , we postulate the discrete integration by parts formula and obtain the derived operator , e.g. , @xmath80 , from it . note that the local derived operator is defined uniquely . these three steps are discussed in section [ sec : mixed : mimetic : formulation ] for the mixed formulation of the diffusion problem . the flexibility of the mimetic framework is exploited in section [ sec : divk ] , where we derive another pair of primary divergence and derived gradient operators for a nonlinear parabolic problem . more examples of mimetic schemes can be found in @xcite . the material properties are often included in the definition of the derived mimetic operator . indeed , the green formula can be rewritten as follows : @xmath81 according to , we can define @xmath56 as an approximation of the combined operator @xmath82 and the inner product @xmath83_{\cf_h}$ ] as an approximation of the right - hand side integral @xmath84 , provided that @xmath53 and @xmath85 are the degrees of freedom of @xmath48 and @xmath86 . for a perfectly conducting medium , we have the following duality relationship for the first - order curl operators : @xmath87 in this case the inner products in spaces @xmath33 and @xmath13 are the weighted inner products . the weights are the magnetic permeability @xmath88 and electric permittivity @xmath89 . the derived curl operator @xmath90 is an approximation of the combined operator @xmath91 . let @xmath92 be a polygonal ( @xmath93 ) or polyhedral ( @xmath94 ) domain with the lipschitz continuous boundary . consider the mixed formulation of the diffusion problem : @xmath95 \div \ , \bu & = & b & \qquad\textrm{in~}\omega , \end{array}\ ] ] subject to the homogeneous dirichlet boundary conditions on @xmath58 . as usual , we will refer to the scalar unknown as the _ pressure _ and to the vector unknown as the _ flux_. we assume that the diffusion tensor is piecewise constant on mesh @xmath5 and we denote its restriction to cell @xmath11 by @xmath96 . if @xmath96 is not constant on @xmath11 , we can take its values at the centroids of the mesh cells without losing the approximation order . in this section , we consider one local mimetic formulation , two fv schemes and two mixed - hybrid fe schemes with the same set of degrees of freedom . moreover , we consider the schemes that use the same discrete divergence operator and the discrete flux balance equation : @xmath97 where @xmath98 is a linear operator and @xmath99 is defined later . the analysis of discretization schemes is typically conducted on a sequence of conformal meshes @xmath5 where @xmath10 is the diameter of the largest cell in @xmath5 and @xmath100 . a mesh is called conformal if the intersection of any two distinct cells @xmath101 and @xmath102 is either empty , or a few mesh points , or a few mesh edges , or a few mesh faces . cell @xmath11 is defined as a closed domain in @xmath103 ( or @xmath104 with flat faces and straight edges . following @xcite , we make a few assumptions on the regularity of 3d meshes . similar assumptions can be derived for 2d meshes by reducing the dimension . let @xmath105 , @xmath106 and @xmath107 denote various mesh independent constants explained below . ( * m1 * ) every polyhedral cell @xmath11 has at most @xmath105 faces and each face @xmath108 has at most @xmath105 edges . ( * m2 * ) for every cell @xmath11 with faces @xmath108 and edges @xmath49 , we have @xmath109 where @xmath110 denotes the euclidean measure of a mesh object . ( * m3 * ) for each cell @xmath11 , there exists a point @xmath111 such that @xmath11 is star - shaped with respect to every point in the sphere of radius @xmath112 centered at @xmath111 . for each face @xmath108 , there exists a point @xmath113 such that @xmath108 is star - shaped with respect to every point in the disk of radius @xmath114 centered at @xmath115 as shown in fig . [ starshaped ] . ( * m4 * ) for every cell @xmath11 , and for every @xmath116 , there exists a pyramid contained in @xmath11 such that its base equals @xmath108 , its height equals @xmath117 and the projection of its vertex onto @xmath108 is @xmath115 . the conditions ( * m1*)-(*m4 * ) are sufficient to develop an _ a priori _ error analysis of various discretization schemes . we recall only two results underpinning this error analysis . the first one is the agmon inequality that uses ( * m4 * ) and allows us to bound traces of functions : @xmath118 the second one is the following approximation result : for any function @xmath119 there exists a polynomial @xmath120 such that @xmath121 hereafter , we will use symbols @xmath122 , @xmath123 , @xmath124 to denote generic constants independent of @xmath10 . the mesh regularity assumptions ( * m1*)(*m4 * ) are not unique . they could be generalized to non - star shaped cells @xcite by splitting each polygonal or polyhedral cell into the finite number of shape regular simplexes . as pointed out at the end of subsection [ subsec : local : mimetic : operators ] , the _ first step _ of the construction of a mimetic scheme consists in the selection of the degrees of freedom . the degrees of freedom are such that the primary divergence operator has a simple form . the discrete space @xmath125 consists of one degree of freedom per cell ; its dimension equals the number of mesh cells ; and for each vector @xmath126 we shall denote the value of @xmath51 associated with cell @xmath11 by @xmath127 furthermore , we denote the vector of degrees of freedom of a smooth function @xmath46 by @xmath128 . in the mfd method @xmath129 is typically the cell average of @xmath46 over cell @xmath11 and this definition is used by the existing superconvergence analysis @xcite . the discrete space @xmath71 consists of one degree of freedom per mesh face , e.g. , @xmath130 ; its dimension equals the number of mesh faces ; and for each vector @xmath131 we shall denote its restriction to cell @xmath11 by @xmath132 . the continuity of local vectors @xmath73 across mesh cells is satisfied automatically . the value @xmath130 can be associated with the value of a smooth function @xmath46 at the face centroid . the discrete space @xmath13 consists of one degree of freedom per boundary face and two degrees of freedom per interior face . for vector @xmath133 , we denote by @xmath25 its restriction to cell @xmath11 , and by @xmath134 its component associated with face @xmath108 of cell @xmath11 . for a smooth function @xmath48 , we denote by @xmath135 the vector of degrees of freedom . the value @xmath136 is defined as the integral average of flux @xmath137 through face @xmath108 , where @xmath138 is the face normal fixed once and for all . hereafter , we consider a subspace of @xmath13 whose members satisfy the flux continuity constraint @xmath139 on each interior face @xmath108 shared by cells @xmath101 and @xmath102 . with a slight abuse of notation , we shall refer to @xmath13 as the space that satisfies condition . the local primary divergence operator is defined using a straightforward discretization of the divergence theorem : @xmath140 where @xmath141 is either @xmath142 or @xmath143 depending on the mutual orientation of the fixed normal @xmath138 and the exterior normal @xmath144 to @xmath145 . observe that this definition remains the same in all coordinate systems . the _ second step _ of the construction of a mimetic scheme is to define accurate inner products in @xmath15 and @xmath14 that satisfy the consistency and stability conditions . the inner product for space @xmath15 is simple : @xmath146_{\cc_{h , c } } = |c|\ , p_c\ , q_c . \label{eq : inner : product : scalar : low : order}\end{aligned}\ ] ] let @xmath147 be the space of constant functions . then , the above inner product implies the obvious result : @xmath148_{\cc_{h , c } } = \int_c p\ , q \dx \qquad\forall p\in \cp^0(c),\ \forall q \in { \cal sc}_c.\ ] ] despite its simplicity , we can use this relation to formulate the general principle that we apply to the derivation of other inner products . we define the consistency condition as the following exactness property : _ the @xmath0 inner product of two smooth functions is equal to the mimetic inner product of their interpolants when one of the functions ( @xmath46 in this case ) is a polynomial of a given degree and the other one belongs to a sufficiently rich space ( possibly infinite dimensional ) that must include polynomial functions . _ the exactness property implies that @xmath0 inner products of a large class of functions can be calculated exactly using the degrees of freedom . in particular , we write the consistency condition for the inner product on space @xmath14 as the exactness property : @xmath149_{\cf_{h , c } } = \int_c \k_c^{-1 } \bu\cdot \bv \dx \qquad\forall \bu\in ( \cp^0(c))^{d},\ \forall \bv \in { \cal sf}_c,\ ] ] where @xmath150 is a specially designed space containing the constant vector - functions : @xmath151 let us show that the right - hand side of ( [ consistency ] ) can be calculated using the degrees of freedom . let @xmath152 be a linear function such that @xmath153 . inserting @xmath48 in ( [ consistency ] ) , integrating by parts , and using the properties of @xmath150 , we obtain @xmath154 let @xmath155 be the inner product matrix and @xmath156 be the vector with components @xmath157 . then , combining the last formulas , we have @xmath158_{\cf_{h , c } } & = \big((\k_c\,\nabla q^1)_c^i\big)^t\ , \m_{\cf , c}\ , \bv_c^i = \sum\limits_{f\in \partial c } \bv\cdot\bn_f\sigma_{cf } \left ( \int_f q^1\dx - \frac{|f|}{|c|}\int_c q^1\dx \right ) \\ & = \big(\br_c(q^1)\big)^t \bv_c^i.\end{aligned}\ ] ] since @xmath86 is any function in @xmath150 , we can show that its interpolant @xmath159 is any vector in @xmath14 . indeed , it is sufficient to define a few functions @xmath86 as the solutions of cell - based pdes with different boundary conditions . hence , the consistency condition gives us the following matrix equations @xmath160 due to the linearity of these equations , it is sufficient to consider only three ( two in two dimensions ) linearly independent linear functions : @xmath161 , @xmath162 , and @xmath163 . let @xmath164 , \quad \r_c = [ \br_c(x)\ \br_c(y)\ \br_c(z ) ] \label{eq : n : r : general : definition}\end{aligned}\ ] ] be two rectangular @xmath165 matrices where @xmath166 is the number of faces in cell @xmath11 . the equations ( [ matrix - equations ] ) are now equivalent to the matrix equation : @xmath167 note that a symmetric positive definite solution @xmath155 ( if it exists ) is not unique even for a tetrahedral cell @xmath11 . the existence of solutions is based on the following result proved in @xcite . [ lem : ntr ] let matrices @xmath168 and @xmath169 be defined as in ( [ eq : n : r : general : definition ] ) . then , @xmath170 this lemma allows us to write the explicit formula for matrix @xmath155 : @xmath171 with a positive factor @xmath172 in front of the projection matrix @xmath173 . a recommended choice for @xmath172 is the mean trace of the first term . a family of mimetic schemes is obtained if we replace @xmath172 by an arbitrarily symmetric positive definite matrix @xmath174 : @xmath175 the stability of the resulting mimetic method depends on the spectral bounds of matrix @xmath174 which should be uniformly bounded by @xmath172 . algebraically , it means that there exists two generic constants @xmath123 and @xmath124 such that @xmath176 holds for every @xmath177 . this formula is called the _ stability condition _ in the mimetic discretization framework . the existence of the mesh independent constants @xmath123 and @xmath124 can be shown using assumptions ( * m1*)-(*m4 * ) . the _ third _ step of the construction of a mimetic scheme is to postulate either the local or global integration by parts formula and derive the gradient operator from it . for instance , given the global discrete operators @xmath18 and @xmath178 , we can write the mimetic scheme as follows : find @xmath133 and @xmath179 such that @xmath180 where @xmath181 . the local mimetic formulation requires us to define @xmath182_{\lambda_{h , c}}$ ] which satisfies condition ( [ assum1 ] ) . let @xmath183_{\lambda_{h , c } } = \sum_{f\in\partial c}\sigma_{c , f}\,|f|\,\lambda_f\,u^c_f.\ ] ] then , the local formulation is to find @xmath184 and @xmath185 in all mesh cells such that @xmath186 subject to the flux continuity conditions ( [ continuity ] ) and the homogeneous dirichlet boundary conditions @xmath187 for @xmath188 . the local derived operator has the explicit form : @xmath189 under assumption ( [ additivity ] ) the local and global mimetic formulations are equivalent . let @xmath85 be an arbitrary vector in @xmath13 and @xmath24 be its restriction to cell @xmath11 . to show that the solution to the local mimetic formulation is the solution to the global one , we first multiply both sides of the local constitutive equations by @xmath24 , then sum up the results over the mesh cells , cancel all internal face terms containing @xmath190 , and finally use the local duality relation ( [ eq : formula : b ] ) between @xmath79 and @xmath80 to obtain : @xmath191_{\cf_h , c } = -\sum\limits_c \big[\gradt_c \left(\!\!\begin{array}{c } p_c \\ { { \bm\lambda}}_c \end{array}\!\!\right),\ , \bv_c\big]_{\cf_h , c } = \sum\limits_c \big[\div_c\,\bv_c,\ , p_c\big]_{\cc_h , c}.\ ] ] from the additivity of the inner products and ( [ dualgrad ] ) we obtain that @xmath192_{\cf_h } = \big[\div\,\bv_h,\ , p_h\big]_{\cc_h } = -\big[\gradt\ , p_h,\ , \bv_h\big]_{\cf_h } \qquad\forall\bv_h\in\cf_h,\end{aligned}\ ] ] which implies that @xmath193 . to show the opposite statement , we repeat the above argument in the reverse order . this proves the assertion of the lemma . let us consider the matrix equation @xmath194 ( compare with ( [ mn = r ] ) ) . to implement the mimetic scheme in a computer program , we need to know only matrix @xmath195 . the general solution to the matrix equation is @xmath196 where @xmath197 is an arbitrary @xmath198 matrix , possibly non - symmetric . this formula leads to a large family of stable and unstable schemes that we refer to as the _ extended mixed - hybrid family _ of schemes . positive definiteness of @xmath199 is necessary for proving the scheme s convergence following the path described in @xcite . if in addition @xmath199 is symmetric , then it is one of the matrices @xmath200 . the classical two - point flux fv scheme is obtained when all matrices @xmath199 are positive definite and diagonal . let @xmath201 have a lipschitz continuous boundary . furthermore , let every cell @xmath11 be shape regular as explained in sec . [ sec : mesh ] . we assume that @xmath111 is the centroid of cell @xmath11 . we use the triple - bar notation , e.g. , @xmath202 , for the norms induced by the mimetic inner products . then , the interpolants of the exact solution , @xmath203 and @xmath204 , satisfy @xcite @xmath205 where @xmath51 and @xmath53 are solutions of the global mimetic formulation . if in addition @xmath201 is convex and @xmath123 in ( [ eq : stab : cond ] ) is sufficiently large , then @xmath206 two examples of fv schemes that fit within the mimetic framework are the mixed finite volume ( mfv ) method @xcite and the hybrid finite volume ( hfv ) method @xcite . both methods give a family of schemes because a stabilization term , which can be suitably parameterized , appears in their formulation . their design principles are based on conditions that imply the mimetic duality principle . both methods use the same degrees of freedom for pressure and flux as the local mimetic formulation and the same discrete flux balance equation , i.e. , @xmath207 . recall that @xmath25 collects the flux unknowns associated with cell @xmath11 . the local pressure unknowns are the cell pressure @xmath27 , which is associated with the interior of @xmath11 , and the interface pressure @xmath130 , which is associated with face @xmath108 . in the formulation of these fv methods , @xmath27 can be associated with the pressure value at any point inside the cell . as in the previous subsection , @xmath208 is the vector whose size is equal to the number of faces in cell @xmath11 . in the hfv method , a discrete gradient in cell @xmath11 is defined by applying the mid - point quadrature rule to the divergence theorem : @xmath209 this formula provides the exact value of the gradient whenever @xmath46 is a linear function since for any constant vector @xmath210 and any position vector @xmath111 , we have the geometric identity @xmath211 where @xmath115 is the face centroid . formula ( [ grad - hfv ] ) is used to define the numerical scheme after an additional _ stabilization term _ @xmath212 is included in the definition of the numerical flux @xmath213 : @xmath214 like in the mimetic framework , the stabilization term is designed very carefully to preserve the consistency of the scheme . let @xmath215 be the diagonal matrix with entries @xmath216 on the main diagonal , @xmath217 . similarly , let @xmath218 be the diagonal matrix with entries @xmath141 . furthermore , let @xmath219 , and @xmath220 note that the last expression is zero on linear pressure functions . the numerical flux is defined implicitly as the solution of @xmath221 for any @xmath222 and @xmath223 . here @xmath224 is a positive parameter and @xmath225 is the distance between @xmath111 and the plane containing face @xmath108 . after a few algebraic manipulations , we obtain : @xmath226\end{aligned}\ ] ] where @xmath227 is the kronecker symbol . it is obvious that the stabilization term is zero when it is calculated using the degrees of freedom of a linear function . the right - hand side of ( [ eq : fv : stab : term : def ] ) is a symmetric bilinear form with respect to the pressure unknowns . it is positive definite when @xmath228 and @xmath187 on @xmath188 . it is uniformly bounded from below when @xmath224 are sufficiently large and the mesh satisfies the regularity conditions described in section [ sec : mesh ] . equation ( [ eq : fv : stab : term : def ] ) is the key design principle . when @xmath27 is associated with the cell centroid , the left - hand side of this equation coincides with the left - hand side of ( [ eq : formula : b ] ) where the divergence is given by ( [ primary - div ] ) and the interface term is defined by ( [ interface ] ) . equation ( [ eq : fv : stab : term : def ] ) is also a representation of the duality principle . formula ( [ grad - hfv ] ) can be rewritten using the mimetic matrix @xmath168 as follows : @xmath229 this formula should be compared with the formula for the discrete gradient in the mfv method : @xmath230 this formula is exact for a linear pressure function and constant flux , but it also has to be stabilized . the vector of numerical fluxes @xmath25 is defined as the solution of @xmath231 where @xmath174 is a symmetric positive definite matrix . this design principle shows even clear connection with the mimetic duality principle ( [ eq : formula : b ] ) . since @xmath232 is invertible , we have that @xmath25 is a linear combination of @xmath27 and @xmath73 , i.e. condition ( [ model ] ) . we can summarize the above discussions in the following lemma . [ lem : equivalence ] any linearity preserving mixed - hybrid scheme of type with the degrees of freedom given by @xmath125 , @xmath71 , and @xmath13 is a member of the extended mixed - hybrid family of schemes . in addition , let @xmath233 and assume that the bilinear form @xmath234 is symmetric , uniformly coercive and uniformly bounded ( with respect to some norm that may depend on the problem ) . then , the resulting scheme belongs to the mimetic family of schemes . the most general form of the constitutive equation in ( [ model ] ) is given by the linear relationship @xmath235 with a rectangular @xmath236 matrix @xmath237 $ ] . since the scheme is exact for constant pressure functions , we have @xmath238 multiplying the last equation by @xmath27 and subtracting from the previous one , we obtain @xmath239 by our assumption , this formula is exact for all linear pressure functions and the corresponding constant flux functions . let us take linearly independent pressure functions @xmath240 , @xmath241 , @xmath242 and use the matrix notations introduced above to derive the following consequence of the linearity preservation property : @xmath243 hence , @xmath244 and the resulting scheme belongs to the extended mixed - hybrid family of schemes . now @xmath245 our symmetry and coercivity assumptions imply that @xmath199 is symmetric and positive definite ; hence , it is invertible and @xmath246 coincides with one of the mimetic inner product matrices @xmath155 . let @xmath247 . then , @xmath248 the uniform coercivity and boundness conditions imply that matrix @xmath246 satisfies the mimetic stability condition . this proves the assertion of the lemma . a detailed comparison of mfd , hfv and mfv methods is performed in @xcite . in particular , the authors have shown that all schemes can be generalized to provide the identical families of numerical schemes . let us consider the family of mimetic schemes , where the mass matrix is given in the form of equation . when the mesh is formed by simplexes , e.g. , triangles in 2d and tetrahedra in 3d , there is a choice of the stabilization matrix @xmath249 that provides the mass matrix from the lowest order raviart - thomas space @xcite . for a simplex , the stabilization matrix is a one - rank matrix , @xmath250\,\bp_c$ ] , where @xmath251 and vector @xmath252 , @xmath253 . the vem was originally introduced as an evolution of the mfd method . in the classical finite element spirit , the duality principle is now incorporated in the weak formulation : _ find @xmath254 and @xmath255 such that _ @xmath256 ( \div_h\,\bu_h , q_h)_{l^2(\omega ) } & = ( b , q_h)_{l^2(\omega)}\phantom{0 } \quad\forall q_h \in \widetilde \cc_h,\label{eq : vem:2 } \end{aligned}\ ] ] where @xmath257 is the space of piecewise constant functions isometric to @xmath125 , @xmath258 is a virtual space built from the local virtual spaces @xmath259 , @xmath260 @xmath261 and @xmath262 are the consistency and stability terms ( to be defined later in the section ) , and operator @xmath263 is defined cell - by - cell as the local @xmath0-orthogonal projection of the continuum divergence operator onto @xmath257 . the virtual space @xmath259 includes polynomial as well as non - polynomial functions . a new design principle is given by the unisolvency property . the virtual space @xmath259 is build to be isomorphic to the mimetic space @xmath14 , i.e. , the space of degrees of freedom . the virtual space is defined as the subspace of @xmath150 : @xmath264 in order to describe the splitting of the bilinear form @xmath265 we need to introduce the problem - dependent @xmath0 projector @xmath266 : @xmath267 where @xmath268 . the projector is computable using only the degrees of freedom @xcite . we define the consistency term as @xmath269 and the stability term as @xmath270 to generate a family of schemes , we can replace the @xmath0 inner product with any other spectrally equivalent bilinear form . like in the mimetic framework , the consistency term is the exactness property that holds when the first argument in the bilinear form is the constant function : @xmath271 hence , the contibution of this term to the local mass matrix is like in the mimetic scheme , @xmath272 . the conventional fe hybridization procedure works for the vem as well . the first equation in the variational formulation is replaced by a set of cell - based equations : @xmath273 where @xmath274 , @xmath275 is the restriction of @xmath263 to cell @xmath11 , and @xmath276 is the lagrange multiplier . it also holds that @xmath277 and @xmath278 where @xmath130 is the constant value of the lagrange multiplier on face @xmath108 ( compare with ( [ interface ] ) ) . the continuity equation is algebraically equivalent to ( [ continuity ] ) . introducing vector @xmath279 and recalling that @xmath51 has the constant value @xmath27 over cell @xmath11 , the hybridized equation can be rewritten as follows : @xmath280 which also provides the relation @xmath281 . the vem is the linearity - preserving method . according to lemma [ lem : equivalence ] , the lowest - order mixed - hybrid virtual element scheme is a member of the mimetic family of schemes . we highlight the new developments that extend the value of the mimetic framework for various physical applications . the development of a high - order mimetic scheme follows the same three steps described above for the low - order mimetic schemes . in the first step , we select the degrees of freedom that are convenient for the definition of the primary divergence operator , still denoted by @xmath55 . with a slight abuse of notation we still use the symbols @xmath125 and @xmath13 for the discrete spaces of pressure and flux unknowns , respectively . the discrete space @xmath125 contains multiple pressure unknowns that can be associated with the solution moments up to order @xmath282 . the discrete space @xmath13 contains multiple flux unknowns both associated with the mesh cells and the mesh faces . the cell - based degrees of freedom represent moments of the flux up to order @xmath282 except for the zero - th order moment . the face - based degrees of freedom represent flux moments up to order @xmath283 , see fig . [ fig : dofs - ho ] . on a polygonal cell ; for each polynomial degree @xmath282 we show the flux degrees of freedom on the left and the scalar degrees of freedom on the right . the edge / face moments of the normal component of the flux are denoted by a vertical line ; the cell moments are denoted by a bullet . ] the discrete divergence operator @xmath284 is defined cell - wise from the commutation property : @xmath285 which is also a useful property for the error analysis . this is the new design principle that can be generalized to other mimetic operators . the right - hand side is computable using only the degrees of freedom of @xmath286 . let @xmath287 be a polynomial of order at most @xmath282 . then , the definition of the moment and integration by parts give @xmath288 in the second step , we define the mimetic inner products in spaces @xmath125 and @xmath13 as accurate approximations of the @xmath0 inner products of pressure and flux functions . the derivation is based on two high - order consistency conditions . since the local space @xmath15 is isomorphic to @xmath289 , the first consistency condition is the obvious generalization of : @xmath148_{\cc_{h , c } } = \int_c p\ , q \dx \qquad\forall p\in \cp^r(c),\ \forall q \in \cp^r(c).\ ] ] the consistency condition in space @xmath14 is defined as the following exactness property : @xmath290_{\cf_{h , c } } = \int_c \k_c^{-1 } \bu\cdot \bv \dx \qquad\forall \bu\in \k_c\,\nabla\cp^{r+2}(c),\ \forall \bv \in { \cal sf}_c,\ ] ] where @xmath150 is a specially designed space containing the vector functions @xmath291 : @xmath292 it is easy to show that the right - hand side of the consistency condition is computable using the degrees of freedom introduced above . let @xmath293 be such that @xmath294 . then , @xmath295 as @xmath86 is in @xmath150 , the arguments of all the integrals in the right - hand side above are polynomials . using the degrees of freedom of @xmath86 it is possible to reconstruct @xmath296 inside @xmath11 and @xmath297 on each @xmath298 , see @xcite , and all these integrals are computable . combining the last formulas , we obtain the algebraic form of the consistency condition : @xmath299_{\cf_{h , c } } = \big((\k_c \nabla q)^i\big)^t\ , \m_{\cf , c}\ , \bv_c^i = \big(\br_c(q)\big)^t\bv_c^i.\ ] ] to find a symmetric positive definite matrix @xmath155 , we need the analog of lemma [ lem : ntr ] , which obviously holds since the function @xmath300 is in the space @xmath150 for any polynomial @xmath301 . in the third step , we formulate the duality formula for the derived gradient operator and use it in the global mimetic formulation . the following error estimates have been shown in @xcite : @xmath302 in the case when the diffusion tensor @xmath303 is no longer constant , the consistency condition has to be modified . let @xmath304 denote the local @xmath0 projector on the space of polynomial functions of order @xmath282 . then , the modified consistency condition reads : @xmath305_{\cf_{h , c } } = \int_c \nabla q \cdot \bv \dx \qquad\forall q\in \cp^{r+2}(c),\ \forall \bv \in { \cal sf}_c.\ ] ] after that , the inner product matrix @xmath155 is derived following the same steps . the consistency term in the formula for matrix @xmath155 ( see ( [ consistency ] ) ) contains the inverse of the diffusion tensor . therefore , numerical difficulties may arise in solving nonlinear parabolic problems of type @xmath306 where function @xmath307 can not be uniformly bounded from below . for instance , on a uniform one dimensional mesh , the numerical flux at mesh point @xmath308 is proportional to the difference of the neighboring pressures and the transmissibility coefficient @xmath309 : @xmath310 if @xmath311 , the numerical flux goes to zero as @xmath312 and may lead to a nonphysical solution as shown by the numerical experiment considered in section [ eq : marshak : heat : equation ] ( see also @xcite ) . to obtain an accurate solution , we have to replace the harmonic average with the arithmetic average in the definition of the transmissibility coefficient . a possible strategy in the mixed finite element framework ( see , e.g. @xcite ) consists in using two velocity variables , @xmath313 and @xmath314 . however , the corresponding weak formulation can not have face - based equations of type @xmath315 which are natural from a physical viewpoint . here @xmath316 is the face - based diffusion coefficient . such face - based equations lead to an algebraic problem that can be symmetrized only on special meshes . we describe a new mimetic scheme that allows us to use different values @xmath316 on different mesh faces . the mimetic framework always guarantees the symmetry of the resulting algebraic problem . let us consider a more general form of the diffusion coefficient , @xmath317 , where @xmath303 is a discontinuous tensor independent of @xmath46 and @xmath318 is a discontinuous scalar function of @xmath46 . the underlying mixed formulation is @xmath319 \frac{\partial p}{\partial t } + ( \div\,\k ) \ , \bu & = & b. \end{array}\ ] ] the combined operators @xmath320 and @xmath321 are dual to each other with respect to the weighted @xmath0 inner products by using @xmath322 as weight : @xmath323 consider again the three - step construction of the mimetic framework . in the _ first _ step we need to specify the degrees of freedom . for the pressure variable , we consider the same discrete space @xmath125 of grid functions that consist of one value per cell and the same discrete space @xmath71 of grid functions that consist of one value per face . the discrete space @xmath324 has the same dimension as in the linear case , but the discrete fluxes in @xmath324 obey a different continuity condition : @xmath325 on each interior face @xmath108 shared by cells @xmath101 and @xmath102 . here , @xmath326 and @xmath327 are accurate one - side approximations of the diffusion coefficient @xmath4 . for example , in regions where function @xmath328 is continuous , we can take @xmath329 as a weighted average of the cell - centered values @xmath330 and @xmath331 calculated using the most recent approximation to solution @xmath46 in the cells @xmath101 and @xmath102 , respectively . the weights are the distances between the cell centers and face @xmath108 . the primary mimetic operator approximates the combined operator @xmath320 . it is defined locally on each mesh cell using a straightforward discretization of the divergence theorem ( compare with formula ( [ primary - div ] ) ) : @xmath332 since @xmath53 is an algebraic vector , it is convenient to think about the discrete divergence operator @xmath333 as a matrix acting between two spaces . this matrix has full rank when @xmath334 . let us introduce a cell - based diagonal matrix @xmath335 formed by coefficients @xmath336 , @xmath217 . then , the primary mimetic operators in ( [ primary - div ] ) and ( [ primary - divk ] ) can be connected as follows : @xmath337 the _ second _ step in the mimetic discretization framework is to define the inner products in spaces @xmath125 and @xmath13 that are accurate approximations of the integrals in ( [ green : duality ] ) . such inner products can be defined again cell - by - cell . moreover , the inner product in space @xmath125 can be defined as in section [ subsec : mimetic : discretization : framework ] by the relation @xmath338_{\cc_{h , c } } = |c|\ , p_c\ , q_c$ ] . the weight in the other @xmath0 inner product is given by @xmath339 . by our assumption , @xmath303 is a piecewise constant tensor on mesh @xmath5 . instead , the scalar coefficient @xmath4 can be a quite general non - negative function . an acceptable first - order error is committed when we replace @xmath4 by the piecewise constant function with value @xmath340 in cell @xmath11 . the consistency condition is expressed through the exactness property : @xmath290_{\cf_{h , c } } = \int_c \k(p_c)\ , \k_c^{-1 } \bu\cdot \bv \dx \qquad\forall \bu\in ( \cp^0(c))^{d},\ \forall \bv \in { \cal sf}_c.\ ] ] since @xmath341 is a constant tensor in cell @xmath11 , the derivation of the inner product matrix ( still denoted by @xmath155 ) proceeds as in section [ subsec : mimetic : discretization : framework ] . the _ third _ step in the construction of the mimetic framework is to obtain the formula for the derived operator , which , in this case , is an approximation of the combined operator @xmath321 . the continuum green formula for cell @xmath11 is given by @xmath342 now , the derived operator @xmath343 satisfies the discrete integration by parts formula @xmath344_{\cc_{h , c } } - \sum\limits_{f\in\p c}\sigma_{c , f}\,|f|\,k_f^c\ , u^c_f \lambda_f = -\big[\bu_c,\ , \gradt_c\ , \big(\!\!\begin{array}{c}q_c \\ { { \bm\lambda}}_c \end{array}\!\!\big ) \big]_{\cf_{h , c}}\end{aligned}\ ] ] for all @xmath345 , @xmath346 , and @xmath347 . this local mimetic formulation gives the following formula for the physical fluxes : @xmath348 note that this formula uses the symmetric matrix @xmath349 . our numerical experiments show that the resulting scheme is second - order accurate . let us consider the modified marshak heat equation @xcite in the rectangular domain @xmath350 with zero source term , @xmath351 , and @xmath352 . the initial value is @xmath353 . we set the time - dependent dirichlet boundary condition @xmath354 on the left side of @xmath201 , the constant boundary condition @xmath355 on the right side , and the homogeneous neumann boundary conditions on the remaining sides . we solve the parabolic equation on the randomly perturbed quadrilateral mesh that have three times more cells in the x - direction than in the y - direction . we use the backward euler time integration scheme and the weighted arithmetic average definition of the face - based diffusion coefficients @xmath356 described above . to reduce the impact of the time integration error , we use small time steps . comparison of pictures in fig . [ fig : marshak - new ] show that the new scheme fixes the deficiencies of the old mfd scheme and leads to the correct speed of propagation of the non - linear wave . for the standard ( left ) and new ( right ) mfd schemes . the solution in the right panel shows the correct and accurate position of the wave front at the chosen time . ] we described a few design principles used in the derivation of mimetic schemes for the numerical solution of pdes . we established the bridges with a few fv and fe methods by showing that three popular discretization frameworks ( mfd , fv and fe ) use the equivalent design principles which leads to algebraically equivalent schemes . we illustrated the flexibility of the mimetic discretization framework to tackle challenging numerical issues in computer modeling of engineering problems with two examples : derivation of higher - order schemes and convergent schemes for nonlinear problems with small diffusion coefficients . this work was carried out under the auspices of the national nuclear security administration of the u.s . department of energy at los alamos national laboratory under contract no . de - ac52 - 06na25396 . the authors acknowledge the support of the us department of energy office of science advanced scientific computing research ( ascr ) program in applied mathematics research . p. bochev and j. m. hyman . principle of mimetic discretizations of differential operators . in d. arnold , p. bochev , r. lehoucq , r. nicolaides , and m. shashkov , editors , _ compatible discretizations . proceedings of i m a hot topics workshop on compatible discretizations _ , i m a volume 142 . springer - verlag , 2006 . r. eymard , t. gallout , and r. herbin . discretization of heterogeneous and anisotropic diffusion problems on general nonconforming meshes . sushi : a scheme using stabilization and hybrid interfaces . , 30(4):10091043 , 2010 . v. gyrya , k. lipnikov , and g. manzini . the arbitrary order mixed mimetic finite difference method for the diffusion equation . technical report la - ur-15 - 00000 , los alamos national laboratory , 2015 . submitted to mathematical modelling and numerical analysis . a. palha , p. p. rebelo , r. hiemstra , j. kreeft , and m. gerritsma . physics - compatible discretization techniques on single and dual grids , with application to the poisson equation of volume forms . , 257 , part b(0):13941422 , 2014 . p. raviart and j. thomas . a mixed finite element method for second order elliptic problems . in i. galligani and e. magenes , editors , _ math . aspects of finite element method , lecture notes in math . 606_. springer - verlag , n.y . , 1977 .

the linear stability of rapid granular flow on a slope under gravity against the longitudinal perturbation is analyzed using hydrodynamic equations . it is demonstrated that the steady flow uniform along the flow direction becomes unstable against the long - wavelength perturbations longitudinal to the flow direction for certain parameter ranges to form the density wave , in contrast with the finite wavelength instability against the transverse perturbation ( ( * ? ? ? * forterre & pouliquen 2002 ) ) . it is shown that the instability can be understood as the the long - wave instability of the kinematic waves in a quasi - one dimensional system . the results are compared with our previous molecular dynamics simulations ( ( * ? ? ? * mitarai & nakanishi 2001 ) ) , where the spontaneous density wave formation has been found .

cond-mat0303543

granular flow exhibits a variety of dynamical phenomena , which have been attracting research interests for years ( for reviews , see e.g. ( * ? ? ? * savage 1984 ) and ( * ? ? ? * jaeger , nagel & behringer 1996 ) ) . its complex behaviors can be seen even in a simple situation like the gravitational flow on a slope . when the inclination angle is large and the slope is rough , a rapid and relatively low - density flow is realized , and the interaction between grains is dominated by inelastic collisions . on the other hand , when the inclination angle is small , the flow becomes dense and slow , and the frictional interaction plays an important role ( ( * ? ? ? * savage 1984 ) ; ( * ? ? ? * mitarai & nakanishi 2003 ) ) . the comprehensive rheology of the granular flow has not been established yet , except for the rapid collisional flow regime , where the hydrodynamic models have been developed with the constitutive relations based on the kinetic theory of inelastic hard spheres ( ( * ? ? ? * jenkins & savage 1983 ) ; ( * ? ? ? * campbell 1990 ) ; ( * ? ? ? * lun 1984 ) ; ( * ? ? ? * goldhirsh 2003 ) ) ; it has been demonstrated that a certain quantitative agreement can be achieved for the steady flow by introducing the spinning motion of each grains ( ( * ? ? ? * mitarai , hayakawa & nakanishi 2002 ) ) . the steady granular flow , however , turns out to be unstable in various ways , and shows rich phenomena . in the experiment on a shallow granular flow on a wide slope , @xcite have observed that there appears a regular pattern of longitudinal streaks along the flowing direction . this phenomenon has been analyzed by means of the hydrodynamic equations for rapid granular flow ( ( * ? ? ? * forterre & pouliquen 2002 ) ) . they have calculated the steady solutions and examined its linear stability numerically . they have found that , at a certain parameter region , the steady flow shows the `` inverted density profiles '' , in which the maximum density appears not at the bottom but at a finite distance from the bottom because of the agitation by the collisions with the rough solid bottom . it has been shown that the solutions with the `` inverted density profiles '' are unstable against the perturbations transverse to the flowing direction , and the instability results in the vortex patterns analogous to the rolls in the rayleigh - brnard convection ; the streaks found in the experiment were interpreted as the result of the rolls of vortices . another instability that has been observed is the density wave formation along the flowing direction ; experimentally , @xcite have observed the jamming patters traveling upstream in the dense chute flow , and ( * ? ? ? * prasad , pal & rmkens ( 2000 ) ) have found the waves develops in the shallow flow as they travel downstream . the present authors have performed the molecular dynamics simulations of two - dimensional granular flow on a slope and found that the steady flows are unstable against the density wave formation when the length of the slope is long enough and/or the particle density is low enough ( ( * ? ? ? * mitarai & nakanishi 2001 ) ) . the purpose of this paper is to perform the linear stability analysis on the hydrodynamic equations to investigate the nature of the density wave formation instability found in the experiments and the numerical simulations . the basic method is the same with that used in @xcite , but we examine the stability against the perturbations longitudinal to the flowing direction in contrast with the case of @xcite , where the transverse stability has been studied . this paper is organized as follows . in [ hydromodel ] , the hydrodynamic model for rapid granular flow is introduced . the steady solutions uniform along the flow direction are numerically obtained in [ secsteady ] , and the results of the linear stability analysis is presented in [ lstaba ] . in [ discussion ] , discussions and the comparison with the molecular dynamics simulations are given . the results are summarized in [ sum ] . the hydrodynamic fields for granular flows in three dimensions are the mass density @xmath0 , the mean velocity @xmath1 , and the granular temperature @xmath2 , where @xmath3 . here , @xmath4 and @xmath5 represents the average over the microscopic scale . under gravity , they follow @xmath6 with the acceleration of gravity @xmath7 , the stress tensor @xmath8 , the heat flux @xmath9 , and the energy loss @xmath10 due to the inelastic nature of interactions between grains . we employ the constitutive relations derived by ( * ? ? * lun ( 1984 ) ) for three dimensional system based on the kinetic theory of the inelastic particles : derived by ( * ? ? ? * lun ( 1984 ) ) is @xmath11 . the coefficient @xmath12 is proportional to @xmath13 , thus disappears in the elastic limit . we checked that the influence of the term @xmath14 on the steady solutions is small in the considered parameter region , therefore neglected this term as @xcite . ] @xmath15 where @xmath16-\frac{1}{3 } ( \nabla\cdot\boldsymbol u)i,\ ] ] and @xmath17 with the material density of particle @xmath18 , the packing fraction @xmath19 , the particle diameter @xmath20 , and the restitution coefficient between particles @xmath21 . here , @xmath22 represents the unit matrix . the dimensionless functions @xmath23 ( @xmath24 ) are given in table [ table ] . for the radial distribution function @xmath25 in these functions , we adopted the form suggested by @xcite : @xmath26 with the maximum solid fraction @xmath27 , for which we use @xmath28 as @xcite . in the following , all the variables are non - dimensionalized by the length unit @xmath20 , the mass unit @xmath29 , and the time unit @xmath30 . the density field is expressed by the packing fraction @xmath31 instead of the mass density @xmath0 . the restitution coefficient between particles is set to be @xmath32 , that is the value used in our previous simulations ( ( * ? ? ? * mitarai & nakanishi 2001 ) ) . l @xmath33 + + @xmath34 + + @xmath35 + + @xmath36 + + @xmath37 + + @xmath38 + + @xmath39 + + the granular flow has a non - zero slip velocity at the solid boundary , where we should impose the momentum and the kinetic energy balances . sophisticated boundary conditions have been proposed based on microscopic calculations of the kinetic theory for specific geometries ( ( * ? ? ? * jenkins & richman 1986 ) ; ( * ? ? ? * richman 1988 ) ; ( * ? ? ? * jenkins 1992 ) ) . we employ , however , a simpler form of the boundary condition obtained from a heuristic approach ( ( * ? ? ? * johnson & jackson 1987 ) ; ( * ? ? ? * johnson , nott & jackson 1987 ) ) ; @xmath40 where the unit vector @xmath41 is normal to the floor , @xmath42 is the slip velocity at the floor , and @xmath43 is the unit vector along the slip velocity . the first equation ( [ eq : bc1 ] ) expresses that the stress at the boundary balances with the momentum transfer due to the collisions between the slope and the flowing grains . the momentum transfer , or rhs of ( [ eq : bc1 ] ) , is assumed to be given by @xmath44 with the collision rate @xmath45 per unit time per unit area . here , the factor @xmath46 comes from the non - dimensionalization of the particle mass @xmath47 . the parameter @xmath48 characterizes the roughness of the boundary , and the expression means that the fraction @xmath48 of the particle momentum is transfered to the boundary in each collision , therefore , the larger value of @xmath48 represents the rougher boundary . for the rough boundary in the two - dimensional system with closed packed disks , @xcite estimated as @xmath49 , but @xcite adopted the smaller value @xmath50 for most of the cases because they expected that a boundary with closed packed spheres in three - dimensional system is smoother on average . in this paper , we mainly use @xmath50 , but the case of @xmath51 is also examined in order to see general tendencies . the second equation ( [ eq : bc2 ] ) represents the energy balance , and means that the heat flux at the boundary comes from two effects , namely , the frictional heating due to the non - zero slip velocity and the energy loss due to the inelastic collision with the floor . the energy loss term @xmath52 in ( [ eq : bc2 ] ) is given by @xmath53 with the collision rate @xmath45 . the parameter @xmath54 represents the rate of energy loss per collision to the restitution coefficient between the floor and the particles @xmath55 in the form @xmath56 , but we adopt ( [ eq : gammas ] ) as a more general expression . ] , and we use @xmath57 in this paper . for the collision rate @xmath45 , we use the form adopted by @xcite , i.e. , @xmath58 : then the expression of @xmath59 and @xmath60 in non - dimensionalized form are @xmath61 where the dimensionless functions @xmath62 and @xmath63 are given in table [ table ] . at infinity , we impose the condition that the stress and the heat flux vanish , namely , @xmath64 where the @xmath65 axis is taken perpendicular to the floor ( figure [ slope ] ) . first , we consider the steady solution which is uniform along the slope for ( [ eq : rho ] ) - ( [ eq : temp ] ) with the boundary conditions ( [ eq : bc1 ] ) , ( [ eq : bc2 ] ) , and ( [ eq : inftycond ] ) in the form @xmath66 where we take the @xmath67 axis along the slope , the @xmath65 axis perpendicular to the floor , and the @xmath68 axis perpendicular to the @xmath69 plane ( figure [ slope ] ) . then , the equations ( [ eq : rho ] ) - ( [ eq : temp ] ) are written as @xmath70 where the superscript @xmath71 denotes that the functions are for the steady solution , namely , @xmath72 , etc . by integrating ( [ eq : steady1 ] ) and ( [ eq : steady2 ] ) over @xmath65 with the stress free condition at infinity , we obtain the condition @xmath73 from ( [ eq : steady1])-([eq : steady4 ] ) and the constitutive relations , we have @xmath74 , \label{eq : t00}\end{aligned}\ ] ] where @xmath75 , @xmath76 , and the prime indicates the derivative by its argument . the boundary conditions ( [ eq : bc1 ] ) and ( [ eq : bc2 ] ) at the floor ( @xmath77 ) for the steady solution can be written as @xmath78 the boundary condition ( [ eq : inftycond ] ) that the stress and the energy flux should vanish at infinity is satisfied when ( ( * ? ? ? * ahn , brennen & sabersky 1992 ) ) @xmath79 in order to obtain the steady solutions , we integrate ( [ eq : nu0 ] ) , ( [ eq : u00 ] ) , and ( [ eq : t00 ] ) numerically using the fourth - order runge - kutta method with the boundary conditions ( [ eq : t0b ] ) and ( [ eq : dt0b ] ) . we employ the shooting method to find the solution which satisfies the condition ( [ eq : bcinf ] ) ( ( * ? ? ? * forterre & pouliquen 2002 ) ) ; for a given inclination angle @xmath80 and a given density at the floor @xmath81 , we search for a solution by adjusting the value of the velocity at the floor @xmath82 . in the actual calculations , we integrate the equations numerically from @xmath77 to a certain height @xmath83 , and search for the solution which gives @xmath84 . the value of @xmath83 is chosen to be large enough in comparison with the relaxation length , that depends on the parameters and can be determined only after the solution is obtained . we use @xmath80 and @xmath81 to specify the solution in the rest of the paper . for a given roughness @xmath85 of the slope , steady solutions are found for a certain range of the inclination angle of the slope @xmath80 ( ( * ? ? ? * forterre & pouliquen 2002 ) ) . we present the steady solutions for the two cases , ( i ) @xmath85=0.05 and ( ii ) @xmath86 ; the most of the results are for the case of @xmath87 , and the case of @xmath86 will be given to examine general trend . in the @xmath88 plane for @xmath87 . the region of the non - monotonic density profiles is shown by a grey region . ( b)the flux @xmath89 vs the one - dimensional density @xmath90 for @xmath87 . the solid line and the dashed line are for @xmath91 and @xmath92 , respectively . , scaledwidth=80.0% ] for an appropriate @xmath80 , there exist steady solutions for a given density at the floor @xmath81 . we numerically find the steady solution for the range @xmath93 for the moderate density ; towards the lower limit of @xmath80 , the length scale of the density decay in the @xmath65-direction tends to zero and becomes smaller than the particle diameter , which is physically unacceptable , while the decay length of density diverges towards the upper limit of @xmath80 . this is consistent with the analysis by @xcite that the steady solution in the high - density limit is allowed for a finite range of @xmath80 . for a given steady solution , we define the `` thickness '' @xmath94 and the `` mean density '' @xmath95 : the thickness @xmath94 is the value of @xmath65 where the density is 1% of the maximum density , and the mean density @xmath96 is given by @xmath97 it has been found numerically by @xcite that there exists a one to one correspondence between @xmath98 and @xmath88 . figure [ steadyphase005](a ) shows the contour lines of @xmath80 in the @xmath88 plane , where @xmath80 increases from left to right . in this plot , it is clear that @xmath94 goes to zero as @xmath80 decreases and @xmath94 diverges as @xmath80 increases . there is a separatrix near the bottom at @xmath99 , where the value of @xmath80 is around @xmath100 ; the contour line for @xmath101 leads to @xmath102 as @xmath95 decreases , while @xmath94 diverges along the contour lines for @xmath103 as @xmath95 becomes small . the difference between @xmath104 and @xmath105 can be seen in the relation between the flux @xmath89 and the one - dimensional density @xmath90 defined as @xmath106 respectively , for a fixed inclination angle . it is found that @xmath89 is an increasing function of @xmath90 for @xmath107 , while @xmath89 has a minimum at a finite @xmath90 for @xmath108 ; the plot of @xmath89 vs @xmath109 is shown in figure [ steadyphase005](b ) for @xmath91 and @xmath92 . the typical profiles of the density , the velocity , and the temperature for @xmath92 are shown in figure [ steady ] for the density at the floor @xmath110 . we see in figure [ steady](a ) that the density decays monotonically when the density at the floor @xmath81 is small enough ( @xmath111 ) , while for higher density ( @xmath112 ) the maximum density appears at a finite height . the region where the maximum density appears at a finite height is shown in figure [ steadyphase005](a ) as a grey region . we focus on the lower density region because the density decays monotonically in our previous simulation ( ( * ? ? ? * ; * ? ? ? * mitarai & nakanishi 2001 ; mitarai 2002 ) ) . for @xmath113 , the higher density flow shows lower flowing speed in the case of @xmath92 , which results in the decrease of the flux @xmath89 as @xmath90 increase for @xmath114 ( figure [ steadyphase005](b ) ) . for higher density ( @xmath112 in figure [ steady ] ) , the flowing speed gets faster with @xmath90 , which causes the increase of the flux @xmath89 for higher @xmath90 . as a result , @xmath89 has a minimum at a finite density . in the case of @xmath107 , the velocity continuously decrease as the density becomes lower , and @xmath89 becomes a increasing function of @xmath90 . the slope is rougher than the previous case , and the steady solution exists for @xmath115 . the contour lines for @xmath80 in the @xmath88 plane are shown in figure [ steadyphase010](a ) . as in the case of @xmath87 , @xmath94 goes to zero as density becomes lower for smaller @xmath80 , while @xmath94 diverges for smaller density when @xmath80 is large enough . figure [ steadyphase010](b ) shows @xmath90 dependence of @xmath89 , which is a monotonically increasing function for @xmath116 and has a minimum for @xmath117 . the typical solutions are shown for @xmath118 in figure [ steady01 ] . for large enough @xmath94 and @xmath95 , the maximum density appears at a finite distance from the floor . the region of the non - monotonic density profile is shown by a grey region in figure [ steadyphase010](a ) . in the @xmath88 plane for @xmath86 . the region of non - monotonic density profiles is shown by a grey region . ( b)relation between @xmath89 vs @xmath90 for @xmath86 . the solid line and the dashed line are for @xmath118 and @xmath119 , respectively . , scaledwidth=80.0% ] we restrict our stability analysis to the perturbation uniform along the @xmath68 direction , because we are interested in the instability along the flow direction . the flow is perturbed around the steady solution as @xmath120 the governing equations and the boundary conditions are linearized with respect to the deviations @xmath121 , @xmath122 , @xmath123 , and @xmath124 , but the resulting expressions are rather lengthy and given in appendix a. now we look for the normal modes for the density , the velocity , and the temperature perturbations of the form @xmath125 . \label{eq : lin}\ ] ] the flow is linearly unstable if @xmath126 . as for the boundary condition at the free surface , the asymptotic behavior of the perturbations at large @xmath65 are used . when @xmath65 is much larger than the decay length of the density and thus @xmath127 is very small , the density perturbation should also decay ( @xmath128 ) , and @xmath129 , @xmath130 , and @xmath131 decay proportional to @xmath132 ( ( * ? ? ? * forterre & pouliquen 2002 ) ) , therefore , we imposed the boundary condition that @xmath133 at @xmath134 ; @xmath83 is a large enough height where @xmath135 , in addition to the condition @xmath136 discussed in [ secsteady ] . we solve the eigenvalue problems of the linearized equations ( [ eq : leqnu])-([eq : leqt ] ) for ( [ eq : lin ] ) numerically using the chebychev collocation method with the discretization in the @xmath65 direction ( ( * ? ? ? * gottilieb , hussaini & orsag 1984 ) ; ( * ? ? ? * canuto 1988 ) ; ( * ? ? ? * boyd 2001 ) ; ( * ? ? ? * foterre & pouliquen 2002 ) ) . it is known that the straightforward discretization of space requires two extra boundary conditions ( ( * ? ? ? * malik 1990 ) ; ( * ? ? ? * foterre & pouliquen 2002 ) ) , for which we use the momentum balance condition in the @xmath65 direction at @xmath77 and the decay condition for the density perturbation , i.e. , @xmath137 at @xmath134 . in actual numerics to solve the generalized eigenvalue problem in the form @xmath138 for the complex eigenvalues @xmath139 and the eigenvectors @xmath140 , we used lapack version 3.0 ( ( * ? ? ? * anderson 1999 ) ) . the number of discretization @xmath141 is about @xmath142 . the numerically obtained eigenmodes contains unphysical modes , called spurious modes , due to the discretization ( ( * ? ? ? * mayer & powell 1992 ) ; ( * ? ? ? * boyd 2001 ) ; ( * ? ? ? * forterre & pouliquen 2002 ) ) . for the spurious modes , it is known that the chebychev coefficients of higher wave number are large , and the eigenvalues are sensitive to small change of @xmath141 . we determine the eigenmodes as physical ones by checking that their chebychev coefficients for higher wave number are small and their eigenvalues varies little upon changing @xmath141 : we confirmed that , for these modes , the highest ten coefficients are less than @xmath143 when the eigenvectors are normalized so that the sum of the absolute values of the real part and the imaginary part of the largest component becomes one , and the variation of the eigenvalues by the small change of @xmath141 are less than @xmath143 . we present the results of the linear stability analysis in the cases of @xmath87 and @xmath86 , for which the steady solutions are shown in [ steadysol ] . vs @xmath95 for @xmath50 , where the unstable ( stable ) regimes are shown by grey ( white ) regions . we find three unstable regimes a , b , and c. the dashed line shows the boundary between the region of the non - monotonic density profiles and that of the monotonic density profiles ( see also figure [ steadyphase005 ] ) . the dispersion relations at the points p1@xmath144p5 are shown in figures [ disp ] @xmath144 [ disp - ac ] . the region of @xmath145 where the stability boundary is shown by a dotted line was not examined in detail because the flow with the decay length less than the particle diameter is physically unacceptable . ] the stability diagram is shown in figure [ phase005 ] in the parameter space of @xmath94 vs @xmath95 . the unstable ( stable ) regimes are shown by grey ( white ) regions , and the boundary between the region of the monotonic density profiles and that of the non - monotonic density profiles is shown by a dashed line . within the investigated region , we find three unstable regimes ; the regime a at small @xmath94 and small @xmath95 region , the regime b at large @xmath94 and small @xmath95 region , and the regime c at large @xmath94 and large @xmath95 region . when we decrease the density with a constant inclination angle ( along a contour in figure [ steadyphase005 ] ) , eventually we will encounter either the regime a or the regime b , namely , the flow with low enough density is always unstable . the two regimes are different in the steady flow behavior as we have seen already : in the regime a at the small @xmath94 side , the flow becomes slower as the density becomes smaller , while the flow in the regime b at the large @xmath94 side flows down faster for the smaller density . on the other hand , the denser flow can be unstable in the regime c , which lies within the region of the non - monotonic density profiles . in this regime , the denser flow goes faster as in the regime a. the dispersion relations of the least stable modes @xmath146 are shown in figure [ disp ] for @xmath147 and @xmath148 ( a ) , @xmath149 ( b ) , and @xmath150 ( c ) , which correspond to the points p1 , p2 , and p3 , respectively , in the stability diagram of figure [ phase005 ] ; p1 lies within the unstable regime b. in all cases , it is found that the least stable mode satisfies @xmath151 . the growth rate @xmath152 is positive for @xmath148 for @xmath153 with @xmath154 ( figure [ disp](a ) ) . the magnification around @xmath155 in figure [ dispmag ] shows that @xmath152 grows quadratically in @xmath156 for small @xmath156 . as @xmath81 is increased , the maximum value of @xmath157 $ ] decreases and becomes negative for all @xmath156 ( figure [ disp](b ) and ( c ) ) . the dispersion relations for the unstable modes at p4 ( in the regime a ) and at p5 ( in the regime c ) are shown in figure [ disp - ac ] for @xmath158 ( a ) and @xmath159 ( b ) , respectively . the instability occurs for the long - wavelength perturbation , and both of the dispersion relations show that @xmath151 and the growth rate quadratic in @xmath156 for small @xmath156 , i.e. , @xmath160\propto k^2 $ ] ; these features are the same as those in the unstable regime b. , @xmath92 , and @xmath161 ( p3 in figure [ phase005 ] ) , whose wave number is @xmath162 . @xmath163 is the wavelength of this eigenmode , @xmath164 . the contour of the density and the temperature eigenfunctions are shown in ( a ) and ( b ) , respectively , by grey scale , where the brighter ( darker ) region corresponds to the larger positive ( negative ) value . the arrows in ( a ) indicate the corresponding velocity eigenfunction . it shows that the grains flow from the brighter region into the darker density region , namely , the density perturbation decays . note that the temperature perturbation is negative at the region of the positive density perturbation . , title="fig : " ] , @xmath92 , and @xmath161 ( p3 in figure [ phase005 ] ) , whose wave number is @xmath162 . @xmath163 is the wavelength of this eigenmode , @xmath164 . the contour of the density and the temperature eigenfunctions are shown in ( a ) and ( b ) , respectively , by grey scale , where the brighter ( darker ) region corresponds to the larger positive ( negative ) value . the arrows in ( a ) indicate the corresponding velocity eigenfunction . it shows that the grains flow from the brighter region into the darker density region , namely , the density perturbation decays . note that the temperature perturbation is negative at the region of the positive density perturbation . , title="fig : " ] , @xmath92 , and @xmath148 ( p1 in figure [ phase005 ] ) , whose wave number is @xmath162 . the contour of the density eigenfunction and the velocity eigenfunction in ( a ) show that the grains flow from the region of negative density perturbation into that of positive perturbation , namely , the perturbation is amplified and results in the formation of the density wave . the temperature perturbation in ( b ) is negative at the region of positive density perturbation . , title="fig : " ] , @xmath92 , and @xmath148 ( p1 in figure [ phase005 ] ) , whose wave number is @xmath162 . the contour of the density eigenfunction and the velocity eigenfunction in ( a ) show that the grains flow from the region of negative density perturbation into that of positive perturbation , namely , the perturbation is amplified and results in the formation of the density wave . the temperature perturbation in ( b ) is negative at the region of positive density perturbation . , title="fig : " ] the least stable eigenmodes for @xmath92 at @xmath162 are shown for the two cases : the stable case of @xmath161 ( p3 in figure [ phase005 ] ) in figure [ eigens ] and the unstable case of @xmath148 in the regime b ( p1 in figure [ phase005 ] ) in figure [ eigenu ] , over one wavelength @xmath164 . the contours in figures [ eigens](a ) and [ eigenu](a ) show the density eigenfunctions ; the brighter ( darker ) regions indicate the positive ( negative ) regions , and the arrows represent the corresponding velocity eigenfunctions . in the both figures near @xmath77 , we see that the velocity perturbations point to the positive ( negative ) @xmath67 direction in the regions of negative ( positive ) density perturbation . the contours for the corresponding temperature eigenfunctions plotted in figures [ eigens](b ) and [ eigenu](b ) show that the regions where the density perturbation is negative ( darker regions in figures [ eigens](a ) and [ eigenu](a ) ) roughly correspond to the positive temperature perturbations ( brighter regions in figures [ eigens](b ) and [ eigenu](b ) ) . the difference between the stable mode ( figure [ eigens ] ) and the unstable mode ( figure [ eigenu ] ) is seen if we focus on the divergence of velocity perturbation . in the case of the stable mode for @xmath161 ( figure [ eigens](a ) ) , the grains flow into the region where the density perturbation is negative ( see the region around @xmath165 and @xmath166 ) , thus the density perturbation has negative feedback . on the other hand , in the case of unstable mode for @xmath148 ( figure [ eigenu](a ) ) , the grains flow into the region where the density perturbation is positive ( see the region around @xmath167 and @xmath168 ) ; as a result , the perturbation grows and eventually causes the nonlinear density wave . the eigenfunctions of the unstable modes at p4 in the regime a and at p5 in the regime c show some different characteristics from those at p1 in the regime b. reflecting the difference in the steady flow between the regimes a and c and the regime b , the denser parts of the density eigenfunctions roughly correspond to the part where the velocity fluctuation has the positive component in the direction parallel to the mean flow . for all three regimes , however , the unstable modes show that the grains flow into the region of the positive density perturbation ; this suggests that the instability leads to the density wave in the regimes a and c as in the regime b. vs @xmath95 with @xmath51 , where the unstable ( stable ) regimes are shown by grey ( white ) regions . we find three unstable regimes a , b , and c. the dashed line shows the boundary between the region of the non - monotonic density profiles and that of the monotonic density profiles ( see also figure [ steadyphase010 ] ) . the dispersion relations at the points q1@xmath144q3 are shown in figure [ disp01 ] . ] the stability diagram for @xmath86 is shown in figure [ phase010 ] . as in the case of @xmath87 , there are three unstable regimes , but the qualitative difference is that the unstable regime c at the large @xmath94 and large @xmath95 region contains a part of the region of the monotonic density profiles as well as that of the non - monotonic density profiles . the dispersion relations around the boundary of the regime c are shown in figure [ disp01 ] for @xmath118 and @xmath161(a ) , @xmath169(b ) and @xmath170(c ) , which correspond to the points q1 , q2 , and q3 , respectively , in the stability diagram [ phase010 ] . it is seen that the instability occurs against the long - wavelength perturbation . the unstable eigenmodes for @xmath118 and @xmath171 ( q3 in figure [ phase010 ] ) with @xmath162 are shown in figure [ eigenu01 ] . the density and the velocity eigenfunctions indicate that the grains flow into the the region of positive density perturbation ( see the region around @xmath165 , @xmath172 ; the magnification is shown ( c ) ) , thus the density perturbation will grow . the difference from the case in figure [ eigenu ] is that the velocity perturbation at the floor ( @xmath77 ) is in the positive @xmath67 direction in the region of the positive density perturbation ; namely , the particles flow faster in the region where they get dense . , @xmath118 , and @xmath171 ( q3 in figure [ phase010 ] ) , whose wave number is @xmath162 . the contour of the density eigenfunction and the velocity eigenfunction in ( a ) show that the grains flow into the the region of the positive density perturbation . in ( b ) , the contour of the temperature perturbation is shown . ( c ) is the magnification of ( a ) around @xmath173 , @xmath174 . , title="fig:",scaledwidth=52.2% ] , @xmath118 , and @xmath171 ( q3 in figure [ phase010 ] ) , whose wave number is @xmath162 . the contour of the density eigenfunction and the velocity eigenfunction in ( a ) show that the grains flow into the the region of the positive density perturbation . in ( b ) , the contour of the temperature perturbation is shown . ( c ) is the magnification of ( a ) around @xmath173 , @xmath174 . , title="fig:",scaledwidth=47.0% ] , @xmath118 , and @xmath171 ( q3 in figure [ phase010 ] ) , whose wave number is @xmath162 . the contour of the density eigenfunction and the velocity eigenfunction in ( a ) show that the grains flow into the the region of the positive density perturbation . in ( b ) , the contour of the temperature perturbation is shown . ( c ) is the magnification of ( a ) around @xmath173 , @xmath174 . , title="fig:",scaledwidth=40.0% ] we have calculated the steady flow solutions and examined their linear stability under the longitudinal perturbation in the cases of @xmath87 and @xmath86 . the linear stability analysis revealed that there are three unstable regimes a , b , and c in the @xmath88 plane in both of the cases . the regimes a and b are at the small @xmath95 region , while the regime c is at the large @xmath95 region . the difference between the regimes a and b is the density dependence of the flow velocity ; the denser flow is faster in the regime a , while the flow with lower density is faster in the regime b. the regime c lies within the region where the density profile is non - monotonic in the case of @xmath87 , while the regime c includes a part of the region of the monotonic density profiles for @xmath86 , although the region of the regime c roughly corresponds with that of the non - monotonic density profiles . in all the regimes , the dispersion relation of the unstable mode shows the features that ( i)@xmath151 and ( ii)re@xmath175\propto k^2 $ ] for small @xmath156 . the obtained unstable eigenmodes suggest that the instability causes the density wave . in this section , we give some discussions on the mechanism of the instability , comparison with simulations , and the relationship with the transverse instability studied by @xcite . the long - wavelength instability which results in the density wave is well known for quasi - one dimensional flows , such as the wave formation in the film flow ( ( * ? ? ? * smith 1993 ) ; ( * ? ? ? * ooshida 1999 ) ) , the density wave formation in the granular flow in a narrow vertical pipe ( ( * ? ? ? * raafat , hulin & herrmann 1996 ) ; ( * ? ? ? * moriyama 1998 ) ) , and the jam formation in the traffic flow ( ( * ? ? ? * kerner & konhuser 1993 ) ; ( * ? ? ? * bando 1995 ) ; ( * ? ? ? * ; * ? ? ? * mitarai & nakanishi 2000a , 2000b ) ) . this instability is closely related to the continuity of `` density '' @xmath176 ( the thickness in the case of incompressible fluid , the density per unit length along the pipe for the granular flow , or the density of cars for the traffic flow ) : @xmath176 obeys the equation of continuity in the form @xmath177 where @xmath178 is the flux . the flux @xmath178 may be expressed in terms of @xmath176 and its spatial derivatives , but may also depend on time especially when the inertia effect exists . in the long - wavelength and the long - time limit , the effect of the spatial derivatives of @xmath176 and the inertia on @xmath178 may be neglected . then the flux is determined by the density ; @xmath179 , and equation ( [ kinematic0 ] ) becomes @xmath180 the wave that can be described by this equation is called the kinematic wave ( ( * ? ? ? * whitham 1974 ) ) . the effect of the inertia and spatial derivatives for the small but nonzero wavelength appears as the growth rate quadratic in @xmath156 ( ( * ? ? ? * whitham 1974 ) ; ( * ? ? ? * ooshida 1999 ) ) , which causes the instability to yield the density wave . the actual form of the flux @xmath178 depends on a physical system , but we call the instability caused through this mechanism `` the long - wave instability '' after @xcite . note that the unstable mode caused by the long - wave instability has the following three features for the complex growth rate @xmath181 : ( i ) @xmath151 , ( ii ) re@xmath175\propto k^2 $ ] for small @xmath156 , and ( iii ) the phase velocity of the least stable mode @xmath182/k$ ] is given by @xmath183 in the long - wavelength limit . the unstable modes obtained in the present analysis of the two - dimensional slope flow satisfy all of the features ( i ) , ( ii ) and ( iii ) ; the features ( i ) and ( ii ) have been already pointed out in the text , and ( iii ) can be seen in figure [ dispmag](b ) in the case of @xmath87 , @xmath148 , and @xmath92 . actually , ( iii ) can be shown using ( i ) and the equation of continuity ( [ eq : rho ] ) as shown in appendix b. these features strongly suggest that the longitudinal instability in the slope flow is the long - wave instability of the kinematic wave in a quasi - one dimensional system for all of the regimes a , b , and c. one may think that , in the regime c , which lies in the region of the non - monotonic density profile in the case of @xmath87 ( figure [ phase005 ] ) , the non - monotonic density profile might play a crucial role in the instability as in the case of the transverse instability of rayleigh - brnard type , where the convection occurs due to the non - monotonic density profile ( ( * ? ? ? * forterre & pouliquen 2002 ) ; ( * ? ? ? * carpen & brady 2002 ) ) ; however , the fact that the regime c contains a part of the region of the monotonic density profile in the case of @xmath86 ( figure [ phase010 ] ) suggests that the shape of the density profile does not determine the instability . it should be also noted that the present longitudinal instability occurs at the long wavelength while the transverse instability appears at the finite wavelength . in order to confirm that the longitudinal instability is the long - wave instability of the one - dimensional kinematic wave , we now try to reduce our two - dimensional model into the one - dimensional model that preserve the major features of the original model . in spite of crudeness of our procedure , we will see the obtained one - dimensional model shows roughly the same stability diagram for the long - wave instability of the kinematic wave . the way we obtain the one - dimensional model is to integrate the original equations in @xmath65-direction from @xmath71 to @xmath184 ; the idea is similar to @xcite , where the one - dimensional model has been obtained for the flow in a vertical chute by integrating the equations across the chute width . we define the one - dimensional density @xmath185 , the average velocity @xmath186 , and the average temperature @xmath187 as @xmath188 the one - dimensional equation of continuity obtained by integrating ( [ eq : rho ] ) is @xmath189 where @xmath190 and @xmath191 represent @xmath192 and @xmath193 , respectively . ( [ conti - one ] ) is in the form of ( [ kinematic0 ] ) with the flux @xmath194 given by @xmath195 . by integrating the @xmath67-component of the equation of motion ( [ eq : u ] ) , we obtain @xmath196 which determines the evolution of @xmath197 . the second term in rhs is the shear stress at the floor and comes from the integration of @xmath198 . to simplify ( [ onedmotion ] ) , we make the following approximations ; ( i ) replace the second term in lhs @xmath199 by @xmath200 , ( ii ) neglect the velocity in the @xmath65-direction , @xmath201 , and use the dilute limit expressions for the the pressure and viscosities in @xmath202 , namely , we replace @xmath202 by @xmath203 , where the first term comes from the pressure and the second term comes from the dilute limit of the shear viscosity . the second viscosity @xmath204 is neglected because it is a higher order quantity in @xmath31 ( see ( [ eq : pmuetc ] ) ) . ( iii ) estimate the integration of the shear viscosity term by multiplying the integrand by the decay length of the density @xmath205 , and replacing @xmath2 by @xmath206 , @xmath207 by @xmath208 , i.e. , replace @xmath209 by @xmath210 , where @xmath211 . then we obtain the equation @xmath212 for the shear stress at the floor @xmath213 , we have the boundary condition of momentum balance ( [ eq : bc1 ] ) , i.e. @xmath214 . to close equations ( [ conti - one ] ) and ( [ onedmotion2 ] ) , we need the relation between ( @xmath215 ) and ( @xmath216 ) , and the equation for the average temperature @xmath206 . we simply assume that @xmath217 and @xmath218 , and we use the empirical relation between @xmath219 and @xmath176 for the steady flows ; namely , we take @xmath220 , with the form of @xmath221 determined from the steady solution obtained numerically for a fixed @xmath80 and different values of @xmath81 by using @xmath222 with @xmath223 . we further assume that @xmath206 is also determined by the one - dimensional density @xmath176 , rather than to use the integrated equation for the temperature . the form of @xmath224 is determined from the steady flows , namely , we assume @xmath225 . now we finally obtain the one - dimensional model in the following form ; @xmath226 & = & a(\bar \rho)\left[u(\bar \rho)-\bar u\right ] -\partial_x(\bar \rho \bar t(\bar \rho ) ) \nonumber \\ & & + ( 4/3 ) f_2(0)\partial_x \left(h(\bar t(\bar \rho ) ) \sqrt{\bar t(\bar \rho)}\partial_x \bar u \right ) , \label{onedv}\end{aligned}\ ] ] where @xmath227 in this model , the steady solution is given by @xmath228 and @xmath229 , and the flux of the steady solution is given by @xmath230 . the equation ( [ onedv ] ) with ( [ oneaandu ] ) shows that the velocity @xmath231 is determined by the balance between the acceleration by the gravity and the drag force from the floor . this model is almost the same form as the traffic flow model proposed by @xcite , but with the different form of the function @xmath232 . in the traffic flow model , @xmath232 is often called the `` optimal velocity function '' , which defines the density dependence of the car velocity and is usually a decreasing function . on the other hand , in the case of the granular flow on a slope , @xmath232 depends on the inclination angle and the boundary condition at the floor , and can take various forms as can be seen in the plot of @xmath90 vs @xmath233 in the one - dimensional model is not exactly the same as the flux @xmath89 in the original two - dimensional flow , due to the approximations used to derive @xmath232 . ] in figures [ steadyphase005](b ) and [ steadyphase010](b ) . the linear stability analysis of the steady solution can be performed analytically ( for the traffic flow model , see e.g. ( * ? ? ? * wada & hayakawa 1998 ) ) . it is easy to show that the instability condition re@xmath234>0 $ ] yields @xmath235 where @xmath236 and the prime represents the differentiation by @xmath109 . the @xmath237 limit of ( [ condition ] ) gives the stability criterion , @xmath238 the explicit form of the long - wavelength expansion of the dispersion relation for the least stable mode is given by @xmath239k^2+o(k^3 ) , \label{disp - oned}\ ] ] and we see that the phase velocity of this mode in the long - wavelength limit is given by @xmath240 , which shows that this is the kinematic wave . the instability arises when the coefficient of the @xmath241 term becomes positive , which occurs if the change of the velocity against density fluctuation is too fast compared to the effect of the pressure that reduces the density fluctuation . note that the criterion of the instability ( [ criterion ] ) does not depend on the shear viscosity term @xmath242 because it only appears in the fourth order term in @xmath156 ( see ( [ condition ] ) ) ; the approximation for that term does not crucially affect on the criterion . the stability diagrams for @xmath87 and @xmath86 obtained from this one - dimensional model are shown in figures [ phaseoned ] . in spite of the crude approximations used in the derivation of the one - dimensional model , the stability diagrams are qualitatively similar to those of the original model . the similarity further indicates that the density wave formation can be understood by the long - wave instability in quasi - one dimensional systems , like in the film flow and the traffic flow . it should be noticed that , in this one - dimensional model , the effect of the parameters in the original model such as @xmath21 , @xmath85 , and @xmath54 are included more or less implicitly in the functional forms of @xmath232 and @xmath243 ( @xmath85 also appears explicitly in ( [ oneaandu ] ) through @xmath244 ) , and the @xmath65-dependences of the variables also affect @xmath232 and @xmath243 through the integration . any changes that affect @xmath245 and @xmath206 result in the changes of the unstable regions ( see the criterion ( [ criterion ] ) ) , although the nature of the instability remains the same . before concluding this subsection , let us make a few comments on the works by ( * ? ? ? * wang , jackson , & sundaresan(1997 ) ) and @xcite on the stability analysis of granular flow in a vertical chute using hydrodynamic models of rapid granular flow ; @xcite performed the linear stability analysis numerically as in the present work , and @xcite analyzed the density wave by deriving a one - dimensional model from hydrodynamic equations for rapid flow . in the analysis of @xcite , a parameter region has been found where the steady flow is unstable against longitudinal long - wavelength perturbation to form the density wave ( figure 9 and 10 in ( * ? ? ? * wang _ _ et al.__1997 ) ) . this instability might be also understood as the long - wave instability observed in the present analysis . on the other hand , they also found the instabilities for finite wavelength perturbations , which has not been observed here . the analysis by @xcite shows clear similarity of the instability in the chute flow to the one in the slope flow . the one - dimensional model they obtained ( we call it the vp model ) has the mathematical structure and physical mechanism similar to those of our one - dimensional model : the velocity of the steady solution is determined by the balance of the gravitational acceleration and the drag force from the wall , and shows the long - wave instability for which the criterion is determined by the change of the velocity against density fluctuation and the pressure term . these results suggest that the instabilities in the vertical chute flow and the slope flow are in the same class . and ( b)@xmath86 obtained from the one - dimensional model . the unstable ( stable ) regimes are shown by grey ( white ) regions . the dashed line shows the boundary between the region of the non - monotonic density profiles and that of the monotonic density profiles for the original model . , title="fig:",scaledwidth=40.0% ] and ( b)@xmath86 obtained from the one - dimensional model . the unstable ( stable ) regimes are shown by grey ( white ) regions . the dashed line shows the boundary between the region of the non - monotonic density profiles and that of the monotonic density profiles for the original model . , title="fig:",scaledwidth=40.0% ] we find some qualitative agreements between the present results and our previous simulations ( ( * ? ? * mitarai & nakanishi 2001 ) ) as follows . our simulations were performed for a fixed inclination angle and a particular roughness of the slope with the periodic boundary condition imposed along the flow direction . within the examined parameter region , the steady flow shows the monotonic density profile ( ( * ? ? ? * mitarai , hayakawa , & nakanishi 2002 ) ) , and the flow with lower density has higher velocity . it has been demonstrated that the density wave appears only in the long system with low enough particle density . we have performed the simulations for several sets of the slope length @xmath246 and the particle number @xmath247 . in the case of the particle density @xmath248 ( single layer ) , the clear density wave is not formed for @xmath249 and @xmath250 , whereas the density wave appears for @xmath251 . upon changing the density with a fixed system length @xmath250 , the density wave is formed when @xmath252 , while the steady flow is stable for the denser cases with @xmath248 and @xmath253 . these tendencies of the simulations agree with the behaviors around the unstable regime b of the present model on the three points : ( i)the flow with lower density is faster , ( ii)the flow with lower density is less stable , and ( iii ) the critical wavelength for instability is very long . regarding ( iii ) , the critical wave length @xmath254 is much longer than the particle diameter ; @xmath255 for @xmath92 and @xmath148 , for example . this seems to be comparable with our simulation results , where the critical slope length @xmath256 was between @xmath257 and @xmath258 for @xmath248 . we do not understand yet how such a small wave number arises , but we suspect that it comes from the long mean free path in the large @xmath65 region where the density is low , namely , the particles flying over the clusters for a long distance prevent the growth of clusters in a smaller length scale . based on the observations ( i ) , ( ii ) , and ( iii ) , the parameters that we have simulated happens to be in the regime b , but if the simulations are performed with different densities , inclination angles , and/or boundary conditions at the floor , the behavior that corresponds to the regime a or c may be found . @xcite examined the linear stability of the granular flow on a wide slope against the perturbation transverse to the flowing direction , in order to understand the regular streak pattern along the flow direction observed in their experiments ( ( * ? ? ? * forterre & pouliqen 2001 ) ) . they mainly focused on the parameter region where the non - monotonic density profile ( or what they call the `` inverted density profile '' ) is observed , because they expected that such a flow would be unstable to form the vortex rolls from the analogy to the rayleigh - brnard instability . they have shown that , the flow is unstable against the transverse perturbation in the large part of the parameter region where the inverted density profile is observed . the unstable mode shows the vortex - like pattern , and they concluded that the streaks observed in the experiments result from the rolls of vortices . they also found that the flow with the monotonic density profile becomes unstable for some parameters , but the detail has not been reported . one of the differences between this transverse rayleigh - brnard type instability and the longitudinal long - wave instability appears in the length scale of the instability . the longitudinal instability occurs against the long - wavelength perturbations in the long - time behavior as can be seen in the dispersion relations , while the transverse instability occurs at a finite length scale comparable with the vortex roll . our analysis shows that there is a parameter region where both of the instability may occur , around the region of the non - monotonic density profiles ( the regime c ) . it should be interesting to investigate how the two instabilities interfere or not by the full three dimensional analysis . in the large inclination angle , @xcite also observed the square lattice pattern . this phenomenon can not be understood by the simple superposition of the long - wave instability and the rayleigh - brnard type instability , because the length scale of the long - wave instability is much longer than the one of the lattice pattern . the steady flows and their linear stability are analyzed for the granular flow on a slope using the hydrodynamic model with the constitutive relations derived from the kinetic theory of inelastic spheres . we mainly focused on the relatively low density region where the density decays monotonically . the stability diagram shows the three unstable regimes a , b , and c in the both cases of @xmath50 and @xmath86 . two of the unstable regimes a and b are in the lower density region , and the regime c is in the high density region . the difference between the regimes a and b is that , the denser flow is faster in the regime a at the small @xmath94 side , while the flow with lower density is faster in the regime b at the large @xmath94 side . the regime c is in the large @xmath94 and large @xmath95 region ; it lies within the region of the non - monotonic density profile in the case of @xmath50 , while it contains a part of the region of monotonic density profiles in the case of @xmath51 , although the region of the regime c roughly corresponds with that of the non - monotonic density profile . in all regimes , the instability occurs for the long - wavelength perturbations and results in the formation of density wave . it has been found that the behaviors around the unstable regime b agree with the tendencies in our previous simulations of density wave formations . the dispersion of the complex growth rate @xmath181 has the features that ( i)@xmath151 , ( ii)re@xmath175\propto k^2 $ ] for small @xmath156 , and ( iii)im@xmath175/k=-\mbox{d}q_0/\mbox{d}\bar \rho_0 $ ] in the long - wavelength limit . these strongly suggest that the instability is the long - wave instability of the kinematic waves , which is often found in quasi - one dimensional flows . this is different from the transverse instability studied by @xcite , where the flow is unstable at the finite wave number . in order to confirm that the instability is the long - wave instability of the kinematic wave , we simplified the original equations rather heuristically into a one - dimensional model , and showed that the long - wave instability occurs in the derived one - dimensional model . the stability diagram obtained from the one - dimensional model corresponds qualitatively to the one obtained from the original equations . n. m. is grateful to ooshida takeshi for informative discussions . this work was partially supported by hosokawa powder technology foundation and grant - in - aid for jsps fellows . in this appendix , the linearized governing equations for the longitudinal perturbations ( [ eq : per1])-([eq : per3 ] ) and the boundary conditions are given ( ( * ? ? ? * forterre & pouliquen 2002 ) , see also ( * ? ? ? * alam & nott 1998 ) ) . the superscript @xmath71 denotes that the quantities are for steady solution . the subscript @xmath31 and @xmath2 denotes the partial derivatives by the variables ; @xmath259 . the subscript @xmath65 indicates the total differential with respect to @xmath65 , namely , @xmath260\nu_{0,y}+$ ] @xmath261t_{0,y}$ ] , @xmath262 , and so on . the expressions of @xmath263 , @xmath264 , @xmath204 , @xmath265 , and @xmath266 are given in ( [ eq : pmuetc ] ) , and @xmath267 . then , by inserting ( [ eq : per1])-([eq : per3 ] ) into ( [ eq : rho])-([eq : temp ] ) through ( [ eq : rho0])-([eq : t0 ] ) , we obtain the following expressions ; @xmath268\nu_1 + [ \nu_0\partial_x ] u_1 + [ \nu_{0,y}+\nu_0\partial_y]v_1=0,\label{eq : leqnu}\\ & & [ \sin\theta - p^0_\nu\partial_x + u_{0,yy}\mu^0_\nu + u_{0,y}\mu_{\nu , y}^0+u_{0,y}\mu_{\nu}^0\partial_y]\nu_1\nonumber\\ & + & [ -\nu_0\partial_t-\nu_0u_0\partial_x + ( \xi^0 + 2\mu^0)\partial_x^2 + \mu_y^0\partial_y+\mu^0\partial_y^2 ] u_1\nonumber\\ & + & [ -\nu_0 u_{0,y}+ \mu_y^0\partial_x+(\xi^0+\mu_0)\partial_x\partial_y ] v_1\nonumber\\ & + & [ -p_t^0\partial_x+u_{0,yy}\mu_t^0 + u_{0,y}\mu_{ty}^0+u_{0,y}\mu_t^0\partial_y ] t_1=0,\\ & & [ -\cos\theta - p^0_{\nu y}-p^0_\nu\partial_y+u_{0,y}\mu^0_\nu\partial_x ] \nu_1\nonumber\\ & + & [ \xi^0_y\partial_x+\xi^0\partial_x\partial_y+ \mu^0\partial_x\partial_y ] u_1\nonumber\\ & + & [ -\nu_0\partial_t-\nu_0u_0\partial_x + \xi_y^0\partial_y+ \xi^0\partial_y^2 + 2\mu_y^0\partial_y+2\mu_0\partial_y^2 + \mu_0\partial_x^2 ] v_1\nonumber\\ & + & [ -p_{ty}^0-p_t^0\partial_y + u_{0,y}\mu_t^0\partial_x ] t_1=0,\\ & & [ \kappa^0_{\nu y}t_{0,y}+\kappa^0_{\nu}t_{0,yy } + \kappa^0_{\nu}t_{0,y}\partial_y-\gamma^0_\nu + u_{0,y}^2\mu_\nu^0]\nu_1\nonumber \\ & + & [ -p^0\partial_x+2\mu^0 u_{0,y}\partial_y ] u_1\nonumber \\ & + & \left [ -\frac{3}{2}\nu_{0}t_{0,y}-p^0\partial_y + 2\mu^0 u_{0,y}\partial_x\right ] v_1\nonumber \\ & + & \left[-\frac{3}{2}\nu_{0}\partial_t-\frac{3}{2}\nu_{0}u_0\partial_x + \kappa^0\partial_x^2 + \kappa_y^0\partial_y+\kappa^0\partial_y^2\right.\nonumber\\ & & \left . + \kappa^0_{ty}t_{0,y}+\kappa^0_{t}t_{0,yy}+\kappa^0_{t}t_{0,y}\partial_y -\gamma_t^0+u_{0,y}^2\mu_t^0 \right]t_1=0.\label{eq : leqt}\end{aligned}\ ] ] the linearized boundary conditions at @xmath77 are obtained from ( [ eq : bc1 ] ) and ( [ eq : bc2 ] ) as @xmath269 , \label{eq : bc0u}\\ \partial_y t_1&= & -f_{6,\nu}^0\left[\frac{1}{3}\phi u_0 ^ 2-\frac{1}{2 } \phi t_0\right]\nu_1 -f_6 ^ 0\left[\frac{2}{3}\phi u_0u_1 -\frac{1}{2 } \phi t_1 \right].\label{eq : bc0t}\end{aligned}\ ] ] we have obtained the dispersion relation of the least stable mode , @xmath146 , and numerically found that it satisfies @xmath270 in this appendix , we show that the phase velocity of this mode , @xmath271 , satisfies @xmath272 in the long - wavelength limit @xmath273 , which suggests the mode is the kinematic wave . here , @xmath90 and @xmath89 are the one - dimensional density and the flux of the steady solution , respectively , defined in ( [ rho0q0 ] ) . note that the derivative @xmath274 is taken within a family of solutions for a fixed inclination angle as below . for a given inclination angle , the density at the floor @xmath275 is a continuous parameter to specify a steady solution . to express @xmath276 dependence explicitly , we rewrite the steady solutions as @xmath277 then @xmath278 is given by @xmath279\mbox{d}y } { \int_{0}^{\infty}[\partial \nu_0(y;\beta)/\partial \beta]\mbox{d}y } . \label{eq : dqdh}\ ] ] in order to show ( [ eq : vp0 ] ) , let us first express @xmath280 by eigenfunctions of the mode . by linearization of the equation of continuity ( [ eq : rho ] ) using ( [ eq : lin ] ) , we obtain the following expression : @xmath281 integrating ( [ linearconti ] ) from @xmath77 to @xmath184 , we have @xmath282 \mbox{d}y . \label{eq : omega}\ ] ] here , @xmath283 , and @xmath284 depend on the wavenumber @xmath156 , and we expands these functions with respect to @xmath156 , i.e. , @xmath285 where @xmath286 do not contain @xmath156 . the long - wavelength expansion of the dispersion relation @xmath146 which satisfies @xmath151 is obtained from ( [ eq : omega ] ) when @xmath287 , and we have @xmath288 \mbox{d}y } { \int_0^\infty \hat \nu_0(y)\mbox{d}y } + o(k^2 ) . \label{eq : omega2}\ ] ] namely , @xmath280 is given by @xmath289 \mbox{d}y } { \int_0^\infty \hat \nu_0(y)\mbox{d}y } \label{eq : vp2}\ ] ] in the long - wavelength limit . now all we have to do is to express @xmath290 with respect to the steady solution @xmath291 . let us write ( [ eq : rho])-([eq : temp ] ) and the boundary conditions in the matrix form : @xmath292 here , @xmath293 , @xmath294 is a constant matrix , and @xmath295 is a nonlinear operator . a steady solution @xmath296 satisfies @xmath297 therefore , from ( [ eq : lin ] ) , @xmath298 satisfies @xmath299 and expanding ( [ eq : hatx ] ) by @xmath156 with @xmath151 , we have @xmath300 for the lowest order of @xmath156 . on the other hand , differentiating ( [ eq : teijo ] ) by @xmath301 , we obtain @xmath302 it is plausible that the mode of zero eigenvalue for @xmath155 does not degenerate , because the mass is the only one conserved quantity ( the momentum is lost at the floor , and the energy is dissipated . ) . thus , from ( [ eq : x01 ] ) and ( [ eq : x02 ] ) , we obtain @xmath303 , or more explicitly , @xmath304 using ( [ eq : vp2 ] ) and ( [ eq : last ] ) , we get @xmath305 \mbox{d}y } { \int_0^\infty \left(\partial{\nu_0(y;\beta ) } /\partial \beta \right ) \mbox{d}y},\ ] ] and by comparing this with ( [ eq : dqdh ] ) , we have @xmath306 which is ( [ eq : vp0 ] ) .

we present near - ir and far - uv observations of the pre - transitional ( gapped ) disk in hd 169142 using nasa s infrared telescope facility and hubble space telescope . the combination of our data along with existing data sets into the broadband spectral energy distribution reveals variability of up to 45% between.5 - 10 m over a maximum timescale of 10 years . all observations known to us separate into two distinct states corresponding to a high near - ir state in the pre-2000 epoch and a low state in the post-2000 epoch , indicating activity within the au region of the disk . through analysis of the pa and br lines in our data we derive a mass accretion rate in may 2013 of m ( 1.5 - 2.7 ) x 10 m yr . we present a theoretical modeling analysis of the disk in hd 169142 using monte - carlo radiative transfer simulation software to explore the conditions and perhaps signs of planetary formation in our collection of 24 years of observations . we find that shifting the outer edge ( au ) of the inner disk by 0.05 au toward the star ( in simulation of accretion and/or sculpting by forming planets ) successfully reproduces the shift in nir flux . we establish that the - 70 au dark ring imaged in the nir by and and at 7 mm by may be reproduced with a 30% scaled density profile throughout the region , strengthening the link to this structure being dynamically cleared by one or more planetary mass bodies .

1410.8606

to study the regions of planetary formation within disks of gas and dust around young stars , we look toward the stars around us in the galaxy to identify those which may offer snapshots of these systems in their various stages of evolution . as these systems age , multiple mechanisms begin to clear their primordial disks of material . because planets may be forming from this material , the details of how and when these disks become cleared are important to understand the origins and evolution of planetary systems . some of these mechanisms for clearing are accretion onto the central star , photo - evaporation , grain growth and coagulation into larger bodies , and large scale sculpting by orbiting companions . in the absence of strong external illumination by hot nearby stars , photo - evaporation may only be responsible for clearing material from the inner regions outward in the disk , and is counteracted by the viscous accretion onto the central star . the other two mechanisms may operate at any radius within the disk . @xcite was first to suggest that we should see disks with cleared inner regions in transition to becoming completely cleared , and leading to the class designation of transitional " disks . these disks have since been confirmed to exist @xcite , and some objects also display evidence in their spectra for the presence of structures inside of these cleared regions , ruling out photo - evaporation as a dominant mechanism for disk clearing . these objects are designated as pre - transitional disks " and are thought to be in earlier stages of evolution compared to the transitional disks @xcite . given their long orbital periods , changes to structures in the outer disk proceed over timescales which may be longer than the lifetime of any single human being . however , material inside of the gap at small radii may orbit on periods short enough to allow for changes to propagate through the disk and become evident in the spectral energy distribution ( hereafter sed ) over timescales shorter than a decade ( see papers i and ii in this series : @xcite and @xcite , along with @xcite , @xcite , and flaherty et al . ( 2011 and 2012 ) ) . hence there may not exist one single model which adequately describes some objects , while multiple models of these objects in their different states allow for speculation on the conditions and mechanisms responsible for producing the variable emission , and perhaps insight on the evolution of planet forming regions at approximately terrestrial orbital radii . additionally , high - contrast imagery has been performed on a number of young stars hosting protoplanetary disks . these images show structures of disks in various stages of evolution such as rings , gaps , shadows , spirals , asymmetric bright / dark regions , and other features which may be indicative of conditions and perhaps companions within the disk . however , it is not always the case that the structures responsible for producing these features are able to be identified through the images alone . in particular , dark bands which appear to be annular gaps may in fact be shadows cast by inner disk structures . to determine the true nature of these features , monte - carlo radiative transfer simulation may first be used to determine the extent of shadows cast by structures which are well constrained by the sed . once these structures ( and their shadows ) are determined , it is possible to explore the density and scale height conditions which reproduce the gap - like features of the images . determining which stellar systems possess gaps in their disks may help to select targets for future studies of young planetary systems , and determining the locations and compositions of these gaps may provide direct insight about the conditions of immediate regions of planet formation . hd 169142 is a 6@xmath10 myr old herbig ae star at a distance of 145 pc which attracts particular interest since it has been identified as hosting a disk with an inner gap comparable to the size of our own solar system ( @xcite , @xcite ) . the disk is in the pre - transitional stage , characterized by emission from warm material close to the central star ( @xmath11 1 au ) as well as emission from cooler material further away from the star , with the inner gap residing between the two rings ( @xcite , @xcite , @xcite ) . the main disk extends to @xmath0250 au from the central star ( @xcite , @xcite ) and is seen at @xmath013@xmath12 from face - on , making it a good target for imagery of finer structures . near infrared ( nir ) polarized light images of hd 169142 were obtained in 2011 by @xcite using subaru / hiciao and in 2013 by @xcite using vlt / naco , showing features of the disk that were previously unresolved . the main structures revealed in these images are depicted in figure 1 . in these images , the inner gap is clearly resolved to a radius of @xmath025 au and appears to be devoid of scattering material , with most of the scattered light emanating from the inner edge of the outer disk . also visible in the images is a dark annular ring , reportedly located between @xmath040 - 70 au of the central star by quanz et al . and osorio et al . , and between @xmath050 - 80 au by momose et al . the interior of this region does not appear to be entirely cleared of scattering material ( unlike the gap inside r=25 au ) , as each team detects emission above instrumental detection levels from within the ring . the feature is confirmed at 7-mm @xcite , indicating that it is at least in part due to a deficit in density extending through the disk mid - plane ( i.e. a second gap ) . we present an investigation of the hd 169142 system through computer modeling using hochunk3d , a monte - carlo radiative transfer simulation software described in @xcite . since not all structures are well constrained by the sed alone , to understand these structures within the disk with greater accuracy we pair sed modeling with modeling to fit the observed properties of the multi - band imagery . we combine modeling of the multi - epoch sed with single epoch multi - band imagery to see what we may learn about the disk and environment of potential planet forming regions in hd 169142 . hd 169142 was observed using the spex spectrograph @xcite on nasa s infrared telescope facility ( irtf ) . the spex observations were obtained using the cross - dispersed ( hereafter xd ) echelle gratings in both short - wavelength mode ( sxd ) covering 0.8 - 2.4 and long - wavelength mode ( lxd ) covering 2.3 - 5.4 . the lxd data were obtained on 2013 may 15 ( ut ) , and the sxd on 2013 may 16 ( ut ) . for both sets of data , a 0.8 arcsec wide slit was used . the spectra were corrected for telluric extinction , and flux calibrated against the a0v star hd 162220 using the spextool software @xcite running under idl . the lxd spectrum was merged to the sxd , and required no re - scaling in the wavelength region where they overlapped . due to the light loss introduced by the 0.8 arcsec slit used to obtain the xd spectra , changes in telescope tracking and seeing between the observations of hd 169142 and a calibration star may result in merged xd spectrum with a net zero - point shift compared to their true absolute flux values . to correct for any zero - point shift that results from the 0.8 arc sec echelle observations , we also observed hd 169142 on 2013 may 16 ( ut ) with the low dispersion prism in spex using a 3.0 arcsec wide slit . on nights where the seeing is 1.0 arc sec or better , past experience shows that this technique yields absolute fluxes that agree with aperture photometry to within 5% or better . at the time of the observations , the seeing was @xmath00.7 arc sec . for the prism observations , hd 162220 was used as the flux standard . the resulting scaling factor derived for the echelle spectrum was 0.95 ( shift to fainter absolute fluxes by 5% ) , which has been applied to the data . we observed hd 169142 with the aerospace corporation s broad - band array spectrograph system ( bass ) on the irtf on 2008 september 4 ( ut ) and 2011 august 1 ( ut ) . bass uses a fixed circular entrance aperture with a diameter of 1 mm . gradual improvements in the internal optics have increased the effective projected diameter on the sky using the irtf from 3.4 arcsec in 2008 to 4.4 arcsec in 2011 . the light then passes through a cold beamsplitter to separate the light into two separate wavelength regimes . the short - wavelength beam includes light from 2.9 - 6 @xmath1 m , while the long - wavelength beam covers 6 - 13.5 @xmath1 m . each beam is dispersed onto a 58-element blocked impurity band ( bib ) linear array , thus allowing for simultaneous coverage of the spectrum from 2.9 - 13.5 @xmath1 m . the spectral resolution @xmath13/@xmath14 is wavelength - dependent , ranging from about 30 to 125 over each of the two wavelength regions @xcite . @xmath15 boo was used as an absolute flux calibration star for both observations @xcite . photometric observations of hd 169142 were obtained on 2005 april 30 ( ut ) using the near infrared camera and multi - object spectrometer ( nicmos ) instrument on the hubble space telescope . images were obtained using the f110w in and f171 m filters . for the f171 m data , the observations were obtained during target acquisition ( acq mode ) . the f110w data were obtained after coronagraphic observations by slewing the target from behind the coronagraphic spot , and observing them in the direct ( multiaccum ) mode , using an 8-point target dither . the f110w were then extracted using a point spread function for an a1 spectral type star . we obtain fluxes of f(@xmath16=1.12 @xmath1m)=5.567@xmath17 w @xmath18 and f(@xmath16=1.60 @xmath1m)=2.836@xmath17 w @xmath18 , each with an uncertainty of 1% . hd 169142 was observed 3 times in the fuv in 2013 using the hubble space telescope : twice by the space telescope imaging spectrograph ( stis , see woodgate et al . 1998 ) , and once using the cosmic origins spectrograph ( cos , see green et al . table 1 is the journal of fuv observations , augmented by archival data for altair using the goddard high - resolution spectrograph ( ghrs ) , which serves as a useful comparison star in the uv . all spectra were processed using the pipeline software appropriate for their date of observation ( cos : holland et al . 2014 ; stis : hernandez , s. et al . 2014 ; and hrs : soderblom et al . 1995 ) . cccccccc + star & date & pid & obs . i d & inst./mode & aper . & cenwave pa & t@xmath19 + hd 169142 & 2013 - 04 - 05 & 13032 & lc0m02010 & cos / g130 m & psa & 1290 & 400 + `` & '' & `` & lc0m02020 & '' & `` & 1309 & 400 + '' & `` & '' & lc0m02030 & `` & '' & 1327 & 760.032 + `` & 2013 - 07 - 18 & 13032 & oc0m03010 & stis / g140 m & 52x0.2 & 1218 & 2040 + '' & `` & '' & oc0m03020 & `` & '' & 1218 & 2750 + `` & 2013 - 07 - 11 & 12996 & oc2h02020 & stis / e140 m & 0.2x0.2 & 1425 & 1732.1 + '' & `` & '' & oc2h02030 & `` & '' & 1425 & 2761.182 + altair & 1996 - 08 - 28 & 6446 & z3gb0104 t & ghrs / g140l & 2.0 & 1425 & 217.6 + in addition to the data presented above , we have collected a multitude of archival and published data to aid in our construction of the broadband multi - epoch sed for hd 169142 ( figure 2 ) . our construction of the sed reveals variability in the nir spectrum , specifically at wavelengths between 1.5 - 10 microns , which is confirmed through multiple observations . measurements fall into two distinct flux states , with those recorded after the year 2000 exhibiting up to 45% lower nir flux than those recorded prior to 2000 , with the change occurring over no more than 10 years as given by the 2mass data ( 1998 july 19 ) and the irtf / bass data ( 2008 september 6 ) . a close - up of the variable near - ir spectrum is presented in figure 3 . in @xmath203.5 we present an investigation on the nature of this variability . data from the infrared spectrograph ( irs ) on the spitzer space telescope were obtained from the spitzer heritage archive . iso pht-40 data were obtained from the infrared space observatory ( iso ) archives @xcite . ultraviolet spectroscopy was obtained from the iue archives . our complete collection of photometric data can be found in appendix b. what we identify as the post-2000 sed in this study consists of photometry from hst / nicmos ( 2005 ) , @xcite , the akari / irc point source catalog , and the akari / fis bright source catalog , along with spectroscopy from hst / cos , irtf / bass , irtf / spex , and spitzer / irs . the pre-2000 sed consists of photometry from @xcite , @xcite , @xcite , @xcite , @xcite ( observations from 1999 ) , the 2mass all sky point source catalog , and the iras point source catalog , along with iue and iso pht-40 spectroscopy . despite the presence of gaps in the inner regions of hd 169142 , it is still likely experiencing a modest amount of gas accretion onto the star . using spectra obtained with the _ international ultraviolet explorer _ , @xcite derived a mass accretion rate in the 1990s of m @xmath21 ( 0.7 - 1.8 ) x 10@xmath6 m@xmath7 yr@xmath8 . the mass accretion rate at the time of the spex data from 15 - 16 may 2013 ( ut ) can be derived using the net emission from the pa @xmath3 and br @xmath4 lines . in order to do so , we modeled the underlying continuum in a manner described by @xcite . here the underlying emission of the stellar photosphere and dust emission were modeled using a spex spectrum of hd 98058 plus a modified blackbody to represent the dust emission , which were subtracted from the observed spectrum of hd 169142 . the results are shown in figure 4 . using a distance of 145 pc , @xmath22 , and @xmath23 , we derive a mass accretion rate of m @xmath5 ( 1.5 - 2.7 ) x 10@xmath6 m@xmath7 yr@xmath8 . for the conversion of line luminosity to mass accretion rate , we used the calibrations of @xcite and @xcite for pa @xmath3 and br @xmath4 , respectively . as noted by grady et al . ( 2007 ) , the fuv spectrum of hd 169142 is dominated by the stellar photosphere for @xmath24 1600 , and then transitions to a continuum + emission - line spectrum at shorter wavelengths . after smoothing by running boxcar filters , net continuum fluxes in excess of 10@xmath25 erg/@xmath26/s / are present over the full spectral range of the cos data . following grady et al . ( 2007 ) , we compare the smoothed cos data with a ghrs spectrum of altair ( a7vn , v=0.76 , b - v=0.22 ) scaled by @xmath27v only ( figure 5 ) . the continuum in hd 169142 traces that of the scaled altair data down to 1350 , consistent with a photospheric origin . at shorter wavelengths , excess continuum light and emission line features are present . the agreement between the scaled altair spectrum and the cos data demonstrates that there is no measurable selective extinction toward hd 169142 and is fully consistent with sed fitting discussed in @xmath203.3 and @xmath203.4 . the strongest emission feature in the fuv spectrum of hd 169142 at all 3 epochs is lyman @xmath15 , similar to other late a herbig stars and classical t tauri stars ( france et al . 2014 ) . lyman @xmath15 emission is detected at all 3 epochs of fuv spectroscopy ( figure 6 ) , with varying amounts of geocoronal contamination affecting @xmath28200 km / s ( cos data ) with much less contamination of the stis data ( within @xmath2840 km / s of line center for the g140 m data ) . the observed emission profile consists of double - peaked emission , with wings extending to @xmath28900 km / s , and absorption which is not centered on the geocoronal emission or the stellar radial velocity . there is a factor of 2 change in flux from the april 2013 to july 2013 data , and @xmath29 variability from 2013 july 11 to 2013 july 13 . variability at this level is similar to that previously reported for other well - studied herbig stars such as hd 163296 and hd 104237 . the stis g140 m data reveal a spatially unresolved source in both the continuum and in the emission lines . the stis data show saturated absorption in the geocoronal feature to + 140 km / s , consistent with interstellar and/or circumstellar h i absorption corresponding to n(h i ) 2.5 - 5@xmath30 @xmath31 , or e(b - v)=0.01 for gas to dust ratios typical of the diffuse ism . none of the lyman @xmath15 profiles show a sharp emission feature similar to the jet emission seen in hd 163296 or hd 104237 , so we interpret the blue shifted absorption as evidence for a stellar wind , but not a jet . we measure the terminal velocity of the wind at the 50% of peak blue emission level as ranging from -364 to -525 km / s . in investigating the structures within the protoplanetary disk in hd 169142 , we present several distinct models corresponding to the two states of the sed . since high contrast and angular resolution imagery dating from the pre-2000 era is unavailable , we explore both states in reverse chronological order , beginning with modeling the post-2000 state of the broadband sed and multi - band imagery , which we describe in detail in this subsection . later , we adapt this model into a set of three models which match the pre-2000 sed and explore several speculative scenarios for disk structures at sub - au radii . the hochunk3d software allows for two separate disks ( each with two coexisting grain populations ) , one gap ( with its own density and scale height profiles ) , and an envelope . usually , one disk is dedicated to small grains and the other dedicated to large , settled grains usually with coplanar geometries . however , to model a dual - gapped nature of hd 169142 , the available disks were constructed as follows : the first disk was adopted for the optically thick inner disk beginning at 0.23 au and ending at 0.27 au . the second disk was adopted as the 25 au ring and outer disk , with the second gap placed between 35 - 70 au . figure 7 details the location of the disk components in the post-2000 model , whose sed is presented in figure 8 . the full scope of the parameter space explored and the values used in construction of the post-2000 model are detailed in appendix a. where available , parameter values were adopted from previous studies , indicated in table a1 by their references . in attempting to fit the sed and imaging available for hd 169142 , we place constraints on several disk geometries and structures . first , models without an inner disk fail to reproduce the observed nir flux levels . this requires the presence of material close to the central star , as suggested by @xcite and others that the protoplanetary disk around hd 169142 belongs to the class of objects known as pre - transitional disks @xcite . for the outer disk grain population , we have used model 1 from @xcite , which includes a mixture of amorphous carbon and astronomical silicates with a maximum grain radius of 1 mm , but which does not produce a strong 10 @xmath1 m silicate emission band . to remain consistent with an inside - out " profile of grain growth ( see @xcite ) , we have used model 3 from @xcite for the mid - plane of our inner ( sub - au ) disk , which similarly has a maximum grain radius of 1 mm , but with a size distribution resulting in a population of larger grains than in the outer regions of the disk . the emission from pahs in the disk do not exhibit a variable emission profile similar to the underlying continuum between 1.5 - 10 @xmath1 m ( see spitzer / irs and iso / pht-40 data , figure 3 ) , suggesting little to no pahs within the variable inner disk . hence we include these molecules in the outer ( 25 - 250 au region ) . the envelope was adopted as a disk - like optically thin body of material suspended above and below the optically thick mid - plane of the inner disk ( see figure 7 ) . in order to explore the source of the nir variability , it was necessary that the changes between pre-2000 and post-2000 models occur primarily within this optically thin structure , as changes to the inner edge of the optically thick mid - plane produce variable shadowing of the outer disk , which is non - evident in the available data in the broadband sed . interestingly , an optically - thin composition of large grains ( as expected of an inside - out " way of grain growth ) fails to reproduce the amount of nir emission without being placed within the dust sublimation radius . however , an optically - thin component of smaller grains ( 0.005 - 1 micron ) succeeds at reproducing the observed emissions while remaining outside of the sublimation radius . the unexpected presence of these small grains at such altitude above the inner disk may hint at impacts or other debris - creating processes occurring within the inner regions of the disk , unless the grain growth and formation of the mid - plane is simply of very recent occurrence . the profile of the outer disk emission very closely resembles the peak of a single temperature black - body of t=120 k , which is consistent with most of the emission originating from the inner wall of the outer disk , and allows us to place constraints on its structure . first , the sed is best fit by placing this wall at 25@xmath282 au , as suggested by previous modeling studies and imaging ( @xcite , @xcite , @xcite , @xcite ) . next , values are obtained for the a and b parameters , respectively the power law exponents for the radial density ( @xmath32 ) and scale height ( @xmath33 ) of the outer disk , which largely impact the width and height of the emission profile . there exist several degenerate solutions for these parameters in combination with the disk mass . out of these , the disk mass is most constrained by the sed . disk masses @xmath110.0015 m@xmath7 fall short of reproducing the observed emissions at @xmath3460@xmath1 m , while disk masses @xmath350.003 m@xmath7 produce too much emission at @xmath3460@xmath1 m . a disk mass of 0.003 m@xmath7 reproduces the 60 @xmath1 m iras , 100 @xmath1 m iras , and 160 @xmath1 m akari / fis , but not the 90 @xmath1 m or 140 @xmath1 m akari / fis . since the density of data in this portion of the spectrum is relatively low , and since the emissions at these wavelengths should not be significantly variable over the period of the observations , we have chosen to fit to the iras and 160 @xmath1 m akari / fis points , as m@xmath36 = 0.003 m@xmath7 is consistent with the lower estimates of other studies ( e.g. @xcite , who obtained [email protected]@xmath280.002 m@xmath7 ) . however , we find that 0.0015 m@xmath38 m@xmath39 0.0025 m@xmath7 may reproduce emissions consistent with the 90 @xmath1 m and 140 @xmath1 m akari / fis points as well as the 50 @xmath1 m photometry in @xcite . since most of the flux between @xmath015 - 40 @xmath1 m is from the 25 au wall , there exists another degeneracy between the scale height normalization and the flare of the disk , as only the height of the wall is most constrained by the sed . the h - band imagery and their corresponding surface brightness profiles allow us to explore beyond this degeneracy , as the regions outside of the 25 au wall are resolved to have observable structure . since the scale height normalization of the 25 - 250 au structures is the same throughout ( given our use of the available disk structures , see above ) , modifying the disk flare exponent in conjunction with the scale height normalization is our only means to adjust the height of the @xmath4070 au structures without affecting the height of the 25 au wall . we find that the sed and images may be simultaneously reproduced with a disk flare @xmath41 @xmath42 for all structures where r@xmath4325 au . unfortunately , without additional disk structures being available in the 2014 - 01 - 31 release of the code , this eliminates our ability to explore the possibility that these regions may have independent geometries , or perhaps ones different than those necessary for our constructions ( e.g. @xcite and @xcite suggest flares proportional to @xmath44 and @xmath45 , respectively ) . nevertheless , our models do maintain agreement with the sed , h - band surface brightness profile , and integrated disk flux percentages , and hence suggest that the values used in this study are entirely plausible . @xcite reports a surface brightness profile behaving as @xmath46 in the regions where r@xmath1152 au and r@xmath3581 au and suggest that a surface density profile proportional to @xmath47 could reproduce the observed surface brightness . @xcite reports a better fit to their data with a broken power law exponent of @xmath48 for the region between 85 - 120 au and an exponent of @xmath49 for the region between 120 - 250 au , suggesting the presence of additional discontinuities to the disk surface . hence there is no exact agreement between the observations , although an approximate @xmath46 fall off seems well established between 80 - 120 au . while maintaining fits to the sed , we find that a disk flare @xmath41 @xmath42 and density profile at the mid - plane @xmath41 @xmath47 reproduce an h - band surface brightness profile of @xmath50 between 80 - 200 au , while we may reproduce a steeper surface brightness profile @xmath51 with a steeper fall off to the mid - plane density @xmath52 . given the available disk structures in the code , we were not able to explore additional discontinuities in the disk surface to match a steeper fall off outward of 120 au . hence we have chosen to match the surface brightness profile of the brightest inner portion by including the @xmath47 radial density profile between 25 - 250 au in the final form of our post-2000 model . note that as a result our model image which appears later in the article ( figure 10 , center ) appears brighter in the outermost regions of the disk than the @xcite image with the same scaling of intensity by @xmath53 . since the surface brightness profile between @xmath040 - 70 au requires a separate in - depth treatment of this region , we present our complete model surface brightness profile later in figure 11 . one of the most striking features present in the multi - band images of the disk in hd 169142 is a second dark ring located between @xmath040 - 70 au ( @xcite , @xcite , @xcite , in addition to the @xmath00.3 - 25 au gap that is heavily discussed in the literature ( e.g. @xcite , @xcite , @xcite ) . such a feature in the nir images may be either the result of shadows cast by the wall at @xmath025 au or the signature of a second gap one which is not so devoid of material as the gap inward of 25 au as @xmath014% of the integrated disk flux emanates from this region @xcite . using the vla , @xcite detect 7 mm emission concentrated between @xmath025 - 40 au , although much more asymmetric than the nir emission ring . as the 7 mm emissions penetrate deep within the disk detecting the thermal emissions of large dust grains , this result indicates that this is a true density feature which extends through the mid - plane of the disk . hence any shadows cast by inner structures must be in addition to effects caused by the density feature . to explore the extent that shadows may be contributing to the nir dark ring , we present three separate models . the first model does not contain any region - specific modifications to the density or scale height profile in the region of the second gap . this model is in attempt to determine the scale height of the 25 au wall and other structures which are constrained by the sed , and to see whether or not the second dark ring appears naturally as a shadow in the model images . although this model reproduces the observed sed , no second dark ring is present in the images ( figure 9 , left ) . thus , we explore the density conditions which may reproduce this feature . we find that the integrated h - band flux percentages ( 49% bright ring - 14% dark ring - 37% outer disk ) reported in @xcite may be reproduced to within 2% by scaling the density between 35 - 70 au by 0.3@xmath54 the value just outside of 70 au ( figure 9 , center ) . however , we also find that decreasing the scale height of the 35 - 70 au region by 0.86@xmath54 may also reproduce the observed nir integrated flux percentages and radial brightness profile . hence there may exist a multitude of models fitting the nir images with a gap density change of @xmath55 and a scale height change of @xmath56 . however , lacking the correct density structure at the mid - plane , many of these models would fail to reproduce the observed lack of 7 mm emission from the gap ( see figure 9 , right ) . this degeneracy too may be overcome by modeling to fit the nir and 7 mm images simultaneously , although this is beyond the scope of our study at this stage . similarly , future acquisition of multi - band imagery will help to constrain the density and location of various grain species throughout the disk . we leave the density / scale height degeneracy to be resolved in future studies and include only a density scaling to the gap in our final model . furthermore , we find that a density profile @xmath57 in the gap produces a surface brightness profile in the region following a power law of @xmath58 , as opposed to the observed profile of @xmath46 reported in @xcite . since a significant amount of radiation is back - scattered off of the 70 au wall , we find that a steeper density profile @xmath52 inside of the gap is necessary to reproduce the @xmath46 surface brightness fall off . thus our final h - band ( 2.2 @xmath1 m ) model image of the post-2000 state ( figure 10 , center ) highly resembles those taken through the actual telescopes with a surface brightness profile ( figure 11 ) which is closely described by a power law of @xmath46 where r@xmath1150 au and @xmath50 where r@xmath3580 au . as detailed in section 2.5 , our construction of the sed reveals variability in the nir spectrum , specifically at wavelengths between 1.5 - 10 @xmath1 m . measurements fall into two distinct flux states , with those recorded after the year 2000 exhibiting up to 45% lower nir flux than those recorded prior to 2000 , with the change occurring over no more than 10 years ( see figure 3 ) . we present three models corresponding to the pre-2000 state of the disk in hd 169142 . model a explains the nir variability by changing the location of the outer edge of the inner disk , model b by changing the altitude and density of the mini - envelope , and model c by the variable density of the mini - envelope alone . we have restricted changes in model structures to the inner regions due to the difficulty for changes to propagate through the outer structures faster than their orbital periods , which would certainly be longer than the observed timescale of @xmath0 10 yr . note that changing the location of the outer edge of the optically thick mid - plane does not affect the shadowing of the outer disk , and contributes to the nir variability since it is being heated by the optically thin envelope . it is also interesting to note that even in our most extreme scenarios , the change to the envelope mass accounts for less than 1% of the mass lost due to accretion ( @xmath59 over the maximum timescale of 10 years ) . hence the optically thick mid - plane is a necessary component of our model constructions to explain the accretion rate . for comparison , the cross - sectional density distribution of the sub - au regions of each model are presented in figure 12 , which we now discuss . in this model the pre-2000 nir sed is achieved through extending the outer edge of the inner disk and mini - envelope to 0.285 au and 0.32 au , respectively ( see model sed in figure 13 and density profile in figure 12 ) . chronologically from this model to the post-2000 model , these correspond to an inward shift of 0.015 au for the optically thick disk and 0.05 au for the optically thin envelope . changes analogous to these might occur in systems in which the inner disk is actively being cleared of material by accretion onto the central star and/or through gravitational sculpting by a planetary mass companion orbiting within the gap ( see 4.2 ) . to keep the density of the inner disk consistent with that of the post-2000 model , the fraction of mass in the inner disk was adjusted accordingly to 7.1@xmath60 m@xmath36 , as opposed to the value of 5.0@xmath60 m@xmath36 in the post-2000 model . these correspond to a net change in mass for the mid - plane of 6.3@xmath61 m@xmath7 and for the envelope of 4.0@xmath62 m@xmath7 . in this model the pre-2000 nir state is achieved through increasing the scale height of the mini - envelope by + 0.07 au and doubling its density to a uniform 5@xmath5410@xmath25 g @xmath63 , as compared to the post-2000 model ( see model sed in figure 14 and density profile in figure 12 ) . these changes correspond to a mass loss of 6.6@xmath62 m@xmath7 from the envelope in going from this model to the post-2000 state . to account for the mass lost due to accretion of @xmath02@xmath61 m@xmath7 ( obtained using our derived accretion rate over 10 years ) , we have adjusted the fraction of mass in the inner disk to 5.58@xmath60 , as opposed to the value of 5@xmath60 used in the post-2000 model . the magnitude of an altitude change alone necessary to reproduce the observed flux levels would effectively constitute a halo in the high flux state , and its unlikely near - complete disappearance in the low flux state . hence in this model we have chosen to vary both the altitude and the density of the mini - envelope simultaneously . changes analogous to these might occur in systems in which material is suspended at higher altitude in the pre-2000 state as opposed to the post-2000 state , perhaps due to an inner disk wind or other phenomena within the inner regions ( see 4.2 ) . in this model the pre-2000 nir state is achieved through increasing the density of the mini - envelope alone to a uniform 10@xmath64 g @xmath63 , which corresponds to a mass loss of 7.2@xmath62 m@xmath7 ( see model sed in figure 15 and density profile in figure 12 ) . again , to account for the mass lost due to accretion of @xmath02@xmath61 m@xmath7 , we have adjusted the fraction of mass in the inner disk to 5.67@xmath60 , as opposed to the value of 5@xmath60 used in the post-2000 model . while this change in density may plausibly be due to accretion onto the central star , such a dramatic change in density to the structures at high altitude produces variable shadowing ( and thus emission ) of the outer disk . we present this model solely as a means of narrowing down where and how the changes must be taking place within the sub - au structures . our primary finding is that the protoplanetary disk in hd 169142 possesses material that is within 1 au the star and undergoing changes to its structure on the order of a decade . a simple way to reduce the nir flux from observed pre-2000 levels to post-2000 levels is to reduce the scale height of the innermost edge of the disk . however , such a change would decrease the shadowing of the outer disk , and hence produce an anti - correlated variability of the inner and outer disk emissions , which is non - evident in the sed . hence , we have included in each of our models an optically thin population of small grains enveloping the optically thick sub - au mid - plane . changes to this optically thin structure do not produce major variations in the shadowing of the outer disk . the presence of such a structure is consistent with the results of @xcite who find that more realistic models of inner dust rims are insufficient to match bump "- like nir emission , as well as the results of @xcite who suggests that the standard puffed - up inner disk rims are insufficient to explain the nir excess of many herbig ae and related stars and that additional sources of nir emission must be present . we find three scenarios for the sub - au structures in the pre-2000 ( high nir flux ) state , one of which we may almost entirely exclude ( model c ) as it produces variable shadowing of the outer disk . we find nothing in the data to prefer either of the two remaining plausible scenarios over the other , although the changes between pre-2000 b @xmath65 post-2000 are much more dramatic than the changes between pre-2000 a @xmath65 post-2000 . in the scenario of our pre-2000 model a @xmath65 post-2000 model , the nir variability is explained through an inward shift of the outer edge of the sub - au structures toward the central star through accretion onto the star and perhaps other means . in the other scenario ( model b @xmath65 post-2000 state ) , the nir variability is explained through the continual vertical evacuation of material from the sub - au structures , leaving the post-2000 sub - au structures with a smaller vertical scale height than those structures in the pre-2000 state . while these two scenarios accurately reproduce observed pre-2000 nir flux levels through physically plausible mechanisms , we stress that the sed is highly degenerate to other structures and there exists a noteworthy potential for other presently unimagined possibilities . furthermore , we find that the dark ring observed between @xmath040 - 70 au in the polarized flux images is not the result of shadows cast by inner disk structures which are well constrained by the sed . instead , we find that this feature may be reproduced in the model images through scaling the density of the region by 0.3@xmath54 the value just outside of the gap , which is consistent with the density structure inferred through the 7 mm observations of @xcite . however , we find that the h - band images may also be reproduced by adjusting the scale height of the region by 0.86@xmath54 the value just outside of the gap . hence , further study of the multi - band imagery is necessary here to eliminate the degeneracy of these parameters and determine the true composition of the various dust grain sizes throughout the disk . extending to a distance comparable to between the orbits of uranus and neptune in our own solar system , the @xmath00.3 - 25 au gap in the protoplanetary disk in hd 169142 is temptingly explained through the clearing of material by planetary mass companions . however , there are other mechanisms which are likely to produce the inferred disk structure and which must be accounted for before we may speculate on which ( if any ) features may be the signature of young and forming planets . additionally , the timescale on which these mechanisms act to clear regions of the disk is of importance as the potential for planetary formation is significantly diminished with the decreasing supply of material . in a general overview of the evolution of protoplanetary disks , @xcite suggest that viscous accretion first dominates over photo - evaporation , until the mass accretion rate drops to @xmath66 . once these two counteracting processes become active on the same order of magnitude , the viscous flow of material from the outer disk will no longer be able to resupply the inner disk of material lost due to photo - evaporation , and an inner hole will be cleared . through analysis of the pa @xmath3 and br @xmath4 lines present in the may 2013 spex observations we derive a mass accretion rate of m @xmath5 @xmath67 . hence this accretion rate suggests that such a scenario may be an active clearing mechanism for the inner ( @xmath68 au ) gap . should this be the case , we may consider these to be ongoing processes within the presently observed state of the disk , as nir flux levels ( as of 2013 ) demand for the presence of material at sub - au radii . such a scenario is consistent with our speculative changes responsible for producing the variable nir emission ( see pre-2000 model a , in which the decrease in nir flux is explained through the shift of material inward toward the star , with a net mass loss of 6.3@xmath61m@xmath7 ) . we may also assume that the disk clearing mechanisms of grain - growth and coagulation into larger bodies , as well as large - scale sculpting by newly formed planets , may also be responsible for producing these observed structures . in fact , the results of our pre-2000 model a are suggestive that such additional clearing may plausibly be occurring , as our derived mass accretion rate accounts for as much as 2.7@xmath61m@xmath7 , or 43% of the total mass cleared over the observed timescale of the decrease in flux . this would leave room for other mechanisms to be actively clearing the region on the same order of magnitude as accretion onto the star . should a scenario such as this actually be underway , we could perhaps expect to see complete disappearance of the inner disk in hd 169142 on the order of another decade or so , prompting a change of class from pre - transitional to transitional disk . while the nir variability may be explained through the above scenario , there exists potential for other possibilities . another scenario in which the nir variability may be reproduced is one in which the optically thin envelope extends to higher altitude in the pre-2000 state than in the post-2000 state , as explored in our pre-2000 model b. this scenario matches pre-2000 nir flux levels with a shifting of the envelope by 0.07 au in the vertical direction and doubling its density as compared to the post-2000 model . while determining the exact mechanism and situations for which such a scenario may be relevant are beyond the scope of this paper , one likely possibility is that reconnection events ( like those present in the model of @xcite ) may be occurring within the magnetic field of a partially ionized inner disk , resulting in material which is lofted at altitude above the mid - plane . such proceedings would also help to explain the presence of the optically thin material which has been necessary throughout our constructions . similar considerations were explored for hd 163296 and hd 31648 in @xcite , and the migration of material at altitude above the inner disk was explored as a mechanism for producing the photometric variability observed in hd 163296 @xcite . however , given the chronology of the observed states , such a scenario would have material decreasing in altitude by 0.07 au and decreasing in density by a factor of 2 between the pre-2000 state and the post-2000 state . these changes , as well as those between our pre-2000 model c and post-2000 model are much more dramatic than those between the pre-2000 a and post-2000 models , and thus by invoking occam s razor we consider the chronology of our pre-2000 model a @xmath65 post-2000 model as being the most favorable of our presented scenarios . we obtain a null result for the dark ring between @xmath040 - 70 au being a shadow cast by inner structures , but rather we find that this feature may be reproduced in the model images by a net deficit in the density of scattering material in the region . hence we are left with the problem of how such structure in the disk might have originated and evolved from the primordial disk . again , large - scale sculpting by planetary mass companions is temptingly invoked as a mechanism for clearing this gap of material especially given its location beyond the snow - line and the large amount of gas available here for giant planet production . another possibility is that grain growth and coagulation may have been active within the region resulting in a net decrease in the density of micron sized grains and a corresponding increase in the density of grains with millimeter cross - sections . however , this seems unlikely as the 7 mm emissions suggest that the largest grains are concentrated within the @xmath025 - 40 au region , and not within the gap @xcite . this result is also consistent with a planet - induced dust filtration process ( as in @xcite ) being responsible for clearing of this gap , as in such a scenario the small grains should remain in the cavity while the millimeter sized grains are trapped at the edge . given the presence of large structures at smaller radii , photoevaporation falls even shorter of explaining the origin of this second gap . likewise , viscous flows of material to smaller radii seem unlikely , as this would raise new questions of why such a mechanism has preferentially cleared this region but not those at immediately larger or smaller radii . hence we are left to consider the intriguing and seemingly likely possibility that this second gap is a direct result of the formation of planets and their ongoing sculpting of this region of the disk . finally , there exists the possibility for the presence of multiple orbiting companions which may be responsible for the clearing of material , as well as sending material into orbit at altitude above the mid - plane ( which may explain the origin of the mini - envelope that we have found to be a necessary component ) . such companions may be on highly eccentric orbits , and at the inner parts of their orbits may travel close enough to perturb the sub - au disk structures on periods consistent with the @xmath010 yr timescale of the change in nir emission . furthermore , @xcite & @xcite have reported the independent l detection of planet - candidates respectively at separations of @xmath00.11 and 0.156"@xmath280.032 from the star . at a distance of 145 pc , these angular separations correspond to physical separations of @xmath016 au and [email protected] au , placing them within the inner gap . both teams present observations lacking corresponding detections of the l hot - spots at shorter wavelengths , thus inferring their origin as sub - stellar . however , each team detects only one hot - spot , at differing radii , magnitude , and position - angle . hence the validation of these planet - candidates presently remains open to skepticism . incidentally , neither team detected l counterpart emission from the compact 7 mm source in @xcite . at 7 mm the disk is optically thin , raising the possibility of a chance coincidence with a background object . to be compelling , the l candidates need to be recovered again , and also seen in other filters ( such as 3.1 microns ) . nevertheless , the structures within the disk provide compelling evidence of planetary interactions and warrant the continued observation of the hd 169142 system as a candidate for the direct imaging of young extra - solar planets . our assembly of the broadband sed from data sets covering a quarter of a century reveals variability of up to 45% at nir wavelengths over a maximum timescale of 10 years . through modeling of the sed we confirm several scenarios where changes to the inner ( sub - au ) structures may reproduce the two distinct states of the nir spectrum . the simplest of these involves a shift of the outer edge of the inner disk by @xmath00.05 au inward toward the star . in this scenario we have kept the density of the mid - plane the same by adjusting the total mass of the inner disk by 6.3@xmath61m@xmath7 to account for the decrease in volume . from the spex observations we derive a mass acrretion rate of m @xmath5 ( 1.5 - 2.7 ) x 10@xmath6 m@xmath7 yr@xmath8 , which accounts for as much as 43% of the total mass change in the above scenario , and is suggestive that other processes ( such as sculpting due to planets near the inner edge of the gap ) may be active in addition to accretion onto the star . however , we find that the pre-2000 sed may be equally reproduced if the density of the inner disk is allowed to increase , allowing for the change in mass to be equivalent to that lost due to accretion . in any event , we do not observe any variable shadowing of the outer disk , requiring that changes must be occurring within the outer edge of the inner disk , or to an optically thin structure above / below the mid - plane . in order to reproduce the level of variability we have found that the latter structure is a necessary component to provide heating of the outer portions of the inner disk , and is similar to components present in the models and predictions in other studies on related objects ( @xcite , @xcite , @xcite ) . through a combination of modeling the broadband sed and multi - band imagery we explore the structures of the outer disk and in particular the nature of the second gap detected in the h - band by @xcite and @xcite , and at 7 mm by @xcite . our models suggest that this ring in the nir is not significantly in part caused by shadows cast by inward disk structures . instead , we find that the dark ring in the nir may be reproduced by a 0.3@xmath54 scaling of density or a 0.86@xmath54 scaling of the height within the region . as a density feature is required to match the 7 mm observations , we find that a model fitting the images at both bands is possible with a density scaling @xmath55 and a height scaling of @xmath56 , which we leave for future studies to explore with the hopeful acquisition of additional mm - continuum imagery . in any event , this structure is highly suggestive of planetary formation within the region . we find that the most constrained outer disk structure in the sed is a wall at @xmath025 au , whose emission resembles a nearly single temperature black - body with t=120k . while maintaining fits to the sed , we find that the h - band polarized surface brightness profile @xmath69 between 85 - 120 au @xcite may be reproduced by a radial density profile @xmath57 and a scale height exponent @xmath70 , while a steeper profile @xmath51 may be reproduced with a steeper fall off to the density @xmath52 . similarly , we find that a density profile @xmath52 in the 35 - 70 au gap reproduces the observed surface brightness profile @xmath51 reported by @xcite for the region where r@xmath1150 au . finally , we conclude that the protoplanetary disk in hd 169142 is home to a multitude of structures which may indicate the activity of planetary formation processes , and perhaps even that the effects of such processes may be visible over a single person s lifetime . at the very least , it offers a unique perspective on the timescales and mechanisms by which such disks become cleared of material , and hence when and how planets have the opportunity to form . in any event , hd 169142 is a highly structured and evolving system , of which continued observation and investigation may reveal further details , as well as answer important questions about the formation and evolution of young planetary systems . this work was supported by nasa adap grant nnx09ac73 g , hubble space telescope grant hst - go-13032 , the ir&d program at the aerospace corporation , and the university of cincinnati honors program . we would like to thank dr . sascha quanz , dr . munetake momose , and the astronomical society of the pacific conference series for permission to reproduce the images from their original publications . we would also like to thank the anonymous journal referee whose comments greatly improved this article . parameters listed here were used in the 2014 - 01 - 31 release of the whitney code in construction of our models of the circumstellar disk in hd 169142 . specific references are given where we have used a value within the bounds given in the referenced material , otherwise it may be assumed that the values were chosen upon for this study in seeking a best fit to the sed and images . the post-2000 model parameters were used as a basis for the pre-2000 models and the changes to each model are detailed below . ccc spectral type & a7v & a + t@xmath71 & 7500 k & b + radius ( r@xmath72 ) & 1.75 r@xmath7 / 1.6 r@xmath7 & c / a + mass & 1.65 m@xmath7 & a + distance & 145 pc & d + disk inclination & 13@xmath12 & e + disk mass ( total ) & 0.003 m@xmath7 & f + fraction of mass in inner disk & @xmath73 & + disk accretion rate & 2.1@xmath74 m@xmath7 & g + inner disk grain file & www005.par & ( 1 ) + inner disk scale height normalization & 0.002 r@xmath72 & + inner disk r@xmath75 - r@xmath76 & 0.23 - 0.27 au & + inner disk radial density @xmath77 & a=1.0 & + inner disk scale height @xmath78 & b=1.12 & + envelope grain file & carbonstardust.par & ( 3 ) + envelope r@xmath75 - r@xmath76 & 0.23 - 0.27 au & + envelope density & 2.5@xmath79 g @xmath63 & + floor density ( outside 0.27 au ) & @xmath80 g @xmath63 & + envelope bipolar cavity opening angle & 15@xmath12 & + bipolar cavity z - intercept & -0.13 au & + outer disk grain file & www003.par & ( 2 ) + outer disk small grains file & draine_opac_new.dat & ( 4 ) + outer disk % mass in small grains & 4% & + outer disk scale height normalization & 0.00675 r@xmath72 & + outer disk r@xmath75 - r@xmath76 & 25 - 250 au & h + outer disk radial density @xmath77 & a=1.0 & i + outer disk scale height @xmath78 & b=1.3 & + gap r@xmath75 - r@xmath76 & 35 - 70 au & h , j + gap radial density @xmath77 & a=2.0 & + gap density scaling & 0.3@xmath54 & + scale for radial exponential density cutoff & 120 au & + number of radial grid cells & 400 & + number of theta ( polar ) grid cells & 197 & + number of phi ( azimuthal ) grid cells & 2 & + + inner disk r@xmath76 & 0.285 au & + envelope r@xmath76 & 0.32 au & + fraction of mass in inner disk & @xmath81 & + bipolar cavity z - intercept & -0.06 au & + envelope density & 5@xmath79 g @xmath63 & + fraction of mass in inner disk & @xmath82 & + envelope density & @xmath83 g @xmath63 & + fraction of mass in inner disk & @xmath84 & + to aid in our construction of the sed , we have gathered the following data from their respective sources . cccccc vanderveen & 0.45 & 1.25e-11 & 1.25e-13 & 1989 & a + vanderveen & 0.55 & 1.11e-11 & 1.11e-13 & 1989 & a + vanderveen & 0.68 & 8.05e-12 & 8.05e-14 & 1989 & a + vanderveen & 0.90 & 5.90e-12 & 5.90e-14 & 1989 & a + nicmos & 1.12 & 5.567e-12 & 5.567e-14 & 2005 - 04 - 30 & b + sylvester & 1.22 & 4.04e-12 & & 1992 - 06 - 12 & c + hales & 1.24 & 4.04e-12 & 8.08e-14 & 2003 - 04 - 29 & d + malfait & 1.24 & 4.07e-12 & & 1992 & e + 2mass & 1.25 & 4.60e-12 & 5.54e-14 & 1998 - 07 - 19 & f + nicmos & 1.60 & 2.836e-12 & 2.836e-14 & 2005 - 04 - 30 & b + malfait & 1.63 & 3.04e-12 & & 1992 & e + 2mass & 1.65 & 3.24e-12 & 1.06e-13 & 1998 - 07 - 19 & f + sylvester & 1.65 & 3.06e-12 & & 1992 - 06 - 12 & c + 2mass & 2.17 & 2.52e-12 & 4.45e-14 & 1998 - 07 - 19 & f + sylvester & 2.18 & 2.05e-12 & & 1992 - 06 - 12 & c + malfait & 2.19 & 2.20e-12 & & 1992 & e + wise & 3.35 & 9.05e-13 & 7.70e-14 & 2010 - 03 - 26 & g + sylvester & 3.55 & 1.18e-12 & & 1992 - 06 - 12 & c + sylvester & 3.76 & 1.00e-12 & & 1992 - 06 - 12 & c + malfait & 3.77 & 1.12e-12 & & 1992 & e + wise & 4.60 & 6.11e-13 & 2.90e-14 & 2010 - 03 - 26 & g + sylvester & 4.77 & 7.40e-13 & & 1992 - 06 - 12 & c + malfait & 4.80 & 6.27e-13 & & 1992 & e + rayjay & 10.8 & 6.58e-13 & 6.583e-14 & 1999 - 05 - 04 & h + wise & 11.56 & 6.28e-13 & 6.30e-15 & 2010 - 03 - 26 & g + iras & 12 & 7.38e-13 & 2.95e-14 & 1983 & i + akari irc & 18 & 3.817e-12 & 3.899e-14 & 2006 & j + rayjay & 18.2 & 1.296e-12 & 1.296e-13 & 1999 - 05 - 04 & h + wise & 22.09 & 2.01e-12 & 1.20e-14 & 2010 - 03 - 26 & g + iras & 25 & 2.21e-12 & 1.11e-13 & 1983 & i + harvey & 50 & 1.374e-12 & 2.46e-13 & 1995 & k + iras & 60 & 1.48e-12 & 1.33e-13 & 1983 & i + akari fis & 65 & 1.128e-12 & 4.371e-15 & 2006 & l + akari fis & 90 & 6.663e-13 & 4.100e-14 & 2006 & l + iras & 100 & 7.03e-13 & 7.73e-14 & 1983 & i + akari fis & 140 & 2.854e-13 & 4.049e-14 & 2006 & l + akari fis & 160 & 2.900e-13 & 7.969e-15 & 2006 & l + acke , b. , & van den ancker , m. e. 2004 , , 426 , 151 biller , b. a. , males , j. , rodigas , t. , et al . 2014 , arxiv:1408.0794 blondel , p. f. c. , & djie , h. r. e. t. a. 2006 , , 456 , 1045 calvet , n. , muzerolle , j. , briceo , c. , hern ndez , j. , hartmann , l. , saucedo , j. l. , and gordon , k. d. 2004 , , 128 , 1294 cushing , m.c . , vacca , w.d . , & rayner , j.t . 2004 , , 116 , 362 clayton , g. c. , de marco , o. , whitney , b. a. , et al . 2011 , , 142 , 54 draine , b.t . , & li , a. 2007 , , 657 , 810 dullemond , c. p. , & monnier , j. d. 2010 , , 48 , 205 ellerbroek , l. e. , podio , l. , dougados , c. , et al . 2014 , , 563 , a87 espaillat , c. , calvet , n. , dalessio , p. , et al . 2007 , , 670 , l135 espaillat , c. , calvet , n. , hernandez , j. , luhman , k. , & muzerolle , j. 2010 , bulletin of the american astronomical society , 42 , # 346.07 espaillat , c. , furlan , e. , dalessio , p. , et al . 2011 , , 728 , 49 flaherty , k. m. , muzerolle , j. , rieke , g. , et al . 2011 , , 732 , 83 flaherty , k. m. , muzerolle , j. , rieke , g. , et al . 2012 , , 748 , 71 france , k. , schindhelm , e. , bergin , e. a. , roueff , e. , & abgrall , h. 2014 , , 784 , 127 grady , c. a. , schneider , g. , hamaguchi , k. , sitko , m. l. , carpenter , w. j. , hines , d. , collins , k. a. , williger , g. m. , woodgate , b. e. , henning , th . , mnard , f. , wilner , d. , petre , r. , palunas , p. , quirrenbach , a. , nuth , j. a. , iii , silverstone , m. d. , and kim , j. s. 2007 , , 665 , 1391 green , j. c. , froning , c. s. , osterman , s. , et al . 2012 , , 744 , 60 guimares , m. m. , alencar , s. h. p. , corradi , w. j. b. , & vieira , s. l. a. 2006 , , 457 , 581 habart , e. , natta , a. , testi , l. , & carbillet , m. 2006 , , 449 , 1067 hackwell , j.h . , warre , d.w . , chatelain , m.a . , dotan , y. , li , p.h . , lynch , d.k . , mabry , d.j . , russell , r.w . , & young , r.m . 1990 , spie , 1235 , 171 hales , a. s. , gledhill , t. m. , barlow , m. j. , & lowe , k. t. e. 2006 , , 365 , 1348 harvey , p. m. , smith , b. , & di franco , j. 1995 , astrophys . space sci . , 224 , 475 honda , m. , maaskant , k. , okamoto , y. k. , et al . 2012 , , 752 , 143 jayawardhana , r. , fisher , r. s. , telesco , c. m. , et al . 2001 , , 122 , 2047 kurucz , r. l. 1979 , , 40 , 1 kurucz , r. l. 1993 , vizier online data catalog , 6039 , 0 maaskant , k. m. , honda , m. , waters , l. b. f. m. , et al . 2013 , , 555 , a64 maaskant , k. m. , min , m. , waters , l. b. f. m. , & tielens , a. g. g. m. 2014 , , 563 , a78 malfait , k. , bogaert , e. , & waelkens , c. 1998 , , 331 , 211 mathis , j.s . , rumpl , w. , & nordsieck , k.j . 1977 , , 217 , 425 meeus , g. , pinte , c. , woitke , p. , et al . 2010 , , 518 , l124 momose , m. , morita , a. , okamoto , y. , et al . 2013 , astronomical society of the pacific conference series , 476 , 389 muzerolle , j. , hartmann , l. , & calvet , n. 1998 , , 116 , 455 muzerolle , j. , flaherty , k. , balog , z. , et al . 2009 , , 704 , l15 osorio , m. , anglada , g. , carrasco - gonzlez , c. , et al . 2014 , iau symposium , 299 , 145 pani , o. , hogerheijde , m. r. , wilner , d. , & qi , c. 2008 , , 491 , 219 quanz , s. p. , avenhaus , h. , buenzli , e. , et al . 2013 , , 766 , l2 raman , a. , lisanti , m. , wilner , d. j. , qi , c. , & hogerheijde , m. 2006 , , 131 , 2290 rayner , j.t . , toomey , d.w . , onaka , p.m. , denault , a.j . , stahlberger , w.e . , vacca , w.d . , cushing , m.c . , & wang , s. 2003 , , 115 , 362 reggiani , m. , quanz , s. p. , meyer , m. r. , et al . 2014 , arxiv:1408.0813 r.w . russell , r.j . rudy , g.s . rossano , d.l . kim , e. laag , k , crawford , m.a . skinner , s.a gregory , & m.l . sitko . 2012 , 21st calibration conference , space dynamics laboratory , utah state university sitko , m. l. , carpenter , w. j. , kimes , r. l. , et al . 2008 , , 678 , 1070 sitko , m.l . , day , a.n . , kimes , r.l . , et al . 2012 , , 745 , 20 shu , f. , najita , j. , ostriker , e. , et al . 1994 , , 429 , 781 sloan , g. c. , keller , l. d. , forrest , w. j. , et al . 2005 , , 632 , 956 strom , k. m. , strom , s. e. , edwards , s. , cabrit , s. , & skrutskie , m. f. 1989 , , 97 , 1451 sylvester , r. j. , skinner , c. j. , barlow , m. j. , & mannings , v. 1996 , , 279 , 915 tanaka , h. , himeno , y. , & ida , s. 2005 , , 625 , 414 vacca , w.d . , cushing , m.c . & rayner , j.t . 2003 , , 115 , 389 van der veen , w. e. c. j. , habing , h. j. , & geballe , t. r. 1989 , , 226 , 108 vinkovi , d. 2014 , , 566 , a117 whitney , b. a. , robitaille , t. p. , bjorkman , j. e. , et al . 2013 , , 207 , 30 williams , j. p. , & cieza , l. a. 2011 , , 49 , 67 wood , k. , wolff , m. j. , bjorkman , j. e. , & whitney , b. 2002 , , 564 , 887 wood , k. , whitney , b. a. , robitaille , t. , & draine , b. t. 2008 , , 688 , 1118 woodgate , b. e. , kimble , r. a. , bowers , c. w. , et al . 1998 , , 110 , 1183 zhu , z. , nelson , r. p. , dong , r. , espaillat , c. , & hartmann , l. 2012 , , 755 , 6 zubko , v.g . , mennella , v. , colangeli , l. , & bussoletti , e. 1996 , , 282 , 1321

starting from the helmholtz free energy we calculate analytically first- and second - order derivatives , as internal energy and specific heats , for the ideal system and the exchange and correlation interactions covering a broad range of degeneracy and relativity . the complex physics of coulomb interactions is expressed by pad approximants , which reflect the actual state of our knowledge with high accuracy . we assume complete ionization and provide a base system of thermodynamical functions from which any other thermodynamical quantities can be calculated . we chose for the base system the free energy , the pressure , the internal energy , the isothermal compressibility ( or density exponent ) , the coefficient of strain ( or temperature exponent ) , and the isochoric specific heat . by means of the latter potentials entropy , isobaric specific heat and adiabatic temperature gradient can be determined . we give comparisons with quantities which are composed by numerical second - order derivatives of the free energy and show that numerical derivatives of the free energy as calculated , for instance , from eos tables , may produce discontinuities for astrophysically relevant quantities as , e.g. , the adiabatic temperature gradient . adiabatic temperature gradients are shown for different chemical compositions ( hydrogen , helium , carbon ) . finally the used formalism of pad approximants allows immediate incorporation of recent results from many particle statistics .

astro-ph0008399

for an accurate modelling of stellar objects we have to compute a complete set of thermodynamical quantities which meets the physical conditions of various evolutionary stages . based on the framework presented in stolzmann & blcker ( 1996a , hereafter paper i ) we derive further thermodynamical potentials in order to provide a base system from which _ any _ other thermodynamical quantity can be calculated . such a system requires six quantities consisting of the helmholtz free energy , two of its single derivatives , one of its second mixed derivatives , and two of its second pure derivatives with respect to temperature and density ( see e.g. dppen et al . we complement the helmholtz free energy with the pressure and the internal energy , with the isochoric specific heat , and with the temperature and density exponents in the equation of state ( eos ) , @xmath0 and @xmath1 . this paper will be devoted to the ideal , exchange and correlation contributions according to the expansions given in paper i , viz . for _ case a _ : weak relativity and arbitrary degeneracy , and _ case b _ : strong degeneracy and arbitrary relativity . these cases cover a large area in the density - temperature plane . the correlations between the charged particles are formulated by the technique of pad approximants as in paper i. here we give improved versions of our earlier applied pad approximants . in particular we have rearranged terms of the quantum virial function . + explicit expressions for the helmholtz free energy and the pressure have already been derived in paper i for ideality as well as for exchange and correlation interaction . this paper aims at providing analytical formulae with high accuracy which supply quick computing with reliable accuracy in practical applications . furthermore the expressions of the thermodynamical functions given here can be easily included as a part of the thermodynamical description of arbitrary degree of ionization . + this paper is organized as follows . in sect . [ thermrel ] we list the set of thermodynamical potentials to be considered . sect . [ theomodel ] gives a brief overview of the concept for the calculation of the eos terms . [ thermppot ] deals with the detailed determination of the ideal , exchange and correlation parts for the eos . numerical results and comparisons are presented in sect . [ results ] , and a summary is given in sect . we summarize briefly some well - known standard relations , which are frequently used to provide the thermodynamics for astrophysical applications . + we have to determine the first and second - order quantities which are related by the ratio @xmath2/@xmath3 of isochoric to isobaric specific heat : @xmath4 the ( isobaric ) thermal expansion coefficient @xmath5 can be expressed by @xmath6 with @xmath7 being the grneisen coefficient @xmath8 @xmath9 is the coefficient of strain @xmath10 and @xmath11 is the isothermal compressibility @xmath12 the so - called temperature and density exponents in the equation of state ( cox & giuli 1968 ) are defined by @xmath13 @xmath14 the adiabatic temperature gradient defined by @xmath15 ( @xmath16 denotes the entropy ) can be expressed by @xmath17 another possibility to calculate the adiabatic gradient is given by the three adiabatic exponents @xmath18 , @xmath19 , and @xmath20 ( cox & giuli 1968 , rogers et al . 1996 ) @xmath21 which can be obtained by @xmath22 in order to calculate the quantities given by eqs . ( [ egamspez])-([egams ] ) we start with the helmholtz free energy @xmath23 and determine by means of standard thermodynamic relations the gibbs energy , @xmath24 the pressure , @xmath25 and the internal energy , @xmath26 leading to the entropy @xmath16 according to @xmath27 and to the isochoric specific heat via @xmath28 moreover , we have to calculate the coefficient of strain and the inverse compressibility ( bulk modulus ) by means of eqs . ( [ ephi])-([echir ] ) . as in paper i we start with the helmholtz free energy @xmath29 of a fully ionized plasma consisting of ideal and coulomb interaction parts @xmath30 where the coulomb term contains the following parts : @xmath31 with x and c marking the exchange and correlation contributions and cq the quantum correction term . the pairs @xmath32 denote the interaction between particles of species @xmath33 and @xmath34 ( electrons , ions ) , respectively . for convenience we introduce dimensionless thermodynamical potentials defined by @xmath35 @xmath36 where @xmath37 refers to the total particle number density of the ions _ or _ the electrons , and @xmath38 denotes the boltzmann constant . ( [ eics ] ) , or correspondingly the potentials summarized by eq . ( [ efdim1 ] ) and eq . ( [ efdim2 ] ) , can be written by @xmath39 + \nonumber \\ & & n_{\rm i}kt \left[\sigma_{\rm i } ^{\rm i d } + \sigma_{\rm ii}^{\rm c } + \sigma_{\rm ii}^{\rm cq } + \sigma_{\rm ie}^{\rm c } \right]\end{aligned}\ ] ] where @xmath40 and @xmath41 symbolize the various thermodynamic functions defined in the previous section . + this section deals in detail with the calculation of the potentials listed in eq . ( [ eicsredu ] ) . by introducing definitions and for the sake of the integrity we repeat here few expressions concerning helmholtz free energy and pressure , which are already given in paper i. the ideality of the nonrelativistic and nondegenerate ions is described by the well - known classical expressions for the helmholtz free energy @xmath42 \;\ ; .\ ] ] the thermal de broglie wavelength for particles of species @xmath33 is @xmath43 . in the classical description we get for the pressure , the compressibilty , and the coefficient of strain @xmath44 and for the internal energy and the isochoric specific heat @xmath45 the ideal pressure of the electrons at any relativity and degeneracy ( chandrasekhar 1939 , cox & giuli 1968 ) can be calculated by @xmath46 \;\ ; , \ ] ] with the correponding particle number density @xmath47 \ ; \;\ ; .\ ] ] the thermodynamical potentials of the electrons are characterized by the relativistic fermi - dirac integrals @xmath48 , which depend on the degeneracy ( @xmath49 ) and the relativity ( @xmath50 ) parameters @xmath51 generally , the degeneracy parameter @xmath49 in ( [ eplaspar ] ) is a function of the density and temperature defined by eq . ( [ ereldens ] ) . in order to evaluate the free energy @xmath52 we have to determine @xmath49 explicitly from eq . ( [ ereldens ] ) by an inversion procedure which has to be performed numerically . note , that @xmath49 in ( [ eplaspar ] ) is identical with the ideal contribution of @xmath53 in ( [ efdim1 ] ) , i.e. @xmath54 . lamb ( 1974 ) and lamb & van horn ( 1975 ) evaluated the thermodynamical potentials applying the parametrizations of eggleton et al . ( 1973 ) for the relativistic fermi - dirac integrals . johns et al . ( 1996 ) improved the accuracy of the polynomials given by eggleton et al . straniero ( 1988 ) calculated the complete set of thermodynamical functions based on the expressions given in eqs . ( [ erelpress])-([erelhelm ] ) by numerical integrations to determine the adiabatic temperature gradient . recently , blinnikov et al . ( 1996 ) and miralles & van riper ( 1996 ) presented parametrizations and used various approximations to evaluate the fully relativistic ideality for the set of thermodynamical potentials , which are listed in sect . [ theomodel ] . however , all calculational schemes are characterized by an immense effort in order to determine temperature- and density derivations even for asymptotic regions . for details , see e.g. miralles & van riper ( 1996 ) . we pursue to include relativistic effects over a broad region of astrophysically relevant densities and temperatures for ideality and exchange . furthermore we evaluate the set of thermodynamical potentials analytically avoiding the well - known noise problems of second - order quantities apparent in purely numerical approaches . + this can be realized by introducing two approximations : + * a ) * arbitrary degeneracy , but weak relativity ( _ case a _ ) and + * b ) * arbitrary relativity , but strong degeneracy ( _ case b _ ) + a numerical study on the density - temperature validity region of these approximations was carried out in paper i. note , that _ case a _ , based on @xmath50-expansions @xmath55 , is limited to @xmath56 and @xmath57 , whereas _ case b _ , based on @xmath58-sommerfeld - chandrasekhar expansions @xmath59 holds for @xmath60 ( see fig.1 in paper i ) . carrying out the expansions ( @xmath55 ) in _ case a _ we get for eqs . ( [ erelpress ] ) and ( [ ereldens ] ) @xmath61 @xmath62 with the abbreviations @xmath63 and @xmath64 ( not to be confused with the internal energy or the volume ) @xmath65\ ] ] and @xmath66\ ] ] taking into account relativistic corrections . @xmath67 are the nonrelativistic fermi - dirac integrals considered by parametrizations and expansions ( see e.g. paper i ) . for the helmholtz free energy in eq . ( [ erelhelm ] ) we have performed the inversion @xmath68 analytically ( see paper i ) . using the maxwell relation eq . ( [ eubas ] ) and eqs . ( [ ephi])-([echir ] ) we obtain for the internal energy , compressibility , and coefficient of strain @xmath69 @xmath70 @xmath71 \;\ ; .\ ] ] obviously , eq . ( [ einna ] ) results for @xmath72 in the well - known nonrelativistic relation @xmath73 . calculating the derivatives of the free energy with respect to the temperature in order to get the isochoric specific heat ( eq . [ ecbas ] ) we have to execute explicitely @xmath74_{v , n_{a } } \ ; . \end{aligned}\ ] ] the temperature derivatives of @xmath49 at constant @xmath75 and @xmath76 must be calculated from eq . ( [ ecane ] ) giving @xmath77 \;\ ; , \ ] ] @xmath78 .\end{aligned}\ ] ] finally , eq . ( [ ecvgena ] ) divided by @xmath79 yields the ideal part of the electronic specific heat @xmath80 \nonumber \\ & & + \ ; \frac{75}{8 } \lambda \left [ \frac{i_{5/2}}{i_{1/2 } } \ ; \frac{u^{\rm a}_{\lambda}}{v^{\rm a } } - \frac{3}{5 } \frac{i_{3/2}}{i_{-1/2 } } \ ; \frac{v^{\rm a}_{\lambda}}{v^{\rm a}_{\psi } } \right ] \nonumber \\ & & + \ ; \frac{105}{64 } { \lambda}^{2 } \left [ \frac{i_{7/2}}{i_{1/2 } } \ ; \frac{u^{\rm a}_{\lambda\lambda}}{v^{\rm a } } - \frac{15}{7 } \ ; \frac{i_{3/2}}{i_{-1/2 } } \ ; \frac{v^{\rm a}_{\lambda}}{v^{\rm a}_{\psi } } \ ; \frac{i_{3/2}}{i_{1/2 } } \ ; \frac{v^{\rm a}_{\lambda}}{v^{\rm a } } \right ] , \end{aligned}\ ] ] where the abbreviations are given by @xmath81 @xmath82 @xmath83 @xmath84 @xmath85\ ] ] @xmath86\ ] ] @xmath87 from these relations we can read off immediately various asymptotics . the nonrelativistic ( @xmath88 ) and nondegenerate limit ( @xmath89 ) of eq . ( [ ecviida ] ) results to the well - known classical value @xmath90 furthermore in the nonrelativistic limit and at arbitrary degeneracy the comparison of eqs . ( [ estraina ] ) and ( [ ecviida ] ) yields @xmath91 _ case b _ describes the regime where the sommerfeld - chandrasekhar expansion ( chandrasekhar 1939 , cox & giuli 1968 ) becomes valid . the degeneracy parameter @xmath49 is related for the strong - degenerate limit by @xmath92 where @xmath93 is the fraction of the fermi- to the relativistic momentum . the sommerfeld - chandrasekhar expansion with @xmath59 provides ( cox & giuli 1968 ) @xmath94 @xmath95 with the abbreviations @xmath96 \end{aligned}\ ] ] and @xmath97 for the helmholtz free energy represented by eq . ( [ erelhelm ] ) the inversion @xmath68 for _ case b _ is already given by eq . ( [ ealphzo ] ) . in order to correct eq . ( [ ealphzo ] ) to the order @xmath59 analytical inversions have been performed by yakovlev & shalybkov ( 1989 ) and paper i. for the internal energy , compressibility , and coefficient of strain we get @xmath98\ ] ] @xmath99\ ] ] @xmath100 \;\ ; .\ ] ] neglecting the order @xmath101 in eq . ( [ einnb ] ) we get the chandrasekhar approximation for the internal energy ( chandrasekhar 1939 , cox & giuli 1968 , lamb 1974 , eliezer et al . 1986 , yakovlev & shalybkov 1989 ) @xmath102 \ ; .\end{aligned}\ ] ] the specific heat eq . ( [ ecbas ] ) can be calculated for _ case b _ via @xmath103_{v , n_{a } } \;\ ; .\end{aligned}\ ] ] the temperature derivatives of @xmath104 at constant @xmath75 and @xmath76 must be calculated from eq . ( [ ecbne ] ) @xmath105 @xmath106 \ ; . \end{aligned}\ ] ] finally , eq . ( [ ecvgenb ] ) divided by @xmath79 yields @xmath107 \right . \nonumber \\ & & \left . - \left ( \frac{t}{\alpha } \frac{{\rm d}\alpha}{{\rm d}t } \right)^{2 } \left [ 6 \left ( \sqrt{1+\alpha^{2}}-1 \right ) \left ( 1 + \frac { v^{\rm b}_{\alpha}}{v^{\rm b } } + \frac{1}{6 } \frac{v^{\rm b}_{\alpha\alpha}}{v^{\rm b } } \right ) \right . \nonumber \\ & & \left . \left . - \frac{3u^{\rm b}}{v^{\rm b } } + \frac{1}{4 } \frac{u^{\rm b}_{\alpha\alpha}}{v^{\rm b } } - \frac{\alpha^{2}}{(1+\alpha^{2})^{3/2 } } \left(1 + 2 ( 1+\alpha^{2 } ) \frac{v^{\rm b}_{\alpha}}{v^{\rm b } } \right)\right ] \right . \nonumber \\ & & \left . + \frac{t^{2}}{\alpha } \frac{{\rm d}^2\alpha}{{\rm d}t^2 } \left [ \left ( \sqrt{1+\alpha^{2}}-1 \right ) \left ( 3 + \frac{v^{\rm b}_{\alpha}}{v^{\rm b } } \right ) \right ] \right\ } \;\ ; . \end{aligned}\ ] ] here the abbreviations labeled by the superscript b are given by the simple derivatives with @xmath108 @xmath109 @xmath110 @xmath111 eq . ( [ ecviidb ] ) yields by neglecting the order @xmath101 ( chandrasekhar 1939 , eliezer et al . 1986 , yakolev & shalybkov 1989 ) @xmath112 which delivers the strong - relativistic ( @xmath113 ) @xmath114 and the nonrelativistic ( @xmath115 ) @xmath116 limiting laws . let us remark that both eqs . ( [ ecviida ] ) and ( [ ecviidb ] ) have in the degenerate , weak - relativistic limit a wide overlap , which we use to combine _ case a _ and _ b _ in order to realize numerical continuity ( see paper i ) . the lowest order exchange ( hartree - fock ) free energy for relativistic electrons is given by @xmath117 where @xmath118 is the relativistic hartree - fock integral ( kovetz et al . 1972 , paper i ) . the @xmath50-expansion in _ case a _ results in @xmath119 where @xmath120 \;.\end{aligned}\ ] ] parametrizations for the nonrelativistic hartree - fock integral @xmath121 are given by perrot & dharma - wardana ( 1984 ) and kalitkin & ritus ( 1986 ) ( see stolzmann & blcker 1996b ) . note , that for @xmath122 the following relation holds @xmath123 the exchange pressure @xmath124 is determined via the exchange gibbs energy ( cf . paper i ) @xmath125 where @xmath126 \end{aligned}\ ] ] for the exchange contribution of the internal energy , compressibility , and coefficient of strain we get @xmath127 @xmath128\ ] ] @xmath129 \ ; .\end{aligned}\ ] ] using ( [ ecvgena ] ) we obtain for the exchange part of the isochoric specific heat @xmath130 \right . \nonumber \\ & & \left . + \frac{75}{8 } \ ; \lambda \left[-\frac{i_{3/2}i_{-1/2}}{i_{1/2 } } \ ; \frac{v^{\rm a}_{\lambda}}{v^{\rm a } } \ ; \left ( \frac{w^{\rm a}_{\psi}}{v^{\rm a}_{\psi } } - \frac{1}{5 } \ ; \frac{w^{\rm a}_{\psi\lambda}}{v^{\rm a}_{\psi } } \right ) \right . \nonumber \\ & & \left . \quad\quad + \frac{3}{5 } \ ; \frac{i_{3/2}i_{-3/2}}{i_{1/2 } } \ ; \frac{v^{\rm a}_{\lambda}}{v^{\rm a}_{\psi } } \ ; \left ( 2 \frac{w^{\rm a}_{\psi\psi}}{v^{\rm a}_{\psi } } - \frac{v^{\rm a}_{\psi\psi}}{v^{\rm a}_{\psi } } \ ; \frac{w^{\rm a}_{\psi}}{v^{\rm a}_{\psi } } \right ) \right . \nonumber \\ & & \left . \left . \quad\quad\quad\quad\quad\quad\quad\quad\quad \quad\quad\quad + \frac{3}{5 } \ ; i_{1/2 } \ ; \frac{v^{\rm a}_{\psi\lambda}}{v^{\rm a}_{\psi } } \ ; \frac{w^{\rm a}_{\psi}}{v^{\rm a}_{\psi } } \right ] \right . \nonumber \\ & & \left . + \frac{225}{64 } \lambda^{2 } \left [ \frac{i_{3/2}^{2}i_{-3/2}}{i_{1/2}i_{-1/2 } } \frac{v^{\rm a}_{\lambda}}{v^{\rm a}{\psi } } \left(2 \frac{w^{\rm a}_{\psi\psi}}{v^{\rm a } } - \frac{v^{\rm a}_{\psi\psi}}{v^{\rm a}_{\psi } } \frac{w^{\rm a}_{\psi}}{v^{\rm a } } \right ) \right \nonumber \\ & & \left . \left . + \ ; 4 \frac{i_{3/2}i_{-3/2}}{i_{1/2 } } \left ( \frac{i_{1/2}}{i_{-1/2 } } \frac{v^{\rm a}_{\lambda}}{v^{\rm a}_{\psi } } \frac{v^{\rm a}_{\psi\lambda}}{v^{\rm a}_{\psi } } - \frac{7}{30 } \frac{i_{5/2}}{i_{3/2 } } \frac{v^{\rm a}_{\lambda\lambda}}{v^{\rm a}_{\psi } } \right ) \right ] \right]\end{aligned}\ ] ] with the relativistic corrections @xmath131\end{aligned}\ ] ] @xmath132\end{aligned}\ ] ] @xmath133\end{aligned}\ ] ] @xmath134 \ ; .\end{aligned}\ ] ] the sommerfeld - chandrasekhar expansion for the exchange free energy in _ case b _ ( salpeter & zapolsky 1967 , kovetz et al . 1972 , lamb 1974 ) delivers for @xmath59 ( see paper i ) @xmath135 where @xmath136 \nonumber \\ & & + \frac{\pi^{4 } \lambda^{4}}{18 } \left [ 1 - \frac{11}{10 \alpha^{2 } } \left ( 1 + \frac{37}{11 \alpha^{2 } } + \frac{63 b}{11 \alpha^{3 } } \right ) \right]\end{aligned}\ ] ] with @xmath137 . + with the gibbs energy ( see paper i ) @xmath138 it is possible to provide the exchange pressure @xmath124 again . in a straightforward manner , the exchange contribution for internal energy , compressibility , and coefficient of strain are represented by @xmath139\end{aligned}\ ] ] @xmath140 \end{aligned}\ ] ] @xmath141 \ ; .\end{aligned}\ ] ] the exchange specific heat can be formulated by @xmath142 \end{aligned}\ ] ] the temperature derivatives of @xmath104 in eq . ( [ ecvxb ] ) are given by ( [ edalpdt ] ) and ( [ edalp2dt ] ) . the derivatives of @xmath143 with respect to @xmath50 and @xmath104 for the exchange contributions are determined by @xmath144 @xmath145 @xmath146 the adiabatic temperature gradient ( [ enabla ] ) expressed by our dimensionless potentials ( [ efdim1 ] ) and ( [ efdim2 ] ) is given by @xmath147 with the isobaric specific heat @xmath148 the figs . [ fnab_h_he_o_ixr_t79 ] and [ fnab_c_ixr_t5_9 ] illustrate the adiabatic temperature gradient ( [ enabdiml ] ) of the ideal gas i m comparison with a gas where additionally the exchange term is taken into account . note , that the representation in figs . [ fnab_h_he_o_ixr_t79 ] and [ fnab_c_ixr_t5_9 ] for extremely high densities serves only to illustrate the asymptotics of the theoretical expressions . processes relevant to such high densities as , e.g. , pycnonuclear reactions or electron captures , are not taken into account . = 0.49 = 0.49 fig . [ fnab_h_he_o_ixr_t79 ] illustrates the course of @xmath149 for the elements hydrogen , helium and oxygen along a nonrelativistic and a weak - relativistic temperature . a comparison with the corresponding data for carbon from lamb ( 1974 ) is given in fig . [ fnab_c_ixr_t5_9 ] . note a disagreement for @xmath150 . this deviation could be caused by an erroneous constant for the exchange contribution used by kovetz et al . ( 1972 ) and lamb ( 1974 ) and is discussed in detail in paper i. a numerical procedure to guarantee a smooth transition between the approximations given by _ case a _ and _ case b _ is developed in paper i. the curves as seen in figs . [ fnab_h_he_o_ixr_t79 ] and [ fnab_c_ixr_t5_9 ] show how exchange effects vanish with increasing temperature . considering the isotherms in the high density region exchange effects are negligible . nevertheless eqs . ( [ efxb])-([ecvxb ] ) guarantee the accurate asymptotics , which is a fundamental problem for the calculation of eos derivatives for quantities with vanishingly small temperature dependence ( miralles & van riper 1996 ) . according to eq . ( [ eicsredu ] ) we have to determine the terms labeled by the superscript c , which are based on correlation effects . as in paper i we aim to present closed - form parametrizations formed by pad approximants to get explicite expressions for the correlation contributions of the thermodynamical potentials . the thermodynamical potentials resulting from the correlation contributions are characterized by the electronic and ionic coulomb - coupling parameters @xmath151 and @xmath152 giving the strength of the coulomb interaction , and by the degeneracy parameter @xmath153 describing the quantum state of the system . the parameter @xmath154 is the ratio of the mean interelectronic distance to the ( electronic ) bohr - radius . the pad approximants for free helmholtz- and gibbs energies of the nonrelativistic electronic subsystem at arbitrary degeneration and coulomb coupling have the general structure @xmath155 / \tau } { 1 + a_{1 } \gamma_{\rm e}^{3/2 } + a_{2 } \gamma_{\rm e}^{6 } } \;\ ; , \ ] ] @xmath156 / \tau } { 1 + s_{1 } \gamma_{\rm e}^{3/2 } + s_{2 } \gamma_{\rm e}^{6 } } \;\ ; , \ ] ] with ground - state energies @xmath157 and @xmath158 derived by vosko et al . ( 1980 ) and low - temperature corrections @xmath159 and @xmath160 , which are given in paper i. the coefficients read @xmath161 @xmath162 @xmath163 \ ; , \quad\quad s_{1 } = \frac{8}{9 } a_{1}\ ] ] @xmath164 note that we have modified the coefficients @xmath165 and @xmath166 as compared to paper i by a rearrangement of the quantum virial function ( ebeling et al . 1976 ) @xmath167\ ] ] in paper i we incorporated for @xmath168 in eq . ( [ ea1kstar ] ) only the first part of eq . ( [ equavirf ] ) as proposed by ebeling ( 1993 ) . the second term was considered by the coefficient @xmath169 of the ion - electron contribution ( see eq . ( 85 ) in paper i ) . the justification to rearrange these summations over the electrons is the omission of the coefficient @xmath170 in eq . ( 53 ) in paper i , which has only been introduced to optimize the interpolation . for @xmath171 and @xmath172 we derived pad approximations as given in paper i with improvements to achieve higher accuracy . @xmath173 @xmath174 with @xmath175 and @xmath176 . a detailed comparison of the pad formula ( [ efeec ] ) with most advanced many body calculations is given in paper i and recently in stolzmann & rsler ( 1998b ) . = 0.49 the electronic correlation contribution for the pressure is obtained by @xmath177 the internal energy can be described by the pad approximant @xmath178 / \tau } { 1 + u_{1 } \gamma_{\rm e}^{3/2 } + u_{2 } \gamma_{\rm e}^{6 } } \;\ ; , \ ] ] with the coefficients @xmath179 @xmath180 the temperature derivative of the quantum virial function ( [ equavirf ] ) is given by @xmath181 \;\ ; .\end{aligned}\ ] ] for the temperature derivatives of @xmath171 and @xmath172 we derive the pad approximants @xmath182 @xmath183 the correlation contribution @xmath184 for the bulk modulus is given by @xmath185 / \tau } { 1 + k_{1 } \gamma_{\rm e}^{3/2 } + k_{2 } \gamma_{\rm e}^{6}}\ ] ] with the coefficients @xmath186 the ground - state bulk modulus is given by @xmath187 using the monte - carlo data - fit from vosko et al . ( 1980 ) for @xmath188 ( see paper i ) we apply @xmath189\ ] ] the coefficient of strain @xmath190 can be expressed by @xmath191 / \tau } { 1 + \phi_{1 } \gamma_{\rm e}^{3/2 } + \phi_{2 } \gamma_{\rm e}^{6}}\end{aligned}\ ] ] with the coefficients @xmath192 @xmath193 and for the isochoric specific heat @xmath194 the pad approximant can be obtained by @xmath195 / \tau } { 1 + v_{1 } \gamma_{\rm e}^{3/2 } + v_{2 } \gamma_{\rm e}^{6}}\ ] ] with the coefficients @xmath196 \ ; , \quad\quad\quad\quad\quad v_{2 } = a_{2 } \;,\ ] ] @xmath197 \;\ ; .\end{aligned}\ ] ] fig . [ fcv_ee_3thel ] shows that the electronic correlation part of the isochoric specific heat can be neglected in the case of strongly degenerate electrons . the classical one - component plasma ( ocp ) is a thoroughly studied plasma system and the results are unified in closed - form parametrizations ( hansen 1973 , graboske et al . 1975 , brami et al . 1979 , ebeling & richert 1985 , ichimaru 1993 , kahlbaum 1996 , chabrier & potekhin 1998 ) . for the ion - ion correlation we modify the pad approximants used in paper i by the new formula ( stolzmann & blcker 1998 , 1999 ) for the free energy @xmath198 + b_{2 } \gamma_{\rm i}^{6}\varepsilon_{\rm ii}(\gamma_{\rm i } ) } { 1 - b_{1 } \gamma_{\rm i}^{3 } g(\gamma_{\rm i } ) + b_{2 } \gamma_{\rm i}^{6 } } \ ; , \ ] ] with @xmath199 and @xmath200 @xmath201 @xmath202 and the coefficients @xmath203 @xmath204 eq . ( [ efiic ] ) is based on the classical @xmath205 result from cohen & murphy ( 1969 ) and for @xmath206 we take into account the most recent monte - carlo fit for the free energy @xmath207 of the liquid ocp from dewitt & slattery ( 1999 ) @xmath208 we remark , that in our notation of eq . ( [ eicsredu ] ) for @xmath209 we have to take for @xmath210 @xmath211 which , with eq . ( [ eionid ] ) rewritten by @xmath212 delivers the thermal ( solid phase ) energy of the ocp given by hansen ( 1972 ) , pollock & hansen ( 1973 ) , slattery et al . ( 1980 ) , and stringfellow et al . the quantity @xmath213 denotes the ionic rydberg energy with the ( electronic ) rydberg unit @xmath214 . the free gibbs energy @xmath215 and the isochoric specific heat @xmath216 are expressed by the pad approximants @xmath217 + b_{2 } \gamma_{\rm i}^{6 } \mu_{\rm ii}(\gamma_{\rm i } ) } { 1 - \frac{5}{3 } b_{1 } \gamma_{\rm i}^{3 } \left(g(\gamma_{\rm i } ) + \frac{2}{15 } \right ) \left [ 1 + \frac{3}{4 } \gamma_{\rm i}^{3/2 } \right ] + b_{2 } \gamma_{\rm i}^{6}}\ ] ] @xmath218 + 6 b_{2 } \gamma_{\rm i}^{6 } \varrho_{\rm ii}(\gamma_{\rm i } ) } { 1 - 7 b_{1 } \gamma_{\rm i}^{3 } \left(g(\gamma_{\rm i } ) + \frac{32}{63 } \right ) + 6 b_{2 } \gamma_{\rm i}^{6}}\ ] ] with @xmath219 the pressure , internal energy , isothermal compressibility , and coefficient of strain ( hansen 1973 , stolzmann & blcker 1999 ) are calculated by @xmath220 @xmath221 @xmath222 @xmath223 = 0.49 fig . [ fk_ii_gam ] compares our pad approximants of the isothermal compressibility with the fit formulae from hansen ( 1973 ) . further comparisons between different interpolation formulae for the classically ionic subsystem , e.g. from hansen ( 1973 ) , ichimaru ( 1994 ) , kahlbaum ( 1996 ) , chabrier & potekhin ( 1998 ) can be found in stolzmann & blcker ( 1999 ) . ionic quantum effects , which were included approximately in our earlier pad approximants for the ionic subsystem ( see paper i ) will be considered separately in the following section . the classical description for the subsystem of ionic ocp becomes inaccurate in the region of high densities at moderate temperatures . a measure for the quantum effects of the ions is the parameter ( weight fraction @xmath224 , molecular weight @xmath225 , proton mass @xmath226 , atomic mass unit @xmath227 ) @xmath228 which is related by @xmath229 and whereas @xmath230 denotes the ( ionic ) plasma frequency and @xmath231 is the ratio of the ( ionic ) wigner - seitz - radius to the ( ionic ) bohr - radius . = 0.49 = 0.49 quantum corrections for a solid ocp over a broad region of @xmath232 were calculated first by pollock & hansen ( 1973 ) . chabrier et al . ( 1992 ) ( see also chabrier 1993 ) and iyetomi et al . ( 1993 ) derived an free energy for a solid ocp model with quantum corrections . nagara et al . ( 1987 ) calculated quantum corrections for the free energy of the ocp valid for fluid and solid phases by a new expansion method . in our model ionic quantum effects according to nagara et al . ( 1987 ) as well as chabrier et al . ( 1992 ) will be included . following paper i , the free energy calculated within the so - called `` sixth reduced moment approximation '' of nagara et al . ( 1987 ) is given by @xmath233\ ] ] with @xmath234 . generally , the reduced moments @xmath235 are dependent of @xmath236 . for simplicity we choose for the solid phase the values for the rigid bcc lattice ( nagara et al . 1987 ) : @xmath237 the functions @xmath238 are given by @xmath239 @xmath240 @xmath241 @xmath242 = 0.49 = 0.49 the free energy expansion ( [ efiicq ] ) reproduces the wigner - kirkwood expansion ( wigner 1932 , kirkwood 1933 ) up to the order @xmath243 . furthermore the expansion of nagara et al . ( 1987 ) delivers the high-@xmath232 asymptotics of the coulomb lattice by @xmath244 with @xmath245 . for the harmonic lattice approximation one obtains @xmath246 ( hansen & vieillefosse 1975 ) . + the ionic quantum effects for our set of potentials are given by the resulting formulae @xmath247 \nonumber \\ & & \quad\quad\quad\quad\quad\quad\quad\quad\quad\quad\quad \ ; -\frac{3}{8}\frac{t}{\gamma_{\rm i } } z_{5}q_{5}(t ) \end{aligned}\ ] ] @xmath248 @xmath249 @xmath250\ ] ] @xmath251\ ] ] @xmath252\ ] ] the functions @xmath253 and @xmath254 are first and second derivatives of the functions @xmath238 with respect to @xmath255 . although the expansion of nagara et al . ( 1987 ) can be applied to estimate quantum effects for fluid ionic phases as well as for solid phases this expansion is restricted to @xmath256 , which is equivalent to @xmath257 or to @xmath258 for the density - temperature regime . for @xmath259 it is convenient to use the quantum crystal models from iyetomi et al . ( 1993 ) or from chabrier et al . ( 1992 ) , which are applicable for a wide region of @xmath232 . + if we extract the ionic quantum corrections for the free energy of a crystal from chabrier et al . ( 1992 ) one obtains ( stolzmann & blcker 1994 ) @xmath260 + \ln\left[1-{\rm e}^{-\gamma \theta_{\rm i}}\right ] \nonumber \\ & & \quad\quad\quad\quad\quad\quad + \beta \theta_{\rm i } - 3 \ln \left(\delta\theta_{\rm i}\right ) \end{aligned}\ ] ] with ( chabrier 1993 ) @xmath261 the last term in ( [ efiicqcad ] ) with @xmath262 guarantees the pure ionic quantum correction in our formalism . @xmath263 is the debye integral , which can be approximated by ( see e.g. iben & tutukov 1984 ) @xmath264 with ( [ efiicqcad ] ) our set of potentials is given by @xmath265 @xmath266 @xmath267 @xmath268 \nonumber \\ & & \quad\quad + \frac{3}{2 } \ ; \frac{\alpha\theta_{\rm i } { \rm { exp}}(-\alpha\theta_{\rm i } ) } { 1 - { \rm { exp}}(-\alpha\theta_{\rm i})}\end{aligned}\ ] ] @xmath269 @xmath270 an explicite expression for ( [ ecvcqiiws ] ) is given by chabrier ( 1993 ) . [ ff_qc_ii_t6]-[fcv_r75_c_nnn_cad ] compare the ionic quantum corrections for the free energy as well as the isochoric specific heat calculated by different approaches . note , the end of the solid lines in the figs . [ ff_qc_ii_t6]-[fcv_r75_c_nnn_cad ] reflects the validity of the moment expansion as pointed out by nagara et al . [ ff_qc_ii_t6 ] shows ionic quantum corrections for @xmath271 calculated with different theories . ionic quantum corrections in the vicinity of phase transition at @xmath272 are shown in fig . [ ff_qc_ii_g180 ] . [ ff_qc_ii_g180 ] illustrates that the quantum corrections calculated by iyetomi et al . ( 1993 ) decrease with increasing @xmath232 . this behaviour is caused by the inclusion of the anharmonic contribution of the zero - temperature oscillations ( see e.g. nagara et al . 1987 , iyetomi et al . 1993 ) , which is neglected by chabrier et al . [ fcv_cqii_r75_c ] shows the ionic quantum corrections for the isochoric specific heat derived from the corresponding free energies . [ fcv_r75_c_nnn_cad ] illustrates the course of the electronic and ionic contributions of the isochoric specific heat vs. temperature and shows how different prescriptions for the quantum corrections affect the results . the discontinuities in the curves for the ionic contribution in fig . [ fcv_r75_c_nnn_cad ] refer to the fluid - solid phase transition at @xmath272 . next we have to include the screening effect between electrons and ions . we adopt the new pad approximant based on the unsld - berlin - montroll asymptotics of stolzmann & ebeling ( 1998 ) and ebeling et al . ( 1999 ) @xmath273 the weak coupling limit ( @xmath274 ) is given here by the debye- hckel two component plasma ( dhtcp ) law . the coefficients @xmath275 are defined by @xmath276 = 0.49 = 0.49 @xmath277\end{aligned}\ ] ] @xmath278 @xmath279 @xmath280 in eq . ( [ ecoeffc1 ] ) is summed up over @xmath281 in contrast to summations made in earlier approaches ( ebeling 1993 , paper i ) . = 0.49 for @xmath282 we apply the expression given by ebeling & richert ( 1985 ) , which consider the high density / large coupling - asymptotics from galam & hansen ( 1976 ) including high - density corrections according to kagan et al . @xmath283\ ] ] @xmath284 with the relativity parameter @xmath285 . @xmath286 as well as @xmath287 , @xmath288 and @xmath289 depending on z are given by galam & hansen ( 1976 ) . recently , chabrier & potekhin ( 1998 ) presented a pad formula for the screening contribution of the free energy , which includes the thomas - fermi approximation in the limit of high densities and large coupling parameters . the results of chabrier & potekhin ( 1998 ) with eq . ( [ efiec ] ) are compared for hydrogen and carbon at five densities in fig . [ ffiec_com_rs_h ] and fig . [ ffiec_com_rs_c ] . + in the following we proceed in the same manner as in the case of the electron - electron subsystem in sect . [ eesec ] . + for the ion - electron gibbs energy we use the pad expression @xmath290 with @xmath291 and @xmath292 . fig . [ fgiec_com_rs_c ] illustrates the gibbbs energy for carbon at five densities . the polarization pressure can be determined by @xmath293 . = 0.49 = 0.49 = 0.49 = 0.49 = 0.49 = 0.49 = 0.49 = 0.49 = 0.49 this section will be devoted to illustrate the behaviour of the adiabatic temperature gradient @xmath149 . the adiabatic temperature gradient is a highly sensitive quantity , because it depends on first and second - order derivations of the model helmholtz free energy . therefore it is well suited to test the quality of the explicit expressions derived in sects . 3 and 4 . the adiabatic gradient is a crucial ingredient for many applications . for instance , @xmath149 determines the schwarzschild criterion for convective instability and an accurate evaluation is required for evolutionary calculations of very - low mass stars . we restrict our comparisons of the adiabatic temperature gradient to the eos tables given by straniero ( 1988 ) , saumon et al . ( 1995 ) , and rogers et al . further detailed comparisons can be found in saumon et al . ( 1995 ) who compared their results for @xmath149 with those of fontaine et al . ( 1977 ) , dppen et al . ( 1988 ) , and magni & mazzitelli ( 1979 ) . [ fcv_he_6 ] shows the total isochoric specific heat according to eq . ( [ eicsredu ] ) as well as their various single contributions ( in units of @xmath294 ) , i.e.electronic exchange and correlation terms ( electronic , ionic and screening part ) . we note that the discontinuity in the ionic correlation is caused by the fluid - solid phase transition at @xmath295 ( stringfellow et al . 1990 ) , which can also be observed for the adiabatic temperature gradient in the high - density region . the influence of quantum effects on the fluid - solid transition of the ocp is discussed in detail by pollock & hansen ( 1973 ) , lamb & van horn ( 1975 ) , nagara et al . ( 1987 ) , chabrier ( 1993 ) and iyetomi et al . we chose for our numerical calculations @xmath295 , which is based on an simulated ocp without quantum effects ( stringfellow et al . the specific heat contribution according to the ionic correlation is monotonously increasing until the phase transition of the ocp model . the first maximum at @xmath296 in the dashed curve in fig . [ fcv_he_6 ] is determined essentially by the electronic exchange and electronic correlation . the influence of electronic exchange and correlation increases with decreasing temperatures . quantum effects for the specific heat according to the ions after chabrier ( 1993 ) are illustrated by the dotted line in fig . [ fcv_he_6 ] , which can be seen also in figs . [ fcv_cqii_r75_c ] and [ fcv_r75_c_nnn_cad ] . this contribution is compensated at very high densities by the contributions of the ideal and the correlated ions . considering ionic quantum corrections calculated by nagara et al . ( 1987 ) the isochoric specific heat will be more strongly reduced as shown in fig . [ fcv_he_6 ] . we point out that the same behaviour appears for the adiabatic temperature gradient as indicated in figs . [ fnab_limes2 ] and [ fnab_554_he_id]-[fnab_7_c ] . note , that although the ionic quantum correction calculated by chabrier et al . ( 1992 ) and chabrier ( 1993 ) are valid only for the solid phase we considered here also the fluid range in order to illustrate the differences to the result of nagara et al . [ fnab_limes2]-[fnab_c50o50_th3 ] display isotherms of the adiabatic temperature gradient evaluated on the basis our analytic expressions given in sect . [ thermppot ] including the photonic contribution ( figs . [ fnab_6_he]-[fnab_c50o50_th3 ] ) over a broad range of densities . as already mentioned , the representations towards very high densities aimed at the purpose to show the asymptotic behaviour of the potentials given by eq . ( [ eicsredu ] ) . the accurate description of plasmas at very high density region requires the consideration of additional effects as electron captures and other processes ( cf . shapiro et al . 1983 ) . [ fnab_limes2 ] shows the high density asymptotics of the adiabatic temperature gradient calculated by different approximations . considering ideal and exchange contributions we obtained the well known value @xmath297 , according to relativistic considerations . the limiting value @xmath298 is determined by inclusion of the classical ionic correlation and @xmath299 is obtained with ionic quantum corrections given by chabrier et al . ( 1992 ) . in fig . [ fnab_554_h_id]-[fnab_7_c ] we compare our formalism with data from other authors ( lamb 1974 , fontaine et al . 1977 , straniero 1988 , saumon et al . 1995 , rogers et al . 1996 ) based on analytical as well as numerical methods and valid for various density and temperature ranges . the radiation contribution is dominant at high temperatures and low densities . the well - known limiting case of @xmath300 in the low - density limit depends on the temperature in a strong manner . the largest deviations for the adiabatic temperature gradient are observed in the intermediate density region , that means for plasma parameter where e.g. correlation effects must be described by accurate expressions . it should be emphasized that the pad technique is well suited to meet the challenging requirements of this - sometimes poorly known - density regime . [ fnab_rad_he_th3 ] shows the adiabatic temperature gradient for helium . in fig . [ fnab_c50o50_th3 ] we draw @xmath149 for a mixture of carbon and oxygen at high temperatures as appropriate for white dwarf interiors . the temperatures and densities are chosen here according to the conditions for a fully ionized plasma . nevertheless , our eos - formalism can be applied for the conditions of a partially ionized plasma ( lower temperatures ) provided additional terms in eq . ( [ eicsredu ] ) are considered and are solved in connection with the formalism of the dissoziation- and ionization equilibrium ( beule et al . 1999 ) . our aim was to provide data for the thermodynamical potentials summarized in eq . ( [ eicsredu ] ) which are applicable to , e.g. , stellar interiors ( fully ionized regions ) . for the helmholtz free energy , the gibbs free energy , and the pressure ( equation of state ) , expressions for each term in eq . ( [ eicsredu ] ) , i.e. ideality , exchange and correlation , have already been presented in paper i. in this paper we have improved our pad approximants for the correlation contributions and continued to present formulae for the explicit calculation of the internal energy , reciprocal compressibility ( bulk modulus ) , coefficient of strain , and the specific heat . further results using our calculational concept for the adiabatic temperature gradient for fully ionized plasmas are presented by stolzmann & blcker ( 1998 ) . the extension of our formalism to the partially ionized region is in progress . one of us ( w.s . ) acknowledges funding by the _ deutsche forschungsgemeinschaft _ ( grant scho 394/18 ) . beule d. , ebeling w. , frster a. , juranek h. , nagel s. , redmer r. , rpke g. , 1999 , phys . b59 , 14177 blinnikov s.i . , dunina - barkovskaya n.v . , nadyozhin d.k . , 1996 , apjs 106 , 171 brami b. , hansen j.p . , joly f. , 1979 , physica 95a , 505 chabrier g. , 1993 , apj . 414 , 695 chabrier g. , ashcroft n.w . , dewitt h.e . , 1992 , 360 , 48 chabrier g. , potekhin a.y . , 1998 , phys . e58 , 4941 chandrasekhar s. , 1939 , _ an introduction to the study of stellar structure _ , university of chicago press , chicago cohen e.g.d . , murphy t.j . , 1969 , fluids 12 , 1404 cox j.p . , giuli r.t . , 1968 , _ principles of stellar structure _ , vol.1 & 2 , gordon and breach , new york dppen w. , hummer d.g . , mihalas d. , weibel - mihalas b. , 1988 , apj . 332 , 261 dewitt h.e . , slattery w.l . , 1999 , plasma phys . 39 , 97 ebeling w. , kraeft w.d . , kremp , d. , 1976 , _ theory of bound states and ionization equilibrium in plasmas and solids _ , akademie verlag , berlin ebeling w. , 1993 , contrib . plasma phys . 33 , 492 ebeling w. , richert w. , 1985 , phys . ( b ) , 128 , 467 ebeling w. , stolzmann w. , frster a. , kasch m. , 1999 , contrib . plasma phys . 39 , 287 eggleton p.p . , faulkner j. , flannery b.p . , 1973 , a&a 23 , 325 eliezer s. , ghatak a. , hora h. , 1986 , _ an introduction to equation of state : theory and applications _ , cambridge university press , cambridge fontaine g. , graboske h.c . , van horn h.m . , 1977 , apjs 35 , 293 galam s. , hansen j.p . , 1976 , a14 , 816 graboske h.c . , olness r.j . , grossman a.s . , 1975 , 199 , 255 hansen j.p . , 1972 , phys . 41a , 18 , 213 hansen j.p . , 1973 , a8 , 3096 hansen j.p . , vieillefosse p. , 1975 , phys . lett . 53a , 187 iben i. , tutukov v.a.,1984 , apj . 282 , 615 ichimaru s. , 1993 , rev . 65 , 255 ichimaru s. , 1994 , _ statistical plasma physics _ , vol . ii : condensed plasmas , addison - wesley publishing company , reading , massachusetts iyetomi h. , ogata s. , ichimaru s. , 1993 , phys . rev . b47 , 11703 johns s.m . , ellis p.j . , lattimer j.m . , 1996 , apj 473 , 1020 kagan yu.m . , pushkarev v.v . , kholas a. , 1977 , zh . 73 , 967 kahlbaum t. , 1996 , in _ physics of strongly coupled plasmas _ , eds . w.d . kraeft and m. schlanges , world scientific , singapore , p. 99 kalitkin n.n . , ritus i.v . , 1986 , zh . vychislit . mat . fiz . 3 , 461 kirkwood j.g . , 1933 , phys . rev . 44 , 31 kovetz a. , lamb d.q . , van horn h.m . , 1972 , apj 174 , 109 lamb d.q . , 1974 , thesis , university of rochester lamb d.q . , van horn h.m . , 1975 , apj 200 , 306 magni g. , mazzitelli t. , 1979 , astron . astrophys . 72 , 134 miralles j.a . , van riper k.a . , 1996 , apjs 105 , 407 nagara h. , nagata y. , nakamura t. , 1987 , phys . a36 , 1859 perrot f. , dharma - wardana m.w.c . , 1984 , a30 , 2619 pollock e.l . , hansen j.p . , 1973 , phys . a8 , 3110 rogers f.j . , swenson f.j . , iglesias c.a . , 1996 , apj 456 , 902 salpeter e.e . , zapolsky h.s . , 1967 , 158 , 876 saumon d. , chabrier g. , van horn h.m . , 1995 , apjs 99 , 713 shapiro s.l . , teukolsky s.a . , 1983 , _ black holes , white dwarfs , and neutron stars _ , wiley , new york slattery w.l . , doolen g.d . , dewitt h.e . , 1980 , a21 , 2087 stolzmann w. , blcker t. , 1994 , in _ the equation of state in astrophysics _ , eds . g. chabrier and e. schatzman , cambridge university press , cambridge , p. 606 stolzmann w. , blcker t. , 1996a , a&a 314 , 1024 ( paper i ) stolzmann w. , blcker t. , 1996b , in _ physics of strongly coupled plasmas _ , eds . w.d . kraeft and m. schlanges , world scientific , singapore , p. 95 stolzmann w. , blcker t. , 1998 , in _ strongly coupled coulomb systems _ , eds . kalman , j.m . rommel , and k. blagoev , plenum press , new york , p. 265 stolzmann w. , rsler m. , 1998 , in _ strongly coupled coulomb systems _ , eds . kalman , j.m . rommel , and k. blagoev , plenum press , new york , p. 673 stolzmann w. , ebeling w. , 1998 , phys . lett . a 248 , 242 stolzmann w. , blcker t. , 1999 , contrib . plasma phys . 39 , 105 straniero o. , 1988 , a&a 76 , 157 stringfellow g.s . , dewitt h.e . , slattery w.l . , 1990 , a41 , 1105 . vosko s.h . , wilk l. , nusair m. , 1980 , can . 58 , 1200 yakovlev d.g . , shalybkov d.a . , 1989 , e. astrophys . space phys . 7 , harwood academic publishers , london wigner e.p . , 1932 ,

a consistent combination of covariant density functional theory ( cdft ) and landau - migdal theory of finite fermi systems ( tffs ) is presented . both methods are in principle exact , but landau - migdal theory can not describe ground state properties and density functional theory does not take into account the energy dependence of the self - energy and therefore fails to yield proper single - particle spectra as well as the coupling to complex configurations in the width of giant resonances . starting from an energy functional , phonons and their vertices are calculated without any further parameters . they form the basis of particle - vibrational coupling leading to an energy dependence of the self - energy and an induced energy - dependent interaction in the response equation . a subtraction procedure avoids double counting . applications in doubly magic nuclei and in a chain of superfluid nuclei show excellent agreement with experimental data .

0909.1276

the understanding of the structure of nuclei far from stability with extreme isospin is one the most exciting challenges of present nuclear physics . new experimental facilities with radioactive nuclear beams make it possible to investigate the nuclear chart to the very limits of nuclear binding . a wealth of structure phenomena in exotic nuclei have been reported and the next generation of radioactive - beam facilities will present new exciting opportunities to study not only the ground states but also excitations and spectra of these strongly interacting many - body systems . this situation has stimulated considerable new efforts on the theoretical side to understand the dynamics of the nuclear many - body problem by microscopic methods . exact solutions of the non - relativistic schrdinger equation based on the bare nucleon - nucleon interaction are used to study very light nuclei with @xmath0 by an ab initio " approach , modern shell - model calculations based on large scale diagonalization techniques and truncation schemes show considerable success in situations where configuration mixing calculations are possible , i.e. in light nuclei or in nuclei with single magic or doubly magic configurations . for the large majority of nuclei , however , a quantitative microscopic description is , so far , only possible by density functional theory ( dft ) and its extensions . although dft can , in principle , provide an exact description of many - body problems@xcite , if the exact density functional is known , in nuclear physics one is far from a microscopic derivation of this functional . in addition , nuclei are self - bound systems . as a consequence of translational invariance the density in the laboratory frame is constant in space . density functional theory in nuclei is therefore based on the intrinsic density , a concept that requires additional approximations @xcite . the most successful schemes of dft in nuclei use a phenomenological ansatz incorporating as many symmetries of the system as possible and adjusting the parameters of the functional to ground state properties of a few characteristic nuclei on the nuclear chart . considerable progress has been reported recently in constructing such functionals . for a recent review see @xcite . one of the underlying symmetries of qcd is lorentz invariance and therefore covariant density functionals @xcite are of particular interest in nuclear physics . this symmetry not only allows to describe the spin - orbit coupling , which has an essential influence on the underlying shell structure , in a consistent way , but it also puts stringent restrictions on the number of parameters in the corresponding functionals without reducing the quality of the agreement with experimental data . most of the nuclei are superfluid systems and therefore the inclusion of pairing correlations is essential for a correct description of structure phenomena in open - shell nuclei @xcite . hartree - bogoliubov theory provides a unified description of @xmath1- and @xmath2-correlations on a mean - field level by using two type of densities , the normal density matrix @xmath3 and antisymmetric pairing tensor @xmath4 @xcite . according to valatin these two densities can be combined to the generalized density matrix @xmath5 of double dimension @xcite . cdft theory for superfluid systems is therefore based on a generalized relativistic hartree - bogoliubov ( rhb ) energy density @xmath6 $ ] . the same is true for the landau - migdal theory and for all the methods discussed in this paper . for simplicity , however , we restrict all our considerations in this article to the case without pairing correlations . pairing correlations can be included on all steps by using super - matrices . details are given in ref . @xcite . only in the applications we present also calculations in isotopic chains of open shell nuclei , that include pairing correlations . a very successful example of a covariant density functional theory is the relativistic hartree - bogoliubov model @xcite . it combines a density dependence through a non - linear coupling between the meson fields @xcite with pairing correlations based on an effective interaction of finite range . a large variety of nuclear phenomena have been described over the years within this model : the equation of state in symmetric and asymmetric nuclear matter , ground state properties of finite spherical and deformed nuclei all over the periodic table @xcite from light nuclei @xcite to super - heavy elements @xcite , from the neutron drip line , where halo phenomena are observed @xcite to the proton drip line @xcite with nuclei unstable against the emission of protons @xcite . in principle density functional theory can be used for the description of all properties depending on the single - particle density . it is therefore not only limited to the description of the ground state properties . the same density functionals have also been applied for a very successful description of excited states , such as rotational bands in normal and super - deformed nuclei @xcite and collective vibrations @xcite . rotations are treated in the cranking approximation providing a quasi - static description of the nuclear dynamics in a rotating frame and for the description of vibrations a time - dependent mean field approximation is used by assuming independent particle motion in time - dependent average fields @xcite . in the small amplitude limit one obtains the relativistic random phase approximation ( rrpa ) @xcite and in superfluid nuclei the relativistic quasiparticle random phase approximation ( rqrpa ) @xcite . this method provides a natural framework to investigate collective and non - collective excitations of @xmath1- ( or @xmath7 ) character . it is successful in particular for the understanding of the position of giant resonances and spin- or / and isospin - excitations as the gamov teller resonance ( gtr ) or the isobaric analog resonance ( iar ) . recently it has been also used for a theoretical interpretation of low lying e1-strengths observed in neutron rich isotopes ( pygmy modes ) @xcite and for low - lying collective quadrupole excitations @xcite . density functional theory in nuclei is based on intrinsic densities and on the mean field approach . therefore it can not provide an exact treatment of the full nuclear dynamics . it breaks down in transitional nuclei , where the intrinsic density is not well defined and where one has to include correlations going beyond the mean field approximation by treating quantum fluctuations through a superposition of several mean field solutions , as for instance in the generator coordinate method ( gcm ) @xcite . but it also provides a poor approximation for the single - particle spectra particularly in ideal shell model nuclei such as @xmath8pb with closed protons and neutron shells . one finds in self - consistent mean field calculations usually a considerably enhanced hartree - fock gap in the single - particle spectrum and a reduced level density at the fermi surface as compared with the experiment . it is well known that this fact is connected with the relatively small effective mass in such models . mahaux and collaborators @xcite have shown that the effective mass in nuclear matter is roughly @xmath9 @xmath10 . in finite nuclei it should be modified by the coupling of the single - particle motion to low - lying collective surface vibrations . this leads , in the vicinity of the fermi surface , to an enhancement of @xmath9 @xmath11 . non - self - consistent models with the bare mass ( @xmath9 @xmath11 ) show indeed a single - particle spectrum with a level density close to the experiment . with a few exceptions , where the quadrupole motion has been studied within the relativistic generator coordinate method ( gcm ) @xcite , applications of covariant density functional theory to the description of excited states are limited to relativistic rpa , i.e. to configurations of @xmath12-nature . none of these methods , however , can be applied to the investigation of the coupling to more complicated configurations , as it occurs for instance in the damping phenomena causing the width of giant resonances . already before density functional theory has been introduced in the sixties for the description of quantum mechanical many - body problems by kohn and sham @xcite landau has developed in the fifties his fermi liquid theory ( flt ) @xcite for infinite systems . it has been extended to the theory of finite fermi systems ( tffs ) @xcite by migdal . this theory provides another very successful method for the description of low - lying nuclear excitations @xcite . it has several general properties in common with density functional theory . first , both theories are know to be exact , at least in principle , but in practice , in nuclear physics , the parameters entering these theories have to be determined in a phenomenological way by adjustment to experimental data . second , both theories are based on a single - particle concept . dft uses the mean field concept with slater determinants in an effective single - particle potential as a vehicle to introduce shell - effects in the exact density functional introduced by hohenberg and kohn @xcite . fermi liquid theory is based on the concept of quasi - particles obeying a dyson equation , which are defined as the basic excitations of the neighboring system with odd particle number . third , in practical applications both theories describe in the simplest versions the nuclear excitations in the rpa approximation , i.e. by a linear combination of @xmath1-configurations in an average nuclear potential . however , there are also essential differences between these two concepts . first , in contrast to dft , tffs does not attempt to calculate the ground state properties of the many - body system , but it describes the nuclear excitations in terms of landau quasi - particles and their interaction . therefore the experimental data used to fix the phenomenological parameters of the theories are bulk properties of the ground state in the case of density functional theory , and properties of single - particle excitations and of the collective excitations in the case of finite fermi systems theory . second , in dft the mean field is determined in a self - consistent way and therefore the rpa spectrum contains goldstone modes at zero energy . this is usually not the case in tffs calculations , which are based on a non - relativistic shell - model potential , whose parameters are adjusted to the experimental single - particle spectra . therefore , apart from a few approximate attempts to treat the goldstone modes by adjusting additional parameters in the effective quasiparticle interaction , there is no self - consistency in the rpa calculations of tffs and the goldstone modes do not separate from the other modes . they are distributed among the low - lying excitations . third , modern versions of tffs go much beyond the mean field approximation . the coupling between the particles and the phonons is investigated with green s function techniques . based on the phonons calculated in the framework of the rpa one has included particle - phonon coupling vertices and an energy dependence of the self - energy in the dyson equation @xcite . this leads also to an induced interaction in the bethe - salpeter equation caused by the exchange of phonons which also depends on the energy . the coupling of particles and phonons has also been derived from nuclear field theory ( nft ) and its extensions @xcite . many aspects of the coupling between the quasi - particles and the collective vibrations have been investigated with these techniques @xcite as well as with other kinds of approaches beyond rpa @xcite over the years . we give here an overview over recent attempts @xcite to find a combination of the basic ideas of covariant density functional theory and landau - migdal theory and show as examples corresponding applications . the concept is similar to earlier work in refs . @xcite , where specific non - relativistic energy functionals have been used to construct a self - consistent theory of finite fermi systems . the starting point is a covariant density functional @xmath13 $ ] widely used in the literature . it is adjusted to ground state properties of characteristic nuclei and , without any additional parameter , it provides the necessary input of finite fermi systems theory , such as the mean field and the single - particle spectrum as well as an effective interaction between the @xmath1-configurations in terms of the second derivative of the same energy @xmath13 $ ] with respect to the density . thus the phenomenological input of landau - migdal theory is replaced by the results of density functional theory . the same interaction is used to calculate the vertices for particle - vibration coupling @xcite . in a second step techniques of landau - migdal theory and its modern extensions are used to describe the coupling of one- and two - quasiparticle configurations . the main assumption of the quasiparticle - phonon coupling model is that two types of elementary excitations two - quasiparticle and vibrational modes are coupled in such a way that configurations of @xmath14 type with low - lying phonons strongly compete with simple @xmath12 configurations close in energy or , in other words , that quasiparticles can emit and absorb phonons with rather high probabilities . in this way a fully consistent description of the many - body dynamics is obtained . as a result an induced additional interaction between single - particle and vibrational excitations provides a strong fragmentation of the pure rqrpa states causing the spreading width of giant resonances and the redistribution of the pygmy strength . two essential approximations are used in this context : first , the _ time - blocking approximation _ ( tba ) @xcite , that has been extended to systems with pairing correlations ( qtba ) in ref . @xcite , blocks in a special time - projection technique the @xmath12-propagation through states which have a more complex structure than @xmath14 . the nuclear response can then be explicitly calculated on the @xmath15 level by summation of an infinite series of feynman s diagrams . second , a special _ subtraction technique _ guarantees , that there is no double counting between the additional correlations introduced by particle - vibration coupling and the ground state correlations already taken into account in the phenomenological density functional . these two tools are essential for the success of this method . tba introduces a consistent truncation scheme into the bethe - salpeter equation and without it it would be hard to solve the equations explicitly . the subtraction method is the essential tool to connect density functional theory so far used only on the level of mean field theory , i.e. on the rpa level , with the extended landau - migdal theory , where complex configurations are included through particle - vibration coupling . the structure of the paper is the following : in section ii we discuss shortly the general formalism of covariant density functional theory , we introduce in section iii the concept of the energy - dependent self - energy @xmath16 and the vertices of particle - vibration coupling in the relativistic framework , and we discuss in section iv the response formalism , the time blocking approximation and the subtraction mechanism for the response function . in section v we present recent numerical applications for the calculation of level densities at the fermi surface and the spreading width of the several nuclei . section vi contains a brief summary and an outlook for future applications . covariant density functional theory uses the walecka model @xcite as a lorentz invariant framework for the formulation of the density functional . in this model the nucleus is described as a system of dirac nucleons coupled to the exchange mesons and the electromagnetic field through an effective lagrangian . the isoscalar scalar @xmath17-meson , the isoscalar vector @xmath18-meson , and the isovector vector @xmath19-meson build the minimal set of meson fields that together with the electromagnetic field is necessary for a quantitative description of bulk and single - particle nuclear properties @xcite . the model is defined by the lagrangian density @xmath20 @xmath21 denotes the lagrangian of the free nucleon @xmath22 where @xmath23 is the bare nucleon mass and @xmath24 denotes the dirac spinor . @xmath25 is the lagrangian of the free meson fields and the electromagnetic field @xmath26 with the corresponding masses @xmath27 , @xmath28 , @xmath29 , and @xmath30 , @xmath31 , @xmath32 are field tensors ( arrows denote isovectors and boldface symbols are for vectors in ordinary space ) . the minimal set of interaction terms is contained in @xmath33 @xmath34 with the vertices@xmath35 with the coupling constants @xmath36 , @xmath37 , @xmath38 and @xmath39 ( @xmath40 or @xmath41 for protons or neutrons ) . already in the earliest applications of the rmf framework it was realized , however , that this simple model with interaction terms only linear in the meson fields , does not provide a quantitative description of complex nuclear systems . an effective density dependence was introduced @xcite by replacing the quadratic @xmath42-potential @xmath43 with a non - linear meson coupling potential @xmath44 , which contains additional parameters . this particular form of the non - linear potential has become standard in applications of rmf models , although additional non - linear interaction terms , both in the isoscalar and isovector channels , have also been considered @xcite . > from the model lagrangian density the classical variation principle leads to the equations of motion:@xmath45 \psi=0 . \label{dirac0}%\ ] ] if one neglects retardation effects for the meson fields , which is well justified because of the large meson masses , a self - consistent solution is obtained when the time - dependent mean - field potentials @xmath46 are calculated at each step in time by the solution of the static klein - gordon equations @xmath47 where the ( @xmath48 ) sign is for vector fields and the ( @xmath49 ) sign for the scalar field . the index @xmath23 denotes mesons and the photon , i.e. @xmath50 , and @xmath51 is derivative of the corresponding potential with respect to the meson field . in applications to nuclear matter and finite nuclei , the relativistic models are used in the _ no - sea _ approximation , i.e. the dirac sea of states with negative energies does not contribute to the densities and currents and one uses @xmath52 where the sum runs only over the occupied states in the fermi sea , i.e. vacuum polarization effects are neglected . in fact , many effects that go beyond the classical mean - field level are apparently neglected in the this models : fock terms , vacuum polarization effects , and the short range brueckner - type correlations . the experimental data to which the meson - nucleon couplings are adjusted , however , contain all these effects and much more . it follows that effects beyond the mean - field level are implicitly included in the rmf approach by adjusting the model parameters to reproduce a selected empirical data set . vacuum effects , chiral symmetry , nucleon substructure , exchange terms , long- and short - range correlation effects are , therefore , effectively included in this approach although neither of them can be accessed separately . the set of coupled equations ( [ dirac0 ] ) and ( [ kg0 ] ) define the relativistic mean field ( rmf ) model . in the stationary case they reduce to a nonlinear eigenvalue problem and in the time - dependent case they describe the nonlinear propagation of the dirac spinors in time @xcite . rmf models can be also formulated without explicitly including mesonic degrees of freedom . meson - exchange interactions can be replaced by local four - point interactions between nucleons . it has been shown that the relativistic point - coupling models @xcite are completely equivalent to the standard meson - exchange approach . in order to describe properties of finite nuclei on a quantitative level , the point - coupling models include also some higher order interaction terms . for instance , six - nucleon vertices @xmath53 , and eight - nucleon vertices @xmath54 and @xmath55^{2}$ ] . these relatively simple models turn out to provide a very successful phenomenological description of the nuclear many - body system all over the periodic table . relatively few parameters are adjusted to ground state properties of a few finite nuclei . at a first glance it is not easy to see how can such a simple approach can be so successful . this can be only understood if one considers that this model represents an approximate implementation of kohn - sham density functional theory ( dft)@xcite , which is successfully employed in the treatment of the quantum many - body problem in atomic , molecular and condensed matter physics . it is evident that equations of motion ( [ dirac0 ] ) and ( [ kg0 ] ) can also be directly derived form a density functional . using the definition of the relativistic single - nucleon density matrix @xmath56 the total energy can be written as a functional of the density matrix @xmath57 and of the meson fields @xmath58=\text{tr}\left [ ( { \mbox{\boldmath $ \alpha$}}\mathbf{p}+\beta m)\hat{\rho}\right ] \pm\int\left [ \frac{1}{2}(\mathbf{\nabla}\phi_{m}% ) ^{2}+u(\phi_{m})\right ] d^{3}r+\text{tr}\left [ ( \gamma_{m}\phi_{m}% ) \hat{\rho}\right ] .\label{e - rmf}%\ ] ] the trace operation involves a sum over the dirac indices and an integral in coordinate space . the index @xmath23 is used as generic notation for all mesons and the photon . in a relativistic many - body theory the motion of single nucleons in the nuclear medium is described by the dyson equation @xmath59 where the self - energy is given by @xmath60 with the scalar part @xmath61 of the self - energy and where the generalized four - vector momentum operator has the form @xmath62 with the vector part @xmath63 of the self - energy @xmath64 the index @xmath65 in the eq . ( [ emass ] ) denotes that the effective mass is described by the scalar @xmath17-meson field . in order to characterize ground state properties the stationary dirac equation has to be solved : @xmath66 in the general case the full self - energy is non - local in space and also in time . this non - locality means that its fourier transform has both momentum and energy dependence . we therefore decompose the total self - energy in a stationary local part and an energy - dependent non - local term : @xmath67 where all the components of the self - energy are involved : @xmath68 and the index `` e '' indicates the energy dependence . the energy - independent parts of the self - energy correspond to the average fields of the walecka model : @xmath69 these fields satisfy the inhomogeneous klein - gordon equations , where the sources are determined by the respective density and current distributions in a system of a nucleons . we assume time - reversal symmetry that means the absence of currents in the nucleus and , thus , we find vanishing space - like components of @xmath70 . the equation of the one - nucleon motion has the form : @xmath71 where @xmath72 is the dirac hamiltonian @xmath73 * * or , in the language of green s functions @xmath74 it turns out to be useful to work in the shell - model dirac basis @xmath75 which diagonalizes the energy - independent part of the dirac equation : @xmath76 in this basis one can rewrite eq . ( [ fg ] ) as follows : @xmath77 where the letter indices @xmath78 denote full sets of the spherical quantum numbers . and the hole @xmath79 components of the relativistic self - energy in the graphical representation . solid and dashed lines with arrows denote one - body propagators for particle ( @xmath80 ) , hole ( @xmath81 ) , and antiparticle ( @xmath82 ) states . wavy lines denote phonon ( @xmath83 ) propagators , empty circles are the particle - phonon coupling vertices @xmath84 in eq . ( [ phonon ] ) . the time direction is from the left to the right . ] obviously , on this stage one needs some model assumptions . the particle - phonon coupling model @xcite provides a rather simple approximation to describe the energy dependence of @xmath85 . within this model @xmath85 is a convolution of the particle - phonon coupling amplitude @xmath86 and the exact single - particle green s function @xcite : @xmath87 where the amplitude @xmath88 is represented in terms of phonon vertexes @xmath84 and their frequencies @xmath89 . they are determined by the following relation : @xmath90 @xmath91 is the transition density and in the linearized version of the model @xmath92 denotes the relativistic matrix element of the static residual interaction , i.e. the second derivative of the energy functional with respect to the density matrix @xmath93}{\delta\hat{\rho}\delta\hat { \rho}}\label{vstatic}%\ ] ] in the linear approximation @xmath91 is not influenced by the particle - phonon coupling and can be computed within relativistic rpa or qrpa . the linearized version implies also that the energy - dependent part of the self - energy ( [ mo1 ] ) contains the mean field green s function @xmath94 instead of the exact green s function @xmath95 . the graphical representation of the self - energy is given in fig . in contrast to the non - relativistic case , where one has occupied states below the fermi surface ( hole states @xmath81 ) and empty states above the fermi surface ( particle states @xmath80 ) we now have according to the no - sea approximation in addition empty states with negative energies in the dirac sea ( anti - particle states @xmath82 ) . particle and hole components are drawn assuming all the possible types of intermediate states . ( [ mo ] ) contains off - diagonal elements of the matrix @xmath96 with relatively large energy denominators . it has been shown by explicit calculations within the non - relativistic approach @xcite that it is justified to use the diagonal approximation : @xmath97 thus , within the diagonal approximation of the self - energy ( [ diagonal ] ) the exact green s function @xmath95 is also diagonal in the dirac basis and the dyson equation forms for each @xmath98 a non - linear eigenvalue equation @xmath99 the poles of the green s function @xmath100 correspond to the zeros of the function @xmath101 in refs . @xcite it is shown how this problem can be solved by a matrix diagonalization . for each quantum number @xmath98 there exist several solutions @xmath102 characterized by the index @xmath103 . because of the coupling to the collective vibrations the single - particle state @xmath98 is fragmented as it will be shown in the application in section [ appl-1 ] . nuclear dynamics of an even - even nucleus in a weak external field is described by the linear response function @xmath104 , where @xmath105 combines the quantum numbers @xmath98 and the time . this response function is the solution of the bethe - salpeter equation ( bse ) in the @xmath1 channel : @xmath106 with the free response @xmath107 , where function @xmath95 is the exact single - particle green s function . the summation over the number indices @xmath108 , @xmath109 implies an integration over the respective time variables . and @xmath110 is the effective interaction irreducible in the @xmath1-channel . this interaction is determined as a variational derivative of the full self - energy @xmath111 with respect to the exact single - particle green s function : @xmath112 since the self - energy in eq . ( [ sig ] ) has two parts @xmath113 , the effective interaction @xmath110 in eq . ( [ bse ] ) is a sum of the static rmf interaction @xmath114 ( [ vstatic ] ) and time - dependent terms @xmath115 after a fourier transformation in time , this time dependence leads to an energy - dependent interaction @xmath116 . in the dirac basis ( [ dirac - basis ] ) it has the form : @xmath117 where @xmath118 for empty states and @xmath119 for occupied states ( for details see ref . @xcite ) . the bethe - salpeter equation ( [ bse ] ) contains the exact greens function @xmath95 . in order to simplify this equation for the further analysis the @xmath95 is expressed it in terms of the mean field green s function @xmath120 . from eq . ( [ mo ] ) we derive the nambu form for it : @xmath121 which reads in fourier space as@xmath122 introducing @xmath123 one can rewrite eq . ( [ bse ] ) as follows : @xmath124 where @xmath125 is a new interaction of the form @xmath126 with @xmath127 the graphical representation of the eq . ( [ bse1 ] ) is shown in fig . [ fig2 ] . -response function @xmath128 in graphical representation . details are given in fig . [ fig1 ] and the small black circle means the static part of the residual @xmath1-interaction ( [ vstatic ] ) . ] in addition to the static interaction @xmath114 the effective interaction @xmath125 contains diagrams with energy - dependent self - energies and an energy - dependent _ induced interaction _ , where a phonon is exchanged between the particle and the hole . as discussed in refs . @xcite the term @xmath129 has to neglected in the time blocking approximation if one neglects backward - going propagators caused by the particle - phonon coupling . this is a reasonable approximation applied and discussed in many non - relativistic models ( see e.g. refs . @xcite and references therein ) . we have to emphasize , however , that all the rpa ground state correlations are taken into account , because it is well known that they play a central role for the conservation of currents and sum rules . it turns out that both the solution eq . ( [ bse1 ] ) @xmath128 and its kernel @xmath125 are singular . another difficulty arises because eq . ( [ bse1 ] ) contains integrations over all time points in the intermediate states . this means that many configurations are contained in the exact response function which are actually more complex than @xmath14 . tselyaev has introduced in the ref . @xcite the time blocking approximation ( tba ) , a special time - projection technique to block the @xmath1-propagation through these complex intermediate states . in this way one obtains after a fourier transformation in time a relatively simple algebraic equation : @xmath130 where @xmath131 and @xmath132 @xmath133 is the mean field propagator : @xmath134 @xmath135 and @xmath136 is the particle - phonon coupling amplitude with the following components : @xmath137 , \label{phiph}%\end{aligned}\]]@xmath138 . \label{phiah}%\end{aligned}\ ] ] as in fig . [ fig1 ] the indices @xmath139 and @xmath81 denote the particles , antiparticles and holes in the dirac basis . the amplitudes @xmath140 , @xmath141 are neglected , because they have only are small effect ( see ref . the amplitudes @xmath142 and @xmath143 are also disregarded within this approximation . therefore , ground state correlations are taken into account only on the rpa level due to the presence of the @xmath144 , @xmath145 terms of the static interaction in the eq . ( [ respdir ] ) . by definition , the response function @xmath146 in eq . ( [ respdir ] ) contains only configurations which are not more complex than @xmath147 . in eq . ( [ respdir ] ) @xmath148 is subtracted from @xmath149 . this corresponds to the subtraction procedure developed by tselyaev in ref . it considers the fact that the effective interaction @xmath114 being adjusted to experimental data of the ground state contains effectively many correlations and , in particular , also admixtures of phonons at the energy @xmath150 . in the present method , all correlations entering through the admixture of phonons are taken care of by the additional interaction term@xmath151 . to avoid double counting in the effective interaction the part @xmath148 is therefore subtracted . this means only the energy dependence of the phonon coupling is effectively taken into account . to describe the observed spectrum of the excited nucleus in a weak external field @xmath152 , as for instance a dipole field , one needs to calculate the strength function : @xmath153 expressed through the polarizability @xmath154 a finite imaginary part @xmath155 of the energy variable is introduced in the calculations for convenience in order to obtain a more smoothed envelope of the spectrum . this parameter has the meaning of an additional artificial width for each excitation . this width emulates effectively contributions from configurations which are not taken into account explicitly in this approach . in order to calculate the strength function it is convenient to convolute eq . ( [ respdir ] ) with an external field operator and introduce the transition density matrix @xmath91 in the external field @xmath152 : @xmath156 using eq . ( [ respdir ] ) we find that @xmath157 obeys the equation @xmath158 and the strength function is expressed as@xmath159}. \label{strf1}%\ ] ] for the following applications discussed in this section the parameter set nl3 @xcite is used for the covariant energy functional . for superfluid nuclei we use in the pairing channel a simple monopole force with the strength parameters @xmath160 ( @xmath161 ) adjusted to experimental gap parameters for protons and neutrons . the cut - off energy in the pairing channel is @xmath162 mev both for protons as well as for neutrons . the parameter set nl3 has been adjusted to ground state properties of a few spherical nuclei more than ten years ago . in numerous applications it has been shown that it provides on the mean field level a very good description of ground states and excited states of nuclei all over the periodic table @xcite . in a recent investigation @xcite its parameters have been slightly modified and several small deficiencies have been eliminated . = 1.25em [ c]ccccccccstate @xmath98 & @xmath163 , mev & & + particle & & @xmath164 & @xmath165 & @xmath1 & @xmath164 & @xmath165 & @xmath1 + 2g9/2 & -2.50 & -2.85 & -3.14 & -2.88 & 0.89 & 0.92 & 0.89 + 1i11/2 & -2.97 & -2.82 & -3.20 & -2.90 & 0.94 & 0.97 & 0.94 + 1j15/2 & -0.48 & -1.16 & -1.33 & -1.21 & 0.70 & 0.74 & 0.70 + 3d5/2 & -0.63 & -0.96 & -1.05 & -0.98 & 0.93 & 0.94 & 0.93 + 4s1/2 & -0.36 & -0.88 & -0.92 & -0.89 & 0.93 & 0.93 & 0.93 + 2g7/2 & -0.56 & -0.71 & -0.90 & -0.76 & 0.92 & 0.94 & 0.92 + 3d3/2 & -0.02 & -0.35 & -0.42 & -0.37 & 0.93 & 0.93 & 0.93 + hole & & @xmath164 & @xmath81 & @xmath1 & @xmath164 & @xmath81 & @xmath1 + 3p1/2 & -7.66 & -7.67 & -7.40 & -7.70 & 0.96 & 0.98 & 0.96 + 2f5/2 & -9.09 & -8.97 & -8.71 & -9.02 & 0.93 & 0.96 & 0.93 + 3p3/2 & -8.40 & -8.20 & -7.87 & -8.22 & 0.90 & 0.94 & 0.90 + 1i13/2 & -9.59 & -9.30 & -9.07 & -9.36 & 0.90 & 0.92 & 0.89 + 2f7/2 & -11.11 & -10.20 & -9.98 & -10.22 & 0.72 & 0.76 & 0.72 + ( 1h9/2)@xmath166 & -13.38 & -13.32 & -13.23 & -13.34 & 0.52 & 0.47 & 0.53 + ( 1h9/2)@xmath167 & & -12.48 & -12.42 & -12.49 & 0.31 & 0.39 & 0.29 + in this section we discuss the changes of the single - particle spectra of the odd mass nuclei @xmath168pb , @xmath169pb , @xmath168tl and @xmath169bi if the coupling to low lying collective vibrations of the surface is taken into account . in order to keep the numerical effort in reasonable limits in a first investigation only the most collective phonons with spin and parity @xmath170 @xmath171 @xmath172 @xmath173 @xmath174 below the neutron separation energy and a reduced number of single - particle states with positive energy ( particles or holes ) is taken into account in the solution of the dyson equation ( [ dyson ] ) . this reduces strongly the number of poles in the self - energy of eq . ( [ mo1 ] ) . the numerical results obtained in these investigations are compiled in the table [ table1 ] . for the first shell of neutron levels above ( particle ) and below ( hole ) the fermi level three versions are given : in the version @xmath164 the index @xmath175 in eq . ( [ mo1 ] ) includes all contributions from intermediate states above the fermi level @xmath80 , below the fermi level @xmath81 and in the dirac sea @xmath82 . version @xmath165 ( for particles ) or @xmath81 ( for holes ) excludes the backward going diagrams , and the third version @xmath1 does not contain antiparticle intermediate states in ( [ mo1 ] ) . in this way , one can see that the effects of ground state correlations ( gsc ) caused by the particle - phonon coupling and neglected in the second version are significant and it is essential to take them into account in a realistic calculation . on the other hand , the contribution of the antiparticle subspace to the self - energy is quantitatively not of great importance . this can be understood by the large values of the energy denominators in eq . ( [ mo1 ] ) for these configurations . thus it is justified to disregard them in the following calculations . notice , however , that version @xmath1 does not eliminate the effects of the dirac sea completely since the phonon vertices still contain this contribution . as it has been discussed in ref . @xcite these terms play an important role in a proper treatment of relativistic rpa . otherwise it is not possible to obtain reasonable properties for the isoscalar modes within rrpa . next we show results where the contribution of the antiparticle subspace to the self - energy are neglected . in this case one is able to enlarge the particle - hole basis considerably by taking into account particle - hole configurations far away from the fermi surface . this increases the collectivity of the phonons and , consequently , the strength of the particle - vibrational coupling . the phonon basis was also enriched by including higher - lying modes up to 35 mev . solving the dyson equation one finds a fragmentation of the single - particle states and a corresponding reduction of the single - particle strength . for the levels one major shell below and one shell above the fermi surface one finds always one dominant level , which is shifted against the corresponding single - particle energy without particle - phonon coupling . almost all the levels are moving downwards providing thus a considerably better agreement with experimental energies then the pure rmf states . in the next shells further away from the fermi surface almost all the single - particle levels turn out to be strongly fragmented due to phonon coupling and it is no longer possible to determine the dominant levels in these shells , in other words , the concept of landau quasi - particles is defined only in the neighborhood of the fermi level and it breaks down at larger distances . ( left panel ) and @xmath176 ( right panel ) states in @xmath169pb obtained in the calculations ( solid lines ) and the experimental strengths of the respective dominant levels ( dashed lines ) . ] in fig . [ fig3 ] we show as an example the 3d@xmath177 and the 1j@xmath178 levels in the nucleus @xmath169pb . in both cases the single - particle strength is distributed over about two thousand states but most of them are vanishingly small . thus only the states with the strength exceeding 10@xmath179 are drown . the state 3d@xmath177 has a pronounced single - particle structure with a single - particle strength close to 0.9 . on the other side the 1j@xmath178 is more fragmented . the experimental strength of the dominant levels are shown with dashed lines . : the pure rmf spectrum ( left column ) , the levels computed within rmf with allowance for the particle - vibration coupling ( center ) and the experimental spectrum ( right ) . ] to illustrate the shifts in the level schemes of the dominant poles as compared to the rmf results we show as an example in figs . [ fig4 ] the single - particle spectrum for neutrons . the spectrum calculated with the energy - dependent correction ( rmf+pvc ) demonstrates a pronounced increase of the level density around the fermi surface of @xmath8pb compared the pure rmf spectra . in some cases the order of levels is inverted and the observed sequence is reproduced as for instance for the@xmath180 and the @xmath181 neutron states . another and more important example is the inversion of the @xmath182 and @xmath183 neutron states which reproduces the spin of the @xmath169pb ground state . in order to quantify these results we calculate the average distance between two levels in the spectrum shown in fig . one finds for the neutrons 1.0 ( rmf ) , 0.83 ( rmf+pcv ) and 0.76 ( exp ) in units of mev . this corresponds to a level density of 1.0 ( rmf ) , 1.20 ( rmf+pcv ) and 1.31 ( exp ) in units of mev@xmath184 . the level density in the neighborhood of the fermi surface is therefore in rmf - calculations by a factor 0.76 smaller than the experimental value . taking into account particle - vibrational coupling we find only a reduction of 0.92 . assuming an effective mass close to 1 for the experiment , and taking into account that the level density at the fermi surface is proportional to @xmath9 , this corresponds to an effective mass @xmath185 for the rmf and @xmath186 for the rmf+pcv calculations . for the protons the situation is similar . jaminon and mahaux have discussed in refs . @xcite the concept of the effective mass in the case of rmf theory . on one side one has the well known dirac mass@xmath187 which is determined by the scalar field @xmath188 . since we do not use an isovector scalar field for the present parameter set nl3 the dirac mass is in these calculations identical for protons and neutrons . however , this quantity should not be compared with the effective mass determined empirically from a non - relativistic analysis of scattering data and of bound states . from a non - relativistic approximation of the dirac equations one finds that the mass @xmath189 should be used for this purpose . here @xmath190 is the time - like component of the lorentz vector field determined by the exchange of @xmath18- and @xmath19-mesons . in symmetric nuclear matter we find for nl3 : @xmath191 and @xmath192 . the latter value is smaller then the values @xmath193 for protons and @xmath185 for neutrons deduced from the calculated spectrum around the fermi surface in simple rmf theory . following similar arguments we would obtain for rmf+pvc calculations an average effective mass of @xmath194 . this is obviously still too low as compared to the experimental value . on the other hand , around the fermi surface where relativistic kinematic effects are not significant the rmf+pvc spectrum can be characterized by the effective mass deduced from the schrdinger equation which is a non - relativistic limit of the dirac equation ( [ deq1 ] ) . in this approximation one can calculate the state - dependent e - mass @xmath195 which is the inverted spectroscopic factor of the dominant level @xmath103:@xmath196^{-1}.\ ] ] for the calculated rmf+pvc spectrum the averaged e - masses are 1.26 for neutrons and 1.41 for protons if one takes into account all the states with spectroscopic factors larger then 0.5 , i. e. good single - particle states . thus , the energy dependence of the self - energy increases the rmf neutron and proton effective masses up to the values 0.96 and 1.0 , respectively . although the problem of particle - vibration coupling in nuclei has a long history and it was considered in a number of works , most of them are based on a non - relativistic treatment of the nuclear many - body problem . only in a relatively recent investigation in ref . @xcite a correction of the rmf single - particle spectrum was undertaken in a phenomenological way assuming a linear dependence of the self - energy near the fermi surface . the corresponding coupling constants were determined by a fit to nuclear ground state properties . despite the fact that the present approach is fully microscopic without any additional parameter adjusted to experiment it shows good agreement with the results of ref . @xcite for the spectrum of @xmath8pb . the shift caused by the phenomenological particle - vibrational coupling in ref . @xcite is only slightly larger than in the present investigation . non - relativistic microscopic investigations of particle - vibrational coupling can be divided into two major groups . the first group @xcite uses a phenomenological single - particle input to reproduce the experimental spectrum and has therefore to exclude the contribution of the particle - vibration coupling from the full self - energy to find the bare spectrum . usually these older approaches take into consideration only a relatively small number of collective low - lying phonons and use a particle - vibration coupling model @xcite . this restriction to only low - lying modes produces shifts less then 1 mev . however , as it was shown in ref . @xcite , enlarging of the phonon space with high - lying vibrations leads to very strong shifts of the single - particle levels up to 4 mev , and no saturation is observed with respect to the dimension of the phonon space . the second group of approaches ( see , for instance refs . @xcite ) starts from a self - consistent hartree - fock description and applies perturbation theory to calculate the particle - vibration contribution to the full self - energy . in such self - consistent methods it is more justified to enlarge the phonon space . it was shown , for instance , in @xcite that the contribution of the isovector modes is noticeably smaller than the isoscalar ones . the detailed investigation of the relative importance of the high multipole states was performed in @xcite . because of the larger phonon space the typical shifts of the single - particle levels in @xmath8pb are about 1 - 2 mev . as for the spectroscopic factors , all the approaches predict similar values because these factors are not very sensitive to the details of the calculation schemes . the solution of the bethe - salpeter equation ( [ respdir ] ) allows to calculate the nuclear response to external multipole fields and the strength functions of the corresponding collective excitations . as in the last section we show applications based on the density functional nl3 with a monopole force in the pairing channel . a small artificial width of 200 kev is introduced as an imaginary part of the energy variable @xmath18 to have a smooth envelope of the calculated curves . the energies and amplitudes of the most collective phonon modes with spin and parity 2@xmath197 , 3@xmath198 , 4@xmath197 , 5@xmath198 , 6@xmath197 are calculated with the same restrictions and selected using the same criterion as in the last section and in many other non - relativistic investigations in this context . only the phonons with energies below the neutron separation energy enter the phonon space since the contributions of the higher - lying modes are found to be small . on all three stages of these calculations the same energy functional , i.e. the same relativistic nucleon - nucleon interaction @xmath114 ( [ vstatic ] ) has been employed . the vertices @xmath199 ( [ phonon ] ) entering the term @xmath149 in eq . ( [ phiph ] ) are calculated with the same force . therefore no further parameters are needed . the scheme is fully consistent . the subtraction procedure developed by tselyaev in the ref . @xcite for the self - consistent scheme removes the static contribution of the particle - phonon coupling from the @xmath1-interaction . it takes into account only the additional energy dependence introduced by the dynamics of the system . it has been found in the calculations of refs . @xcite as well as in the calculations of the ref . @xcite that within the relatively large energy interval ( 0 - 30 mev ) the subtraction procedure provides a rather small increase of the mean energy of the giant dipole resonance ( 0.8 mev for lead region ) and gives rise to the change by a few percents in the sum rule . this procedure restores the response at zero energy and therefore it does not disturb the symmetry properties of the rrpa calculations . the zero energy modes connected with the spontaneous symmetry breaking in the mean field solutions , as for instance the translational mode in the dipole case , remain at exactly the same position after the inclusion of the particle - vibration coupling . in practice , however , because of the limited number of oscillator shells in the calculations this state is found already in rrpa without particle - vibration coupling at a few hundreds kev above zero . in cases , where the results depend strongly on a proper separation of this spurious state , as for instance for investigations of the pygmy dipole resonance in neutron rich systems @xcite one has to include a large number of @xmath1-configurations in the rrpa solution . pb and @xmath200sn obtained within two approaches : rrpa ( dashed line ) and rrpa with particle - phonon coupling rrpa - pc ( solid line ) . both computations have been performed with relativistic hartree ( rh ) mean field and employ nl3 parameter set for rmf forces . ] .[table2 ] = 2.15em [ c]cccc & & @xmath201e@xmath202 ( mev ) & @xmath86 ( mev ) + & rrpa & 14.16 & 1.71 + @xmath8pb & rrpa - pc & 14.05 & 2.36 + & exp . @xcite & 13.73(20 ) & 2.58(20 ) + & rrpa & 16.10 & 2.63 + @xmath200sn & rrpa - pc & 16.01 & 3.09 + in fig . [ fig5 ] we show the calculated strength functions for the isoscalar monopole resonance in @xmath8pb and @xmath200sn . the fragmentation of the resonance caused by the particle - phonon coupling is clearly demonstrated although the spreading width of the monopole resonance is not large because of a strong cancellation between the self - energy diagrams and diagrams with the phonon exchange ( see fig . [ fig2 ] ) . this fact has also been discussed in detail in refs . @xcite and it is not disturbed by the subtraction procedure because this cancellation takes place as well in @xmath149 as in @xmath148 . in order to compare the spreading of the theoretical strength distributions with experimental data we show in table [ table2 ] mean energies @xmath203 and widths parameters @xmath86 obtained by fitting the theoretical strength distribution in a certain energy interval to a lorentz curve in the same way as it has been done in the experimental investigations . the experimental values shown in table [ table2 ] are derived in the ref . @xcite from the evaluation of a series of data obtained in different experiments for the isoscalar monopole resonance in @xmath8pb . details are given in the text . ] but for @xmath200sn . ] [ fig6 ] and [ fig7 ] present calculated photoabsorption cross sections @xmath204 for the isovector dipole resonance in @xmath8pb and in @xmath200sn . the left panels give the results obtained within the non - relativistic approach with a woods - saxon ( ws ) single - particle potential and landau - migdal ( lm ) forces described in ref . they are compared in the right panel with the relativistic fully consistent theory of ref . ( r)rpa calculations are shown by the dashed curves , ( r)rpa extended by the phonon coupling ( ( r)rpa - pc ) calculations by the thick solid curves . in fig . [ fig6 ] we have also displayed experimental data with error bars taken from ref . @xcite . in both calculations , relativistic ( right panel ) and non - relativistic ( left panel ) , the continuum is taken into account only in a discrete approximation , which is very reliable for heavy nuclei . as discussed in ref . @xcite , the dirac equation ( [ dirac - basis ] ) is solved in an oscillator basis . to make the comparison reasonable calculations within the non - relativistic framework have been performed with box boundary conditions for the schrdinger equation in @xmath205-space which ensures completeness of the single - particle basis . the corresponding lorentz fit parameters in the two energy intervals : @xmath206 mev and @xmath207 mev ( @xmath208 is the neutron separation energy ) are included in table [ table3 ] and they are compared with the data of ref . we notice that the inclusion of particle - phonon coupling in the rrpa calculation induces a pronounced fragmentation of the photoabsorption cross sections , and brings the width of the gdr in much better agreement with the data , both for @xmath8pb and @xmath200sn . the fragmentation of the resonance introduced by the particle - phonon coupling is clearly demonstrated in both cases . also , one finds more or less the same level of agreement between theory and experimental data for these two calculations . in the case of the isovector e1 resonance in @xmath200sn this is , however , not so clear because the cross section and the integral characteristics of the resonance obtained in the experiment of ref . @xcite are given with relatively large error bars . in @xmath8pb the self - consistent relativistic approach reproduces the shape of the giant dipole resonance much better than the non - relativistic one although the whole resonance is about 0.5 mev shifted to lower energies with respect to the experiment . as one can see from the fig . [ fig6 ] and table [ table3 ] , we observe some shift already in the rrpa calculation , which is determined by the properties of the nl3 forces . improvement of the forces , for instance , the use of the density dependent versions @xcite of the rmf should bring the e1 mean energy in better agreement with the data . = 1.1em [ c]cccccccc & & & + & & @xmath201e@xmath202 & @xmath86 & ewsr & @xmath201e@xmath202 & @xmath86 & ewsr + & & ( mev ) & ( mev ) & ( % ) & ( mev ) & ( mev ) & ( % ) + & rrpa & 13.1 & 2.4 & 121 & 12.9 & 2.0 & 128 + @xmath8pb & rrpa - pc & 12.9 & 4.3 & 119 & 13.2 & 3.0 & 128 + & exp . @xcite & 13.4 & 4.1 & 117 & & & 125(8 ) + & rrpa & 14.7 & 3.3 & 116 & 14.5 & 2.6 & 126 + @xmath200sn & rrpa - pc & 14.4 & 4.0 & 112 & 14.6 & 3.2 & 126 + & exp . @xcite & 16.1(7 ) & 4.7(2.1 ) & 125(32 ) & & & + however , there is an essential difference between the fully self - consistent relativistic calculations and the non - relativistic approach : in non - relativistic approach discussed in ref . @xcite one introduces on all three stages of the calculation phenomenological parameters , which have to be adjusted to experimental data : first , the woods - saxon parameters as , for instance , the well depth are varied to obtain single - particle levels close to the experimental values , second , one of the parameters of the landau - migdal force is adjusted to get phonon energies at the experimental positions ( for each mode ) and , third , another landau - migdal force parameter is varied to reproduce the centroid of the giant resonance . although the varying of the parameters is performed in relatively narrow limits , it is necessary to obtain realistic results . in contrast , in the relativistic fully consistent approach no adjustment of additional parameters is necessary . of course , the underlying energy functional has been determined in a phenomenological way by a fit to experimental ground state properties of characteristic nuclei . however , it is of universal nature and the same parameters are used for investigations of many nuclear properties all over the periodic table . the predictive power of this scheme is therefore much higher than that of the present semi - phenomenological approach discussed , for instance , in ref . @xcite . pairing correlations play an essential role in all open shell nuclei and apart from the vicinity of the very few doubly magic configurations nuclei show superfluidity all over the periodic table . in a theoretical description this fact can be taken into account by bogoliubov s quasiparticles . many - body theories for normal systems are thus relatively easily extended to the case of superfluid nuclei . combining creation and annihilation operators @xmath209 and @xmath210 to a two - component operator , operators of the type @xmath211 , @xmath212 , @xmath213 are replaced by super - matrices of rank 2 and the form of the equations stays nearly unchanged . this very elegant method has been introduced already half a century ago in ref . @xcite and over the years it has been used for various many - body approximation schemes in non - relativistic systems as for instance in refs . response theory with the time blocking approximation ( tba ) introduced for normal non - relativistic systems in ref . @xcite has been extended in refs . @xcite to qtba for superfluid systems and in ref . @xcite to rqtba for relativistic superfluid systems . in the following we discuss several applications of rqrpa and of rqtba in the chain of spherical even - even semi - magic nuclei with @xmath214 . we show calculations of the isovector dipole spectrum in the giant dipole resonance region and in the low - lying energy region in the two approximations . as discussed before the effective interaction @xmath149-@xmath148 takes into account only the additional energy dependence introduced by the dynamics of the system . it has been found in relativistic @xcite as well as in non - relativistic calculations @xcite that within a relatively large energy interval ( 0 - 30 mev ) the subtraction procedure provides a rather small but noticeable increase of the mean energy of the giant dipole resonance ( about 0.7 mev for tin region ) and gives rise to changes by a few percents in the sum rule . the absolute value of the energy shift produced by the subtraction of @xmath148 in eq . ( [ w - omega ] ) is comparable with but not exactly equal to the absolute value of the shift produced by the dynamical part of the interaction amplitude @xmath149 which always reduces the mean energy of the resonance . the subtraction procedure restores the response at zero energy and , therefore , it does not disturb the symmetry properties of the rqrpa calculations . the zero energy modes connected with the spontaneous symmetry breaking in the mean field solutions , as , for instance , the translational mode in the dipole case , remain at exactly the same positions after the inclusion of the quasiparticle - vibration coupling . sn , @xmath215sn , @xmath216sn , compared to data of ref . @xcite for @xmath217sn . right panels ( b , d , f ) : photo absorption cross sections computed with the artificial width 200 kev . left panels ( a , c , e ) : the low - lying portions of the corresponding spectra in terms of the strength function , calculated with 20 kev smearing . calculations within the rqrpa are shown by the dashed curves , and the rqtba - by the solid curves . ] in fig . [ fig8 ] we show dipole spectra for the tin isotopes @xmath218sn , @xmath215sn , @xmath216sn . the right panels show the photo absorption cross section ( [ cs ] ) which is determined by the dipole strength function @xmath219 . it is calculated , analogously to eq . ( [ strf1 ] ) , with the usual isovector dipole operator . the left panels show the low - lying parts of the corresponding spectrum in terms of the strength function . a small imaginary part of 20 kev is used for the energy variable , in order to see the fine structure of the spectrum and sometimes individual levels in this region . rqrpa calculations are shown by dashed curves and the rqtba by the solid curves . experimental data are taken from the exfor database @xcite . these figures clearly demonstrate how the two - quasiparticle states , which are responsible for the spectrum of the rqrpa excitations , are fragmented through the coupling to the collective vibrational states . the effect of the particle - vibration coupling on the low - lying dipole strength below and around the neutron threshold within the presented approach is shown in the left panels of the fig . such calculations give us an example how the low - lying strength develops with the increase of the neutron excess . it is also found that the presence of pairing correlations causes a noticeably stronger fragmentation of both the gdr and the pdr modes as compared to the case of a normal system discussed above . this effect has the two reasons . first , pairing correlations lead to a diffuseness of the fermi surface and , thus , increase the number of possible @xmath220 configurations , and second , pairing correlations cause a considerable lowering of the energies and increased transition probabilities of the lowest 2@xmath197 states . in spherical open - shell medium mass nuclei the highly collective first 2@xmath197 states appear at energies around 1 mev ( and they are usually well reproduced in rqrpa @xcite ) whereas in magic nuclei and often in nuclei near the shell closures they appear much higher , at about 3 - 4 mev and have considerably reduced transition probabilities . this causes a strong configuration mixing in the case of presence of very low - lying vibrational states . these modes admix to others , in particular , to the gdr and the pdr and the lower their energies and the higher their transition probabilities are , the stronger fragmentation they cause . a systematic analysis of the transition densities of the rqrpa and the rqtba states shows that the @xmath7 transition densities in the broad low - lying energy region dominated by the fragmentation of the rqrpa pygmy mode have a very similar behavior as the initial rqrpa state : proton and neutron components oscillate in phase in the nuclear interior and neutron components dominate on the surface in nuclei with noticeable neutron excess . we have given an overview over recent efforts to combine two theoretical methods for the description of the quantum - mechanical many - body problem of nuclear physics , covariant density functional theory ( cdft ) and landau - migdal theory of finite fermi systems ( tffs ) . both methods are very successful and they are claimed to provide in principle an exact description . in practice , however , there are limitations . both methods use phenomenological input . dft can only be applied to physical quantities , which can be expressed in terms of the single - particle density and in self - bound systems such as nuclei dft is based on the intrinsic density , a concept , which requires additional approximations . in particular the self - energy used in dft theory does not depend on the energy . landau - migdal theory on the other side restrains itself from calculating ground state properties , but it goes far beyond the mean field approach and takes into account couplings to complex configurations . there are also similarities for these two methods . both are based on a single - particle description , i.e. on the motion of independent particles . in density functional theory an average field in introduced as a vehicle in order to take into account shell effects . however , the single - particle energies themselves are not observables in the strict sense . landau - migdal theory uses quasiparticles as the exact eigenstates of the @xmath221-systems . in both cases the self energies are given as the first derivatives of the total energy with respect to the density and the effective interactions between the particles are the second derivatives of this quantity . the combination of cdft and tffs described in this manuscript starts from the covariant density functional . no further parameters are needed . this functional is used to describe the ground state properties and the parameters of the functional are adjusted to experimental data of ground states of several nuclei . on this level the self - energy does not depend on the energy and landau - migdal theory is used to introduce an energy dependence with a particle - vibrational scheme . the properties of the phonons needed for the calculation of the energy - dependent part of the self - energy are phonon energies and phonon - nucleon vertices . they are calculated with the static effective interaction obtained as the second derivative of the density functional . no additional parameters are needed . the essential equation of the landau - migdal theory is the response equation . the effective interaction to calculate the full response is the derivative of the self - energy with respect to the density and this means that one obtains in addition to the static interaction resulting from the energy - independent part an induced interaction resulting from the energy - dependent part of the self - energy . of course there would be double counting , because many of the correlations induced by the coupling of virtual phonons have also contributions at the ground state energy . therefore a subtraction method is introduced , which removes from the induced interaction at finite energy its value at zero energy , i.e. after this subtraction , the induced interaction vanishes at the ground state and takes into account only its energy dependence . therefore this subtraction procedure guarantees that one does not need to readjust the parameters of the density functional , because the effects of particle - vibrational coupling vanish at the ground state of the even - even system , where the parameters are adjusted . we have discussed several applications of this method , as the fragmentation of single - particle energies in odd - mass nuclei in the vicinity of a double magic configuration . close to the fermi surface there is always a dominant pole with a reduced single - particle strength and many other poles with rather small strength . the dominant pole is shifted in the direction of the fermi surface , i.e. the level density at the fermi surface is increased . the effective mass derived from this level density in the pb region is considerably increased , but there is still room for an additional energy dependence of the self - energy not taken into account by the coupling to surface vibrations . in addition we discussed several solutions of the response equations for nuclei in an external field . this allows to calculate the strength functions with respect to an external operator and the photo absorption cross sections in the correlated system . if one takes into account only the static part of the interaction one finds the usual rpa or qrpa results of time - dependent density functional theory , which reproduces the position of the resonances rather well , but it can not account for the width that has its origin in the coupling to more complicated configurations . the energy - dependent part of the interaction includes this coupling and therefore it induces again a fragmentation of the rather sharp resonance peaks in rpa of qrpa over many complex configurations . because of the subtraction procedure the position of the resonances is not changed very much , but the width is considerably increased , in excellent agreement with experimental data . of course there is room for additional improvements . so far , for numerical simplicity , exist only applications of this theory with the parameter set nl3 , which has no density dependence in the is isovector channel . at present there exist more modern parameter sets with the density dependent meson exchange @xcite which give already on the mean field level improved results for characteristic properties such as the neutron skin of neutron - rich nuclei or the density dependence of the symmetry energy . they should also be implemented in the theoretical investigations of the type discussed in this article . a further improvement can be achieved by using a more realistic pairing force . so far there are only investigations available with a monopole pairing force . it is therefore highly desirable to implement in the pairing channel a density dependent zero range force or the finite range gogny force . all the calculations presented here have been done in a discrete basis , i.e. in the spectral representation of the response equation and therefore the coupling to the continuum is not taken into account properly so far . this might have a strong influence on application to light nuclei and therefore relativistic continuum rpa should be extended to relativistic continuum qrpa and relativistic continuum qtba . finally , so far ground state correlations have been taken into account only on the rqrpa level . in the non - relativistic case there exist investigations going beyond this limitations . they should be extended also to the relativistic calculations . valuable discussions with v. tselyaev are gratefully acknowledged . this work has been supported in part by the bundesministerium fr bildung und forschung under project 06 mt 246 and by the dfg cluster of excellence origin and structure of the universe(www.universe - cluster.de ) . acknowledge financial support from the hessian loewe initiative through the helmholtz international center for fair and the russian federal agency of education , project no .

sparse coding algorithms trained on natural images can accurately predict the features that excite visual cortical neurons , but it is not known whether such codes can be learned using biologically realistic plasticity rules . we have developed a biophysically motivated spiking network , relying solely on synaptically local information , that can predict the full diversity of v1 simple cell receptive field shapes when trained on natural images . this represents the first demonstration that sparse coding principles , operating within the constraints imposed by cortical architecture , can successfully reproduce these receptive fields . we further prove , mathematically , that sparseness and decorrelation are the key ingredients that allow for synaptically local plasticity rules to optimize a cooperative , linear generative image model formed by the neural representation . finally , we discuss several interesting emergent properties of our network , with the intent of bridging the gap between theoretical and experimental studies of visual cortex .

1109.2239

a central goal in systems neuroscience is to determine what underlying principles might shape sensory processing in the nervous system . several coding optimization principles have been proposed , including redundancy reduction @xcite , and information maximization @xcite , which have both enjoyed some successes in predicting the behavior of real neurons@xcite . closely related to these notions of coding efficiency is the principle of sparseness @xcite , which posits that few neurons are active at any given time ( population sparseness ) , or that individual neurons are responsive to few specific stimuli ( lifetime sparseness ) . sparseness is an appealing concept , in part because it provides a simple code for later stages of processing and it is in principle more quickly and easily modifiable by simple learning rules compared with more distributed codes involving many simultaneously active units @xcite . there is some experimental evidence for sparse coding in the cortex @xcite , but there are also reports of dense neural activity @xcite and mixtures of both @xcite as well . compounding this , it is not obvious what absolute standard should be used to assess the degree of sparseness in cortex , but it is notable that the relative level of sparseness of cortical responses to natural images increases when a larger fraction of the visual field is covered by the stimulus @xcite , as a result of inhibitory interneuronal connections @xcite . interestingly , the correlations between the neuronal activities also decreases when a larger area is stimulated , as a result of these inhibitory connections @xcite . in a landmark paper , olshausen and field @xcite reproduced several qualitative features of the receptive fields ( rfs ) of neurons in primary visual cortex ( vi ) without imposing any biological constraints other than their hypothesis that cortical representations simultaneously minimize the average activity of the neural population while maximizing fidelity when representing natural images . however , agreement with measured v1 simple cell receptive fields was not perfect @xcite . recently , a more sophisticated version of olshausen and field s algorithm @xcite has been developed that is capable of minimizing the number of active neurons rather than minimizing the average activity level across the neural population . this algorithm , called the sparse - set coding ( ssc ) network @xcite , learns the full set of physiologically observed rf shapes of simple cells in v1 , which include small unoriented features , localized oriented features resembling gabor wavelets , and elongated edge - detectors . we note that , under certain conditions @xcite not necessarily satisfied by the natural image coding problem , minimizing the average activity across the neural population ( @xmath0-norm minimization ) , as is done by olshausen and field s original sparsenet algorithm , can be equivalent to minimizing the number of active units ( @xmath1-norm minimization ) , as is achieved by rehn and sommer s ssc algorithm . the ssc model is the only sparse coding algorithm that has been shown to learn , from the statistics of natural scenes alone , rfs that are in quantitative agreement with those observed in v1 . it has also been found that sufficiently overcomplete representations ( 4 times more model neurons than image pixels ) that minimize the @xmath0 norm can display the same qualitative variety of rf shapes , but these have not been quantitatively compared with physiologically measured rfs @xcite . unfortunately , the lack of work on biophysically realistic sparse coding models has left in doubt whether v1 could actually employ a sparse code for natural scenes . indeed , it is not clear how olshausen s original algorithm @xcite , the highly overcomplete @xmath0-norm minimization algorithm @xcite , or that of rehn and sommer @xcite , could be implemented in the cortex . rather than employing local network modification rules such as the synaptic plasticity that is thought to underly learning in cortex @xcite , all three of these networks rely on learning rules requiring that each synapse has access to information about the receptive fields of many other , often distant , neurons in the network . furthermore , both the ssc sparse coding model that has successfully reproduced the full diversity of v1 simple cell rfs @xcite as well as the @xmath0-norm minimization algorithm that achieved qualitatively similar rfs @xcite involve non - spiking computational units : continuous - valued information is shared between units while inference is being performed . in cortex , however , information is transferred in discrete , stereotyped pulses of electrical activity called action potentials or spikes . particularly for a sparse coding model with few or no spikes elicited per stimulus presentation , approximating spike trains with a graded function may not be justified . spiking image processing networks have been studied @xcite , but none of them have been shown to learn the full diversity of v1 rf shapes using local plasticity rules . it remains to be demonstrated that sparse coding can be achieved within the limitations imposed by biological architecture , and thus that it could potentially be an underlying principle of neural comptutation . here we present a biologically - inspired variation on a network originally due to fldik @xcite that performs sparse coding with spiking neurons . our model performs learning using only synaptically local rules . using the fact that constraints imposed by such mechanisms as homeostasis and lateral inhibition cause the units in the network to remain sparse and independent throughout training , we prove mathematically that it learns to approximate the optimal linear generative model of the input , subject to constraints on the average lifetime firing rates of the units and the temporal correlations between the units firing rates . this is the first demonstration that synaptically local plasticity rules are sufficient to account for the observed diversity of v1 simple cell rf shapes , and the first rigorous derivation of a relationship between synaptically local network modification rules and the twin properties of sparseness and decorrelation . finally , we describe several emergent properties of our image coding network , in order to elucidate some experimentally testable hallmarks of our model . interestingly , we observe a lognormal distribution of inhibitory connection strengths between the units in our model , when it is trained on natural images ; such a distribution has previously been observed in the excitatory connections between neurons in rat v1 @xcite , but the inhibitory connection strength distribution remains unknown . our primary goal is to develop a biophysically inspired network of spiking neurons that learns to sparsely encode natural images , while employing only synaptically local learning rules . towards this end , we implement a network of spiking , leaky integrate - and - fire units @xcite as model neurons . as in many previous models @xcite , each unit has a time dependent internal variable @xmath2 and an output @xmath3 associated with it . the simulation of our network operates in discrete time . the neuronal output at time @xmath4 , @xmath3 , is binary - valued : it is either 1 ( spike ) or 0 ( no spike ) , whereas the internal variable @xmath2 is a continuous - valued function of time that is analogous to the membrane potential of a neuron . when this internal variable exceeds a threshold @xmath5 , the unit fires a punctate spike of output activity that lasts for one time step . this thresholding feature plays the role of neuronal voltage - gated ion channels ( represented , as in hopfield s @xcite circuit model , by a diode ) whose opening allows cortical neurons to fire . other units in the network , and the inputs @xmath6 , which are pixel intensities in an image , modify the internal variable @xmath2 by injecting current into the model neuron . the structure of our network , and circuit diagram of our neuron model , are illustrated in fig . 1 . the dynamics of sailnet neurons are discussed in detail in the methods section . we assess the computational output of each neuron in response to a stimulus image @xmath7 by counting the number of spikes emitted by that neuron , @xmath8 , following stimulus onset for a brief period of time lasting five times the time constant @xmath9 of the rc circuit . our simulation updates the membrane potential every @xmath10 , thus there are 50 steps in the numerical integration following each stimulus presentation . consequently , at least in principle , 50 is the maximum number of spikes we could observe from one neuron in response to any image . we note that one could instead use first - spike latencies to measure the computational output @xcite ; these two measures are highly correlated in our network , with shorter latencies corresponding to greater spike counts ( data not shown ) . the network learns via rules similar to those of fldik @xcite . these rules drive each unit to be active for only a small but non - zero fraction of the time ( lifetime sparseness ) and to maintain uncorrelated activity with respect to all other units in the network : @xmath11 where @xmath12 is the target average value for the number of spikes per image , which defines each neuron s lifetime sparseness , and @xmath13 , @xmath14 , and @xmath15 are learning rates small positive constants that determine how quickly the network modifies itself . updating the feed - forward weights @xmath16 in our model is achieved with oja s implementation @xcite of hebb s rule ; this rule is what drives the network to represent the input . note that because the firing rates are low ( @xmath17 spikes per image , for the results shown in this paper ) , and spikes can only be emitted in integer units , our model implicitly allows only small numbers of neurons to be active at any given time ( so called hard " sparseness , or @xmath1 sparseness ) , similar to what is achieved by other means in some recent non - spiking sparse coding models @xcite . these learning rules can be viewed as an approximate stochastic gradient descent approach to the _ constrained _ optimization problem in which the network seeks to minimize the error between the input pixel values @xmath18 , and a linear generative model formed by all of the neurons @xmath19 , while maintaining fixed average firing rates and no firing rate correlations . this constrained optimization interpretation of our learning rules , and the approximations involved , are discussed in the methods section . in fig . 2 , we demonstrate that the activity of the sailnet units can be linearly decoded to recover ( approximately ) the input stimulus . the success of linear decoding in a model that encodes stimuli in a non - linear fashion is a product of our learning rules , and it has been observed in multiple sensory systems @xcite and spiking neuron models optimized to maximize information transmission @xcite . our learning rules encourage all neurons to have the same average firing rate of @xmath12 spikes per image , which may at first appear to be at odds with the observation @xcite that cortical neurons display a broad distribution of activities firing rates vary from neuron to neuron . however , when trained on natural images , neurons in sailnet can actually exhibit a fairly broad range of firing rates . moreover , the mean firing rate distribution ranges from approximately lognormal to exponential in response to natural image stimuli , depending on the mean contrast of the stimulus ensemble with which they are probed . we discuss this further in the firing rates section below . we emphasize here that each of our learning rules is synaptically " local : the information required to determine the change in the connection strength at any synaptic junction between two units is merely the activity of the pre- and post - synaptic units . the inhibitory lateral connection strengths , for example , are modified according to how many spikes arrived at the synapse , and how many times the post - synaptic unit spiked . the information required for the unit to modify its firing threshold is the unit s own firing rate . finally , the rule for modifying the feed - forward connections requires only the pre - synaptic activity @xmath6 , the post - synaptic activity @xmath20 , and the present strength of that connection @xmath21 . this locality is a desirable model feature because learning in cortex is thought @xcite to occur by the modification of synaptic strengths and thus by necessity should depend only upon information available locally at the synapse . by contrast , much previous work @xcite has used a different learning rule for the feed forward weights : @xmath22 . this rule is non - local because the update for the connection strength between input pixel @xmath23 and unit @xmath24 requires information about the activities and feed - forward weights of every other unit in the network ( indexed by @xmath25 ) . it is unlikely that such information is available to individual synapses in cortex . interestingly , in the limit of highly sparse and uncorrelated neuronal activities , our local learning rule approximates the non - local rule used by previous workers @xcite , when averaged over several input images ; we provide a mathematical derivation of this result in the methods section . this suggests an additional reason why sparseness is beneficial for cortical networks , in which plasticity is local , but cooperative representations may be desired . we trained a 1536-unit sailnet with sparseness @xmath17 on @xmath26 pixel image patches drawn randomly from whitened natural images from the image set of olshausen and field @xcite . the network is six - times overcomplete with respect to the number of input pixels . this mimics the anatomical fact that v1 contains many more neurons than does lgn , from which it receives its inputs . owing to the computational complexity of the problem there are @xmath27 parameters to be learned in a sailnet model containing @xmath28 neurons we found it prohibitive to consider networks that are much more than @xmath29 overcomplete . our six - times overcompleteness is in a sense analogous to the three - times overcompleteness of the ssc network described by rehn and sommer @xcite , since the outputs of their computational units could be either positive or negative , while our model neurons can output only one type of spike . thus , each of their units can be thought of as representing a pair of our neurons , with opposite - signed receptive fields . the rfs of 196 randomly selected units from our sailnet are shown in fig . 3 , as measured by their spike - triggered average activity in response to whitened natural images . these are virtually identical to the feed - forward weights of the units ; in the methods section , we discuss why this must be the case . to facilitate a comparison between the sailnet rfs , and those measured in macaque v1 ( courtesy of d. ringach ) , we fit both the sailnet , and the macaque rfs to gabor functions . as in the ssc study of rehn and sommer @xcite , only those rfs that could be sensibly described by a gabor function were included in fig . 3 ; for example , we excluded rfs with substantial support along the square boundary , suggesting that the rf is only partly visible . in the methods section , we discuss the gabor fitting routine and the quality control measures we used to define and identify meaningful fits . our sailnet model rfs show the same diversity of shapes observed in macaque v1 , and in the non - local ssc model @xcite . they consist of three qualitatively distinct classes of neuronal rfs : small unoriented features , localized and oriented gabor - like filters , and elongated edge - detectors . our sailnet learning rules approximately minimize the same cost function as the ssc model @xcite , albeit with constraints as opposed to unconstrained optimization , which explains how it is possible for sailnet to learn similar rfs using only local rules . furthermore , in our model , the number of co - active units is small , owing to the low average lifetime neuronal firing rates , and the fact that spikes can only be emitted in integer numbers . this feature is similar to the l@xmath30-norm minimization used in the ssc model of rehn and sommer @xcite and the lca model of rozell and colleagues @xcite . this is the first demonstration that a network of spiking neurons using only synaptically local plasticity rules applied to natural images can account for the observed diversity of v1 simple cell rf shapes . our learning rules ( eq . [ eq : learning ] ) encourage every unit to have the same target value , @xmath12 , for its average firing rate , which might appear to be inconsistent with observations @xcite that cortical neurons exhibit a broad distribution of mean firing rates . however , we find that sailnet , too , can display a wide range of mean rates , as we now describe . to determine the distribution of mean firing rates across the population of model neurons in our network , we first trained a 1536-unit sailnet on @xmath26 pixel patches drawn from whitened natural images , and then presented the network with @xmath31 patches taken from the training ensemble . our measurement was performed with all learning rates set to zero , so that we were probing the properties of the network at one fixed set of learned parameter values , rather than observing changes in network properties over time . we then counted the number of spikes per image from each unit to estimate each neuron s average firing rate , as it might be measured in a physiology experiment . the distribution of these mean firing rates is fairly broad and well - described by a lognormal distribution ( fig . this distribution is strongly non - monotonic , clearly indicating that it is poorly fit by an exponential function . subsequently , we probed the same network ( still with the learning turned off , so that the network parameters were identical in both cases ) with @xmath31 low - contrast images consisting of patches from our training ensemble with all pixel values multiplied by @xmath32 . we found that the firing rate distribution was markedly different than what we found when the network was probed with higher - contrast stimuli . in particular , it became a monotonic decreasing function that was similarly well - described by either a lognormal or an exponential function ( fig . 4b ) . from the dynamics of our leaky integrate - and - fire units , it is clear that the low contrast stimuli with reduced pixel values will cause the units to charge up more slowly and subsequently to spike less in the allotted time the network is given to view each image . consequently , the firing rate distribution gets shifted towards lower firing rates . however , negative firing rates are impossible , so in addition to being shifted , the low - firing - rate tail of the distribution is effectively truncated . note that truncating the lognormal distribution anywhere to the right of the peak results in a distribution that looks qualitatively similar to an exponential . mean firing rates in primary auditory cortex ( a1 ) have been reported by one group @xcite to obey a lognormal distribution , whether spontaneous or stimulus - evoked in both awake and anesthetized animals . however , exponentially distributed spontaneous mean firing rates have also been reported in awake rat a1 @xcite . although several groups have measured the distribution of firing rates over time for individual neurons @xcite , we are unaware of a published claim regarding the distribution of mean firing rates in visual cortex . recall that our learning rules encourage the neurons to all have the same average firing rate . this fact may be puzzling at first given the spread in mean firing rates apparent in the distributions shown in fig . there are two main effects to consider when making sense of this : finite measurement time , and non - zero step - sizes for plasticity . the first effect relates to the fact that there is intrinsic randomness in the measurement process which randomly selected image patches happen to fall in the ensemble of probe stimuli so that the measured distribution tends to be broader than the true " underlying distribution of the system . to check that this effect is not responsible for the broad distribution in firing rates , we computed the variance in the measured firing rate distribution after different numbers of images were presented to the network . the variance decreased until it reached an asymptotic value after approximately @xmath33@xmath34 image presentations ( data not shown ) . thus , the @xmath31 image sample size in our experiment is large enough to see the true distribution ; finite sample - size effects do not affect the distributions that we observed . the other , more interesting , effect that gives rise to a broad distribution of firing rates is related to learning . while the network is being trained , the feed - forward weights , inhibitory lateral connections , and firing thresholds get modified in discrete jumps , after every image presentation ( or every batch of images , see the methods section for details ) . since those jumps are of a non - zero size as determined by the learning rates @xmath13 , @xmath14 , and @xmath15 there will be times when the firing threshold gets pushed below the specific value that would lead to the unit having exactly the target firing rate , and the unit will thus spike more than the target rate . similarly , some jumps will push the threshold above that specific value , and the unit will fire less than the target amount . even after learning has converged , and the parameters are no longer changing _ on average _ in response to additional image presentations , the network parameters are still bouncing around their average ( optimal ) values ; any image presentation that makes a neuron spike more than the target amount results in an increased firing threshold , while any image that makes the neuron fire less than the target amount leads to a decreased firing threshold . recent results @xcite suggest that the sizes of these updates ( jumps ) are quite large for real neurons . interestingly , this indicates that the observed broad distributions in firing rate @xcite do not rule out the possibility that homeostatic mechanisms are driving each neuron to have the same average firing rate . reducing the sailnet learning rates @xmath13 , @xmath14 and @xmath15 does reduce the variance of the firing rate distributions , but our qualitative conclusions non - monotonic , approximately lognormal firing rate distribution in response to images from the training set , and monotonic , exponential / lognormal distribution in response to low contrast images are unchanged when we use different learning rates for the network ( data not shown ) . recent experimental work @xcite has shown that neurons in visual cortex tend to have small correlations between their firing rates . in order to facilitate a comparison between our model , and the physiological observations , we have measured the ( pearson s ) linear correlation coefficients between spike counts of sailnet units , in response to an ensemble 30,000 natural images . these correlations ( fig . 5 ) tend to be near zero , as is observed experimentally @xcite , while the experimental data show a larger variance in the distribution of correlation coefficients than we observe with sailnet . we note that , like the firing rate distribution ( discussed above ) , the distribution of correlation coefficients is affected by the update sizes ( learning rates ) in the simulation , with larger update sizes leading to a larger variance of the measured distribution . in fig . 5 , the distribution appears truncated on the left . this effect arises because there is a lower bound on the correlation between the neuronal firing rates that arises when the two neurons are _ never _ co - active . the low mean firing rate of @xmath35 used in our simulation means that this bound is not too far below zero . several previous studies of sparse coding models @xcite have focused on the receptive fields learned by adaptation to naturalistic inputs , but we are aware of only one published study @xcite that investigated the connectivity in sparse coding models , albeit with a model that lacked biological realism . one previous study @xcite investigated synaptic mechanisms that could give rise to the measured distribution of connection strengths , but this work was not performed in the context of a sensory coding model . no prior work has studied the connectivity learned in a biophysically well - motivated sensory coding network , which would provide additional testable predictions for physiology experiments . 6 shows the distribution of non - zero connection strengths ( non - zero elements of the matrix @xmath36 ) learned by a 1536-unit sailnet with @xmath17 trained on @xmath26 pixel patches drawn from whitened natural images ( the same network whose receptive fields are shown in fig . when trained on natural images , sailnet learns an approximately lognormal distribution of inhibitory connection strengths ; a gaussian best fit to the histogram of the logarithms of the connection strengths accounts for @xmath37 of the variance in the data . despite this close agreement , sailnet shows some systematic deviations from the lognormal fit , especially on the low - connection - strength tail of the distribution . interestingly , the experimental data @xcite show an approximately lognormal distribution of excitatory connection strengths , with similar systematic deviations ( fig . 5b of song _ et al . _ by contrast , prior theoretical work @xcite has employed learning rules tailored to create exactly lognormal connection strength distributions , and thus show no such deviations . note also that neither of these previous studies addressed the issue of how neurons might represent sensory inputs , nor how they might learn those representations . whereas the experimental data of song _ et al . _ @xcite show a roughly lognormal distribution in the strengths of excitatory connections between v1 neurons , our model makes predictions about the strengths of _ inhibitory _ connections in v1 . the @xmath38 time window for measuring post - synaptic potentials in the experiment of song _ et al . _ @xcite ensured that they measured only direct synaptic connections . however , suppressive interactions between excitatory neurons in cortex are mediated by inhibitory interneurons . consequently , the inhibitory interactions between pairs of excitatory neurons in v1 must involve two or more synaptic connections between the cells . thus , our model predicts that the inhibitory functional connections between excitatory simple cells in v1 , like the excitatory connections measured by song _ et al . _ @xcite , should follow an approximately lognormal distribution ( fig . 6 ) , but it does not specify the extent to which this is achieved through variations in strength among dendritic or axonal synaptic connections of v1 inhibitory interneurons . one recent theoretical study @xcite has uncovered some interesting relationships between coding schemes and connectivity in cortex , but it did not make any statements about the anticipated distribution of inhibitory connections . interestingly , there is a clear correlation between the strengths of the inhibitory connection between pairs of sailnet neurons , and the overlap ( measured by vector dot product ) between their receptive fields : neurons with significantly overlapping receptive fields tend to have strong inhibitory connections between them ( fig . this correlation is expected because cells with similar rf s receive much common feed - forward input . thus , in order to keep their activities uncorrelated , significant mutual inhibition is required . this same feature was assumed by the lca algorithm of rozell @xcite and colleagues , but is naturally learned by sailnet , in response to natural stimuli . our connectivity predictions are amenable to direct experimental testing , although that testing may be challenging , owing to the difficulty of measuring functional connectivity mediated by two or more synaptic connections between pairs of v1 excitatory simple cells . the present work represents the first demonstration that synaptically local plasticity rules can be used to learn a sparse code for natural images that accounts for the diverse shapes of v1 simple cell receptive fields . our model uses purely synaptically local learning rules connection strengths are updated based only on the number of spikes arriving at the synapse and the number of spikes generated by the post - synaptic cell . by contrast , the local competition algorithm ( lca ) of rozell and colleagues @xcite assumes that @xmath39 , so that the strength of the inhibitory connection between two neurons is equal to the overlap ( _ i.e. _ , vector dot product ) between their receptive fields . this non - local rule requires that individual inhibitory synapses must somehow keep track of the changes in the receptive fields of many neurons throughout the network in order to update their strengths . moreover , the lca network does not contain spiking units , even though cortical neurons are known to communicate via discrete , indistinguishable , spikes of activity @xcite . similarly , the units in the networks of falconbridge _ et al . _ @xcite and fldik @xcite communicate via continuous - valued functions of time . although these two models @xcite do use synaptically local plasticity rules , neither of these groups demonstrated that such local plasticity rules are sufficient to explain the diversity of simple cell rf shapes observed in v1 . we note that , independent of the present work , rozell and shapero have recently implemented a spiking version of lca @xcite that uses leaky integrate - and - fire units ( s. shapero , d. brderle , p. hasler , and c. rozell , cosyne 2011 abstract ) . however , that work does not address the issue of how to train such a network using synaptically local plasticity rules . some groups have used spiking units to perform image coding @xcite , but those studies did not address the question of whether synaptically local plasticity rules can account for the observed diversity of v1 rf shapes . interestingly , it has been demonstrated @xcite that orientation selectivity can arise from spike timing dependent plasticity rules applied to natural scenes . previous work @xcite has also explored the addition of homeostatic mechanisms to sparse coding algorithms and found it to improve the rate at which learning converges and to qualitatively affect the shapes of the learned rfs ; homeostasis is enforced in our model via modifiable firing thresholds . finally , we note that one previous group @xcite has demonstrated that independent component analysis ( ica ) can be implemented with spiking neurons and local plasticity rules . that work did not , however , account for the diverse shapes of v1 receptive fields , although they did also demonstrate that homeostasis ( a mean firing rate constraint ) was critical to the learning process . our model attempts to be biophysically realistic , but it is not a perfect model of visual cortex in all of its details . in particular , like many previous models @xcite , our network alternates between brief periods of inference ( the representation of the input by a specific population activity pattern in the network ) and learning ( the modification of synaptic strengths ) , which may not be realistic . indeed , it is unclear how cortical neurons would know " when the inference period is over and when the learning period should begin , though it is interesting to note that these iterations could be tied to the onset of saccades , given the @xmath40 inference period between learning " stages in our model . as in previous models , the inputs to our network @xmath41 are continuous - valued , whereas the actual inputs from the lateral geniculate nucleus to primary visual cortex ( v1 ) are spiking . as mentioned above , suppressive interactions between pairs of units in our model are mediated by direct , one - way , inhibitory synaptic connections between units , rather than being mediated by a distinct population of inhibitory interneurons . we do not include the effects of spike - timing dependent plasticity @xcite , although this has been shown to have interesting theoretical implications for cortex @xcite in general and for image coding in particular @xcite . we are currently developing models that incorporate spike timing dependent learning rules , applied to time - varying image stimuli such as natural movies . finally , the neurons in our model have no intrinsic noise in their activities , although that noise may , in practice , be small @xcite . interestingly , since our model neurons require a finite amont of time to update their internal variables @xmath2 , there is a hysteresis effect if one presents the network with time - varying image stimuli the content of previous frames affects how the network processes and represents the current frame . even if the features in a movie change slowly , the optimal representation of one frame can be very different from the optimal representation of the next frame in many coding models , so this hysteresis effect can provide stability to the image representation compared to other models such as ica @xcite or olshausen and field s sparsenet @xcite . this effect has previously been studied by rozell and colleagues @xcite , encouraging our efforts to apply sailnet to dynamic stimuli . though it is highly simplified , our model does captures many qualitative features of v1 , such as inhibitory lateral connections @xcite , largely uncorrelated neuronal activities @xcite , sparse neuronal activity @xcite , a greater number of cortical neurons than input neurons ( over - complete representation ) , synaptically local learning rules , and spiking neurons . importantly , this model allows us to make several falsifiable experimental predictions about interneuronal connectivity and population activity in cortex . we hope that these predictions will help uncover the coding principles at work in the visual cortex . each of the neurons in our sailnet follows leaky integrate - and - fire dynamics @xcite . the neurons , indexed by subscript @xmath24 , each have a time - dependent , continuous - valued internal variable @xmath2 , analogous to a neuronal membrane potential . we explicitly model each neuron as an rc circuit ( fig . 1 ) , where the internal variable @xmath2 corresponds to the voltage across the capacitor . whenever this internal variable exceeds a threshold value @xmath5 specific to that neuron , the neuron emits a punctate spike of activity . the unit s external variable @xmath3 , which represents the spiking output that is communicated to other neurons throughout the network , is @xmath42 for a brief moment . at all other times , the unit s external variable is @xmath43 . since the thresholds @xmath5 are adapted slowly compared to the time scale of inference , they are approximately constant during inference . the same is true for the feed - forward weights @xmath21 and the lateral connection strengths @xmath36 , discussed below . we model the effects of the input image @xmath18 and the activities of other neurons in the network @xmath44 on the internal variable as a current , @xmath45 , that is impinging on the rc circuit ; here the feed - forward weights @xmath21 and lateral connection strengths @xmath36 describe how much a given input ( either an image pixel value , or a spike from another neuron in the network ) should modify the neuron s internal variable . the internal variable evolves in time via the differential equation for voltage across our capacitor , in response to the input current @xmath46 . we simulate these dynamics in discrete time , performing numerical integration of the differential equation @xmath46 . whenever the internal variable @xmath2 exceeds the threshold ( at time @xmath47 ; @xmath48 ) , the output spike occurs at the next time step : @xmath49 . in the subsequent time step , the external variable @xmath50 returns to zero , unless the internal variable @xmath2 has again crossed the threshold . after the unit spikes , the internal variable returns to its resting value of @xmath43 , from whence the unit can again be charged up . for simplicity , our differential equation assumes that the rc time constant of the model neuron is one unit " of time . our simulated dynamics are allowed to run for five such units of time ( with the time step of numerical integration being 0.1 units in duration ) , in response to each input image . at the start of these dynamics , the internal variables of all neurons are set to their resting values : @xmath51 . unlike previous work @xcite , which performed unconstrained optimization on a cost function penalizing both reconstruction error and network activity , our learning rules can be viewed as a gradient descent approach to a _ constrained _ optimization problem . given the neuronal activities @xmath20 in response to an image , and their feed - forward weights @xmath21 , one can form a linear generative model @xmath52 of the input stimulus @xmath19 . the mean squared error between that model @xmath52 and the true input @xmath7 is @xmath53 , and the creation of a high fidelity representation suggests that this error function @xmath54 , or one like it , be minimized by the learning process . let us suppose that the neuronal network is not free to choose any solution to this problem ; instead it must satisfy constraints that require the neurons to have a fixed average firing rate of @xmath12 and minimal correlation between neurons . indeed , neurons tend to have low mean firing rates when averaged across many different images , and those firing rates span a finite range of values @xcite , motivating our first constraint . the second constraint is justified by observations that neural systems tend to exhibit little or no correlation between pairs of units @xcite , and that the correlation between the activity of v1 neurons decreases significantly as one increases the fraction of the visual field that is stimulated @xcite . we use the method of lagrange multipliers to solve this problem , allowing our learning rules to adapt the network so as to minimize reconstruction error while approximately satisfying these constraints . to do this , we perform gradient descent on a lagrange function @xmath55 that contains both the error function and the constraints : @xmath56 where the sets of values @xmath57 and @xmath58 are our ( unknown ) lagrange multipliers . to perform constrained optimization , gradient descent is performed with respect to all of the free parameters in @xmath55 : namely , the set of feed - forward weights @xmath59 , and the lagrange multipliers @xmath60 and @xmath61 : @xmath62 the first two equations lead to our learning rules for inhibitory connections and firing thresholds , once we identify @xmath63 and @xmath64 ; these network parameters correspond to the lagrange multipliers of the constrained optimization problem . this reflects the fact that the role of the variable thresholds and inhibitory connections is to enforce the sparseness and non - correlation constraints in the network , which is the same as the role of the lagrange multipliers in the lagrange function . we emphasize that the terms of our objective function that effectively enforce these constraints are critical for our algorithm s success . by contrast , consider the situation in which the model units had no other possibility but to maintain their fixed firing rate and lack of correlation , due to some clever parameterization of the model s state space . in that case , one could simply minimize the reconstruction error , via gradient descent , and the existence of these extra terms , or even of the analogous lagrange multipliers , would be redundant . however , in our model , each change of the feed - forward weights ( @xmath21 ) could change the neuron s firing rate , and the correlation between its activity and those of other neurons , unless something forces the network back towards the constraint surface . the variable firing thresholds and inhibitory inter - neuronal connection strengths in our model perform this function . the last equation from our gradient descent calculation gives the update rule for the feed - forward weights @xmath65 . this rule , as written , is unacceptable for our sailnet because we wish to interpret the strengths of connections in that network as the strengths of synaptic connections in cortex . in that case , learning at any given synapse should be accomplished using only information available locally , at that synapse . for updating connection strength @xmath21 , this could include the pre - synaptic activity @xmath6 , the post - synaptic activity @xmath20 , and the current value of the connection strength @xmath21 , but should not require information about the receptive fields of other neurons in the network , nor their activities , because it is not clear that that information is available at each synapse . hence , the @xmath66 term that arises from gradient descent on our objective function is a problem for the biological interpretation of these learning rules . we will now show that , in the limit that the neuronal activity is sparse and uncorrelated , when averaged over several input images , the non - local gradient descent rule @xmath65 is approximately equivalent to a simpler rule , originally due to oja @xcite , that is synaptically local . consider the non - local update rule @xmath65 . expanding the polynomial , and averaging over image presentations , we find @xmath67 if the learning rate @xmath14 is small , such that the feed - forward weights change only slowly over time , then we can approximate that they are constant over some ( small ) number of image presentations , and take them outside of the averaging brackets ; @xmath68 now , so long as the neuronal activities are uncorrelated , and all units have the same average firing rate ( recall these constraints are enforced by our lagrange multipliers ) , @xmath69 , and thus the learning rule is @xmath70 this last term is small compared to the first two for a few reasons . first , the neurons in the network have _ sparse _ activity , meaning they are selective to particular image features , and thus @xmath71 . this can be easily seen by that fact that we use small values for @xmath12 , meaning that the neurons fire , on average , much less than one spike per image . the spikes , however , can only be emitted in integer numbers , so the neurons are silent in response to most image presentations , and are thus highly selective . furthermore , the last term , @xmath72 , involves a sum over the receptive fields of many neurons in the network . some of the rfs will be positive for a given pixel , whereas others will be negative . these random signs mean that the sum @xmath73 tends towards zero . thus , in the limit of sparse and uncorrelated neuronal activity ( the limit in which our network operates ) , gradient descent on the error function @xmath54 yields approximately @xmath74 which is equivalent to the average update from oja s implementation of hebbian learning @xcite , which we use for learning in sailnet . thus , sailnet learns to approximately solve the same error minimization problem as did previous , non - local sparse coding algorithms @xcite . interestingly , our result suggests that , despite the highly non - linear way in which our model neurons outputs ( spikes @xmath20 ) are generated from the input , a linear decoding of the network activity should provide a good match to the input : @xmath75 . this linear decodability has previously been observed in physiology experiments @xcite , as well as models designed to maximize the information rate about input stimulus conveyed by individual spiking neurons @xcite , and it is indeed a property of sailnet . we summarize the learning rules for sailnet here . @xmath76 the first two rules enforce the sparseness and correlation constraints , and arise from the lagrange multipliers in our lagrange function . the final rule drives the sailnet representation to form a better match to the input stimulus , as it adapts to the ensemble of training images . consider the oja - hebb @xcite learning rule for the feed - forward weights in our model , @xmath77 once the learning has converged over some set of training stimuli , the feed - forward weights are , on average , no longer changing in response to repeated presentations of examples from the training set . thus , @xmath78 expanding the middle term in this expression , we find that @xmath79 where the second equality occurs because the learning has converged , and thus the feed - forward weights are constant over repeated image presentations . thus , we find that @xmath80 ; the spike - triggered average ( sta ) stimulus is equivalent to the set of feed - forward weights , up to a multiplicative scaling factor that can be calculated from the spike train . we start out each simulation with all inhibitory connection strengths @xmath36 set to zero , all firing thresholds @xmath5 set to @xmath81 , and the feed - forward weights @xmath21 initialized with gaussian white noise . to train the network , batches @xcite of 100 images with zero mean , and unit standard deviation pixel values , are presented , and the number of spikes from each neuron are counted separately for each image . after each batch , the average update for the network properties is computed ( following our learning rules ) over the 100-image batch . this batch - wise training lets us use matrix operations for computing the updates , which dramatically speeds up the training process . after each update , all negative values for inhibitory connections @xmath36 ( which would correspond to excitatory connections ) are set to zero , as in the previous work by fldik @xcite . relaxing this constraint , and allowing the recurrent weights to change sign does not affect our qualitative conclusions . in that case , some of the recurrent connections become excitatory , while the majority remain inhibitory , the rf s are qualitatively the same as those shown in fig . 3 , and the distributions of inhibitory and excitatory connection strengths are both approximately lognormal ( data not shown ) . the relative values of @xmath82 and @xmath15 were chosen based on fldik s @xcite observation that @xmath14 must be much less than @xmath13 or @xmath15 so that the neurons activities remain sparse and uncorrelated , even in the face of changing feed - forward weights . we study the network after the properties stop changing macroscopically over time . however , as noted in the firing rates section of this paper , the network parameters continue to bounce around the final target " state , with the size of the bounces determined by the learning rates in the network . empirically , we find that it takes on the order of @xmath83 image presentations ( @xmath84 steps of 100 image presentations per step ) for this dynamic equilibrium to be established . for the results presented in this paper , we let the network train for roughly @xmath85 image presentations . to speed up the simulation , we start the training with large values for the learning rates , and these are eventually reduced . for the last @xmath86 batches of training ( @xmath87 image presentations ) , the learning rates were @xmath88 . all of the computer codes used to generate the results presented in this paper are available upon request . a gabor function ( @xmath89 ) is a common model for visual cortical receptive fields @xcite , which consists of a two - dimensional gaussian multiplied by a sinusoid : @xmath90 , \\ x_p & = & ( x - x_0 ) \cos(\theta ) + ( y - y_0 ) \sin(\theta ) , \nonumber \\ y_p & = & -(x - x_0 ) \sin(\theta ) + ( y - y_0 ) \cos(\theta ) . \nonumber\end{aligned}\ ] ] the center of the shape is defined by the coordinates @xmath91 and @xmath92 , while the amplitude and orientation of the pattern are defined by the parameters @xmath93 and @xmath94 , respectively . @xmath95 defines the phase of the sinusoid , relative to the center of the gaussian envelope , which has spatial extent @xmath96 and @xmath97 in the direction along , and perpendicular to , the direction in which the sinusoid oscillates ( with frequency @xmath98 ) , respectively . given a neuronal rf , our code performs unconstrained optimization to choose the gabor parameters such that the mean squared error @xmath99 is minimized . we then perform several quality control measures to ensure that our analysis only contains sensible gabor parameters that accurately describe our rfs . the first such measure is to exclude any rf for which the deviation between the rf and the gabor fit is large ; cells with @xmath100 were excluded . this is equivalent to placing a ( fairly mild ) restriction on the minimum allowable signal - to - noise ratio . the second quality control measure is to exclude those rfs for which the center of the pattern @xmath101 lies either outside the @xmath26 pixel patch , or within one standard deviation ( of the gaussian envelope ) of the patch edge . as described by other workers @xcite , when the center of the pattern is outside of the visible @xmath102 pixel patch , it is not clear that the shape of the rf itself is well - described by the gabor parameters , or even well - constrained , for that matter . our ( more stringent ) restriction also avoids the problem of biased shape estimates , when fitting gabors to rfs that are truncated by the edge of the patch ; the model rfs essentially tile the available space , so some of them will , by necessity , have centers that lie right along or outside of the edges of the patch . indeed , in a @xmath26 pixel space , many pixels are near the edge , thus this cut excludes many rfs . after making all of these cuts , we were left with 299 rfs , on which to perform subsequent shape analysis . we performed the same fitting and quality - control analysis on both the sailnet and the macaque physiology rfs , although we used a gentler goodness - of - fit restriction on the macaque data , since the macaque rfs , as measured , have fairly large regions of zero support , in which any measurement noise reduces the apparent goodness - of - fit . for the macaque data , we excluded those rfs with @xmath103 , leaving 116 of the 250 macaque rfs for subsequent analysis . one reason for the relatively low yield of well - fit rfs is that not all rfs are actually well - described by gabor functions . for example , there is no choice of gabor parameters that will accurately describe a center - surround receptive field ; that rf is much better described by a difference of gaussians function , for example . we leave for future work the issue of determining the best family of functions with which to describe visual cortical receptive fields . 1 attneave , f. ( 1954 ) . some informational aspects of visual perception . 61 , 18393 . barlow , h.b . possible principles underlying the transformations of sensory messages . in sensory communication , rosenblith , ed . ( cambridge , ma : mit press ) , pp . 21734 . atick , j.j . and redlich , a.n . ( 1992 ) . what does the retina know about natural scenes ? neural comput . 4 , 196 - 210 . bialek , w. , deweese , m. , rieke , f. and warland , d. ( 1993 ) . bits and brains : information fow in the nervous system . physica a 200 , 581 - 593 . bialek , w. , ruderman , d.l . and zee , a. ( 1991 ) . optimal sampling of natural images : a design principle for the visual system . in advances in neural information processing 3 , lippman , r.p . , moody , j.e . and touretzky , d.s . ( san mateo , ca : morgan kaufmann ) , pp . 363 - 369 . linsker , r. ( 1989 ) . an application of the principle of maximum information preservation to linear systems . in advances in neural information processing 1 . , touretzky , d. ed . ( san mateo , ca : morgan kaufmann ) , pp.186 - 194 . atick , j.j . could information theory provide an ecological theory of sensory processing ? in princeton lectures on biophysics , bialek , w. ed . ( singapore : world scientific ) . bell , a.j . and sejnowski , t.j . ( 1997 ) the independent components " of natural scenes are edge filters . 37 , 3327 - 3338 . hyvrinen , a. and hoyer , p.o . a two - layer sparse coding model learns simple and complex cell receptive fields and topography from natural images . 41 , 2413 - 2423 . tkaik , g. , prentice , j.s . , balasubramanian , v. and schneidman , e. ( 2010 ) . optimal population coding by noisy spiking neurons . usa 107 , 14419 - 14424 . laughlin , s.b . a simple coding procedure enhances a neuron s information capacity . z. 36c , 910 - 912 rieke , f. , warland d. , de ruyter van steveninck , r. , and bialek , w. ( 1999 ) . spikes : exploring the neural code ( cambridge , ma : mit press ) . dan , y. , atick , j.j . and reid , c.r . efficient coding of natural scenes in the lateral geniculate nucleus : experimental test of a computational theory . j. neurosci . 16 , 3351 - 3362 . deweese , m. ( 1996 ) . optimization principles for the neural code . network 7 , 325 - 31 . olshausen , b.a . and field , d.j . emergence of simple - cell receptive field properties by learning a sparse code for natural images . nature 381 , 607 - 609 . olshausen , b.a . and field , d.j . sparse coding with an overcomplete basis set : a strategy employed by v1 ? 37 , 3311 - 3325 . fldik , p. ( 1990 ) . forming sparse representations by local anti - hebbian learning . 64 , 165 - 170 . graham , d.j . and field , d.j . sparse coding in the neocortex . in evolution of nervous systems , iii , j. h. kass et al . ( oxford : academic press ) , pp . 181 - 187 . hromdka , t. , deweese , m.r . and zador , a.m. ( 2008 ) . sparse representation of sounds in the unanesthetized auditory cortex . plos biol . 6 , 124 - 137 . vinje , w.e . and gallant , j.l . natural stimulation of the nonclassical receptive field increases information transmission efficiency in v1 . j. neurosci . 22 , 2904 - 2915 . vinje , w.e . and gallant , j.l . sparse coding and decorrelation in primary visual cortex during natural vision . science 287 , 1273 - 1276 . haider , b.a . , krause , m.r . , duque , a. , yu , y. , touryan , j. , mazer , j.a . and mccormick , d.a . synaptic and network mechanisms of sparse and reliable visual cortical activity during nonclassical receptive field stimulation . neuron 65 , 107 - 121 . lennie , p. ( 2003 ) . the cost of cortical computation . 13 , 493 - 497 . tolhurst , d.j . , smyth , d. and thompson , i.d . the sparseness of neuronal responses in ferret primary visual cortex . j. neurosci . 29 , 2355 - 2370 . sakata , s. and harris , k.d . laminar structure of spontaneous and sensory - evoked population activity in auditory cortex . neuron 64 , 404 - 18 . ringach , d.l . spatial structure and symmetry of simple - cell receptive fields in macaque primary visual cortex . j. neurophysiol . 88 , 455 - 63 . rehn , m. and sommer , f.t . a network that uses few active neurones to code visual input predicts the diverse shapes of cortical receptive fields . j. comput . 22 , 135 - 146 . donoho , d.l . compressed sensing . ieee tran . inform . theory . 52 , 1289 - 1396 . olshausen , b.a . , cadieu , c.f . , and warland , d.k . learning real and complex overcomplete representations from the statistics of natural images . spie 7446 , 74460s-1 - 74460s-11 . dayan , p. and abbott , l. ( 2001 ) . theoretical neuroscience : computational and mathematical modelling of neural systems ( cambridge , ma : mit press ) . perrinet , l. , sameulides , m. and thorpe , s. ( 2004 ) . sparse spike coding in an asynchronous feed - forward multi - layer neural network using matching pursuit . neurocomputing 57 , 125 - 134 . delorme , a. , perrinet , l. and thorpe , s.j . networks of integrate - and - fire neurons using rank order coding b : spike timing dependent plasticity and emergence of orientation selectivity . neurocomputing 38 , 539 - 545 . perrinet , l. ( 2010 ) . role of homeostasis in learning sparse representations . neural comput . 22 , 1812 - 1836 . masquelier , t. and thorpe , s.j . unsupervised learning of visual features through spike timing dependent plasticity . plos comput . 3 , e31 . vanrullen , r. and thorpe , s.j . surfing a spike wave down the ventral stream . 42 , 2593 - 2615 . savin , s. , joshi , p. and triesch , j. ( 2010 ) . independent component analysis in spiking neurons . plos comp . biol . 6 , e1000757 . falconbridge , m.s . , stamps , r.l . and badcock , d.r . a simple hebbian / anti - hebbian network learns the sparse , independent components of natural images . neural comput . 18 , 415 - 429 . song , s. , sjstrm , p.j . , reigl , m. , nelson , s. and chklovskii , d.b . highly nonrandom features of synaptic connectivity in local cortical circuits . plos biol . 3 , 0507 - 0519 . rozell , c.j . , johnson , d.h . , baraniuk , r.g . and olshausen , b.a . sparse coding via thresholding and local competition in neural circuits . neural comput . 20 , 2526 - 2563 . hopfield , j. ( 1984 ) . neurons with graded response have collective computational properties like those of two - state neurons . usa 81 , 3088 - 3092 . oja , e. ( 1982 ) . a simplified neuron model as a principal component analyzer . 15 , 267 - 273 . bialek , w. , rieke , f. , de ruyter van steveninck , r.r . and warland , d. ( 1991 ) . reading a neural code . science 252 , 1854 - 1857 . gaese , b.h . and ostwald , j. ( 2003 ) . complexity and temporal dynamics of frequency coding in the awake rat auditory cortex . eur . j. neurosci . 18 , 26382652 . baddeley , r. , et al . responses of neurons in primary and inferior temporal visual cortices . b. 264 , 1775 - 1783 . abeles , m. , vaadia , e. and berman , h. ( 1990 ) . firing patterns of single units in the prefrontal cortex and neural network models . network 1 , 13 - 25 . clopath , c. , bsing , l. , vasilaki , e. and gerstner , w. ( 2010 ) . connectivity reflects coding : a model of voltage - based stdp with homeostasis . 13 , 344 - 352 . dan , y. and poo , m.m . spike timing - dependent plasticity : from synapse to perception . 86 , 1033 - 1048 . ecker , a.s . et al . ( 2010 ) . decorrelated neuronal firing in cortical microcircuits . science 327 , 584 - 587 . renart , a. et al . the asynchronous state in cortical circuits . science 327 , 587 - 590 . garrigues , p.j . and olshausen , b.a . learning horizontal connections in a sparse coding model of natural images . in advances in neural information processing systems platt , d. koller , y. singer and s. roweis eds . ( san mateo , ca : morgan kaufmann ) , pp . 505 - 512 . koulakov , a.a . , hromdka , t. and zador , a.m. ( 2009 ) . correlated connectivity and the distribution of firing rates in the neocortex . j. neurosci . 29 , 3685 - 3694 . mainen , z.f . and sejnowski , t.j . ( 1995 ) . reliability of spike timing in neocortical neurons . science 286 , 1503 - 1506 . the authors are grateful to bruno olshausen for his role in inspiring this work and to bruno olshausen , jascha sohl - dickstein , fritz sommer , and the other members of the redwood center for many helpful discussions . we are also very grateful to dario ringach for providing the macaque physiology data , and to chris rozell and jascha sohl - dickstein for constructive comments on the manuscript . the work of jz is supported by funds from the university of california and a fulbright international science and technology phd fellowship . mrd gratefully acknowledges support from the national science foundation , the mcknight foundation , the mcdonnell foundation , and the hellman family faculty fund . jz initiated the project , designed and implemented the simulations , and performed the analytic calculations . jtm implemented the gabor fitting routine . jz performed the remainder of the data analysis . jz and mrd wrote the manuscript . to the network ( from image pixels ) contact the neuron at connections ( synapses ) with strengths @xmath21 , whereas inhibitory recurrent connections between neurons @xcite in the network have strengths @xmath36 . the outputs of the neurons are given by @xmath3 ; these spiking outputs are communicated through the recurrent connections , and also on to subsequent stages of sensory processing , such as cortical area v2 , which we do not include in our model . ( b ) circuit diagram of our simplified leaky integrate - and - fire @xcite neuron model . the inputs from the stimulus with pixel values @xmath6 , and the other neurons in the network , combine to form the input current @xmath45 to the cell . this current charges up the capacitor , while some current can leak to ground through a resistor in parallel with the capacitor . the resistors are shown as cylinders to highlight the fact that they model the collective action of ion channels in the cell membrane . the internal variable evolves in time via the differential equation for voltage across our capacitor , in response to input current @xmath104 : @xmath46 , which we simulate in discrete time . once that voltage exceeds threshold @xmath5 , the diode , which models neuronal voltage - gated ion channels , opens , causing the cell to fire a punctate action potential , or spike , of activity . for sake of a complete circuit diagram , the output is denoted as the voltage , @xmath105 , across some ( small : @xmath106 ) resistance . after spiking , the unit s internal variable returns to the resting value of @xmath43 , from whence it can again be charged up.,width=288 ] pixel patches , each of which was preprocessed so as to have zero - mean and unit variance of the pixel values ( like the training set ) . each patch was presented to sailnet , and the number of spikes were recorded from each unit in response to each patch . a linear decoding of sailnet activity for each patch @xmath19 was formed by multiplying each unit s activity by that unit s rf and summing over all neurons . the preprocessing was then inverted , and the patches were tiled together to form the image in panel ( b ) . the decoded image resembles the original , but is not identical , owing to the severe compression ratio ; on average , each @xmath26 input patch , which is defined by 256 continuous - valued parameters , is represented by only 75 binary spikes of activity , emitted by a small subset of the neural population . linear decodability is a product of our learning rules , and it is an observed feature of multiple sensory systems @xcite and spiking neuron models optimized to maximize information transmission @xcite.,width=576 ] , whereas those to the left of it span @xmath107 . ( b ) rfs of 196 randomly selected model neurons from a 1536-unit sailnet trained on patches drawn from whitened natural images . the gray value in all squares represents zero , whereas the lighter pixels correspond to positive values , and the darker pixels correspond to negative values . all rfs are sorted by a size parameter , determined by a gabor function best fit to the rf . the sailnet model rfs show the same diversity of shapes as do the rfs of simple cells in macaque monkey v1 ( a ) ; both the model units and the population of recorded v1 neurons consist of small unoriented features , oriented gabor - like wavelets containing multiple subfields , and elongated edge - detectors . ( c ) we fit the sailnet and macaque rfs to gabor functions ( see methods section ) , in order to quantify their shapes . shown are the dimensionless _ width _ and _ length _ parameters ( @xmath108 and @xmath109 , respectively ) of the 299 sailnet rfs and 116 ( out of 250 rfs in the dataset ) macaque rfs for which the gabor fitting routine converged . these parameters represent the size of the gaussian envelope in either direction , in terms of the number of cycles of the sinusoid . the sailnet data ( open blue circles ) span the space of the macaque data ( solid red squares ) from our gabor fitting analysis ; sailnet is accounting for all of the observed rf shapes . we highlight four sailnet rfs with distinct shapes , which are identified by the large triangular symbols that are also displayed next to the corresponding rfs in panel ( b ) . , width=528 ] image patches drawn from the training ensemble for each of the 1536 units of a sailnet trained on whitened natural images . all learning rates were set to zero during probe stimulus presentation . a wide range of mean rates was observed , but as expected , the distribution is peaked near @xmath35 spikes per image , the target mean firing rate of the neurons . the paucity of units with near - zero firing rates suggests that this distribution is closer to lognormal than exponential . accordingly , the lognormal least - squares ( solid red curve ) fit accounts for @xmath110 of the variance in the data , whereas the exponential fit ( black dashed curve ) accounts for only @xmath111 . ( b ) in response to low contrast stimuli , the firing rate distribution across the units ( every unit fired at least once ) in the same network as in panel ( a ) was similarly well fit by either an exponential ( dashed black curve ; accounting for @xmath112 of the variance in the data ) or a lognormal function ( solid red curve ; accounting for @xmath113 of the variance ) . the low - contrast stimulus ensemble used to probe the network consisted of images drawn from the training set , with all pixel values reduced by a factor of three.,width=528 ] ) learned by a 1536 unit sailnet trained on @xmath102 pixel patches drawn from whitened natural images . the measured values ( blue points ) are well - described by a gaussian distribution ( solid line ) , which accounts for @xmath114 of the variance in the dataset . this indicates that the data are approximately lognormally distributed . note that there are some systematic deviations between the gaussian best fit and the true distribution , particularly on the low - connection strength tail , similar to what has been observed for excitatory connections within v1 @xcite . this plot was created using the binning procedure of hromdka and colleagues @xcite . the histogram was normalized to have unit area under the curve . ( b ) the strengths of the inhibitory connections between pairs of cells are correlated with the overlap between those cells receptive fields : cells with significantly overlapping rfs tend to have strong mutual inhibition . data shown in panel ( b ) are for 5,000 randomly selected pairs of cells . pairs of cells with significantly negatively overlapping rfs tend not to have inhibitory connections between them , hence the apparent asymmetry in the rf overlap distribution obtained by marginalizing over connection strengths in panel ( b).,width=528 ]

when a spinning system experiences a transient gravitational encounter with an external perturber , a _ quasi - resonance _ occurs if the spin frequency of the victim matches the peak orbital frequency of the perturber . such encounters are responsible for the formation of long tails and bridges of stars during galaxy collisions . for high - speed encounters , the resulting velocity perturbations can be described within the impulse approximation . the traditional impulse approximation , however , does not distinguish between prograde and retrograde encounters , and therefore completely misses the resonant response . here , using perturbation theory , we compute the effects of quasi - resonant phenomena on stars orbiting within a disk . explicit expressions are derived for the velocity and energy change to the stars induced by tidal forces from an external gravitational perturber passing either on a straight line or parabolic orbit . comparisons with numerical restricted three - body calculations illustrate the applicability of our analysis .

1009.3927

astrophysical objects dominated by disks supported against gravity by rotation are ubiquitous . they range from planetary rings , to planetary systems , to protoplanetary disks , to accretion disks around young stars and compact objects , to spiral galaxies . given that astrophysical disks are common , it is not surprising that gravitational encounters involving them and other objects occur frequently . in particular , the numerous examples of peculiar galaxies ( e.g. , * ? ? ? * ; * ? ? ? * ) suggest that tidal interactions between spirals are a key driver of the morphological transformations of galaxies . beginning in the 1960s , various studies established that many peculiar galaxies are , in fact , disks in collision @xcite . however , it was @xcite ( hereafter , tt72 ) who performed the first systematic study of this process , using a restricted three - body technique . in their approach , each galaxy was modeled as a point - particle , representing the potential well , surrounded by a disk of non - interacting test particles ( stars ) on circular orbits . when two such model galaxies pass by one another , their mutual gravitational tidal forcing distorts the disks as traced by the orbital motion of the test particles in the combined potential of the two point - particles . in this manner , tt72 showed that , contrary to prevailing wisdom , the narrow bridges and tails seen in many peculiar galaxies could be produced by gravity alone , and argued that these features are essentially kinematic in nature . tt72 s analysis further demonstrated that the bridges connecting pairs of colliding galaxies and the tidal tails that develop continue to lengthen and thin out after an encounter . the well - known systems ngc 4038 ( the `` antennae '' ) , ngc 4676 ( the `` mice '' ) , and ngc 7252 ( the `` atoms for peace galaxy '' ) all have tails extending 50 - 100 kpc in length @xcite . the superantennae ( iras19254 - 7245 ) is an extreme case in which the tails span 350 kpc from tip to tip @xcite . if the galaxies merge , some tail stars may escape the system entirely , but most eventually fall back into the merger remnant , resulting in the formation of `` shells '' and other fine structures associated with elliptical galaxies @xcite through `` phase - wrapping '' @xcite . tt72 also showed that the efficiency of `` tail - making '' depends sensitively on the inclinations of the disks relative to the orbit plane . this is illustrated in figure 1 , which shows the outcome of two of tt72 s simulations in which disks are perturbed by the passage of equal - mass companions on parabolic orbits . the collisions are co - planar , meaning that the victim disks lie exactly in the orbit plane , but in the top row of figure 1 the encounter is prograde , so that the internal spin of the disk is aligned with the direction of the orbital angular momentum , whereas in the bottom row the two are anti - parallel , defining a precisely retrograde interaction . it is obvious from visual inspection of figure 1 that prograde encounters can do much more violence to spinning disks than retrograde ones . tt72 interpreted this difference physically as owing to a `` near - resonance or matching of their [ internal ] orbital speeds with the peak angular motion of the companion . '' in other words , a relatively strong response follows if the spin angular frequency of the disk is aligned with and of similar magnitude to the orbital angular frequency of the collision at pericenter . if the stars in the disk are on circular orbits , but the trajectory of the interaction is non - circular ( as in the examples of parabolic orbits shown in figures 1 ) , then the orbital angular frequency varies with time and the resonance is only temporary ; hence , a `` near- '' or `` quasi - resonance . '' mathematically , the condition for a strong response is expressed by @xmath0 where @xmath1 and @xmath2 are the internal rotation velocity and characteristic size of the disk , respectively , @xmath3 and @xmath4 are the orbit velocity and separation at pericenter , and @xmath5 the eccentricity of the orbit @xcite . the work of @xcite and @xcite generalized tt72 s modeling by treating the internal gravity of colliding galaxies self - consistently . these and related studies verified tt72 s claim that tail - making is a kinematic process ( see figure 3 of * ? ? ? * ) and examined the sensitivity of tail - making efficiency to the distribution of stars in interacting and merging galaxies . subsequently , @xcite and @xcite demonstrated empirically that the lengths and kinematics of tidal tails also depend critically on the distribution of dark matter surrounding each galaxy . for the same radial stellar profiles , longer ( shorter ) and more ( less ) prominent tidal tails form in shallower ( deeper ) dark matter potential wells . in particular , @xcite and @xcite showed that the depth of the dark matter potential is as important in determining the lengths of tidal tails as the orbit ( see e.g. figure 4 in * ? ? ? * ) . similar conclusions have been reached in studies of local dwarf spheroidal galaxies . most recently , it has been shown that resonant stripping of stars in disks can alter the mass to light ratios of dwarf galaxies when they encounter more massive systems by removing luminous material more efficiently than dark matter . the minimal response of the dark matter is expected if its particles move on random orbits , in which case the net perturbation on the halo mostly averages out @xcite . a similar resonant phenomenon has been suggested to cause the lmc disk to thicken by interactions with the milky way @xcite or more generally in the context of heating and disruption of satellites @xcite . in addition , the origin of the magellanic stream might have a tidal origin @xcite through interactions between dwarfs in groups @xcite , provided that the magellanic clouds are on their first pericentric passage @xcite . while these works have elucidated the physics of tail - making and proven that observed peculiar galaxies are indeed a consequence of collisions and mergers , they have left open a number of important issues . most previous studies of tidal interactions between disk galaxies have been empirical in nature . thus , they have identified broad conditions for tidal features to be produced , but have not provided simple criteria for isolating the efficiency of this process on specific parameters of an encounter . therefore , the circumstances under which very long tails could result , like those in the superantennae , remain uncertain . moreover , even nearly 40 years after tt72 , there is still considerable confusion in the community regarding galactic bridges and tails , especially in regards to the role played by resonances in their origin . finally , attempts to reproduce the morphology and kinematics of individual observed systems is greatly complicated by the sensitivity of tail - making to the distribution of dark matter around galaxies . one approach for dealing with this last issue is to survey the vast parameter space that allows for variations in the orbit , the relative masses and spatial sizes of the luminous galaxies , and the dark matter potentials . a promising scheme along these lines has been developed recently by @xcite with their identikit algorithm . in what follows , we pursue a complementary technique by formulating a simple analytic description of tail - making to understand it from an alternate perspective . our methodology necessarily entails various approximations and is hence less general than simulation - based procedures , but makes it possible to identify scalings of , and interpret physically , the response of rotating objects to gravitational tidal perturbations . while our immediate attention is devoted to galactic disks , our formalism is general with respect to the system under consideration , so we anticipate that our analysis will be relevant to a wide range of tidal phenomena . in 2 , we describe our method for analyzing tidal responses of rotating systems . we adopt a variant of the impulse approximation in which we allow particles in the perturbed system to move along their internal , unperturbed orbits during the encounter . in 3 , we derive analytic expressions in terms of special functions for the velocity perturbations delivered to spinning disks in these interactions , in the case of coplanar collisions . we do this first for high - speed , straight - line trajectories , in 3.1 , and then for parabolic orbits in 3.2 , the latter based partly on the work of @xcite for tidal interactions between stars . we present numerical examples and contrast the straight - line and parabolic cases in 3.3 . in 4 , we generalize our formalism to non - coplanar encounters , for both straight - line and parabolic orbits . we compare the results of our analytic prescriptions to simulations of tidal interactions in 5 . finally , we summarize and conclude in 6 . our approach is based on a variant of the impulse approximation for studying gravitational perturbations on systems . in some respects , our formalism has similarities to that of @xcite , who considered tidal interactions between stars as a means for forming close binaries . @xcite calculated the response of gas spheres to external gravitational perturbations and the energy deposited into non - radial oscillations . later , similar analytic estimates of the energy and angular momentum exchange between a circumstellar disk and a passing star on a near - parabolic orbit were inferred by @xcite in the context of accretion disks . consider a flat , rotationally supported disk of stars , perturbed gravitationally by a passing object . in the usual impulse approximation @xcite , it is assumed that the stars in the perturbed system remain strictly stationary during the course of the encounter . the tidal force from the perturber relative to the center of mass of the perturbed body is calculated at each location within it from each point along the relative orbit of the interaction . the total velocity impulse delivered to each star in the perturbed object is then calculated by integrating the force over the entire orbit . in the simplest application of this method , the perturber follows a straight - line trajectory , as for a high - speed encounter , although as we show in what follows , it is possible to generalize the technique also to parabolic collisions , following @xcite . the analytic expressions that result from the impulse approximation @xcite give reasonably accurate results for the energy deposited in objects during tidal encounters for systems that are dominated by internal random motions @xcite . however , this method is not appropriate for capturing the essentials of responses like those in figures 1 because , by construction , it does not distinguish between prograde and retrograde interactions , since the stars in the perturbed body are held fixed during the encounter . to qualitatively capture the influence of resonances during tidal interactions , we employ the following variant of the impulse approximation . during the course of the encounter , stars in the perturbed system are allowed to move along unperturbed orbits within their host . for simplicity , we assume that the unperturbed orbits are strictly circular , although modest departures from these paths could , in principle , be handled using epicyclic theory . in this manner , the response of a given star will depend not only on its spatial location within the perturbed system , but also its velocity . as we demonstrate explicitly below , this approach makes it possible to characterize the resonant aspects of tidal interactions that distinguish between prograde and retrograde collisions , as in figure 1 . if the orbit of the encounter is non - circular , as in the examples that follow , the orbital angular frequency varies with time and , so , a given star within the perturbed system will be in precise resonance with the motion of the perturber for only a limited time . for this reason , we will refer to the formalism herein as describing _ quasi - resonant _ behavior , meaning that the response only resembles that characteristic of a true resonant interaction . this meaning should be taken to be equivalent to tt72 s description of tidal interactions between spinning objects as displaying `` near - resonant '' qualities . to be specific , consider a disk of stars on circular orbits comprising the _ victim _ interacting with an object which we will refer to as the _ perturber_. employ a coordinate system with origin at the center of mass of the victim and assume that the disk is razor thin and orient the coordinate system so that the disk is in the @xmath6 plane . then , the coordinate vector to any star within the victim is given by @xmath7 where @xmath2 is constant , because we assume that the unperturbed orbits internal to the victim are circular , and the position vector to the perturber will be denoted by @xmath8 in the sections below , we will consider various choices for the trajectory @xmath9 , which will fix the time - dependence of the components @xmath10 . we will adopt the convention that the unperturbed disk always lies in the @xmath6 plane . thus , for non - coplanar encounters , we will incline the orbit plane defined by the trajectory @xmath9 so that @xmath11 will be non - zero . the acceleration of each star relative to that on the center of mass of the perturbed body is @xmath12 \ , , \ ] ] where @xmath13 is the interaction potential between the perturber and the victim , @xmath14 is the mass of the victim , and the integral is over the density profile of this object . expand @xmath15 about the origin in a taylor series using @xmath16 after algebra , taking into account that the origin is located at the center of mass of the victim , the @xmath17-th component of the acceleration becomes @xmath18 \ , + \ , \ldots \ , \biggr \rbrace \ , , \ ] ] where @xmath19 and @xmath20 is the moment of inertia tensor @xcite : @xmath21 for simplicity , treat the force on each star in the disk from the perturber as that from a point mass . then , the interaction potential is @xmath22 performing the derivatives required in the above expression , we obtain the acceleration of a particular star at a given time . the velocity impulse delivered by the encounter can then be obtained by integrating over time . thus , the leading term in the series is @xcite @xmath23 dt \ , .\ ] ] likewise , the next order correction term can be written @xmath24 \biggr \rbrace dt \ , . \label{correction}\ ] ] in principle , this procedure can be extended to even higher order terms in the series . note that the terms in this last equation are all @xmath25 since the elements in the matrix @xmath26 involve integrals over squares of the internal coordinates of the victim , while the terms in eq . ( [ impulse ] ) are @xmath27 . in what follows , we will employ eq . ( [ impulse ] ) as the starting point for our analysis . unlike as in the usual impulse approximation we will allow both @xmath9 _ and _ @xmath28 to vary in time . we will , however , assume that the trajectory of the interaction , specified by @xmath9 , is prescribed ( i.e. orbital decay is not accounted for ) , and that the orbital motion within the victim , set by @xmath28 , is such that the stars follow their unperturbed motions throughout the course of the interaction . to illustrate our approach , we begin by considering a coplanar interaction between a perfectly thin , rotating disk of stars and a passing perturber . as noted earlier , the origin of the coordinate system will be at the center of mass of the victim and the coordinate system will be oriented so that the disk lies in the @xmath6 plane . while not general , this case suffices to characterize the dynamics of such encounters ; we will generalize to non - coplanar collisions later . denote the mass of the victim disk by @xmath14 and that of the perturber by @xmath29 . the case of a perturber moving along a straight line relative to the victim is the simplest one to analyze and is appropriate for high - speed encounters , as in clusters of galaxies . for definiteness , take the orbit path to be @xmath30 where , as indicated in figure [ stripstraight ] , @xmath31 is the distance of closest approach ( the impact parameter ) , which occurs at time @xmath32 , and @xmath33 is the velocity of the encounter , which is constant for a straight - line trajectory . the internal motions of the disk particles are given by eq . ( [ vicorb ] ) with @xmath34 where @xmath35 the internal angular frequency of the victim and @xmath36 is the phase at the minimum distance from the perturber at @xmath32 . note that while @xmath35 is assumed to be constant in time , it can vary spatially according to @xmath37 , depending on the shape of the rotation curve of the victim . we will adopt the convention that @xmath31 and @xmath33 are non - negative and distinguish prograde versus retrograde collisions for co - planar encounters by the sign of @xmath38 so that @xmath39 we further define the non - negative parameter @xmath40 by @xmath41 the orbital angular frequency varies with time along the trajectory according to @xmath42 where @xmath43 . at the distance of closest approach , when @xmath32 , @xmath44 and so physically @xmath45 describes the condition for a resonance , with a strong response expected for @xmath46 when @xmath47 . employing the usual trigonometric relations @xmath48 @xmath49 and substituting into eq . ( [ impulse ] ) gives @xmath50\rm{du}\end{aligned}\ ] ] @xmath51\rm{du}\end{aligned}\ ] ] and @xmath52 . in these expressions , for the terms with @xmath53 or @xmath54 symbols the upper and lower signs correspond to prograde and retrograde cases , respectively . these integrals can be evaluated in terms of the modified bessel functions @xmath55 and @xmath56 @xcite . employing the recursion relation @xmath57 we arrive at @xmath58 \label{coplxstrufinal}\ ] ] @xmath59 , \label{coplystrufinal}\ ] ] where the upper and lower signs are for prograde and retrograde encounters , respectively , and @xmath60v@xmath61=0 for coplanar collisions . the structure of the expressions in terms of the modified bessel functions @xmath62 and @xmath55 is reminiscent of the results describing the perturbations of orbits of stars within disks owing to passing molecular clouds @xcite . it is of interest to consider various limiting cases for these expressions . when @xmath63 , corresponding to a slowly rotating system , the bessel functions asymptote to @xmath64 and @xmath65 @xcite , and we find @xmath66 @xmath67 which can be obtained from the usual result for the impulse approximation ( eq . ( 7 - 54 ) in * ? ? ? * ) , as this limit describes the situation when the stars in the disk remain nearly stationary during the collision . note that this limit is insensitive to the sign of @xmath35 and hence does not distinguish between prograde and retrograde encounters . the limit @xmath68 corresponds to an interaction where the response should be weak because , for example , the encounter is a distant one or the spin and orbital frequencies are highly mismatched . employing the asymptotic expansion for the bessel functions @xcite , @xmath69 \,,\ ] ] where @xmath70 , we find for the prograde and retrograde cases separately : @xmath71 @xmath72 @xmath73 @xmath74 we note that the response is exponentially suppressed in the limit @xmath68 , which demonstrates explicitly that the perturbed system is protected by adiabatic invariance . it is also interesting that in this limit the prograde and retrograde cases are simply related : @xmath75 and for the y - component : @xmath76 thus , the prograde response diverges relative to the retrograde one , by a factor of @xmath40 , and , for a given @xmath40 the magnitude in the velocity perturbation is exactly a factor of four larger . from the above expressions , the change in the energy of the perturbed system can be determined from : @xmath77 d\phi \ , , \ \ \ \ \ \ \ \ \ \ \ \ \\end{aligned}\ ] ] where it is assumed that the disk is axisymmetric and has zero vertical thickness , and @xmath78 denotes the surface mass density distribution . doing the angular integral gives @xmath79 \ , . \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \label{ener_sl}\end{aligned}\ ] ] in the limit @xmath63 , the energy change is @xmath80 which agrees with the corresponding expression in @xcite using the impulse approximation , as expected , while in the limit @xmath68 for the prograde case @xmath81 and for a retrograde encounter @xmath82 to evaluate these , or eq . ( [ ener_sl ] ) , it is necessary to specify the radial dependence of @xmath83 , i.e. , the shape of the rotation curve of the victim . we have also evaluated the next order correction term given by equation ( [ correction ] ) , yielding : @xmath84 \biggr \rbrace \label{impulse_corr_sl_x}\ ] ] @xmath85 where the upper and lower signs terms refer to prograde and retrograde encounters , respectively , and @xmath60v@xmath86 because the encounter is assumed to be coplanar . we note that the @xmath87 term does not couple to the phase @xmath36 of an individual star , because it describes the overall reaction of the victim to the perturber ; i.e. the gain of angular momentum of the entire disk . on top of this global contribution , each star also receives a resonant sensitive and @xmath36 dependent contribution . a similar effect occurs in linear order . there , the overall contribution leads to a motion of the center of mass of the victim . but since our calculation is performed in the center of mass reference frame , this @xmath36 independent contribution does not show up explicitly in the linear order results . there all terms depend on the phase @xmath36 . we further note that the argument of the trigonometric functions involving @xmath36 is different in first and second order : @xmath36 in linear order and @xmath88 in second order . this implies , that the coupling of both orders leads to an asymmetric perturbation in @xmath89 and @xmath90 , whereas the linear term alone always produces symmetric results as can be seen from the velocity increments in the @xmath89 and @xmath90 directions . next , consider an encounter from a parabolic trajectory , which is the case analyzed by @xcite in their study of tidal interactions between stars . this situation is relevant for collisions between galaxies in the field or in loose groups , where the orbits are highly elongated . for this reason , tt72 focused on this geometry in particular , as in the examples shown in figure 1 . the relative velocity between the victim and perturber in this case is their mutual escape velocity , treating the interaction as that between two point masses , and is given by : @xmath91^{1/2 } \ , . \label{vparab}\ ] ] at the minimum separation between the victim and the perturber the relative velocity attains its maximum value v@xmath92 : @xmath93^{1/2}\ ] ] and the relative velocity can be written as @xmath94 we orient the disk as in the earlier derivation , so the orbits within the victim disk are given again by eqs . ( [ vicorb ] ) and ( [ phaset ] ) . to specify the orbit path , employ polar coordinates and write @xmath95 the orbit is then specified parametrically by the relations @xcite @xmath96 and @xmath97 so that the trajectory can be written as @xmath98 a schematic illustration of this case is shown in figure [ stripparab ] . the time - dependence of the orbit is given implicitly by the relation @xmath99^{1/2 } ( \xi+\frac{1}{3}\xi^3)= \frac{2\rm{b}}{\rm{v}_0}(\xi+\frac{1}{3}\xi^3 ) \ , .\ ] ] we again introduce the parameter @xmath100 with @xmath101 , where the plus and minus signs are for prograde and retrograde coplanar encounters , respectively . the orbital angular frequency is @xmath102 so that , as in the case of the straight - line trajectory @xmath103 and the largest response is expected for @xmath46 . to evaluate the velocity perturbations in this case , we begin with eq . ( [ impulse ] ) and adopt the time dependence of the motion within the victim from eq . ( [ vicorb ] ) . the phase angle of the stellar orbits within the victim is given by @xmath104 expand @xmath105 and @xmath106 using trigonometric relations analogous to eqs . ( [ trigcos ] ) and ( [ trigsin ] ) and parameterize the orbit of the encounter as in the polar form of eq . ( [ polorb ] ) . finally , in the resulting integrals , make a change of time coordinate from @xmath107 according to eq . ( [ timeorb ] ) . eliminating the integrals that are odd in @xmath108 we find , after algebra : @xmath109 \ , \mp \ , 3 \sin \phi \cos \phi \sin \left [ 2\alpha(\xi+\frac{\xi^3}{3})\right ] \biggr \rbrace\ ] ] @xmath110 \ , \pm \ , 3 \sin \phi \cos \phi \sin \left [ 2\alpha(\xi+\frac{\xi^3}{3})\right ] \biggr \rbrace .\ ] ] in these expressions , @xmath111 depends implicitly on time through eq . ( [ timeorb ] ) according to @xmath112 and the upper and lower signs in the @xmath53 and @xmath54 terms refer to prograde and retrograde coplanar encounters , respectively . these can be put into a form amenable to further analysis by repeated application of trigonometric relations of the form @xmath113 \ , , \ ] ] giving @xmath114 \ , + \ , 3 \cos \left [ 2\alpha(\xi+\frac{\xi^3}{3 } ) \mp 2 \phi \right ] \biggr \rbrace\ ] ] @xmath115 \ , - \ , 3 \cos \left [ 2\alpha(\xi+\frac{\xi^3}{3 } ) \mp 2 \phi \right ] \biggr \rbrace.\ ] ] finally , these results can be written in terms of the `` generalized '' airy functions of @xcite , who defined @xmath116\rm{d}\xi,\ ] ] giving @xmath117\ ] ] @xmath118 \ , , \ ] ] where , as usual , the upper and lower signs in the @xmath53 factors refer to prograde and retrograde coplanar encounters , respectively . computation of these expressions is facilitated using the recursion relations @xmath119 @xmath120 which are proven in @xcite . equation ( [ rec1 ] ) makes it possible to obtain all the @xmath121 s from the @xmath122 s while equation ( [ rec2 ] ) allows the @xmath122 s for @xmath123 to be computed from @xmath124 . in particular : @xmath125 and , so @xmath126 \ \ \\end{aligned}\ ] ] @xmath127 , \ \ \ \ \ \ \ \ \ \end{aligned}\ ] ] where the upper and lower signs in the @xmath54 terms refer to prograde and retrograde co - planar encounters , respectively , and @xmath60v@xmath86 . for moderate arguments , the functions @xmath124 can be computed by numerical integration of eq . ( [ ilmpt77 ] ) or from the rational function approximations provided by @xcite , which are reproduced in the appendix . we again consider limiting cases of these expressions . when @xmath63 , corresponding to a slowly rotating system , it can be shown straightforwardly from the defining relation eq . ( [ ilmpt77 ] ) that @xmath128 and @xmath129 . thus , from eqs . ( [ dvxpt77 ] ) and ( [ dvypt77 ] ) @xmath130 @xmath131 as earlier , this limit is insensitive to the sign of @xmath35 and hence does not distinguish between prograde and retrograde encounters . the change in energy of the victim in the limit @xmath63 is then @xmath132 comparing this change in energy to that produced by a perturber on a straight - line trajectory , eq . ( [ inputenalpha0 ] ) , we obtain @xmath133 thus , if the velocity in the straight - line case is chosen to be equal to the velocity at closest approach for a parabolic encounter , the straight - line example provides a good approximation to the damage done , at least when @xmath63 . the limit @xmath68 can be analyzed using eqs . ( [ impxparab ] ) and ( [ impyparab ] ) . we note that in this limit the rational function approximations provided by @xcite for the functions @xmath124 are not sufficiently accurate to estimate the velocity perturbations for precisely retrograde encounters because of exact cancellations between dominant terms in the defining relations eqs . ( [ impxparab ] ) and ( [ impyparab ] ) . likewise , direct numerical integration of eq . ( [ ilmpt77 ] ) to sufficient accuracy for large values of @xmath40 is not possible because of the highly oscillatory behavior of the integrand . instead , we employ asymptotic expansions for these functions as @xmath68 obtained using the method of steepest descent . the results of this analysis are provided in the appendix . @xcite previously obtained asymptotic expansions for the generalized airy functions to leading order , but we carried out our analysis to higher orders since we consider limits in which the leading order terms cancel exactly . combing eqs . ( [ delta vx parab])([delta vy parab ] ) with eqs . ( [ i 0l infinity])([i 3l infinity ] ) , we find : @xmath134 @xmath135 @xmath136 @xmath137 as for a straight - line trajectory , the response is exponentially suppressed in the limit @xmath68 , owing to adiabatic invariance . the prograde and retrograde cases are again simply related : @xmath138 and for the y - component : @xmath139 this is similar to the result obtained earlier for straight - line collisions , but now with a numerical coefficient of 16 rather than four . the corresponding changes in energy in the limit @xmath68 are , for the prograde case @xmath140 and for a retrograde encounter @xmath141 quantitatively , the relation of these results to those for a straight - line trajectory will depend in detail on the rotation curve of the victim through the radial dependence of @xmath142 . we have also evaluated the next order correction term given by eq . ( [ correction ] ) , yielding : @xmath143 \biggr \rbrace \label{impulse_corr_par_x}\ ] ] @xmath144 \biggr \rbrace , \ \label{impulse_corr_par_y}\ ] ] where the upper and lower signs in the @xmath54 terms refer to prograde and retrograde co - planar encounters , respectively , and @xmath60v@xmath86 . we now compare the perturbations in velocities of stars in a thin disk in the x - y plane owing to a coplanar tidal interaction with a system passing on either a straight line path or a parabolic orbit . to isolate the dependence on the parameter @xmath40 , we define , for a given relative velocity @xmath3 , impact parameter @xmath31 , and mass of the perturber m@xmath145 the quantities : @xmath146 @xmath147 for the case of encounters on straight line paths : @xmath148\end{aligned}\ ] ] @xmath149 \ , , \end{aligned}\ ] ] while for parabolic encounters : @xmath150\end{aligned}\ ] ] @xmath151.\end{aligned}\ ] ] figure [ alpha ] shows the velocity perturbations on stars within a spinning disk from either straight - line or parabolic encounters . for the prograde cases , the response is a maximum near @xmath46 , where there is a matching between the spin frequency of the rotating victim and the orbital frequency around the perturber , @xmath152 characteristic of a resonance . however , the resonance is broad , with fwhm @xmath153 ; hence , we refer to this phenomenon as a quasi - resonance . the width of the resonance can be understood from the fact that , owing to the non - circular nature of the perturber s orbit , its effective orbital frequency as perceived by the victim changes with time . as a result , the resonance is `` maximal '' only for a short period of time . furthermore , at each point along the orbit the instantaneous orbital frequency can be in resonance with stars located at different radii within the victim , and can therefore act more efficiently on different parts of it at different times . in a true resonant interaction , it is assumed that a periodic perturbation is applied for an indefinite length of time . in that event , the resonance will be narrow because perturbations off - resonance will cancel out in the limit of infinitely many cycles . for a tidal encounter from an unbound trajectory , as we consider here , this is not necessarily the case and off - resonant driving forces can result in a non - zero perturbation . this incomplete cancellation explains why the response for prograde encounters shown in figure [ alpha ] is broad . for retrograde interactions , the cancellation is more complete and the response is suppressed . for both prograde and retrograde cases , in the limit @xmath68 , a star in the victim can complete many orbits while the perturber is near a particular orbital frequency , resulting in an averaging out of the perturbation , yielding an exponentially suppressed response . figure [ alpha ] also illustrates the maximal response in the different cases for @xmath1541 . near this value of @xmath40 , the velocity increments owing to the quasi - resonant phenomenon are greater when the victim experiences a parabolic encounter than in the case of a straight line path , since the curvature of the orbit implies that the resonance is applied for a longer time . for non - rotating systems ( @xmath63 ) , the velocity increment gained by the stars in the victim is described by the usual impulse approximation . in quasi - resonance , the velocity perturbations in the prograde cases are a factor @xmath155 higher than in this limit , while the response is suppressed of a similar factor for a retrograde collision for @xmath46 . note that in our analysis , the disk is assumed to be infinitely thin . however the large resonance width shown in figure [ alpha ] would make the results applicable even to rotating systems with a finite velocity dispersion . qualitatively we expect that for a `` hot '' disk victim the resonant response would still occur , although the efficiency will be somewhat reduced . in order to generalize the formulation presented earlier to non - coplanar encounters , we need to consider situations where the orbit plane is arbitrarily inclined relative to that of the victim disk . to facilitate interpretation of the results , especially vertical heating of the disk , we choose to perform this operation by actively rotating the orbit plane , leaving the unperturbed disk in the @xmath6 plane . thus , as above , the coordinates of stars in the victim disk are still given by eq . ( [ vicorb ] ) and the trajectory of the perturber , @xmath9 , will be specified by rotating its coplanar path , which we will refer to as @xmath156 in the following . to describe the rotation , we introduce the @xmath157 matrix @xmath158 , given by : @xmath159 .\ ] ] then the rotated trajectory of the encounter is given by @xmath160 rotating the path of the perturber does not alter the magnitude of @xmath9 and so @xmath161 the convention for the euler angles that specify the rotation matrix @xmath158 is not unique . defining @xmath162 we take the rotation matrix to be @xmath163,\ ] ] corresponding to a first rotation about the @xmath164 axis by an angle @xmath165 , a second rotation about the @xmath90 axis by an angle @xmath166 , and a third rotation about the @xmath164 axis again by an angle @xmath167 . in other words , @xmath168 , where @xmath169 is a standard matrix describing a rotation by an angle @xmath170 about the @xmath171 axis . the angles @xmath165 , @xmath166 , and @xmath167 can take arbitrary values ( positive or negative ) , with the sense of rotation determined by the right hand rule . for example , the orbital angular momentum for a coplanar collision is @xmath172 , implying @xmath173 we consider first the case of non - coplanar encounters in which the perturber moves along a straight - line , generalizing the results of 3.1 . the coplanar trajectory is given by eq . ( [ pathsl ] ) , the phase angle of the stars along their orbits in the victim is , as earlier , defined by eq . ( [ phaset ] ) , and we employ the parameter @xmath40 , as defined by eq . ( [ alphapar ] ) . note , however , that now the sign of @xmath35 does not determine if the encounter is prograde or retrograde , but merely fixes the direction of the spin angular momentum of the victim , either along the @xmath174 axis ( positive @xmath35 ) or @xmath175 axis ( negative @xmath35 ) . for non - coplanar collisions , the condition of whether the encounter is mainly prograde versus retrograde is set by @xmath176 , where @xmath177 and @xmath178 are the orbital angular momentum of the encounter and the spin angular momentum of the victim disk , respectively . the quantity @xmath179 lies between @xmath180 ( purely prograde ) and @xmath181 ( purely retrograde ) and takes the value @xmath182 for a polar orbit . starting from the expressions for the velocity perturbations in eq . ( [ impulse ] ) we find , after algebra : @xmath183 \notag \\ \ , & + \ , \alpha k_1(\alpha ) \left [ ( 1 - 2a_{11}^2-a_{12}^2 ) \cos \phi_0 - ( a_{12}a_{22 } + 2 a_{11}a_{21 } ) \sin \phi_0 \right ] \ , \notag \\ \ , & \mp \alpha^2 k_1(\alpha ) \left [ ( a_{11}a_{22 } + a_{21 } a_{12 } ) \cos \phi_0 -2a_{11}a_{12 } \sin \phi_0 \right ] \biggr \rbrace\end{aligned}\ ] ] @xmath184 \ , \notag \\ & + \ , \alpha k_1(\alpha ) \left [ ( 1 - 2a_{21}^2-a_{22}^2 ) \sin \phi_0 - ( a_{12}a_{22 } + 2 a_{11}a_{21 } ) \cos \phi_0 \right ] \notag \\ \ , & \pm \ , \alpha^2 k_1(\alpha ) \left [ ( a_{21}a_{12 } + a_{22 } a_{11 } ) \sin \phi_0 -2a_{21}a_{22 } \cos \phi_0 \right ] \biggr \rbrace\end{aligned}\ ] ] @xmath185 \notag \\ \ , & + \ , \alpha k_1(\alpha ) \left [ - ( a_{32}a_{22}+2a_{31}a_{21 } ) \sin \phi_0 - ( a_{32}a_{12 } + 2 a_{11}a_{31 } ) \cos \phi_0 \right ] \notag \\ \ , & \pm \ , \alpha^2 k_1(\alpha ) \left [ ( a_{31}a_{12 } + a_{32 } a_{11 } ) \sin \phi_0 - ( a_{31}a_{22}+a_{21}a_{32 } ) \cos \phi_0 \right ] \biggr \rbrace .\end{aligned}\ ] ] here , the factors @xmath186 are the relevant components of the rotation matrix @xmath158 , and the upper signs in the @xmath53 and @xmath54 terms refer to the case with @xmath47 while the lower signs are for the case with @xmath187 . it is straightforward to show that these expressions reduce to the appropriate ones for coplanar encounters given in 3.1 if the rotation angles are set to zero . as a simple illustration , consider a straight - line path where the original orbit is in the @xmath6 plane and is given by @xmath188 , and employ the euler angle convention summarized above . rotate this path by an angle @xmath166 around the @xmath90-axis , so that @xmath189 in this case , @xmath190 and the components of the rotation matrix are given by @xmath191 .\ ] ] thus , from the equation about for @xmath192 , the vertical heating can be estimated as a function of @xmath166 and is @xmath193 it is straightforward to show that the dependence on @xmath166 in this expression is such that the response leads to a vertical warping of the disk , with an efficiency depending on @xmath166 . we now adopt the coplanar trajectory given in eqs . ( [ polorb ] ) - ( [ timeorb ] ) and apply the rotation matrix @xmath158 to the orbit . following a similar procedure as for the straight - line path , the velocity perturbations can be written in the form : @xmath194 i_{20}(\sqrt{2 } \alpha ) \ , - \ , 3 b_x i_{22}(\sqrt{2 } \alpha ) \ , - \ , 3 c_x i_{2 - 2}(\sqrt{2 } \alpha ) \biggr \rbrace\ ] ] @xmath195 i_{20}(\sqrt{2 } \alpha ) \ , - \ , 3 b_y i_{22}(\sqrt{2 } \alpha ) \ , - \ , 3 c_y i_{2 - 2}(\sqrt{2 } \alpha ) \biggr \rbrace\ ] ] @xmath196 where the terms @xmath197 , and @xmath198 are given by : @xmath199 @xmath200\ ] ] @xmath201\ ] ] @xmath202 @xmath203\ ] ] @xmath204\ ] ] @xmath205 @xmath206\ ] ] @xmath207 .\ ] ] as earlier , the factors @xmath186 are the relevant components of the rotation matrix @xmath158 , and the upper signs in the @xmath53 and @xmath54 terms refer to the case with @xmath47 while the lower signs are for the case with @xmath187 . in particular , whether the encounter is mainly prograde or retrograde is set by the sign of @xmath179 , as given above . in the equations for @xmath208 , the functions @xmath209 and @xmath210 can be simplified to expressions involving @xmath122 using the recursion relations eqs . ( [ rec1 ] ) and ( [ rec2 ] ) . to test the reliability of the quasi - resonant approximation , we carry out simulations of encounters between a spinning system and an external perturber passing either on a straight line path or a parabolic orbit . for simplicity , we adopt a restricted three - body method for following the development of tidal tails during collisions . the restricted three - body scheme is appropriate for this application because the formation of tidal tails is essentially a kinematic process @xcite . in our experiments , the victim is a point mass surrounded by a flat annular disk of test particles . initially , the particles are on circular , keplerian orbits around the central point mass . we employ a system of units in which newton s constant , @xmath211 , and the maximum radius and mass of the victim are all set equal to unity . for each victim , we place test particles on 400 annuli surrounding the point mass . the annuli are linearly spaced in radius , between @xmath212 and @xmath213 around the central mass point with a total number of approximately 1.3@xmath21410@xmath215 particles . the annuli have an angular spacing such that the particles are equidistantly distributed along each ring , with a spacing @xmath216 . this choice ensures a fair sampling of the outermost rings , a better visual comparison , and an accurate sampling of the energy distribution of the stars in the victim . the perturber is modeled as a point mass passing either on a straight line or parabolic orbit . in this way , the problem of the encounter is restricted to a three - body problem since each test particle ( `` star '' ) feels only the gravitational force from the massive particles ( the central point mass of the victim and the perturber ) . we evolve the models for many dynamical times and compute the energy distribution of the particles long after the passage of the perturber , at a time where the energy configuration of the system reaches a steady state . these estimates are compared to analytical estimates using the approximations presented earlier . the comparison to the quasi - resonant impulse gained from each star in the victim is done as follows . we mimic the effect of encounters by setting up the victim as a point mass surrounded by a flat annular disk of particles . we then assign to each test particle a circular , keplerian velocity plus a velocity increment estimated from the analytic approximation for a given perturber mass and impact parameter , and allow the disk to evolve . the total energy of individual test particles is thus always conserved . while our formalism can describe encounters of any mass ratio , we focus here on equal mass situations with @xmath217 , and an impact parameter @xmath31=2 , which is twice as large as our disk of test particles . for the straight line case , the choice of the relative velocity @xmath33 is arbitrary , so first we test our approximations for a fast encounter ( i.e. a weak perturbation ) , which is the case where our formalism is expected to be most accurate . in addition , we test the validity of the approximations for slow encounters , which characterize stronger perturbations that are less impulsive . in what follows , we present numerical tests of coplanar , straight line encounters with fast and slow relative velocities , as well as parabolic encounters at linear and second order approximations . the non - coplanar case has been also tested to ensure that our analytic formulation is reasonable , but for brevity we present results only for co - planar encounters . more general tests of the validity of our analytic treatment will be considered in due course in applications to observed systems . figure [ tail_copl_pro ] gives a visual comparison of the evolution of a victim under the tidal influence of the fast passage of a perturber on a straight line with relative velocity @xmath33=7.5 . the top panels illustrate the system with test particles _ kicked _ by velocity increments according to the quasi - resonant approximation described in the previous sections . the bottom panels display the evolution of the victim when an external perturber passes close to the disk using a restricted three - body treatment . figures [ tail_copl_pro]-[epro ] illustrate a coplanar prograde encounter , while the corresponding retrograde coplanar case is displayed in figures [ tail_copl_retro]-[eretro ] . colors are assigned to rings according to their initial distance from the central point mass . first , as expected , tails are produced for prograde encounters ( figure [ tail_copl_pro ] ) , but are suppressed in retrograde cases ( figure [ tail_copl_retro ] ) . the visual agreement in the extents and shapes of the tails between the analytic ( upper panels ) and the numerical ( lower panels ) approaches is good . further insight is provided by comparing the distribution of specific energies of the disk particles as shown in figures [ epro ] and figure [ eretro ] for the prograde and retrograde cases , respectively . the different panels show the time evolution of the specific energies for the restricted three - body simulation ( blue ) and the analytic calculation ( red ) , which is static . fast encounters correspond to relatively modest tidal perturbations , leading to a weak resonant response . the amount of energy transferred during the encounter is small and the resulting tails are less spatially extended than ones produced for stronger tidal perturbations ( see figure [ tail_copl_pro_corr_sl ] , discussed below ) . as expected , prograde and retrograde encounters lead to very different energy distributions , since prograde encounters transfer much more energy to the victim than retrograde ones , producing much more massive tails in the prograde case , as demonstrated in figures [ tail_copl_pro ] and [ tail_copl_retro ] . indeed , according to our analytic results the maximum resonant response is obtained for rings placed at distances corresponding to the parameter @xmath46 ( figure [ alpha ] ) . in this particular example , the outermost rings which contribute significant material to the tails are only weakly resonant because their spin frequencies are such that they have a value of @xmath218 . note that our lowest - order approximations yield symmetric tails as shown in the top panels of figure [ tail_copl_pro ] , whereas the numerical simulation yields tails that are slightly asymmetric ( bottom panels in figure [ tail_copl_pro ] ) . this effect increases for stronger perturbations , and is connected to higher order corrections as described in the sections above . for retrograde encounters , the orbital and spin angular momenta are misaligned and , as shown in figure [ eretro ] for the evolution of the energy distribution , little energy is transferred during the encounter . a resonance does not develop and tail formation is suppressed , as shown in figure [ tail_copl_retro ] . the quasi - resonant approximation at lowest order , eq . ( [ impulse ] ) , becomes less accurate for slow encounters , where the tidal forces are less impulsive and where the effective duration of the encounter is of the order of the dynamical time of the particles ( `` stars '' ) in the outermost rings . indeed , this is the case where the restricted three - body simulation shows that some mass from the victim is captured by the perturber , introducing a large asymmetry and greater changes in the energy configuration of the final system . the lack of symmetry in tail - making owes to non - linear effects . we can account for some of these by including the next order correction to the response , as given by eqs . ( [ impulse_corr_sl_x]-[impulse_corr_sl_y ] ) and eqs . ( [ impulse_corr_par_x]-[impulse_corr_par_y ] ) for the straight - line and parabolic cases , respectively . we emphasize , however , that even when we include these corrections we assume that stars in the victim move along unperturbed orbits throughout the course of the encounter , thereby treating the orbital variations as small and implicitly ignoring the non - linear response . figure [ tail_copl_pro_corr_sl ] shows a prograde straight - line encounter for the relative velocity @xmath219 , as predicted by our analytic approximation when the second order correction is included in the evaluation of the velocity increments given to the disk particles ( top panels ) . this evolution is compared to a simulation with the same disk perturbed by an external body ( bottom panels ) . as noted above , the tail connecting the victim to the perturber is not reproduced by our analytic formalism in detail because we ignore the non - linear response of the victim during the encounter . nevertheless , the trailing tail , displayed in the region of the plane with @xmath220 , is similar in shape and orientation to the one generated in the restricted three - body simulation . indeed our analytic expression predicts the correct shape of the tail , although it is slightly more extended compared to the one in the actual simulation . the energy distribution of the particles contained in the trailing tail in both models is shown in figure [ en_sl_slow ] , where the energy is shown in a logarithmic scale . allowing for the fact that our analytic method can not entirely account for non - linear effects , we find the comparison encouraging . indeed , as the energy scale in figure [ en_sl_slow ] is logarithmic , the regions showing obvious disagreement contain few stars . next , we show an encounter from a parabolic trajectory where the relative velocity of victim and perturber at closest approach is @xmath221 for the parameters adopted in the numerical experiments . figure [ tail_copl_pro_par ] shows a prograde parabolic encounter as predicted by our analytic approximation when the second order correction is included in the evaluation of the velocity increments ( top panels ) . the evolution is compared to a restricted three - body simulation with the same disk as perturbed by an external body passing on a parabolic trajectory ( bottom panels ) . this encounter , as well as the slow straight line case , leads to strong resonant effects . in fact , the parameter @xmath40 , for both cases , is around 1.3 for the outermost ring . hence these rings should have a near - maximal response . this is demonstrated in figure [ tail_copl_pro_par ] , which shows longer tails as compared to the fast , straight - line encounter . the asymmetry is also reflected in the energy distribution of the slow straight - line encounter ( figure [ en_sl_slow ] ) and of the parabolic case ( figure [ en_par ] ) which appear to be bimodal . we have formulated a simple analytic theory describing the resonant response of rotating objects to gravitational tidal perturbations . we estimated the velocity perturbations on a spinning disk resulting from an encounter with a system on either a straight - line or a parabolic orbit . our main conclusions are the following : * the resonant phenomenon described here is an incomplete or `` quasi- '' resonant process because the perturbation is applied at a given frequency for only a finite period of time . consequently , the resonance is broad and is maximal when the spin frequency of the rotating victim matches the peak orbital frequency of the encounter . the resonance is broad because the perturbation at each frequency is not applied over many cycles , as for a strictly periodic forcing , so perturbations at slightly off - resonant frequencies do not cancel out completely . however , because the resonance is broad , at each point along the orbit the orbital frequency can be in resonance with parts of the victim disk located at different radii and can therefore act on various locations in the disk at different times . this feature is interesting for galaxy encounters because it means that the victim can loose baryonic mass from different points in the disk by resonant stripping at various times , making the `` quasi - resonant '' response more damaging than if only certain regions in the disk were being affected . * in the quasi - resonant regime , the velocity increments gained by the stars in the victim are greater during a parabolic encounter than in the case of a straight - line interaction . for non - rotating systems , the velocity perturbations are described by the usual impulse approximation . in prograde cases ( both straight or parabolic ) the velocity perturbations owing to quasi - resonant phenomenon are at least factor @xmath222 higher than in the impulse approximation while the response is suppressed for retrograde encounters . * when compared to numerical simulations we find that the quasi - resonant approximation provides a simple physical model for describing the nature of the tails in interacting disk galaxies and gives well - defined and reasonable values for the changes in velocities as a result of tidal encounters . in particular , the quasi - resonant approximation gives accurate results for high speed encounters , where the effective duration of the encounter is less than or of order the dynamical time of the particles ( `` stars '' ) of the spinning system . the main reason is that high velocity encounters generate small perturbations of otherwise steady - state systems and keep them in the linear regime , where the analytic approximation is valid . the quasi - resonant approximation is less accurate for slower encounters , where the tidal forces act over a longer period of time . this is particularly true for the case of a perturber passing on a parabolic trajectory . indeed , in these events the numerical simulations show that some mass of the victim galaxy is captured by the perturber . however , a better match can be obtained by including higher order corrections for the velocity perturbations in the analytic formalism . * the lengths of the tails depends on the amount of energy transferred to the victim during an encounter . prograde encounters with more massive perturbers pump larger amounts of energy into the victims , yielding more extended tails but also introducing asymmetries with the leading tails being more elongated than the trailing ones . retrograde encounters transfer much less energy into the victim , suppressing the development of tails . finally , we note that our findings for tail - making during galaxy - galaxy encounters may apply to gas as well . thus , our calculations may be used to understand the geometry of the gas streams and bridges of stars associated with dwarf galaxies , as recently discovered in the panda survey of m31 @xcite . other potential applications include warps and heating of galactic disks by tidal perturbations , understanding the conditions for long tails to be produced in galaxy collisions depending on the dark matter distribution in the halos , and identifying situations where stars can be unbound in galaxy interactions and be ejected into the intergalactic medium or populate the field of groups and clusters . moreover , our theory may be relevant to other applications , such as to studies of the stability of binary stars perturbed by a third body or to investigations of protoplanetary disks perturbed by close passages of stars . we thank alar toomre and avi loeb for valuable advice . ed acknowledges support from the keck foundation . cafg is supported by a fellowship from the miller institute for basic research in science , and received further support from the harvard merit fellowship and fqrnt during the course of this work . our analysis of parabolic tidal encounters results in expressions that involve the generalized airy functions of @xcite , defined by eq . ( [ ilmpt77 ] ) . the first one , @xmath223 , is a modified bessel function or ( conventional ) airy function , @xmath224 \ , . \label{ioo}\ ] ] @xcite provide the following rational function approximations to @xmath225 and @xmath226 , making it possible to straightforwardly estimate the velocity perturbations in a parabolic tidal encounter : to evaluate the limit @xmath232 of the parabolic encounter case studied in this work , it is necessary to know the asymptotic behavior of the generalized airy functions as @xmath233 . in what follows , we give these asymptotic expansions and outline the key steps of their derivation . while the @xmath234 case corresponds to the usual airy function , the @xmath235 cases have not been extensively studied before , and the numerical fits provided by @xcite are not sufficiently accurate to capture exact cancellations in the results . @xcite previously calculated these asymptotic expansions to leading order using similar techniques , but some of the limits we take involve exact cancellations at this order , so that it is necessary to compute higher - order terms . the starting point for the asymptotic expansions is the integral expression ( eq . [ ilmpt77 ] ) @xmath236 } d\xi.\ ] ] the main technical difficulty in determining the asymptotic behavior of @xmath237 arises from the fact that the cosine term oscillates extremely rapidly as @xmath233 , and that the value of the integral depends on exactly how the positive and negative parts cancel each other . it is however possible to circumvent this difficulty by expressing equation ( [ i l0 integral ] ) in terms of a contour integral in the complex plane , and by using the method of steepest descent ( e.g. , * ? ? ? specifically , we note that @xmath238 } d\xi \right\}\ ] ] and consider the integration contour shown in figure [ contour ] . making the change of variable @xmath239 , the integral within the curly brackets is seen to be equal to an equivalent integral along the imaginary axis : @xmath240 } ds.\ ] ] writing the complex integrand as @xmath241 , the residue theorem implies that @xmath242 , where the @xmath243 are the poles within @xmath244 . since the integrands along @xmath245 and @xmath246 are exponentially suppressed in modulus , @xmath247 as the contour is expanded to infinity , yielding @xmath248 as the left hand side is simply related to equation ( [ i l 0 after change of var ] ) by the multiplicative factor @xmath249 , the problem is essentially reduced to evaluating @xmath250 . the key is to choose the path @xmath251 so that the integral is tractable in the limit @xmath233 . according to the method of steepest descent , the integral will in this limit receive most of its contribution in the neighborhood of a saddle point , if the integration path is chosen to follow the direction of steepest descent of the exponential argument . setting @xmath252/ds=0 $ ] , the relevant saddle point is found to be @xmath253 , and we may parametrize the path of steepest descent by @xmath254 chosen so that @xmath253 at @xmath32 . then , @xmath255^{l } } \frac{ds}{dt } dt,\ ] ] provided that the parametrization @xmath256 is oriented in the same sense as @xmath251 . once an explicit expression for @xmath256 is obtained by solving the cubic equation ( [ steepest descent path ] ) , we expand the product @xmath257^{-l } ( ds / dt)$ ] in a taylor series around @xmath32 . for instance , the @xmath234 case yields @xmath258 purely real terms , where the purely real terms do not enter in the final result . the integral on the right hand side of equation ( [ c3 t integral ] ) becomes a term - by - term sum of simple gaussian integrals of the form @xmath259 , where @xmath260 and the @xmath261 coefficients are determined by the taylor expansion . these integrals are easily evaluated and we thus obtain an analytic expansion for @xmath250 valid in the limit @xmath233 . this procedure can be carried out as described for @xmath234 . in this case , there is no residue and we find @xmath262,\ ] ] in agreement with the known result for the ordinary airy function . for @xmath263 , we however encounter the difficulty that the saddle point @xmath253 is also a pole of order @xmath264 . for the simple pole of the case @xmath265 , the basic procedure remains valid but we must account for the @xmath266 contribution in equation ( [ residue theorem applied ] ) . since the pole is located directly on the contour ( rather than contained within it ) it contributes only half of its value , namely @xmath267 rather than @xmath268 , as can be shown by considering infinitesimally modified contours that avoid and enclose the pole . we find @xmath269}$ ] and @xmath270.\ ] ] + for the @xmath271 and @xmath272 cases , the procedure yields divergent integrals . these can however be avoided by going back to equation ( [ i l 0 after change of var ] ) and using a combination of integration by parts and partial fraction expansions to rewrite it in terms of @xmath273 , @xmath274 , and other integrals that are regular at @xmath253 and everywhere within @xmath275 . the latter integrals can then be evaluated in the limit @xmath233 using the method of steepest descent , in close analogy to the @xmath234 case . we finally find : @xmath276\ ] ] and @xmath277.\ ] ] , a. w. , irwin , m. j. , ibata , r. a. , dubinski , j. , widrow , l. m. , martin , n. f. , ct , p. , dotter , a. l. , navarro , j. f. , ferguson , a. m. n. , puzia , t. h. , lewis , g. f. , babul , a. , barmby , p. , bienaym , o. , chapman , s. c. , cockcroft , r. , collins , m. l. m. , fardal , m. a. , harris , w. e. , huxor , a. , mackey , a. d. , pearrubia , j. , rich , r. m. , richer , h. b. , siebert , a. , tanvir , n. , valls - gabaud , d. , & venn , k. a. 2009 , , 461 , 66

we present high signal - to - noise , moderate spectral resolution ( ) near - infrared ( m ) spectroscopy of the nearby t tauri star tw hya . by comparing the spectrum and the equivalent widths of several atomic and molecular features with those for stars in the irtf near - infrared library , we revise the spectral type to m2.5v , which is later than usually adopted ( k7v ) . this implies a substantially cooler stellar temperature than previously assumed . comparison with various pre - main sequence models suggests that tw hya is only myr old ; much younger than the usually adopted myr . analysis of the relative strengths of the h lines seen in the spectrum yields estimates for the temperature and density of the emitting region of k and . the thickness of the emitting region is km and the covering fraction is . our derived physical parameter values agree with the predictions of the magnetospheric accretion scenario . the highest signal - to - noise h lines have profiles that indicate multiple emission components . we derive an excess spectrum ( above that of the m2.5v template ) that peaks in the h band . although our derived veiling values , , agree with previous estimates , the excess spectrum does not match that of current models in which this flux is generated by an inner optically thin disk . we suggest that the excess flux spectrum instead reflects the differences in atmospheric opacity , gravity , and age between tw hya and older , higher gravity field m2.5 dwarfs .

1102.0535

the optical and near - infrared ( nir ) spectra of classical t tauri stars ( cttss ) are characterized by strong h emission lines ( see e.g. , * ? ? ? * ; * ? ? ? * ) . in the currently accepted paradigm for these pre - main sequence objects , these emission lines arise from gas being accreted onto the stellar surface from the inner regions of a circumstellar disk along the stellar magnetic field lines . in this magnetospheric accretion flow model , the inflowing material is heated and ionized as it passes through a standing shock close to the stellar photosphere @xcite . attempts to confirm observationally the predictions of these models have generally focussed on only a few of the strongest h lines , usually h@xmath11 , pa@xmath12 , and br@xmath13 ( e.g. * ? ? ? * ; * ? ? ? furthermore , much of the theoretical work to date has focussed on modeling individual emission line profiles ( at high resolution ) and the overall spectral energy distributions ( at low resolution ) of cttss , with relatively little emphasis on the fluxes and flux ratios of a series of emission lines . only recently have the relative strengths of the higher order members of various h series been combined with their stronger counterparts to estimate the density and temperature of the emission regions in cttss @xcite and thereby constrain theoretical models . a complication in carrying out such comparisons of relative line fluxes to model predictions arises from the notorious variability of the spectra of cttss ( e.g. * ? ? ? * and references therein ) , and therefore the need to acquire spectra covering all useful lines simultaneously . consequently , the location and physical conditions of the h line emitting regions are not well constrained observationally . at a distance of @xmath14 51 pc @xcite , tw hya is one of the nearest t tauri stars . although it is a member of a small group of stars known as the tw hydra association ( twa ; * ? ? ? * ) , tw hya resides surprisingly far from any known molecular cloud . @xcite originally identified this object as a possible post - t tauri star with a spectral class of k7 ve . however , its optical spectrum exhibits a very strong and variable h@xmath11 emission line and emission lines of he i , [ ] , , and , features that suggest that it is a ctts @xcite . @xcite noted that tw hya was a strong iras source , suggesting that it is surrounded by a dusty accretion disk . this was confirmed by @xcite , who found tw hya to be a strong sub - millimeter source . the accretion disk around tw hya was subsequently directly imaged at various wavelengths by @xcite , @xcite , @xcite , @xcite and @xcite and is seen approximately face - on . analysis of the spectral energy distribution ( sed ) of tw hya has indicated that the system contains a so - called transitional disk " @xcite with two separate components : a largely dust free , optically thin , hot , inner region , extending from @xmath14 0.06 - 4 au @xcite , and a cold outer region extending to a radius of @xmath14 200 au . tw hya is actively accreting material from its disk , with an accretion rate of @xmath15 m@xmath16 yr@xmath17 @xcite . the age of the twa has been has been estimated from a variety of methods to be between 5 and 30 myr @xcite . for a k7v spectral type , comparison of the location of tw hya on the h - r diagram with pre - main sequence isochrones gives an age of tw hya of @xmath18 myr @xcite , although a few substantially younger age estimates can be found in the literature ( e.g. , * ? ? ? * ; * ? ? ? summaries of the properties of tw hya and the twa have been given in the general reviews by @xcite and @xcite . in this paper we present new high signal - to - noise medium resolution ( @xmath19 ) nir ( @xmath20 m ) spectroscopy of tw hya and use these data to reassess its spectral classification . we derive a new spectral type , which is substantially later than the canonical value . the later spectral type implies a much cooler stellar temperature , and hence a much younger age than usually adopted . subtraction of a scaled , matching spectral template from our spectrum reveals numerous ( primarily h ) emission features . because the spectra simultaneously cover both the pa and br series , we avoid the ambiguity introduced by temporal variability of the line fluxes and we use the relative line fluxes to determine the physical conditions of the emitting gas . the derived temperature , density , and thickness of the emission region are found to be in good agreement with the predictions of magnetospheric accretion models . we observed tw hya at the nasa infrared telescope facility ( irtf ) on mauna kea on 2010 feb 26 and feb 27 ( ut ) with spex , the facility near - infrared medium resolution cross - dispersed spectrograph @xcite . ten individual exposure of tw hya , each lasting 100 s , were obtained using the short - wavelength cross - dispersed ( sxd ) mode of spex on 2010 feb 27 11:26 ut . this mode yields spectra spanning the wavelength range 0.8 -2.4 @xmath2 m divided into 6 spectral orders . twenty exposures , each lasting 30 s , were obtained in the long wavelength cross - dispersed ( lxd2.1 ) mode on 2010 feb 26 11:20 ut . this mode yields spectra covering [email protected]@xmath2 m in 6 spectral orders . the observations were acquired in pair mode " , in which the object was observed at two separate positions along the 15-long slit . the slit width was set to 0@xmath223 , which yields a nominal resolving power of 2000 for the sxd spectra and 2500 for the lxd2.1 spectra . ( at the distance of tw hya of 51 pc , the spex @xmath23 slit spans @xmath24 au . ) the slit was set to the parallactic angle during the observations . the airmass was about 1.74 for both sets of observations . observations of hd 92845 , an a0@xmath25v star , used as a telluric standard " to correct for absorption due to the earth s atmosphere and to flux calibrate the target spectra , were obtained immediately preceding the observations of tw hya . the airmass difference between the observations of the object and the standard was 0.07 . the seeing was estimated to be @xmath26 and @xmath27 at 2.2 @xmath2 m on 26 feb and 27 feb , respectively , and conditions were clear on both nights . a set of internal flat fields and arc frames were obtained immediately after the observations of tw hya for flat fielding and wavelength calibration purposes . the data were reduced using spextool @xcite , the idl - based package developed for the reduction of spex data . the spextool package performs non - linearity corrections , flat fielding , image pair subtraction , aperture definition , optimal extraction , and wavelength calibration . the sets of spectra resulting from the individual exposures were median combined and then corrected for telluric absorption and flux calibrated using the extracted a0 v telluric standard spectra and the technique and software described by @xcite . the spectra from the individual orders were then spliced together by matching the flux levels in the overlapping wavelength regions , and regions of poor atmospheric transmission were removed . the final sxd and lxd spectra were merged by scaling the sxd spectra by 1.05 in order to match the flux levels of the lxd data in the region of wavelength overlap between 2.2 and 2.4 @xmath2 m and then computing a weighted average . the weights were determined from the signal - to - noise ( s / n ) values in the two spectra . the final @xmath1 @xmath2 m spectrum is shown in fig . [ fig1 ] . the s / n varies across the spectral range ; it is of the order of several hundred across the sxd wavelength range , @xmath28 across the @xmath29 @xmath2 m range and about @xmath30 in the @xmath31 @xmath2 m range . we computed synthetic nir magnitudes in various filters from our final spectrum . the estimated 2mass magnitudes are @xmath32 , @xmath33 , and @xmath34 . comparison with the 2mass point source catalogue indicates that our synthetic magnitudes are systematically lower by @xmath35% . given the variable nature of tw hya , it is not clear if this difference reflects a real change in the source brightness or a systematic error in our absolute flux calibration . however our synthetic @xmath36 and @xmath37 colors agree to within 0.02 mags with the 2mass values , which indicates that our relative flux calibration is very accurate ( see also * ? ? ? * where it is demonstrated that the adopted method of flux calibration produces spectra with relative fluxes accurate to a few percent ) . although a spectral type as early as k6ve has been given for tw hya @xcite , the most commonly quoted spectral type for this object is k7ve following @xcite . however , determinations of the spectral type of t tauri stars based on optical spectra can be biased by the substantial excess optical continuum produced by the accretion process ( see e.g. , * ? ? ? * ) as well as any differential extinction due to dust either locally or along the line of sight . in the nir , however , the contribution from the excess continuum is believed to be small for most cttss , and this spectral region is much less affected by dust than the optical , while still containing numerous diagnostic lines that can be used for spectral classification . with the recent publication of the irtf spectral library @xcite , it is possible to re - evaluate the spectral types for cttss by directly comparing their nir spectra with those for a large stellar sample . therefore , we have re - examined the spectral type for tw hya based on our spex spectrum , which was obtained with the same instrumentation and reduced in exactly the same manner as those in the irtf spectral library . we attempted to determine the spectral classification of tw hya using a variety of methods . comparison of the overall shape @xmath1 @xmath2 m spectrum of tw hya ( fig . [ fig1 ] ) with those presented in the irtf spectral library @xcite immediately suggests that this object has a spectral type between m1v and m4v , and certainly later than k7v . in particular , the broad shallow absorption at @xmath38 m , due to tio bandheads , and the prominent continuum ` bumps ' in the h and k bands , due to water absorption at the edges of the bands at @xmath39 , @xmath40 , and @xmath41 m , suggest a spectral type later than about m1.5v but earlier than about m3.5v ( see * ? ? ? such a comparison should not be affected by extinction along the line of sight to tw hya because , at a distance of only 51 pc and a location well outside of any molecular cloud , as well as a nearly pole - on orientation , tw hya is believed to have @xmath42 ( * ? ? ? * see 3.2 below ) . furthermore , the m stars in the irtf spectral library either have zero extinction or their spectra have been corrected for extinction . nevertheless , t tauri stars are known to exhibit veiling " , in which emission from the surrounding accretion disk provides additional flux above that of the photosphere in the ir , while the accretion shock contributes additional flux in the uv / optical to the total spectrum from the object . as tw hya is a transitional disk " system , the nir veiling is believed to be rather low ( @xmath43% ; * ? ? ? * see also 3.3 below ) . even if the veiling were substantial , the smooth excess flux spectrum expected from the disk emission should not produce the broad absorption features that characterize the spectra of early m subtypes . we then independently examined various spectral regions , and compared them with the corresponding regions of the spectra in the irtf spectral library . we carried out this comparison both `` by eye '' and using a least - squares fitting procedure ( to determine the best - fit scaling for each spectrum in the irtf spectral library ) . as can be seen in figs . [ fig2a ] and [ fig2b ] , both the spectral slope and the depths of various spectral features in the @xmath44 @xmath2 m region indicate that tw hya has a spectral type between m1v and m3v . in particular , the presence of weak tio bands at @xmath45 and @xmath46 @xmath2 m , a weak absorption due to feh at @xmath47 @xmath2 m , and the absorption feature due to h@xmath48o at 1.33 @xmath2 m suggest a spectral type of early m , rather than k. a similar result is found for the k - band region ( fig . [ fig2d ] ) , in which the spectral shape of tw hya appears to be very similar to that of an m2.5v star . for the h - band , however , we find a slight disagreement between the spectral shape , which indicates a spectral type of m2v or later , and the strengths of various discrete features ( particularly the complex of fe and si lines between 1.60 and 1.65 @xmath2 m ) , which indicate a spectral type of @xmath14 m1v . the @xmath49 comparison over the wavelength region between 0.95 and 1.34 @xmath2 m ( where veiling is expected to be minimal ; see below ) yields low @xmath49 values for the m0.5v , m1.5v , m2.5v , and m3v spectra from the library , with a formal best match with the m1.5v spectrum . however , the overall spectrum does not match that of the m0.5v star and the shape of the k band spectral region does not match that of the m1.5v star . there is very little difference between the m2.5v and m3v spectra . we also measured the equivalent widths ( ews ) of numerous features in our spectrum and compared them with values determined from direct measurements of the spectra of k and m stars available in the irtf spectral library @xcite . ( since our spectrum has the same resolution and comparable s / n as those in the irtf spectral library , a comparison with these values should provide a better indication of the spectral type than those given by other authors with different stellar libraries . ) the ews of some of these absorption features ( due to , , , , and feh ) in the spectrum of tw hya are presented in table [ abslines ] . a comparison between a subset of these values deemed to be the most sensitive diagnostics of spectral type and those derived in an identical manner from the k and m stars in irtf spectral library is shown in figure [ fig3 ] as a function of spectral type . in this figure , the horizontal dashed lines denote the range in the ew ( @xmath50 ) for tw hya . the results presented in fig . [ fig3 ] , particularly the feh and @xmath51 m ews , clearly suggest a spectral type of m2.5v - m3.5v for tw hya . because of veiling , one must be careful in interpreting the ews as spectral type diagnostics : lines may appear to be too weak simply because of the additional continuum flux from the disk . therefore , the effect of veiling on a spectral type determination depends on the trend of the ew of a feature with spectral type for normal stars . if the ew of a feature decreases with spectral type , veiling will cause a feature to appear weaker and will result in a later spectral classification . if the ew increases with spectral type , veiling will result in an earlier spectral classification . however , veiling can not produce lines that are too _ strong _ for a given spectral type . furthermore , the veiling in tw hya is expected to be negligible in the y and j bands @xcite . for these reasons , we relied primarily on the diagnostic lines in these bands and on those lines whose strength increases with later spectral type . the latter set of lines should yield a limit on the spectral type . nevertheless , even for the h and k bands , the effect of veiling on the spectral type we determine from the ews should be small . the true ew ( @xmath52 ) can be computed from the observed ew ( @xmath53 ) with @xmath54 where @xmath55 is the veiling value . @xcite give a k band veiling value for tw hya of @xmath56 . even a correction of 10% to the observed ew values will not yield a spectral type substantially different from what we determine from the observed ( uncorrected ) ews . furthermore , such an increase in the ews will tend to yield later , not earlier , spectral types from the most sensitive lines . the effects of veiling can be mitigated to a large extent by considering pairs of closely spaced diagnostic lines sensitive to spectral class . the ratio of the ews of two such lines is given by @xmath57 we can express the veiling of one line in terms of the other @xmath58 and if the lines are suitably close in wavelength then @xmath59 . therefore , @xmath60 and the observed ew ratios will reflect the true ews . if this ratio is sensitive to spectral type , it can be used to classify the star . fig . [ ewratio ] demonstrates the use of this technique for tw hya using the 1.14 @xmath2 m and 1.18@xmath2 m lines . this figure demonstrates again that the spectral type for tw hya is approximately m2.5v , independent of any possible veiling . the agreement between this determination and that derived from the individual lines demonstrates that the veiling around 1 @xmath2 m must be negligible in tw hya . based on the results of these three comparison methods , we conclude that tw hya has a spectral type of m2.5v or m3v , with a possible range between m1.5 v and m3.5 v. for this spectral type , the calibrations for m stars given by @xcite and @xcite yield an effective temperature of @xmath61 k and a mass of @xmath62 . to estimate the stellar radius , we adopted the 2mass j band magnitude of 8.22 , a distance of 51 pc , and the j - band bolometric corrections from the @xcite models to compute the bolometric luminosity . the radius was then determined from @xmath63 with this procedure , we find @xmath64 and @xmath65 . the derived luminosity is in good agreement with other estimates ( @xmath66 ; * ? ? ? * ; * ? ? ? as can be seen in fig . 1 , the nir spectrum of tw hya consists of numerous emission lines atop a photospheric spectrum . the emission lines presumably arise from the accretion process . we generated the spectrum of this emission component by scaling the spectrum of an m2.5v star ( gl 381 ) from the irtf spectral library to that of tw hya and subtracting it . the scale factor was determined to be 0.29 via a least squares calculation between 0.95 and 1.34 @xmath2 m . this wavelength region was chosen for two reasons : ( 1 ) as stated above , previous investigations have found that the veiling in the y band ( i.e. , around 1 @xmath2 m ) is negligible @xcite ; ( 2 ) matching a library spectrum at any longer or shorter wavelengths resulted in substantial regions with negative ` excess ' fluxes and veiling factors . since the extinction toward tw hya is negligible , the latter result is not physically justified . the result of the subtraction of the scaled m2.5v template spectrum is shown at the bottoms of figs . [ fig2a]-[fig2d ] . numerous lines of the hydrogen paschen ( pa ) , brackett ( br ) , pfund ( pf ) , and humphreys ( hu ) series , as well as emission lines of and appear in the subtracted spectrum . we see no evidence of the h@xmath48 emission at 2.1218 @xmath2 m reported by @xcite . the measured line fluxes are presented in table [ emisslines ] . we now discuss the properties of the h , he , and o lines . [ [ section ] ] as shown by @xcite for a sample of 15 classical t tauri stars and by @xcite for the case of tw hya itself , the relative h line strengths in the spectra of these objects can be well reproduced with the case b recombination approximation . although the relative line strengths within any given series can be used to estimate the temperature and density of the emitting region , a more sensitive diagnostic is provided by the inter - series ratios of line strengths ( e.g. , * ? ? ? therefore , we computed the flux of each line relative to that of pa @xmath12 , the strongest line in our spectrum . we then used the theoretical h line fluxes for case b as provided by @xcite and @xcite to compute the corresponding line ratios for electron temperatures between 500 and 30000 k and electron densities between @xmath67 and @xmath68 @xmath7 . for each temperature and density pair we computed a @xmath49 value between the theoretical and observed line ratios . contours of the @xmath49 values are shown in fig . [ chicont ] . the minimum @xmath49 was found for @xmath69k and @xmath70 @xmath7 and the corresponding theoretical line ratios for these values are plotted along with the observed line flux ratios as a function of upper level quantum number in fig [ lineratio ] . however , as seen in fig . [ chicont ] , the electron temperature is not tightly constrained by the line ratios and we find that any temperature @xmath5 k can provide an adequate representation of the data points . for temperatures @xmath71 k , the best fit electron density is @xmath72 @xmath7 , while for temperatures @xmath73 k , the best fit yields @xmath70 @xmath7 . in our analysis below we will use both @xmath72 @xmath7 and @xmath70 @xmath7 , with the corresponding temperature values . for comparison , we also plot in fig [ lineratio ] the theoretical values for @xmath74 k and @xmath75 @xmath7 derived by @xcite . this comparison clearly demonstrates that the values found by @xcite can not reproduce the observed line ratios for tw hya . we also analyzed the h line strengths seen in the _ spitzer _ irs spectra of tw hya presented by @xcite in the same manner . because of the large error bars and relatively few data points , the h flux ratios relative to hu@xmath11 in the sh2 spectrum provide only weak constraints on the electron temperature and density . although the formal best fit for the sh2 line fluxes is found for @xmath76 k and @xmath77 @xmath7 , there is very little difference between the theoretical line flux ratios for this set of parameters and the values we found from our nir line analysis . for the sh1 spectrum of @xcite , the best fit was found for @xmath78 k and @xmath75 @xmath7 , close to the values advocated by @xcite . as stated above , it is the combination of multiple h line series that provide the strongest constraints on the temperature and density . with just the pa h series alone , for example , it would be nearly impossible to distinguish lines produced in a region with @xmath79 @xmath7 from those produced in a region with @xmath70 @xmath7 . however , the fluxes of the h lines in the spectra of t tauri stars are known to vary , and our lxd data were not acquired simultaneously with the sxd data ( the time difference was almost exactly 24 hours ) . therefore , we also carried out our analysis using only the pa and br lines in our sxd 0.8 - 2.5 @xmath2 m spectrum . we obtained results similar to those given above : a temperature of @xmath80 k and a density @xmath81 @xmath7 provided the best fit in this case , but @xmath69 k , and @xmath70 @xmath7 were nearly equally acceptable . this , along with the fact that the observed flux levels of the sxd and lxd spectra differed by only 5% between the two nights , suggests that tw hya did not vary substantially between the two observations , and confirms the results found by including the pf lines in the analysis . we note that the h line strengths in the spectrum of tw hya obtained by @xcite in 2008 are only about 15 - 20% larger than the values we measure . the measured h lines contain several sets for which the @xmath82 is the same ( e.g. , pa @xmath12 and br @xmath11 , or pf @xmath12 and br @xmath13 and pa @xmath83 ) . the flux ratios of these lines are fairly insensitive to the density and temperature of the emitting gas , particularly at densities above @xmath84 @xmath7 , and hence can be used to estimate the extinction . the agreement with the case b ratios shown in fig . [ lineratio ] indicates that the assumption of zero reddening , and hence low dust content in the emitting region , is correct . the volume of the ionized region is given by @xmath96 or @xmath97 where we have assumed @xmath98 . the emitting volume was found to be surprisingly small , @xmath99 @xmath100 for @xmath72 @xmath7 and @xmath101 @xmath100 for @xmath70 @xmath7 . we derive a mass of the emitting gas , @xmath102 , of between @xmath103 g ( for @xmath72 @xmath7 ) and @xmath104 g ( for @xmath70 @xmath7 ) . these values are substantially lower ( factors of @xmath105 ) than that derived by @xcite , who assumed a density of @xmath106 @xmath7 . assuming the emitting volume is located at or near the surface of the star , we can estimate the thickness of the emitting region from @xmath107 where @xmath108 is the radius of the star and @xmath109 is the fraction of the surface area covered by the accretion spot . combining this with eqn . [ veqn ] , we have @xmath110 if we require that the optical depth in the pa @xmath12 line @xmath111 , as suggested by the good fit of the theoretical case b values to the observed line ratios , we can place a lower limit on @xmath109 . using the line opacity factors , @xmath112 , given by @xcite , we have @xmath113 and we find @xmath114 for @xmath72 @xmath7 and @xmath115 for @xmath70 @xmath7 . these limits are in reasonable agreement with the values of @xmath109 derived by @xcite . using their relation between @xmath109 and @xmath116 @xmath117 we derive @xmath118 from the pa @xmath12 luminosity . using this value in eqns . [ dr ] and [ drscale ] , we derive the thicknesses corresponding to the emission volumes of @xmath119 km for @xmath72 @xmath7 and @xmath120 km for @xmath70 @xmath7 . the ranges of thicknesses in turn imply column densities of @xmath121 and @xmath122 @xmath123 , respectively . since tw hya is seen nearly pole - on , the relatively small volumes and thicknesses we derive suggests that the h lines arise primarily from a fairly thin region , consisting of one or more hot spots located presumably at the base of the accretion column at the surface of the star near the magnetic pole . in fig . [ hlineprof ] we show the profiles of the pa @xmath12 and br @xmath13 emission lines . overplotted is a gaussian with a fwhm of 150 km s@xmath124 . ( recall that our spectra have @xmath125 in this wavelength range , which corresponds to a velocity resolution of @xmath126 km s@xmath17 . ) the high s / n afforded by our spectra allows us to discern multiple components in these profiles , which are replicated in many of the stronger h lines : a narrow ( unresolved ) core at the rest wavelength , a weaker component that extends over a fairly large velocity range ( from roughly @xmath127 to @xmath128 km s@xmath17 ) , and an inverse p cygni absorption feature on the red side at @xmath129 km s@xmath17 . we note that extended wings are not seen in the ( substantially weaker ) and line profiles , whereas they are present on even the weak , higher order paschen lines , a result that suggests that they are not due to our point spread function . in addition , the pa @xmath83 line was used to generate the telluric correction ( see * ? ? ? * ) and so should be the least affected by any residuals from the telluric correction process , yet it clearly exhibits broad wings . also , the broad emission wings are not seen on many of the h lines in the reduced spectrum of tw hya , but appear only after subtraction of the m2.5v template . furthermore , broad emission over a range of velocities up to @xmath130 km s@xmath17 is not uncommon in cttss ( e.g * ? ? ? the h line profiles suggest that the strongest emission arises at the base of the accretion column ( zero velocity ) , with additional emission from an outflowing wind with velocities up to about @xmath131 km s@xmath17 , and absorption from gas that is falling onto the star with a velocity of @xmath129 km s@xmath17 . [ [ section-1 ] ] exhibiting two absorption troughs on either side of an emission peak at the line center , the he i line profile ( fig . [ helineprof ] ) is more complicated than those of the h lines . the blue shifted absorption component has a minimum at @xmath132 km s@xmath17 and extends to about @xmath133 km s@xmath17 , while the weaker redshifted component has a minimum at @xmath129 km s@xmath17 and extends to about @xmath134 km s@xmath17 . the velocity of the red absorption feature is the same as that seen in the h line profiles , which suggests that the h and he absorption features are formed in the same region of the accretion flow or infall onto the star . our 1.083 @xmath2 m line profile is unlike any of those presented by @xcite , who observed the 1.083 @xmath2 m line in a sample of six ysos . nor is it similar to the 1.083 @xmath2 m line profile presented by @xcite from their observations of tw hya in 2002 . in the @xcite sample of 39 cttss ( which includes tw hya ) , the only profile the current observations of 1.083 @xmath2 m resembles is that of cy tau , although the extent in velocity of the blueshifted absorption we observe ( with our limited velocity resolution of @xmath126 km s@xmath17 ) agrees with that of the blueshifted emission seen in the tw hya spectrum presented by @xcite . these comparisons indicate that the he i line profile in tw hya is remarkably variable . this conclusion is strengthened by the profile of 1.083 @xmath2 m in the recently published spectrum of tw hya obtained by @xcite in 2008 using spex at the irtf . the 1.083 @xmath2 m profile in their spectrum exhibits a considerably stronger emission peak , a somewhat shallower p cygni absorption trough , and almost no evidence of the red absorption feature . this change in the line profile suggests a denser outflow during the observations of @xcite , which would be consistent with their somewhat stronger h lines . [ [ section-2 ] ] as discussed by @xcite and summarized by @xcite , the permitted lines seen in the spectrum of tw hya can arise from recombination , fluorescence excited by the stellar continuum , or resonance fluorescence excited by ly @xmath12 . because continuum fluorescence should generate an 0.8446 @xmath2 m line that is considerably stronger than the 1.1287 @xmath2 m line , while the observations indicate that these two lines have approximately equal strength , this excitation mechanism can be ruled out . the relative line strengths produced by recombination can be derived from the volume emission coefficients , @xmath135 @xmath136 and @xmath137 because and have very similar ionization potentials , we might expect o and h to have similar spatial distributions and ionization fractions in the emission regions . in this case , the strength of the 0.8446 @xmath2 m line relative to pa @xmath12 can be expressed as @xmath138 this is provides an upper limit to the flux ratio because it assumes all the o is in the form of ; some o may be in and higher ionization stages . extrapolating the effective recombination coefficients of @xcite for to electron temperatures of @xmath139 k , and using the h line emissivities given by @xcite and a solar oxygen abundance @xcite , we estimate line strength ratios resulting from recombination of @xmath140 and @xmath141 . the observed line ratios are found to be @xmath142 and @xmath143 . although it could be argued that the former ratio could be affected by reddening , which could be responsible for reducing the observed ratio below the expected value , the latter ratio is already @xmath144 times stronger than predicted and would only increase with any corrections for reddening . furthermore , to bring the former line ratio into agreement with the predictions would require a reddening correction corresponding to @xmath145 , which seems far too large for tw hya . therefore , under the assumption that the h and emission lines are generated in the same region , we can rule out recombination as the source of the emission . this leaves fluorescence by ly @xmath12 as the excitation mechanism for the lines . this mechanism should produce equal numbers of photons in the 0.8446 and 1.1287 @xmath2 m lines , which is exactly what we observe . the fact that we do observe equal numbers of photons in the lines again confirms that there is no differential reddening along the line of sight to the emitting region and very little dust within the emission region itself , as found previously by many others ( e.g. * ? ? ? * ) . in fig . [ veilplot ] we present the @xmath146 m spectrum of tw hya with the scaled irtf library spectra of a k7v star ( hd 237903 ) and an m2.5v star ( gl 381 ) overplotted . the middle panels show the result of subtracting the scaled library spectra from the observed spectrum , while the bottom panels present the ratios of this excess spectrum to the scaled library spectral templates . both library spectra were scaled to match the observed spectrum of tw hya between 0.95 and 1.34 @xmath2 m . template spectral types of m2v - m3.5v yield very similar excess emission spectra to that shown in the middle right panel of fig . [ veilplot ] . as seen in the middle left of fig . [ veilplot ] , earlier template spectral types yield excess emission spectra with several broad features " , particularly between the j , h , and k photometric bands , and somewhat stronger excess emission longwards of @xmath3 @xmath2 m , but wholly unphysical negative values shortward of @xmath147 m . the excess flux spectrum we derive from subtracting the m2.5v star exhibits a rise to shorter wavelengths ( @xmath148 m ) that is most likely due to the ( possibly optically thick ) continuum emission produced by the accretion process ( `` optical veiling '' ) . the excess at longer wavelengths could be due to the thermal emission from the accretion disk , either the optically thick disk at @xmath149 au @xcite or an optically thin inner disk @xcite . however , the excess spectrum can not be fit with any simple disk model ; the model of @xcite is shown as a dotted line in this panel and clearly does not match the shape of the derived spectrum , greatly underestimating the flux in the h and k bands and overestimating the flux at longer wavelengths . the peak in the excess flux spectrum in the h band requires a dust temperature of @xmath150 k , but optically thin dust disk models , with power law distributions for the temperature and surface density , generally can not reproduce the sharp drop off in flux on either side of this peak that is seen in our data . an ` excess flux ' spectrum that peaks in the h band is very reminiscent of the spectral difference observed between late type giants and dwarfs . as can be seen in the series of spectra presented in the irtf spectral library @xcite , for a given spectral type a late type giant star exhibits a prominent h band excess over that of a dwarf star due to the decrease in the @xmath151 opacity at the lower density and gravity @xcite . to demonstrate this , in fig . [ giant_dwarf ] we plot the ` excess ' spectrum derived by scaling the spectrum of an m2.5v star to that of an m2.5iii star , both taken from the irtf spectral library , over the wavelength range @xmath152 m , and subtracting the former from the latter . the resulting difference spectrum was then scaled to match the level of the h band excess in the spectrum of tw hya and overplotted on the tw hya excess spectrum . as can be seen from this figure , the shapes of the excesses are nearly identical , with the giant @xmath21 dwarf excess extending from the h band through the k band and into the m band . clearly , of course , tw hya is not an m giant . nevertheless , the mass and radius values determined above yield a gravity of @xmath153 , far below the value of @xmath154 estimated for field m2.5 dwarfs from the parameters given by @xcite and @xcite . therefore , we believe the tw hya h and k band excess is in fact not indicative of disk emission , but rather is a signature of tw hya s youth ( relative to field m dwarfs ) , as it contracts and evolves down to the main sequence . this conjecture might be tested with higher resolution observations of atomic absorption lines , whose widths would reflect the value of @xmath155 , in much the same manner as described by @xcite for l dwarfs . if the excess could be calibrated as a function of @xmath155 , it could provide a separate estimate of the age of the system . under the assumption that the excess spectrum is due to emission arising from the accretion process , the ratio of the excess spectrum to the library template yields the veiling @xmath55 . as seen in the bottom right panel of fig . [ veilplot ] , the veiling derived from the m2.5v spectral template is fairly small : @xmath156% between 0.8 and 0.9 @xmath2 m , @xmath157% between 1.4 and 1.8 @xmath2 m , @xmath18% between 2.0 and 2.4 @xmath2 m , and @xmath158% between 3.0 and 4.0 @xmath2 m . our derived veiling values are in good agreement with the estimate of @xcite for the k band and those of @xcite for @xmath159 m . however , if the suggestion given above for the shape of the excess spectrum is correct , only a fraction of this excess emission arises from the accretion process and therefore our veiling estimates for the h , k and l bands are strict upper limits . although it is uncertain how much of the excess longwards of @xmath160 @xmath2 m is due to the effects of the lower gravity of tw hya relative to the comparison field m dwarf used for the subtraction , it is possible that this difference in gravity is responsible for nearly all of the observed excess in the h and k bands , and a significant fraction of the excess in the l band . the wavelength region around 1 @xmath2 m is ideal for spectral classification of cttss : ( 1 ) the excess emission arising from the accretion process , which is very strong in the uv and optical wavelengths , is relatively small ; ( 2 ) the emission from the accretion disk itself , which can dominate the spectrum at k band and longer wavelengths , is minimal ; ( 3 ) the effects of any extinction are mitigated substantially compared to the optical ; and ( 4 ) there are numerous strong photospheric absorption features present that can be used for classification . the last point is particularly important because , while the contributions from the accretion shock and circumstellar disk ( uv and ir excesses , respectively ) are at a relative minimum in the y and j bands , they are not necessarily negligible and evidence for significant j band veiling has been found in some sources @xcite . however , the ratio of absorption ews , as discussed in 3.1 , provides a veiling independent method of determining the spectral type of a source . although the relatively ` clean ' nature of this spectral region has been pointed out before @xcite and exploited to study the emission associated with the accretion process in cttss @xcite , this is the first time of which we are aware that the absorption features in this region have been used for spectral classification of a ctts . the absorption line ews we measure , in comparison with those derived from the irtf library spectra , provide strong evidence that the underlying star in tw hya has a spectral type of @xmath14 m2.5v . this result suggests that the values for the stellar mass and radius typically adopted in the literature ( @xmath161 k , @xmath162 , and @xmath163 ; e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) are incorrect , as they are based on an incorrect spectral type ( k7v ) . we suggest that the appropriate values for tw hya are @xmath164 k , @xmath165 , and @xmath166 , based on an m2.5v spectral type , the temperature and mass calibrations given by @xcite and @xcite , and the estimated bolometric luminosity . the spectral type we determine , and the stellar parameters we derive , are clearly inconsistent with those found by others ( e.g. * ? ? ? * ; * ? ? ? examination of the spectra shown by @xcite indicates that the k7v spectral classification is appropriate , as tw hya clearly has an earlier type than those of the m stars shown . @xcite and @xcite fit atmosphere models to high resolution optical spectra and derived parameters again consistent with a k7v star . on the other hand , the mass estimated for tw hya from the @xcite stellar parameters is far too large for even a k7v spectral type . furthermore , the moderate resolution optical spectrum presented by @xcite ( their figure 9 ) , with its strong rise in the continuum to the red and its sharp and prominent tio bandheads near 7100 , is more suggestive of an early m dwarf than a late k dwarf ( see below ) and does not seem to be completely consistent with the spectrum presented by @xcite . can these two discrepant spectral type classifications be reconciled ? the optical spectrum of tw hya suffers from considerable and variable veiling ( see e.g. , * ? ? ? * ) , so it is possible that veiling has contributed to mis - classification , especially if the spectral comparison is restricted to very small wavelength regions lacking strong lines diagnostic of spectral type . to investigate this possibility , we selected an m2v star from the indo - us spectral library @xcite and added veiling in an attempt to reproduce a k7v star spectrum between 6400 and 7600 ( the spectral region displayed in fig . 1b of @xcite ) . we found that it was not possible unless the veiling was substantially larger ( @xmath167 ) than that estimated by either @xcite or @xcite in this wavelength range . therefore , veiling can not be completely responsible for the classification differences . we also considered the possibility that tw hya is an unresolved binary . however , given the relative luminosities of k and m stars , it is not physically possible for a coeval k7v+m2.5v binary system to generate a spectrum in which the k7v component dominates the optical spectrum but is nearly invisible in the nir spectrum . because both the continuum and line emission are know to vary substantially in tw hya , it is tempting to consider the possibility that the spectrum ( and spectral type ) of the object varies as a function of both accretion rate and wavelength . a physically plausible explanation for this variation is that the spectrum of tw hya consists of two separate components , one due to the star ( m2.5v ) and one arising from an accretion hot spot , either on the surface of the star or located in the inner disk , or from the inner disk itself . in this scenario , the hot spot or inner disk would have a temperature of @xmath168 k and would generate a spectrum similar to that of a k7v star . this component would also vary in strength with the accretion rate , dominating the ( optical ) spectrum and completely overwhelming the underlying m2.5v stellar spectrum when the accretion rate is high . support for this suggestion can be found by comparing the observed optical spectra with the reported h@xmath11 equivalent widths , a proxy for the accretion rate . the optical spectrum presented by @xcite is clearly that of a k7v star and yields an @xmath169 . similarly , the optical spectrum shown by @xcite is consistent with that of a k7v star and has @xmath170 ( c. m. johns - krull , private communication ) . the optical spectrum presented by @xcite , however , corresponds to spectral type intermediate between a k7v and an early m star , and has @xmath171 between @xmath172 and @xmath173 . in fact , when we attempted to reproduce the features seen in this spectrum we found that a combination of a k7v and an m2.5v , with roughly equal contributions at 0.8 @xmath2 m , provides a much better representation than any single k or early m spectrum alone . the shape and depth of the tio bands at 7100 and the simultaneous presence of these bands and a strong line at 7148 indicates that the observed spectrum is a composite of both a k7v and an early m type spectrum . in our nir spectrum of tw hya , the region shortward of 1@xmath2 m can also be well fit with an m2.5v spectrum and a small contribution from a k7v spectrum , comprising @xmath144% of the total flux at 0.8 @xmath2 m . this is shown in fig . [ m+k ] , where it can be seen that at wavelengths shorter than @xmath174 m we would expect that , during our observations of tw hya , the k7v spectrum contributed roughly half of the total flux , but comprised only a very small fraction of the flux longward of @xmath175 m . unfortunately a more satisfactory test of this suggestion can not be carried out with the spectra and accretion rates or ew values found in the literature . ( surprisingly few optical and nir spectra of tw hya are actually shown in the voluminous literature on this object . ) frequent simultaneous monitoring of tw hya in both the optical and nir wavelength regimes would be able to confirm or rule out our suggestion . our results for the spectral type and effective temperature require a re - evaluation of the age of tw hya . the age adopted for tw hya under the assumption of a k7v spectral type with @xmath176 k is often @xmath18 myr , derived by comparing the location of tw hya in the hr diagram with model isochrones ( e.g. * ? ? ? * ) . in fig . [ hrd ] , we plot the location of tw hya on the h - r diagram , using our estimated @xmath177 , along with the model tracks and isochrones of @xcite , @xcite , @xcite , and @xcite . despite the considerable variation in the model predictions , all of the models yield masses smaller than our adopted value of @xmath178 and ages substantially younger than the canonical value of 10 myr . large underestimates of both mass and age by the various models have been pointed out previously by @xcite . however , even accounting for an underestimate of the age by a factor perhaps as large as two @xcite would be insufficient to make the star 10 myr old . the models of @xcite provide a mass of @xmath179 , closest to our adopted value of @xmath178 . for these models , the age of tw hya estimated from the isochrones is @xmath3 myr . we note that the new location of tw hya on the h - r diagram actually brings this source into better agreement with other members of the twa , as can be seen from fig . 3 of @xcite , which indicates that , aside from twa 6 and twa 9 a and b , the locations of the members of the twa on the h - r diagram are consistent with an isochrone age for the association of @xmath180 myr . this age is also in reasonable agreement with the expansion age of the twa of @xmath181 myr derived by @xcite , and therefore alleviates to a considerable degree the previous discrepancy between the isochrone age of twa and dynamical age estimates . furthermore , compared to the usually adopted age of 10 myr , an age of 3 - 5 myr for tw hya is in much better agreement with estimates of the half life of accretion disks in low mass stellar systems of @xmath182 myr @xcite . by 10 myr the disks in such systems have usually transformed into debris disks with masses of @xmath183 @xcite . the large mass of the disk in tw hya ( @xmath184 * ? ? ? * ) therefore suggests an age of considerably less than 10 myr @xcite , in agreement with our much younger estimate . we also note that tw hya exhibits many of the same emission features , and has exactly the same li 6708 absorption ew , as the m2.5 star gn tau @xcite . the latter star is a low - mass t tauri star with an estimated age of @xmath3 myr , again providing support for our age estimate for tw hya . the determination of an accurate spectral type permits the subtraction of the correct spectral template to reveal the emission from the accreting gas and the circumstellar disk . under the assumption that the h lines are optically thin and follow the case b predictions , our analysis of the h line flux ratios yields a temperature of @xmath185 k , a density of @xmath186 @xmath7 and a thickness of the emission column of @xmath187 km . comparison of our results with those presented by @xcite , who modeled the emission from the accreting column of gas in the magnetospheric infall scenario , reveals surprisingly good agreement for some of the physical parameters . the models of @xcite include emission from the shock and post - shock region , the pre - shock region , and the heated stellar atmosphere . the emission properties of the models are parametrized in terms of the energy flux carried by the accretion column into the shock , @xmath188 , given by @xmath189 the h line emission in these models arises from the pre - shock region , which is predicted to have a density given by @xmath190 for the stellar parameters for tw hya given above , these equations give @xmath191 erg @xmath123 s@xmath17 and @xmath192 @xmath7 . the latter value is fairly close to our empirical results ( under the assumption that all of this material becomes fully ionized in the pre - shock region ) . the models also predict a temperature for this region of @xmath193 k , and thickness of the fully ionized zone of @xmath194 km , again in good agreement with our findings . recently @xcite has used interferometry to determine the stellocentric radii of the br @xmath13 emitting gas in t tauri and herbig ae / be stars . they find that the emission in general arises from gas within 0.01 au ( @xmath195 km ) , an upper limit that our results certainly satisfy . the infall velocity in the models of @xcite is predicted to be @xmath196 or @xmath197 for the mass and radius we derive for tw hya , in excellent agreement with the velocities corresponding to the observed inverse p cygni absorption features seen in our h and 1.083@xmath2 m line profiles . using our derived density and placing a limit on the optical depth of the pa @xmath12 line , we derive a lower limit to the fractional area covered by the accretion column of @xmath198 . our limit is well within the range of previous estimates of the covering fraction , for both tw hya itself as well as other cttss @xcite . although our limit is a factor of 7 larger than that estimated by @xcite , we also derive substantially larger ( factors of 4 - 5 ) values of the accretion luminosity and rate than these authors . ( @xcite also derive a substantially larger value of the accretion energy flux @xmath188 than we do , largely because of this difference in the values of the covering fractions @xmath109 . ) if the covering fraction is correlated with the accretion rate , as claimed by @xcite , then the considerably larger accretion rates we measure compared to @xcite should be associated with a larger covering fraction . it is interesting to note that the values of the energy flux and covering fraction we derive for tw hya are very similar to those found by @xcite for cy tau , the star whose 1.083 @xmath2 m line profile resembles that in our spectrum of tw hya most closely . despite the good agreement between our temperature and density estimates and the model predictions of @xcite , our values do not agree with those of @xcite for their sample of cttss , and our analysis rules out their values for tw hya with high confidence . @xcite do not present results for individual stars , so it is unclear how much variation they find from object to object . the mass accretion rates listed for the objects in their sample are substantially higher than that measured for tw hya . ( the minimum value listed in their table is more than double the value derived for tw hya in 3.2.1 from the pa @xmath12 line and five times larger than that derived from the br @xmath13 line . ) however , one would expect that higher mass loss rates would generate larger volume densities . at the moment , we have no satisfactory explanation for this discrepancy . the high densities we derive provide a reasonable explanation for why other lines typical of ionized regions ( e.g. , forbidden lines of and ) are weak or absent in optical and nir spectra of tw hya obtained with narrow slits centered on the star ( e.g. , * ? ? ? * ) . since these ionization states are found primarily within the region where h is fully ionized and the density is well above the critical density for these transitions , the ground - state levels giving rise to these transitions are collisionally de - excited . spectra obtained with wide slits , encompassing lower density regions farther from the star , might be expected to reveal the usual forbidden lines ( see e.g. , * ? ? ? our analysis suggests that the lines seen in the nir spectrum are produced by ly @xmath12 fluorescence . it is not immediately obvious , however , where the emission is generated . @xcite demonstrated that the profile of the 0.8446 line correlates with that of pa 11 in a small sample of cttss , and therefore suggested that the lines were formed in a common region . furthermore , the lines in our spectrum of tw hya are unresolved at our resolution , which indicates that they are not arising from a region experiencing substantial infall or outflow , and therefore it is tempting to suggest that they are formed at the base of the accretion column , along with the h lines . however , production of the lines by ly @xmath12 fluorescence requires a high ly @xmath12 flux density in a region that contains neutral oxygen . the latter requirement is inconsistent with completely ionized h gas in the pre - shock region in the models of @xcite . a partially ionized region at the outer edge of the pre - shock region could be responsible for the emission . this region could also give rise to the [ ] 6300 , 6363 lines that are seen in the optical spectrum @xcite . however , fluorescence also requires a large optical depth in the balmer transitions . using the equation given by @xcite for the escape probability of h@xmath11 , @xmath199 and that by @xcite for the optical depth @xmath200 in terms of @xmath199 , we derived a rough estimate of the h@xmath11 optical depth of @xmath201 . such a large optical depth in h@xmath11 seems incompatible with the finding that the observed h line strength ratios are well reproduced by the theoretical case b values ( see also * ? ? ? * ) , which assumes that all transitions above the ly series are optically thin . a more detailed model of the emission line region and its spectrum will be needed to resolve this puzzle . finally , our derived nir excess flux spectrum , usually interpreted as arising from the hot inner accretion disk , does not match that from any recent model of tw hya . while the h - band peak in our excess flux spectrum suggests a dust temperature of @xmath150k , which would be expected for dust in the inner disk , the sharp drops in flux at both shorter and longer wavelengths are not reproduced by current optically thin disk models . in fact , as discussed above , we suspect that most of this excess emission does not arise from the inner accretion disk , but is rather due to the lower h@xmath202 opacity in the star s atmosphere relative to that of an older , higher gravity field dwarf . therefore , during our observations , any nir emission from the hot inner accretion disk must have been very small , much smaller than either previous estimates or the predictions of current models . it is possible that previous values of the nir excess have been overestimated as a result of both the adoption of an incorrect spectral type for tw hya , as well as the spectral differences due to differences in the opacity , gravity , and age between tw hya and any comparison field dwarf . matching a library template with a spectral type earlier than m2v to the short wavelength end of the nir spectrum would yield substantially more excess emission at wavelengths @xmath203 @xmath2 m than we estimate . in addition , scaling earlier spectral type templates to match the observed tw hya flux levels between 0.95 and 1.34 @xmath2 m yields an excess spectrum for which synthetic photometry is in reasonable agreement with some of the models ( e.g. * ? ? ? nevertheless , in these cases the resulting excess flux spectra appear to be unphysical ( because of negative values ) , and are certainly not the smooth blackbody functions characterized by a single dust temperature expected from the models or observed in other ctts sources ( e.g. * ? ? ? we caution that any attempts to determine the nir excess emission arising from accretion disks around young stellar objects similar to tw hya must account for any possible differences in age and gravity between the target objects and the spectral templates . failure to do so could lead to spurious results . as we have shown here , the use of an old , higher gravity field dwarf as a spectral template can lead to an excess spectrum that primarily reflects differences in the h@xmath202 atmospheric opacity between the ctts and the template , not the emission from an accretion disk . we have presented new high signal - to - noise , moderate resolution nir spectra of tw hya obtained with spex at the irtf . these data have enabled us to revise the spectral classification of this object . we find that the spectrum is more consistent with a spectral type of m2.5v , rather than the commonly adopted k7v . based on this spectral classification , we derive new values for the effective temperature ( @xmath204 k ) , mass ( @xmath178 ) , radius ( @xmath205 ) , and age ( @xmath206 myr ) of the system . subtraction of an m2.5v template drawn from the irtf cool star spectral library reveals emission lines of h , he , ca , and o. the h lines consist of a narrow component , a much weaker broad component , and a red absorption feature . analysis of the strengths of the narrow h lines indicates that they arise from a hot ( @xmath207 k ) , extremely dense ( @xmath208 @xmath7 ) region with a filling factor of @xmath115 and a thickness of @xmath209 km . the temperature and thickness values agree remarkably well with the predictions from the model of magnetospheric accretion developed by @xcite . we interpret the structure in the line profile as a reflection of the relative locations of the emitting regions : the narrow core arises from the base of the accretion shock column at the stellar photosphere , while the inverse p cygni absorption arises from the infall of gas in the accretion column and the extended wings arise from the outflow of material in a disk or coronal wind . analysis of the permitted lines indicates that they are generated by ly @xmath12 fluorescence . the location of the mostly neutral region generating the lines , however , is uncertain . we also find very little excess continuum emission arising from the optically thin hot inner accretion disk . the spectrum of the excess emission we derive does not appear to agree with the model of the inner disk presented by @xcite . rather the shape of the excess suggests that it results from the lower h@xmath202 opacity in the atmosphere of tw hya , due to the lower @xmath155 and younger age , compared to the older , higher @xmath155 value for field m2.5v stars used for the spectral comparison . wdv thanks m. c. cushing for providing software useful for some of the analysis and plots presented in this paper , as well as for discussions about m stars . we also thank j. najita for discussions regarding tw hya , c. johns - krull for providing h @xmath11 equivalent width measurements , and c. howard for comments on the manuscript . mora , a. , mern , b. , solano , e. , montesinos , b. , de winter , d. , eiroa , c. , ferlet , r. , grady , c. a. , davies , j. k. , miranda , l. f. , oudmaijer , r. d. , palacios , j. , quirrenbach , a. , harris , a. w. , rauer , h. , cameron , a. , deeg , h. j. , garzn , f. , penny , a. , schneider , j. , tsapras , y. , & wesselius , p. r. 2001 , , 378 , 116 @xmath2 m spectrum of tw hya with those of late k and early m dwarfs stars in the irtf library . note the variation in the strength of the feh absorption with spectral type . the identification of other features can be found in @xcite . ( bottom ) residual spectrum obtained by subtracting the scaled m2.5v spectrum from that of tw hya . , height=672 ] @xmath2 m spectrum of tw hya with those of late k and early m dwarfs stars in the irtf library . note the variation in the strength of the k i and h@xmath48o absorption features with spectral type . the identification of other features can be found in @xcite . ( bottom ) residual spectrum obtained by subtracting the scaled m2.5v spectrum from that of tw hya . , height=672 ] @xmath2 m spectrum of tw hya with those of late k and early m dwarfs stars in the irtf library . note the variation in the strength of the mg and al absorption features , as well as the complex of fe and si lines between @xmath210 and @xmath211 @xmath2 m , with spectral type . the identification of other features can be found in @xcite . ( bottom ) residual spectrum obtained by subtracting the scaled m2.5v spectrum from that of tw hya . , height=672 ] @xmath2 m spectrum of tw hya with those of late k and early m dwarfs stars in the irtf library . note the variation in the strength of the mg absorption feature with spectral type . the identification of other features can be found in @xcite . ( bottom ) residual spectrum obtained by subtracting the scaled m2.5v spectrum from that of tw hya . , height=672 ] values ( relative to @xmath212 ) as a function of electron density @xmath213 and temperature @xmath214 from fits of the theoretical h line fluxes to the values derived from the continuum subtracted observations of tw hya . contour levels are 1.05 , 1.1 , 1.2 , 1.3 , 1.5 , 2.0 , and 3.0 @xmath212 . pluses denote the temperature and density combinations for which theoretical line ratios were computed and compared to the observations . the solid triangle denotes the best fit . , height=480 ] k and @xmath215 @xmath7 ) relative to pa @xmath12 . we detected only a single line of the humphreys series ( hu 11 ) ; the solid green square denotes the theoretical value for this line . the dotted lines and open green square denote the theoretical line fluxes for the temperature and density found by @xcite in their study of 15 ctts systems ( @xmath216 k and @xmath217 @xmath7).,width=576 ] m. ( middle ) excess flux spectra generated by subtracting the scaled k7v spectrum ( left ) and the scaled m2.5v spectrum ( right ) from the observed spectrum of tw hya . the red dashed line denotes the excess flux predicted by the model of @xcite . ( bottom ) the percentage excess relative to the scaled template spectra.,height=720 ] & 1.313 & 2.38 & 0.08 + & 1.675 & 4.13 & 0.17 + & 2.110 & 0.80 & 0.08 + & 2.117 & 1.12 & 0.08 + & 1.253 & 0.72 & 0.04 + & 1.183 & 0.97 & 0.03 + & 1.504 & 4.81 & 0.14 + & 1.488 & 1.51 & 0.13 + & 1.711 & 2.61 & 0.21 + & 2.107 & 0.50 & 0.07 + & 1.140 & 2.94 & 0.34 + & 2.206 & 4.41 & 0.08 + & 2.336 & 2.12 & 0.16 + feh & 0.99 & 3.10 & 0.08 + l c c l 0.8183 & 8.21 & 0.63 & + 0.8195 & 11.18 & 0.71 & + 0.8448 & 13.61 & 0.58 & + 0.8502 & 17.03 & 0.83 & + pa 16 + 0.8545 & 19.16 & 0.70 & + pa 15 + 0.8602 & 6.01 & 0.82 & pa 14 + 0.8630 & 1.85 & 0.56 & ? + 0.8665 & 18.52 & 0.75 & + pa 13 + 0.8753 & 12.02 & 0.84 & pa 12 + 0.8864 & 16.83 & 0.80 & pa 11 + 0.9018 & 13.99 & 0.79 & pa 10 + 0.9232 & 32.16 & 0.88 & pa 9 + 0.9549 & 47.15 & 1.09 & pa 8 ( @xmath218 ) + 1.0051 & 64.00 & 1.00 & pa 7 ( @xmath83 ) + 1.0406 & 9.01 & 1.04 & [ ] ? + 1.0941 & 92.11 & 0.92 & pa 6 ( @xmath13 ) + 1.1289 & 9.99 & 0.40 & + 1.2821 & 176.89 & 1.79 & pa 5 ( @xmath12 ) + 1.6812 & 2.45 & 0.87 & br 11 + 1.7368 & 8.69 & 1.33 & br 10 + 1.9449 & 10.11 & 1.11 & br 8 ( @xmath218 ) + 2.1661 & 22.32 & 1.37 & br 7 ( @xmath13 ) + 3.0395 & 3.27 & 0.32 & pf 10 ( @xmath218 ) + 3.2971 & 4.96 & 0.30 & pf 9 ( @xmath83 ) + 3.7404 & 7.70 & 0.47 & pf 8 ( @xmath13 ) + 4.0523 & 62.38 & 0.55 & br 5 ( @xmath11 ) + 4.6538 & 12.57 & 0.84 & pf 7 ( @xmath12 ) + 4.6727 & 1.46 & 0.58 & hu 11 ( @xmath218 ) +

in this article we extend the results presented in ref . [ phys . rev . a * 76 * , 032101 ( 2007 ) ] to treat quantitatively the effects of reservoirs at finite temperature in a bosonic dissipative network : a chain of coupled harmonic oscillators whichever its topology , i.e. , whichever the _ _ way _ _ the oscillators are coupled together , the _ _ strength _ _ of their couplings and their _ natural frequencies_. starting with the case where distinct reservoirs are considered , each one coupled to a corresponding oscillator , we also analyze the case where a common reservoir is assigned to the whole network . master equations are derived for both situations and both regimes of weak and strong coupling strengths between the network oscillators . solutions of these master equations are presented through the normal ordered characteristic function . we also present a technique to estimate the decoherence time of network states by computing separately the effects of diffusion and the attenuation of the interference terms of the wigner function . a detailed analysis of the diffusion mechanism is also presented through the evolution of the wigner function . the interesting collective diffusion effects are discussed and applied to the analysis of decoherence of a class of network states . finally , the entropy and the entanglement of a pure bipartite system are discussed .

0905.1107

the subject of networks of interacting quantum systems has acquired an important role in the emerging field of quantum information theory . since a realistic quantum logic processor must ultimately be composed of a large number of interacting quantum systems , it becomes mandatory to understand processes as like as perfect state transfer from one to another system of the network , and even to compute the fidelity of such a state transfer when the action of the environment is taken into account . a significant amount of result has recently been derived on the subject of perfect state transfer in optical lattices @xcite and networks of spin @xcite and harmonic oscillators @xcite . perfect state transfer has also been considered in networks of arbitrary topology and coupling configuration @xcite and even under random fluctuations in the couplings of a quantum chains @xcite . apart from state transfer , the process of decoherence of a network state has also attracted attention and interesting properties of collective damping effects , as the nonadditivity of decoherence rates , have been discussed in different contexts as in superconducting qubits @xcite , two - atom systems @xcite , and chains of dissipative harmonic oscillators @xcite . still regarding the process of collective decoherence , the emergence of decoherence - free subspaces ( dfss ) has also instigated several interesting results when considering the particular case of a composite system interacting with a common reservoir @xcite , or the more realistic situation where each system interacts with its own reservoir @xcite . we call the attention to the fact that all refs . @xcite envisage such realistic cases of networks where each oscillator interacts with its own reservoir , also addressing the particular case where a common reservoir is considered . to better understand the results in refs . @xcite , which are crucial to introduce the subject of the present work , we remember that , apart from the distinct reservoirs , the network of @xmath0 dissipative harmonic oscillators could present direct an indirect dissipative channels . through the direct channels each oscillator loses excitation to its own reservoir , whereas through the indirect channels it loses excitation to all the other reservoirs but to its own . when we consider distinct reservoirs for each network oscillator , the indirect dissipative channels intrinsically associated with the nonadditivity of decoherence rates and the emergence of dfss @xcite are significant only in the strong coupling regime where @xmath1 , i.e. , the number of network oscillators @xmath0 multiplied by their coupling strengths @xmath2 are about their natural frequencies @xmath3 . therefore , the strong coupling regime , which brings together the collective damping effects , depends on the number of network oscillators as much as on their coupling strengths . for markovian white noise reservoirs , however , where the spectral densities of the reservoirs are invariant over translation in frequency space , the indirect channels becomes null , except for the case @xmath4 @xcite . in the weak coupling regime where @xmath5 , the indirect channels always disappears . however , these indirect channels , coming from the strong coupling regime , remains in the case where all network oscillators interacts with a common reservoir @xcite , even assuming a common markovian white noise reservoir . this is due to the fact that a common reservoir induces an additional correlation between the network oscillators , restoring the indirect decay channels . recently , a generalization of refs . @xcite has been presented through a comprehensive treatment of networks of dissipative quantum harmonic oscillators , whichever its topology , i.e. , whichever the _ _ way _ _ the oscillators are coupled together , the _ _ strength _ _ of their couplings and their _ natural frequencies _ @xcite . focusing on the general more realistic scenario where each oscillator is coupled to its own reservoir , the case where all the network oscillators are coupled to a common reservoir was also addressed . however , after deducing the master equation for the case where all the reservoirs are at finite temperatures , all further analysis of the dynamics of the network states was restricted to the case where the reservoirs are at @xmath6 k. whereas a quantitative analysis of the decoherence and the evolution of the linear entropy of representative states of the network were given at @xmath7 k , only a brief qualitative analysis of the equilibrium states of the network was presented at finite temperatures . in the present manuscript we extend the treatment in ref . @xcite given a detailed analysis of the temperature effects on networks of dissipative quantum harmonic oscillators . the present extension of ref . @xcite that accounts for the temperature effects coming from thermal reservoirs is not only interesting due to its more realistic approach but also from the mathematical development here achieved . in fact , we present an alternative approach to previous results in the literature @xcite regarding the obtainment of the solution of the master equation and the estimation of decoherence times through the wigner distribution function . to circumvent noise effects , many of the nowadays experiments demonstrating quantum logic operations through atom - field interactions occur in cryogenic environments where temperature effects are negligible . in cavity quantum electrodynamics , the setup is cooled to around 0.5 k by a @xmath8he-@xmath9he refrigerator to avoid blackbody radiation in the high - q superconducting cavity . under such a specific condition , the temperature effects on the decoherence process are almost negligible . however , when the setup is scaled from one single cavity to a network of @xmath0 high - q cavities , major questions arise due to temperature effects . first of all , would the dfss survive despite the temperature effects ? apart from the special class of states composing the dfss , how the temperature affects other states of the network , as for example initial entangled states ? evidently , these questions present no obvious answers , even under the assumption that all network cavities are cooled at low temperatures . on this regard , we expect that the collective damping effects coming from the indirect dissipative channels to play a major role for the answers to the above questions . apart from providing the mathematical treatment of the temperature effects on a network of @xmath0 dissipative harmonic oscillators , in the present manuscript we also analyze the role played by temperature on the evolution of particular states of the network other than those composing dfss . we reserve the analyses of the emergence of dfss under temperature effects to an specific work @xcite where the mechanism of construction of such privileged states is also discussed along with decoherence . this paper is organized as following : in section ii we revisit our model of a bosonic dissipative network @xcite and present the derivation of the master equation governing the dynamics of the associated density operator . in section iii we present the solution of the normal ordered characteristic equation obtained from the master equation for the density operator of the network . in section iv we analyze the evolution of two general classes of initial states of the network , given by mixtures of coherent and number states , through the normal ordered characteristic equation , the glauber - sudarshan @xmath10-function , and the wigner distribution . a detailed analyses of the diffusion processes is presented in section v and the collective decoherence rates of a family of states of the network is analyzed in section vi . in section vii we discuss the entropy and the entanglement degree of a pure bipartite system and , finally , in section viii we present our concluding remarks . we present here a brief review of the steps for the derivation of the master equation of a bosonic network , as developed in ref . we start from the general case of a network of @xmath0 interacting oscillators , where each one interacts with each other , from which all other topologies can be recovered . as depicted in fig . 1 , we also consider the case where each oscillator interacts with its own reservoir due to this more realistic approach for most of the physical systems . however , as pointed in ref . @xcite , despite the realistic scenario of the case of distinct reservoirs , the case of a common reservoir is more general from the technical point of view . in fact , as discussed at the end of this section , the master equation for the case of distinct reservoirs can be deduced from the case of a common reservoir . we start by considering a general hamiltonian for a bosonic network , @xmath11 , composed by a set of @xmath0 coupled oscillators @xmath12 \text { , } \label{1}\ ] ] @xmath0 distinct reservoirs , modeled by a set of @xmath13 modes,@xmath14 and the coupling between the network oscillators and their respective reservoirs@xmath15 where @xmath16 ( @xmath17 ) is the creation ( annihilation ) operator for the @xmath18th bath mode @xmath19 coupled to the @xmath20th network oscillator @xmath21 whose creation ( annihilation ) operator reads @xmath22 ( @xmath23 ) . the coupling strengths between the oscillators are given by the set @xmath24 , while those between the oscillators and their reservoirs by @xmath25 . we assume , from here on , that @xmath26 and @xmath27 run from @xmath28 to @xmath0 . before addressing the dissipative process through hamiltonian ( [ 3 ] ) , we focus first on hamiltonian @xmath29 to show how to derive different topologies of a nondissipative network of coupled harmonic oscillators . rewriting @xmath29 in a matrix form@xmath30{ccc}a_{1}^{\dag } & \cdots & a_{n}^{\dag}\end{array } \right ) \left ( \begin{array } [ c]{ccc}\mathcal{h}_{11 } & \cdots & \mathcal{h}_{1n}\\ \vdots & \ddots & \vdots\\ \mathcal{h}_{n1 } & \cdots & \mathcal{h}_{nn}\end{array } \right ) \left ( \begin{array } [ c]{c}a_{1}\\ \vdots\\ a_{n}\end{array } \right ) \text { , } \label{3.5}\ ] ] we identify the elements of the matrix @xmath31 as@xmath32{ccc}\omega_{m } & \text{for } & m = n\\ \lambda_{mn } & \text{for } & m\neq n \end{array } \right . \text { , } \label{4}\ ] ] whose values characterize whichever the network topology , i.e. , whichever the way the oscillators are coupled together , the set of coupling strengths @xmath24 , and their natural frequencies @xmath33 . to obtain the master equation of the network we first diagonalize the hamiltonian @xmath34 ( within the physical regime where the normal modes are assumed to be positive ) through the canonical transformation@xmath35 where the coefficients of the @xmath20th line of matrix @xmath36 define the eigenvectors associated to the eigenvalues @xmath37 of matrix @xmath34 . with @xmath36 being an orthogonal matrix , its transposed @xmath38 turns out to be exactly its inverse @xmath39 , resulting in the commutation relations @xmath40 = \delta_{mn}$ ] and @xmath41 = 0 $ ] , which enable the hamiltonian @xmath42 to be rewritten as a sum @xmath43 , where [ 6]@xmath44 with the diagonalized hamiltonian @xmath45 we are ready to introduce the interaction picture , defined by the transformation @xmath46 , where @xmath47 \text { , } \label{7}\ ] ] and @xmath48 b_{mk}$ ] . next , we assume that the interactions between the resonators and the reservoirs are weak enough to allow a second - order perturbation approximation . we also assume a markovian reservoir such that the density operator of the global system can be factorized as @xmath49 . under these assumptions the reduced density operator of the network of @xmath0 dissipative coupled resonators satisfy the differential equation@xmath50 \right ] \text{. } \label{8}\ ] ] since for a thermal reservoir @xmath51 , we have to solve the integrals appearing in eq . ( [ 8 ] ) , related to correlation functions of the form@xmath52 \right\ } . \label{9}\ ] ] considering that the reservoir frequencies are very closely spaced to allow a continuum summation , we obtain @xmath53 where we have defined the average excitation of the reservoir associated to the @xmath20th oscillator as @xmath54 through the relation @xmath55 , apart from the damping rates [ 11]@xmath56 ^{2}e^{-i\left ( \nu-\omega\right ) \left ( \tau - t\right ) } \text{,}\label{11a}\\ \tilde{\gamma}_{m}(\omega ) & = \int_{0}^{t}d\tau\int_{0}^{\infty}\frac{d\nu } { n\pi}\left [ v_{m}(\nu)\sigma_{m}(\nu)\right ] ^{2}\frac{\bar{n}_{m}(\nu ) } { \bar{n}_{m}(\omega)}e^{-i\left ( \nu-\omega\right ) \left ( \tau - t\right ) } \text { , } \label{11b}\ ] ] with @xmath57 being the density of states of the @xmath20th reservoir . in the context of markov approximation , where @xmath58 and @xmath59 are slowly varying functions around the normal modes @xmath60 we can simplify the expressions ( [ 11 ] ) to their usual forms @xmath61 ^{2}\text{. } \label{11.5}\ ] ] back to the schrdinger picture and to the original field operators @xmath23 , we finally obtain from the steps outlined above , the master equation@xmath62 + \sum_{m , n}\left [ \frac{\gamma_{mn}+\upsilon_{mn}}{2}\mathcal{l}_{mn}\rho _ { s}(t)+\frac{\upsilon_{mn}}{2}\mathfrak{l}_{mn}\rho_{s}(t)\right ] \text { , } \label{12}\ ] ] where we have defined the damping and the diffusion matrix elements @xmath63 and @xmath64 in the forms [ 13]@xmath65 whereas the liouville operators accounting for the _ direct _ ( @xmath66 ) and _ indirect _ ( @xmath67 ) dissipative channels , are given by [ 14]@xmath68 + \left [ a_{m},\rho_{s}(t)a_{n}^{\dag}\right ] , \label{14a}\\ \mathfrak{l}_{mn}\rho_{s}(t ) & \equiv\left [ a_{n}^{\dag}\rho_{s}(t),a_{m}\right ] + \left [ a_{m}^{\dag},\rho_{s}(t)a_{n}\right ] . \label{14b}\ ] ] as mentioned in the introduction and discussed in ref . @xcite , the oscillators lose excitation to their own reservoirs through the direct dissipative channels , whereas through the indirect channels they lose excitation to all the other reservoirs but not to their own . although in ref . @xcite we have obtained the master equation for the general case of reservoirs at finite temperatures , all further analysis was carried out for reservoirs at @xmath7 k where the diffusion matrix elements @xmath64 are null . next , considering the case of reservoirs at finite temperatures , we must discuss the master equation ( [ 12 ] ) under the weak and strong coupling regimes between the network oscillators . we first remember that the weak coupling regime for a network of @xmath0 coupled oscillator is defined by the relation @xmath5 . ( however , if a specific coupling @xmath69 between two oscillators , @xmath20 and @xmath70 , fails to satisfy the relation @xmath5 , the network dynamics is necessarily described by the strong coupling regime , with some normal - mode frequencies far beyond their natural values . ) in the weak coupling regime , the interaction between the network oscillators , described by @xmath71 , could be directly introduced into the von neumann term of the master equation to a good approximation , circumventing the necessity to diagonalize the hamiltonian @xmath34 through a canonical transformation @xmath72 . this is equivalent to approximate the matrix @xmath36 by a identity matrix @xmath73 , implying that [ 15]@xmath74 where , evidently , we have also approximated the normal modes by their original natural frequencies . under the above considerations , the master equation ( [ 12 ] ) becomes @xmath75 + n\sum_{m}\left [ \frac{\gamma_{m}(\omega_{m})+\tilde{\gamma}_{m}(\omega _ { m})\bar{n}_{m}(\omega_{m})}{2}\mathcal{l}_{mm}\rho_{s}(t)\right . \nonumber\\ & + \left . \frac{\tilde{\gamma}_{m}(\omega_{m})\bar{n}_{m}(\omega_{m})}{2}\mathfrak{l}_{mm}\rho_{s}(t)\right ] \text { , } \label{16}\ ] ] where , essentially , the indirect dissipative channels disappear , establishing the additivity of the decoherence rates . the strong coupling regime means that @xmath76 , i.e. , at least one of the coupling @xmath24 between two network oscillators must be of the order of any natural frequency @xmath77 . in this case , the indirect dissipative channels become effective , inducing collective damping and diffusion effects that we must investigate . as pondered in the introduction , would the collective effects of nonadditivity of the decay rates and the emergence of dfss still survive despite the temperature effects ? it must be mentioned that markovian white noise reservoirs washes out the collective damping effects introduced by the strong coupling regime since the spectral densities are invariant over translation in frequency space , i.e. , @xmath78 , rendering the same matrix elements @xmath63 as in eq . ( [ 15a ] ) . however , markovian white noise reservoirs do not washes the collective diffusion effects . in fact , only under the _ additional _ assumption that @xmath79 for whatever normal mode @xmath60 , we recover eq . ( [ 15b ] ) for the diffusion matrix @xmath80 , erasing the collective effects completely . next we discuss the case where the whole network is under the action of a common reservoir . when all the network oscillators are coupled to a single reservoir , the master equation , derived in ref . @xcite , is similar to that in eq . ( [ 12 ] ) , replacing the damping and the diffusion matrix elements by @xmath81 where the damping rates @xmath82 and @xmath83 for the case of a single common reservoir are given by @xcite [ 18]@xmath84 as mentioned above and discussed in ref . @xcite , the master equation for the case of distinct reservoirs can be deduced from the case of a single common reservoir . the above deduction of the master equation ( [ 12 ] ) , where we started from the case of distinct reservoirs , was entirely due to its broad application in many physical systems . to demonstrate how to derive the case of distinct reservoir from that of a common one , we remember that @xmath85 gives the distribution function of the reservoir modes coupled to the @xmath20th oscillator . therefore , in the absence of overlap between the distribution functions , i.e. , @xmath86 for @xmath87 , eqs . ( [ 18 ] ) reduce to those in eqs . ( [ 11 ] ) . in this case , the occurrence of the indirect - decay channels follows entirely from the strong coupling between the oscillators , as discussed in the subsections presented above . when there is a significant overlap between the distribution functions , i.e. , @xmath88 @xmath7 for at least one @xmath67 , we get the indirect - decay channels even when the network oscillators do not interacts at all . the strength of the damping and the diffusion matrix elements being defined by the amount of the overlap , i.e. , when the overlap between the distributions @xmath89 and @xmath90 is maximum , the strengths @xmath63 and @xmath64 equals @xmath91 and @xmath92 . to analyze the dynamics of the network states for the case where the reservoirs are at finite temperatures , we consider the evolution of the ( normal ordered ) characteristic function , derived from the master equation ( [ 12 ] ) ( suitable for all cases discussed in the previous section ) as @xmath93 \chi(\{\eta _ { m}\},t)\text { , } \label{19}\ ] ] where we defined the matrix elements @xmath94 as noted in ref . @xcite , the matrix @xmath95 is an extension of the free evolution @xmath34 in eq . ( [ 4 ] ) , which takes into account the dissipative mechanisms of the network . starting with the assumption that eq . ( [ 19 ] ) admits a solution of the form @xmath96 , we obtain two differential equations , one accounting for the dynamic process , given by@xmath97 \phi(\{\eta_{m}\},t)\text { , } \label{21}\ ] ] and the other accounting for the stationary solution of the characteristic function , given by@xmath98 \varphi(\{\eta_{m}\})=0\text{. } \label{22}\ ] ] if we perform the substitution @xmath99 in the first differential equation ( [ 21 ] ) , it turns out to be exactly that appearing in ref . @xcite for the derivation of the solution of the glauber - sudarshan @xmath10-function . therefore , following the steps outlined in ref . @xcite , the solution of eq . ( [ 21 ] ) can be written as@xmath100 where we employed the diagonal form of @xmath95 following from the transformation @xmath101 . note that writing the solution ( [ 23 ] ) in a matrix form , it becomes@xmath102 \nonumber\\ & = \mathbf{\eta}\left ( 0\right ) \bullet\exp\left [ -\left ( \mathcal{h}^{d}\right ) ^{\ast}t\right ] \label{24}\ ] ] such that@xmath103 or , equivalently , @xmath104 representing a system of coupled differential equations which follows from eq . ( [ 21 ] ) under the assumption that @xmath105 , with @xmath106 @xcite . the second differential equation ( [ 22 ] ) can be solved assuming a general gaussian form@xmath107 where the elements of matrix @xmath108 are the coefficients to be determined . substituting ( [ 27 ] ) into eq.([22 ] ) and changing conveniently the labels @xmath20 and @xmath70 of the involved matrices , we verify that the differential equation ( [ 22 ] ) reduces to a matrix equation of the form @xmath109 which is explicitly written as@xmath110 with the superscript @xmath111 in eq . ( [ 28 ] ) standing for transposed . it is worth noting that for identical reservoirs , where @xmath112 and so @xmath113 , we obtain a symmetric dissipative matrix @xmath95 , i.e. , @xmath114 , making eq . ( [ 28 ] ) the well - known lyapunov equation . the solution of eq . ( [ 28 ] ) , namely the determination of @xmath108 , can be obtained by converting the matrix equation into a system of @xmath115 algebraic equations , i.e. , into a new matrix equation of the simplified form @xmath116 , with the elements of matrix @xmath117 being the @xmath115 unknown variables . to this end , it is useful to define the column vector@xmath118{ccccccccccc}\pi_{11 } & \pi_{21 } & \cdots & \pi_{n1 } & \pi_{12 } & \cdots & \pi_{n2 } & \cdots & \pi_{1n } & \cdots & \pi_{nn}\end{array } \right ) ^{\top},\label{30}\ ] ] where the first @xmath0 elements of @xmath119 correspond to the first column of matrix @xmath108 , whereas the next @xmath0 elements correspond to the second column of @xmath108 and so on . as so , the matrix equation ( [ 28 ] ) can be rewritten into the form @xcite @xmath120 \bullet\operatorname{vec}\left ( \pi\right ) = \operatorname{vec}\left ( \mathbf{\upsilon}+\mathbf{\upsilon } ^{\top}\right ) \text{,}\label{31}\ ] ] where @xmath73 is an @xmath121 identity matrix . from the mathematical properties presented in appendix a for the matrix @xmath122 $ ] , we verify that the elements of matrix @xmath108 can be written as@xmath123 finally leading to the solution of eq . ( [ 22 ] ) through eq.([27 ] ) . in fact , substituting the expression ( [ 32 ] ) into the left hand side of eq . ( [ 29 ] ) , and using the relation @xmath124 , we obtain@xmath125 = \sum_{m^{\prime},n^{\prime}}\left ( \upsilon_{m^{\prime}n^{\prime}}+\upsilon_{n^{\prime}m^{\prime}}\right ) \delta_{mn^{\prime}}\delta_{nm^{\prime}}=\upsilon_{mn}+\upsilon _ { nm},\label{33}\ ] ] which is exactly the right hand side of eq . ( [ 29 ] ) . we thus verify that the solution of the characteristic equation is of the form@xmath126 since for @xmath127 we get @xmath128 , such that @xmath129 , we end up with the solution of the characteristic function @xmath130 given in terms of its initial state . an interesting point to be noted is that the dynamics of the problem , given by @xmath131 , takes into account only the dissipative rates @xmath63 together with the free evolution hamiltonian @xmath34 in eq . ( [ 4 ] ) , leaving aside the diffusive process associated with @xmath64 . such a diffusive process , appearing in the ratio @xmath132 is however , modified , by the dissipative mechanisms . starting from two general classes of initial network states , given by mixed superpositions of coherent and number states , we next analyze the evolution of such states through the characteristic function , the glauber - sudarshan @xmath10-function , and the wigner distribution . we also compute the network density operator for the case of mixed superposition of fock states . considering that the initial network state comprehends a mixture of superpositions of coherent states like @xmath133 , the initial density operator becomes @xmath134 where @xmath135 is the probability associated with the state @xmath136 . the parameters @xmath137 ( @xmath138 ) represent a set of variables defining the probability density function @xmath139 , and @xmath140 stands for a product of coherent states , where @xmath141 represents the state associated with the @xmath20th network oscillator . in the particular case where @xmath142 , the pure state @xmath143 becomes the discrete superposition @xmath144 where we have defined @xmath145 . through the definition of the time - dependent vector elements @xmath146 and matrix elements [ 38]@xmath147 we verify , after a rather lengthy calculation , that the evolution of the initial network state ( [ 36 ] ) can be described either through the characteristic function @xmath148 -\frac{1}{2}\sum_{m , n}\eta_{m}j_{mn}\left ( t\right ) \eta_{n}^{\ast}\right\ } \text { , } \label{39}\ ] ] either by the glauber - sudarshan @xmath10-function@xmath149 \left [ \xi_{n}-k_{n}\left ( r_{\jmath};t\right ) \right ] ^{\ast}\right\ } \text { , } \label{40}\ ] ] or even by the wigner distribution function @xmath150 \left [ \xi_{n}-k_{n}\left ( r_{\jmath};t\right ) \right ] ^{\ast}\right\ } \text{. } \label{41}\ ] ] note that the difference between the glauber - sudarshan @xmath10-function and the wigner distribution comes from the time - dependent function associated with the width of the their gaussian function . consequently , the wigner function can be obtained from the glauber - sudarshan @xmath10-function through the substitution @xmath151 . whereas the glauber - sudarshan @xmath10-function diverges when there is no diffusion process such that @xmath152 ( with all the reservoirs at @xmath7 k ) , the width of the wigner function presents an additional term @xmath73 inhibiting any singularity . for the case of @xmath7 k reservoirs @xcite , the density operator of the network , to be used below , is given by@xmath153 we now assume the initial network state to be a mixture of superposition of fock states @xmath154 , where the parameter @xmath155 indicates the number of photons in the @xmath20th oscillator while the coefficient @xmath156 represents the probability amplitude associated with each state @xmath157 composing the whole superposition . the initial density operator is thus given by @xmath158 where @xmath135 is the probability associated with the state @xmath159 . since the fock state @xmath160 , of the @xmath20th oscillator , can be expanded as a superposition of coherent states of the form @xmath161 it is easy to note that the initial state ( [ 42 ] ) can be obtained by eq . ( [ 36 ] ) , identifying [ 42.5]@xmath162 such that we can use the results of the previous subsection to obtain the characteristic function , the glauber - sudarshan @xmath10-function and wigner distribution for a mixed superposition of pure fock states . alternatively , such functions may be directly computed from the initial state ( [ 42 ] ) . their expressions are presented in appendix b , where the density operator for a mixed superposition of pure fock states is also presented . to analyze the diffusion mechanism , due to the finite temperature of the reservoirs , we start by computing the wigner distribution associated with the normal - mode oscillators . as depicted in fig 2 , these oscillators , described by hamiltonian @xmath43 [ eqs . ( [ 6a ] ) and ( [ 6b ] ) ] , do not interact with each other , but they do interact with all the reservoirs . we , thus , rewrite the wigner distribution ( [ 41 ] ) in a new coordinate frame @xmath163 , obtained through the diagonalization of matrix @xmath164 . this new framework follows from the rotation @xmath165 where the unitary operation @xmath166 satisfies @xmath167 . from this matrix relation , we obtain the evolved diffusion coefficients@xmath168 as the elements of diagonal matrix @xmath169 . in this framework , the rotated wigner distribution , written as @xmath170 are composed by diagonal ( @xmath171 ) and off - diagonal ( @xmath172 ) elements defined by @xmath173 \left [ \tilde{\xi}_{m}-\tilde{k}_{m}\left ( r_{\jmath};t\right ) \right ] ^{\ast}\right\ } \text { , } \label{55}\ ] ] where @xmath174 and @xmath175 . the vector @xmath176 gives the excitation intensity of the @xmath20th normal - mode oscillator through @xmath177 . we stress that the larger or smaller values of @xmath178 depend on the network topology ( contained within the matrix elements @xmath179 ) , apart from the regime of coupling strengths between the oscillators ( contained within the matrix elements @xmath180 ) . as a particular example of this dependence , we consider a degenerate symmetric network , i.e. , a degenerate network of @xmath0 oscillators , all of them interacting with each other , where @xmath181 , @xmath182 and @xmath183 . in this case , in the strong coupling regime , we obtain the expression @xmath184 and the diffusion coefficients@xmath185{cllll}\tilde{j}_{mm}\left ( t\right ) -\tilde{j}_{mn}\left ( t\right ) & = & 1 & \text{for } & m\neq n\text{,}\\ \tilde{j}_{mm}\left ( t\right ) + \left ( n-1\right ) \tilde{j}_{mn}\left ( t\right ) & = & 1 + 2\bar{n}\left ( 1-e^{-n\gamma t}\right ) & \text{for } & m = n\text{,}\end{array } \right . \label{58}\ ] ] showing that @xmath186 can reduce or enhance the strength of the diffusion coefficients @xmath178 associated with the normal - mode oscillators . from the above td diffusion coefficients ( [ 58 ] ) we define the directional diffusion time @xmath187 displaying a tendency to a significant spread of the peak common to all elements ( the diagonal and off - diagonal ) of the wigner function associated with the @xmath20th normal - mode oscillator . since each normal - mode oscillator defines a direction in the coordinate frame @xmath163 , we are naturally led to define the mean diffusion time , associated with all the dimensions of the space , as the average value@xmath188 the average diffusion time becomes useful to compute the decoherence time of any network state when complemented with the estimated time for a significant decay of the peaks associated with the interference terms of the wigner function ( @xmath172 ) , to be defined below as @xmath189 , as an illustrative example of the above theory , below we analyze the diffusion coefficients @xmath178 for the weak and strong coupling regimes considering the case of a degenerate symmetric network . in the weak coupling regime , the matrix @xmath190 , already in a diagonal form , is defined by the elements @xmath191 \delta_{mn}$ ] , such that @xmath192 . in this regime , all the diffusion coefficients equal to@xmath193 the average diffusion time becomes @xmath194 showing , as expected , that the larger the temperature , the smaller the time required for a significant diffusion rate . in this case , the coefficients @xmath178 are mode independent and assume a common value , such that the spreads of the peaks associated with the diagonal terms of the wigner function occurs homogeneously in all directions . in the strong coupling regime , the elements of matrix @xmath164 are given by eq . ( [ 57 ] ) and the diffusion coefficients by eq . ( [ 58 ] ) , showing that only the @xmath0th normal - mode oscillator undergoes the diffusion process . for all the normal - mode oscillators but the @xmath0th , the diffusion coefficients @xmath195 are counterbalanced by the diffusion rates @xmath196 and @xmath180 coming from the direct- and indirect - decay channels , respectively . the diffusion coefficients in this regime lead to the same mean diffusion time as that in eq . ( [ 62 ] ) , showing that the average diffusion effect comes entirely from the temperatures of the reservoirs . as to be demonstrated in the next section , this interesting result is not limited to the degenerate symmetric topology . starting from eq . ( [ 60 ] ) and noting that @xmath197 ( @xmath198 ) , with the elements of matrix @xmath199 given by eq . ( [ 38c ] ) , we obtain the general expression@xmath200 applicable to whatever the network topology and the strength coupling regime between the oscillators , where@xmath201 we note that the information regarding the topology of the network is contained only in the product @xmath202 which acts as a normalized distribution function ( @xmath203 ) when computing the average value of the diffusion rate given by eq . ( [ 64 ] ) . we identify two general situations where , as in the case of a degenerate symmetric network , the diffusion mechanism becomes independent of the topology of the network . the first situation occurs @xmath204 when identical reservoirs are assumed , such that @xmath205 and , consequently , @xmath206 , making the mean diffusion@xmath207 independent of the network topology . the second situation @xmath208 arises from the assumptions of markovian white noise reservoirs and low - temperature regime , where the normal - mode frequencies satisfy the relation @xmath209 , @xmath210 being the boltzmann constant . in this case we obtain @xmath211 , such that @xmath212 , @xmath213 being computed around the average value of the normal - mode frequencies . the mean diffusion time , independent of the network topology , becomes @xmath214 both situation @xmath204 and @xmath208 were considered in order to demonstrate that the mean diffusion time for both , weak and strong coupling regimes , is the same when considering a degenerate symmetric network . for any other situation , apart from @xmath204 and @xmath208 , the average diffusion rate becomes dependent on the network topology , apart from the reservoirs temperatures . since an analysis of decoherence through the density operator of the network is hard to derive when temperature effects are present , it becomes appropriate to use the wigner distribution function of the system , instead of the density operator , to estimate the decoherence time of a family of superposition states which are particular cases of the general state given by eq . ( [ 36 ] ) . this family of states is given by @xmath215 where @xmath216 ( @xmath217 ) indicates the number of oscillators in the coherent state @xmath218 ( @xmath219 ) in the first term of the superposition and @xmath219 ( @xmath218 ) in the second term of the superposition . the remaining @xmath220 oscillators are in the coherent state @xmath221 . we again stress that we are considering a degenerate symmetric network where all the oscillators are indistinguishable . therefore , swapping the states of any two oscillators @xmath20 and @xmath70 , we obtain a state which is completely equivalent to eq . ( [ 68 ] ) . we also note that when @xmath222 and @xmath223 , we obtain from ( [ 68 ] ) the superposition @xmath224 where a schrdinger cat - like state is prepared in oscillator @xmath28 while all the remaining oscillators are prepared in the coherent states @xmath221 . we start our calculation noting that for a pure two - level state @xmath225 , whose density matrix is given by @xmath226 , the ratio of the products between the diagonal and off - diagonal elements equals unity , i.e. , @xmath227 . for an open system described by a mixed density matrix , however , this ratio decrease from unity . bearing this in mind , we rewrite the wigner function ( [ 54 ] ) , to the discrete case where @xmath228 , in a form@xmath229 with its diagonal ( @xmath230 ) and off - diagonal ( @xmath231 ) elements given by@xmath232 \left [ \tilde{\xi}_{m}-\tilde{k}_{m}^{r}\left ( t\right ) \right ] ^{\ast } \right\ } , \label{71}\ ] ] where @xmath233 and @xmath234 ( running from @xmath28 to @xmath235 ) label the product states composing the superposition ( [ 68 ] ) . now , through the diagonal and off - diagonal elements of the wigner function , we define the ratio@xmath236 \text{. } \label{72}\ ] ] which turns to be independent on the variables @xmath163 of the wigner function , as desired . moreover , for @xmath127 , such that @xmath237 and @xmath238 , we obtain @xmath239 . in analogy with the above observation concerning the ratio of the products between the diagonal and off - diagonal elements of a pure or mixed density matrix , the above defined ratio @xmath240 offers a measure of the decoherence rate which follows from the function @xmath241 \text { , } \label{73}\ ] ] which equals unity for @xmath127 . the above deduction of the decay function ( [ 73 ] ) can also be developed for the general case of an initial continuous superposition state , instead of a discrete one . this subsection is devoted to demonstrate that the measure of the decoherence rate offered by eq . ( [ 73 ] ) is equivalent to the one coming from the interference terms of the density operator , which is commonly used for the case of 0@xmath242 reservoirs . in fact , for reservoirs at 0@xmath242 , where @xmath243 , it is simple to verify that eq . ( [ 73 ] ) reduces to@xmath244 \nonumber\\ & = \left\vert \frac{\left\langle \left\ { \beta_{m}^{r}\right\ } \left\vert \left\ { \beta_{m}^{s}\right\ } \right . \right\rangle } { \left\langle \left\ { k_{m}^{r}\left ( t\right ) \right\ } \left\vert \left\ { k_{m}^{s}\left ( t\right ) \right\ } \right . \right\rangle } \right\vert ^{4}\text { , } \label{73.5}\ ] ] where the coefficients @xmath245 are those coming from the interference terms of density operator ( [ 41.5 ] ) , when considering a discrete case . therefore , considering that decoherence times are usually estimated through the relation @xmath246 , for the case of reservoirs at 0@xmath242 , we obtain from eq . ( [ 73.5 ] ) the equivalent relation @xmath247 , which gives the estimative of the decoherence time through the wigner function . now we are able to define the time @xmath189 for a significant decay of the peaks associated with the interference terms of the wigner function ( @xmath172 ) . this is done , by generalizing the relation @xmath248 , for the case of reservoirs at finite temperatures , to the equality @xmath249 \text{,}\ ] ] that corresponds to measure the decay of the interference terms of the wigner function by deducting their spreadings , common to all the diagonal and off - diagonal elements , as we can see in eq.([71 ] ) . in other words , it is similar to analyze the decay of the interference terms in a frame where the diagonal terms are frozen . finally , to define a decoherence time @xmath250 , which take into account both the diffusion and decay of the interference terms , we must consider both the above defined times : the mean diffusion time @xmath251 and the decay time of the interference terms of the wigner functions @xmath189 . we thus define the relation @xmath252 where @xmath253 only becomes relevant for particular initial states whose interference terms of the wigner function are null , as occur , for example in the case @xmath254 , to the coherent state @xmath255 , or when the excitation of the components of a superposition state is significantly smaller than unity . this will become clear in the example to be analyzed below for the degenerate symmetric network . in the first case , it is well - known that a coherent state remains as such , even under a dissipative process , when considering a reservoir at @xmath7 k. however , when considering a reservoir at finite temperature , the decoherence time of a coherent state @xmath256 can be estimated through our defined eqs . ( [ 64 ] ) and ( [ 65 ] ) . the wigner function associated with the pure state ( [ 68 ] ) in the weak coupling regime , is obtained from eq . ( [ 71 ] ) with @xmath192 and @xmath257 . our defined decay function ( [ 73 ] ) thus becomes@xmath258 \text{. } \label{76}\ ] ] we estimate the decoherence time @xmath250 of the family of states ( [ 68 ] ) through the equality @xmath259 $ ] . the obtained result for the decay time and so for the decoherence time reads@xmath260 which recover the results in ref . @xcite for 0@xmath242 reservoirs ( @xmath261 ) . in eq . ( [ 77 ] ) we have disregarded the mean diffusion time @xmath262 since we assumed that the excitation @xmath263 is significantly larger than unity . note that in the case where @xmath264 ( @xmath223 ) or @xmath265 ( @xmath266 ) , given the initial entangled state @xmath267 , the decoherence time decreases as the number of network oscillators increases . for the case of the schrdinger cat - like state in eq . ( [ 69 ] ) , we obtain the result @xmath268 summarizing , the temperature effect decreases the decoherence time when the weak coupling regime is considered . from the wigner function associated with the state @xmath269 , derived from eq . ( [ 71 ] ) and using the coefficients ( [ 58 ] ) , we obtain in the strong coupling regime@xmath270 \text{. } \label{78}\ ] ] the estimated decay time @xmath189 of the interference terms of the wigner functions is established through the inequality @xmath271 \right\ } \leq\exp\left\ { -4\mathcal{d}_{n}^{-1}\left ( \tau_{int}\right ) \right\ } $ ] , such that@xmath272 showing that the interference terms of the wigner distribution decay at a fastest rate than in the weak coupling regime . for the schrdinger cat - like state , eq . ( [ 69 ] ) , we obtain exactly the result shown in eq . ( [ schr ] ) . we finally note that , considering only the usual decay of the interference terms , given by @xmath273 , the estimation of the decoherence time leads to inconsistent results which present negative values apart from singularities . for example , for the schrdinger cat - like state in eq . ( [ 69 ] ) , in the particular case @xmath254 , we obtain @xmath274 } , \ ] ] which has a singularities at @xmath275 and becomes negative for @xmath276 @xmath277 . therefore , the procedure adopted in eq . ( [ 74 ] ) to estimate the decoherence time by separating both effects of diffusion and decay of the wigner function interference terms , is in fact more sound than the cruder approach where only the interference effects present in the decay function ( [ 73 ] ) are considered . another example refers to the decoherence of a coherent state @xmath256 , where the result @xmath278 computed though the our technique , account exactly for the diffusion effect , apart from the decay rate @xmath279 , as expected . the usual procedure fails to give such an account . as mentioned above , the analysis of the emergence of dfss with the reservoirs at finite temperature is addressed in another work @xcite , where both , collective effects of damping and diffusion , are managed together with the network topology to build up desired dfss . the computation of the density operator of the network for the case of reservoirs at finite temperatures becomes a difficult task for the majority of the initial network states . therefore , similarly to our procedure to the analysis of decoherence , we next compute the entropy of the network using the wigner functions as given by@xmath280 where the factor @xmath281 was introduced to produce a null lower bound for the entropy . using the integral result @xmath282 and the wigner function given by eq . ( [ 41 ] ) , or eq . ( [ 55 ] ) , we obtain the general expression@xmath283 which is applicable to any initial network state , where@xmath284 \right\ } \text{,}\label{p}\ ] ] and @xmath285 for the case where dissipation is absent , i.e. , @xmath286 , we verify that @xmath287 and , consequently , @xmath223 . oppositely , when @xmath288 and @xmath289 , the purity loss follows from the decay of @xmath290 which reduces to the function @xmath291eq . ( [ 73 ] ) , that enters in the calculation of the decoherence time , under the conditions @xmath292 ( @xmath6 k@xmath293 , @xmath294 , and @xmath295 . as expected , the purity loss mechanism is intimately related to the decoherence one . focusing on the case when @xmath286 , the entanglement degree of a bipartite system , described by a _ pure _ density operator @xmath296 @xmath297 and @xmath298 standing for two complementary sets of network oscillators , can be computed through the reduced entropy ( concurrence)@xmath299 ^{2}=1-\operatorname{tr}_{b}\left [ \operatorname{tr}_{a}\rho_{ab}\right ] ^{2}\text{,}\ ] ] which is given , through the joint wigner function @xmath300 , as@xmath301 ^{2}\nonumber\\ & = 1-\pi^{n_{b}}\int_{-\infty}^{\infty}d^{2}\left\ { \xi_{b}\right\ } \left [ \int_{-\infty}^{\infty}d^{2}\xi_{a}w\left ( \left\ { \xi_{a}\right\ } , \left\ { \xi_{b}\right\ } , t\right ) \right ] ^{2}\text{,}\ ] ] where @xmath302 and @xmath303 refer to the numbers of oscillators composing the sets @xmath297 and @xmath298 , respectively . when the subsystems @xmath297 and @xmath298 are uncorrelated , such that @xmath304 , the wigner function is factorized as @xmath305 and , consequently@xmath306 as expected . in the present work we have analyzed the effects of temperature in a network of dissipative quantum harmonic oscillators . starting from a previous work where a general treatment of such a bosonic dissipative network was presented @xcite , in the case of reservoirs at 0@xmath242 , here we considered reservoirs at finite temperatures . through the solution obtained for the normal - ordered characteristic function , we did compute formal expressions for the glauber - sudarshan @xmath10-function , the wigner distribution function , and the density operator for whichever the initial network state . an important point to be stressed is the relevance played by the wigner function in the present context where the reservoirs are at finite temperature . in fact , it becomes hard to identify the main features associated with the dynamic of the network states through the density operator which results to be an intricate expression . through the wigner function , however , the diffusion coefficients of the normal - mode oscillators are clearly identified as well as the decay of its interference terms . we also showed how to compute the entropy and the entanglement degree through the wigner function . we demonstrated that the diffusion coefficients @xmath195 associated with the normal - mode oscillators present completely different behaviors in both weak and strong coupling regimes . in the former case , where the indirect - decay channels do not take place , the diffusion coefficients are entirely related to the dissipative processes of the oscillators to their own reservoirs . in this case the collective damping and diffusion effects are dismissible . however , in the later case , the diffusion coefficients @xmath178 are counterbalanced by the diffusion rates @xmath307 and @xmath308 coming from the direct- and indirect - decay channels , respectively . in this case , the collective damping and diffusion effects emerges from the fact that all network oscillators interact with all the reservoirs due to the strong coupling between each other . in fact , in the strong coupling regime , the individual oscillators can not account for the dynamic of the whole network , which must be described through the collective normal - mode oscillators . differently , in the weak coupling regime , the network dynamic follows directly from those of the individual oscillators . in sum , we have presented an analysis of the mechanisms for handling the diffusion coefficients @xmath178 in the strong coupling regime , by manipulating the diffusion rates @xmath196 and @xmath180 through the nature and the temperature of the reservoirs , apart from the network topology . such approach was explored in ref . @xcite to demonstrate the possibility of the emergence of dfss in a network of dissipative oscillators even with the reservoirs at finite temperatures . we have also present a technique to estimate the decoherence time of network states which separates effects of diffusion from the decay of the interference terms in the wigner distribution function . our technique overcomes the difficulties that show up with negative values and singularities arising from the usual definition of the decoherence time based only on the decay of interference terms . we have computed the decoherence time for some particular states of the network , leaving for another work @xcite the analysis of the emergence of dfss under temperature effects . the solution of an arbitrary matrix equation of the form @xmath309 ( for an unknown @xmath117 ) can be obtained through the solution of the system@xmath310 \bullet\operatorname{vec}\left ( \mathbf{x}\right ) \equiv\operatorname{vec}\left ( \mathbf{p}\right ) \text { , } \label{89}\ ] ] following from the inverse of @xmath311 $ ] , given by @xmath312 ^{-1}\bullet\operatorname{vec}\left ( \mathbf{p}\right ) \text { , } \label{90}\ ] ] where the notation @xmath313 was defined in eq.([30 ] ) . before computing the elements of the inverse matrix @xmath314 , it is useful to observe some important properties of @xmath315 : @xmath204 the eigenvalues of matrix @xmath315 , defined by @xmath316 , are obtained through the direct sum of the eigenvalues @xmath317 and @xmath318 of matrices @xmath319 and @xmath320 , such that@xmath321 @xmath208 the eigenvectors of matrix @xmath315 are obtained through the tensor product @xmath322 where @xmath323 and @xmath324 describe the eigenvector associated to the eigenvalue @xmath317 and @xmath318 . in fact , knowing the eigenvalues and eigenvectors of matrices @xmath319 and @xmath320 , we can easily verify that @xmath325 defines the desired eigenvector , associated with the eigenvalue @xmath316 , since @xmath326 \bullet\left ( \tilde{\nu}^{\left ( i\right ) } \otimes\nu^{\left ( j\right ) } \right ) & = \tilde{\nu}^{\left ( i\right ) } \otimes\left ( \mathbf{m}\bullet\nu^{\left ( j\right ) } \right ) + \left ( \mathbf{n}\bullet\tilde{\nu}^{\left ( i\right ) } \right ) \otimes \nu^{\left ( j\right ) } \nonumber\\ & = \left ( \tilde{\epsilon}_{i}+\epsilon_{j}\right ) \left ( \tilde{\nu } ^{\left ( i\right ) } \otimes\nu^{\left ( j\right ) } \right ) \nonumber\\ & = \varepsilon_{ij}\left ( \tilde{\nu}^{\left ( i\right ) } \otimes \nu^{\left ( j\right ) } \right ) \text{. } \label{93}\ ] ] we verify that the evolution of the initial state ( [ 42 ] ) can be characterized , using the same definitions ( [ 38a ] ) and ( [ 38c ] ) , through the characteristic function@xmath327 , \label{43}\ ] ] which leads to the wigner distribution function given in terms of derivatives as@xmath328 . \label{44}\ ] ] the above distribution can also be given explicitly in the form @xmath329 \nonumber\\ & \times\left ( { \displaystyle\prod\limits_{\ell } } \lambda_{\ell}(\left\ { r_{n , n^{\prime}}\right\ } , \left\ { \xi_{p}\right\ } ; t)\frac{\left [ 2\sum_{m , n}\tilde{j}_{mn}^{-1}\left ( t\right ) \theta _ { m\ell}(t)\xi_{n}^{\ast}\right ] ^{r_{\ell n}}}{r_{\ell n}!}\right ) \exp\left ( -2\sum_{m , n}\xi_{m}\tilde{j}_{mn}^{-1}\xi_{n}^{\ast}\right ) , \label{45}\ ] ] where we have defined [ 46]@xmath330 ^{r_{n , m-1}-r_{n , m}}\text{,}\label{46b}\\ \lambda_{m}(\left\ { r_{\ell,\ell^{\prime}}\right\ } , \left\ { \xi _ { p}\right\ } ; t ) & = \frac{\left [ 2\sum_{\ell,\ell^{\prime}}\tilde{j}_{\ell\ell^{\prime}}^{-1}\left ( t\right ) \xi_{\ell}\theta_{\ell^{\prime}m}^{\ast}(t)\right ] ^{y_{m}-x_{m}+q_{m}-\sum_{\ell}\left ( r_{\ell , m-1}-r_{\ell , m}\right ) } } { \left [ y_{m}-x_{m}+q_{m}-\sum_{\ell}\left ( r_{\ell , m-1}-r_{\ell , m}\right ) \right ] ! } \text { , } \label{46c}\ ] ] with as noted in section iv , we remember that the glauber - sudarshan @xmath10-function @xmath332 can be derived from the wigner distribution by replacing @xmath333 by @xmath334 . using such a @xmath10-function we obtain a compact expression of the evolved density operator associated with the initial state ( [ 42 ] ) , given by@xmath335 \left\vert i_{n}\right\rangle \left\langle j_{n}\right\vert \right\ } \exp\left ( \frac{1}{2}\sum_{m , n}\varepsilon _ { m}j_{mn}\left ( t\right ) \varepsilon_{n}^{\ast}\right ) \text{. } \label{48}\ ] ] @xmath337 ^{k_{\ell,\ell^{\prime}}}\left [ \theta_{\ell\ell^{\prime}}^{\ast}(t)\right ] ^{r_{\ell,\ell^{\prime}-1}-r_{\ell,\ell^{\prime}}}}{\left ( r_{\ell , \ell^{\prime}-1}-r_{\ell,\ell^{\prime}}\right ) ! \left ( s_{\ell,\ell ^{\prime}-1}-s_{\ell,\ell^{\prime}}-k_{\ell,\ell^{\prime}}\right ) ! k_{\ell,\ell^{\prime}}!}\right ] \delta\left ( q_{\ell}-\sum_{\ell^{\prime}}k_{\ell^{\prime},\ell}\right ) \nonumber\\ & \times\left . \delta\left [ r_{\ell , n}-\sum_{\ell^{\prime}}\left ( s_{\ell^{\prime},\ell-1}-s_{\ell^{\prime},\ell}-k_{\ell^{\prime},\ell}\right ) \right ] \delta\left [ y_{\ell}-x_{\ell}+q_{\ell}-\sum_{\ell^{\prime}}\left ( r_{\ell^{\prime},\ell-1}-r_{\ell^{\prime},\ell}\right ) \right ] \left\vert i_{\ell}\right\rangle \left\langle j_{\ell}\right\vert \right\ } \text{. } \label{49}\ ] ] for the case where all the reservoirs are at @xmath7 k , so that @xmath338 , only the terms with @xmath339 survive in the summation over @xmath340 of expression ( [ 49 ] ) . therefore , at @xmath7 k , the density operator ( [ 49 ] ) reduces to the expression @xmath341 as already presented in ref . @xcite , where we have defined , with @xmath342 , the superposition of product states @xmath343 ^{s_{\ell,\ell^{\prime}-1}-s_{\ell , \ell^{\prime}}}}{\left ( s_{\ell,\ell^{\prime}-1}-s_{\ell,\ell^{\prime}}\right ) ! } \right ) \nonumber\\ & \times\delta\left ( q_{\ell}-\sum_{\ell^{\prime}}\left ( s_{\ell^{\prime } , \ell-1}-s_{\ell^{\prime},\ell}\right ) \right ) \left\vert j_{\ell } \right\rangle \text{.}\label{52}\ ] ] p. zanardi and m. rasetti , phys . 79 * , 3306 ( 1997 ) ; d. a. lidar , i. l. chuang , and k. b. whaley , phys . lett . * 81 * , 2594 ( 1998 ) ; e. knill , r. laflamme , and l. viola , _ ibid _ * 84 * , 2525 ( 2000 ) ; a. r. bosco de magalhes and m. c. nemes , phys . a * 70 * , 053825 ( 2004 ) ; d. a. lidar and k. b. whaley , quant - ph/0301032 .

we prove a discrete jordan - brouwer - schoenflies separation theorem telling that a-sphere embedded in a-sphere defines two different connected graphs in such a way that and and such that the complementary graphs are both-balls . the graph theoretic definitions are due to evako : the unit sphere of a vertex of a graph is the graph generated by . inductively , a finite simple graph is called contractible if there is a vertex such that both its unit sphere as well as the graph generated by are contractible . inductively , still following evako , a-sphere is a finite simple graph for which every unit sphere is a-sphere and such that removing a single vertex renders the graph contractible . a-ball is a contractible graph for which each unit sphere is either a-sphere in which case is called an interior point , or is a-ball in which case is called a boundary point and such that the set of boundary point vertices generates a-sphere . these inductive definitions are based on the assumption that the empty graph is the unique-sphere and that the one - point graph is the unique-ball and that is contractible . the theorem needs the following notion of embedding : a sphere is embedded in a graph if it is a subgraph of and if any intersection with any finite set of mutually neighboring unit spheres is a sphere . a knot of co - dimension in is a-sphere embedded in a-sphere .

1506.06440

the jordan - brouwer separation theorem @xcite assures that the image of an injective continuous map @xmath19 from a @xmath0-sphere @xmath1 to a @xmath2-sphere @xmath3 divides @xmath3 into two compact connected regions @xmath4 such that @xmath20 and @xmath5 . under some regularity assumptions , the schoenflies theorem assures that @xmath21 and @xmath12 are @xmath2-balls . hypersphere embeddings belong to knot theory , the theory of embedding spheres in other spheres , and more generally to manifold embedding theory @xcite . while @xmath1 is compact and homeomorphic to the standard sphere in @xmath22 , already a @xmath23-dimensional jordan curve @xmath24 can be complicated , as artwork in @xcite or osgood s construction of a jordan curve of positive area @xcite illustrate . the topology and regularity of the spheres as well as the dimension assumptions matter : the result obviously does not hold surfaces @xmath3 of positive genus . for codimension @xmath25 knots @xmath1 in a @xmath26-sphere @xmath3 , the complement is connected but not simply connected . alexander @xcite gave the first example of a topological embedding of @xmath27 into @xmath28 for which one domain @xmath21 is simply connected while the other @xmath12 is not . with more regularity of @xmath1 , the mazur - morse - brown theorem @xcite assures that the complementary domains @xmath4 are homeomorphic to euclidean unit balls if the embedding of @xmath1 is locally flat , a case which holds if @xmath1 is a smooth submanifold of @xmath3 diffeomorphic to a sphere . in the smooth case , all dimensions except @xmath29 are settled : one does not know whether there are smooth embeddings of @xmath28 into @xmath30 such that one of the domains is a @xmath31-ball homeomorphic but not diffeomorphic to the euclidean unit ball . related to this open schoenflies problem is the open smooth poincar problem , which asks whether there are is a smooth @xmath31-sphere homeomorphic but not diffeomorphic to the standard @xmath31-sphere . if the smooth poincar conjecture turns out to be true and no exotic smooth 4-spheres exist , then also the schoenflies conjecture would hold ( a remark attributed in @xcite to friedman ) as a schoenflies counter example with an exotic @xmath31-ball would lead to an exotic @xmath31-sphere , a counter example to smooth poincar . + even in the particular case of jordan , various proof techniques are known . jordan s proof in @xcite which was unjustly discredited at first @xcite but rehabilitated in @xcite . the schoenflies theme is introduced in @xcite . brouwer @xcite proves the higher dimensional theorem using @xmath32-dimensional nets `` defined in euclidean space . his argument is similar to jordan s proof for @xmath33 using an intersection number is what we will follow here . the theorem was used by veblen @xcite to illustrate geometry he developed while writing his thesis advised by eliakim moore . the jordan curve case @xmath33 has become a test case for fully automated proof verifications . its deepness in the case @xmath33 can be measured by the fact that ' ' 4000 instructions to the computer generate the proof of the jordan curve theorem " @xcite . there are various proofs known of the jordan - brouwer theorem : it has been reduced to the brouwer fixed point theorem @xcite , proven using nonstandard analysis @xcite or dealt with using tools from complex analysis @xcite . alexander @xcite already used tools from algebraic topology and studied the cohomology of the complementary domains when dealing with embeddings of with finite cellular chains . in some sense , we follow here alexander s take on the theorem , but in the language of graph theory , language formed by a.v . evako in @xcite in the context of molecular spaces and digital topology . it is also influenced by discrete morse theory @xcite . when translating the theorem to the discrete , one has to specify what a sphere " and what an embedding " of a sphere in an other sphere is in graph theory . we also need notions of intersection numbers " of complementary spheres as well as workable notions of homotopy deformations " of spheres within an other sphere . once the definitions are in place , the proof can be done by induction with respect to the dimension @xmath2 . intersection numbers and the triviality of the fundamental group allow to show that the two components in the @xmath0-dimensional unit sphere of a vertex in @xmath3 lifts to two components in the @xmath2-dimensional case : to prove that there are two complementary components one has to verify that the intersection number of a closed curve with the @xmath0-sphere @xmath1 is even . this implies that if a curve from a point in @xmath21 to @xmath12 in the smaller dimensional case with intersection number @xmath23 is complemented to become a closed curve in @xmath3 , also the new connection from @xmath21 to @xmath12 has an odd intersection number , preventing the two regions to be the same . the discrete notions are close to the parity functions " used by jordan explained in @xcite . having established that the complement of @xmath1 has exactly two components @xmath4 , a homotopy deformation of @xmath1 to a simplex in @xmath21 or to a simplex @xmath12 will establish the schoenflies statement that @xmath4 are @xmath2-balls . we will have to describe the homotopy on the regularized simplex level to regularizes things . the jordan - schoenflies theme is here used as a test bed for definitions in graph theory . indeed , we rely on ideas from @xcite . + lets look at the jordan case , an embedding of a circle in a 2-sphere . a naive version of a discrete jordan theorem is the statement that a simple closed curve in a discrete sphere divides the complement into two regions " . as illustrated in figure ( [ octahedron ] ) , this only holds with a grain of salt . take the octahedron graph @xmath3 and a closed hamiltonian path which visits all vertices , leaving no complement . it is even possible for any @xmath34 to construct a discrete sphere @xmath3 and a curve @xmath1 in @xmath3 such that the complement has @xmath35 components as the curve can bubble off regions by coming close to itself without intersecting itself but still dividing up a disk from the rest of the sphere . as usual with failures of discrete versions of continuum results , the culprit is the definition , in this case , it is the definition of an embedding " . the example of a hamiltonian path is more like a discrete peano curve in the continuum as it visits all vertices but without hitting all directions or area forms of the plane . we need to make sure that also the closure of the embedding is the same curve . in the graph theoretical case , we ask that the graph generated in @xmath3 by the vertex set of @xmath1 is still a sphere . for curves in a 2-sphere for example we have to ask that the embedded curve features no triples of vertices forming a triangle in @xmath3 . there is an other reinterpretation to make the theorem true for simple closed curves as we will see in the proof : there is a regularized picture on the simplex level , where two complementary domains always exist . we need the simplex regularization because embeddings do not play well with deformations : when making homotopy deformation steps , we in general lose the property of having an embedded sphere . already in discrete planar geometry , where we work in the flat @xmath25-dimensional hexagonal lattice @xcite one has to invoke rather subtle definitions to get conditions which make things work . discrete topological properties very much depend on the definitions used . it would be possible to build a homotopy deformation process which honors the embeddability , but it could be complicated . the construction of a graph product @xcite provided us with an elegant resolution of the problem : we can watch the deformation of an enhanced embedding " @xmath36 in @xmath37 , where @xmath36 is the graph obtained from @xmath1 by taking all the complete subgraphs of @xmath1 as vertices and connecting two of them if one is a subgraph of the other . it turns out that even if @xmath1 is only a subgraph of @xmath3 , the graph @xmath36 is an embedding of @xmath37 . this holds in particular if @xmath1 is a hamiltonian path in @xmath3 , the closure of @xmath36 , the graph generated by the vertex set of @xmath36 remains geometric in @xmath37 . the cartesian product @xcite allows also to look at homotopy groups geometrically : a deformation of a curve in a @xmath25-sphere @xmath3 for example is now described as a geometric surface in the @xmath26-dimensional solid cylinder @xmath38 , where @xmath39 is the @xmath23-dimensional line graph with @xmath40 vertices and @xmath41 edges . this is now close to the definition of homotopy of a curve using a function @xmath42 in two variables so that @xmath43 is the first curve and @xmath44 the second . + this paper is not the first take on a discrete jordan theorem as various translations of the jordan theorem have been constructed to the discrete . they are all different from what we do here : @xcite uses notions of discrete geometry , @xcite looks at jordan surfaces in the geometry of digital spaces , @xcite proves a jordan - brouwer result in the discrete lattice @xmath45 , @xcite extends a result of steinhaus on a @xmath46 checkerboard @xmath3 , a minimal king - path @xmath1 connecting two not rook - adjacent elements of the boundary divides @xmath3 into two components . a variant of this on a hexagonal board @xcite uses such a result to prove the brouwer fixed point theorem in two dimensions . @xcite deals more generally with graphs which can have multiple connections . the result essentially establishes what we do in the special case @xmath33 . @xcite looks at two theorems in @xmath47 or @xmath48 , where @xmath49 is the usual cartesian product and @xmath50 the tight product . in both cases , the complement of @xmath51 has exactly two path components . the paper @xcite deals with graphs , for which every unit sphere @xmath10 is a hamiltonian graph . also here , a closed simple path @xmath51 in a connected planar graph @xmath3 divides the complement into exactly two components . @xcite deals with graphs which can have multiple connections . on @xmath25-spheres , it resembles the jordan case @xmath33 covered here as a special case . + the main theorems given here could readily be derived from the continuum by building a smooth manifold from a graph , and then use the jordan - brouwer - schoenflies rsp . mazur - morse - brown theorem . the approach however is different as no continuum is involved : all definitions and steps are combinatorial and self - contained and could be accepted by a mathematician avoiding axioms invoking infinity . sometimes , constructabily can be a goal @xcite . new is that we can prove jordan - brouwer - schoenflies entirely within graph theory using a general inductive graph theoretical notion of sphere " @xcite . papers of jordan , brouwer or alexander show that the proofs in the continuum often deal with a combinatorial part only and then use an approximation argument to get the general case . as the alexander horned sphere , the open schoenfliess conjecture or questions about triangulations related to the hauptvermutung show , the approximation part can be difficult within topology and we do nt go into it . our proof remains discrete but essentially follows the arguments from the continuum by defining intersection numbers and use induction with respect to dimension . the induction proof is possible because of the recursive definition of spheres and seems not have been used in the continuum , nor in discrete geometry or graph theory . but what really makes the theorem go , is to watch the story on the simplex level , where geometric graphs @xmath3 remain geometric and where sub - spheres @xmath1 of @xmath3 can be watched as embedded spheres @xmath36 in @xmath37 . the step @xmath52 can be used in the theory of triangulations because it has a regularizing effect " . given a triangulation @xmath3 described as a graph , the new triangulation @xmath37 has nicer properties like that the unit sphere of a point is now a graph theoretically defined sphere . a subset @xmath53 of the vertex set @xmath54 of a graph @xmath3 * generates * a subgraph @xmath55 of @xmath3 , where @xmath56 is defined as the subset @xmath57 of the edge set of @xmath58 . the * unit sphere * @xmath10 of a vertex @xmath7 in a graph is the subgraph of @xmath3 generated by all neighboring vertices of @xmath7 . the * unit ball * @xmath59 is the subgraph of @xmath3 generated by the union of @xmath60 and the vertex set of the unit sphere @xmath10 . if @xmath1 is a subgraph of @xmath3 , one can think of the graph generated by @xmath1 within @xmath3 as a closure " of @xmath1 within @xmath3 . it is in general larger than @xmath1 . for example , the closure of the line graph @xmath61 within the complete graph @xmath3 with vertex set @xmath62 is equal to @xmath3 . starting with the assumption that the one point graph @xmath15 is contractible , recursively define a finite simple graph @xmath63 to be * contractible * if it contains a vertex @xmath7 such that its unit sphere @xmath10 as well as the graph generated by @xmath11 are both contractible . a complete subgraph @xmath64 of @xmath3 will also be denoted * @xmath17-dimensional simplex*. if @xmath65 is the set of @xmath17-dimensional simplices in @xmath3 , then the * euler characteristic * of @xmath3 is defined as @xmath66 . * examples*. + * 1 ) * the euler characteristic of a contractible graph is always @xmath23 as removing one vertex does not change it . one can use that @xmath67 and use inductively the assumption that unit balls as well as the spheres @xmath10 in a reduction are both contractible . + * 2 ) * also by induction , using that a unit sphere of a @xmath2-sphere is @xmath0-sphere , one verified that @xmath68 for a @xmath2-sphere . this holds also in the case @xmath69 as the euler characteristic of the empty graph is @xmath16 . the euler characteristic of an octahedron for example is @xmath70 as there are @xmath71 vertices , 12 edges and 8 triangles . the cube graph @xmath3 is not a sphere as the unit sphere at each vertex is @xmath72 . its euler characteristic is @xmath73 . @xmath3 is a sphere with @xmath71 holes punched in , leaving only a @xmath23-dimensional skeleton . a @xmath25-dimensional cube can be constructed as the boundary @xmath13 of the solid cube @xmath74 defined in @xcite . removing a vertex @xmath7 from @xmath3 for which @xmath10 is contractible is called a * homotopy reduction step*. the inverse operation of performing a suspension over a contractible subgraph @xmath1 of @xmath3 by adding a new vertex @xmath7 and connecting @xmath7 to all the vertices in @xmath1 is called a * homotopy extension step*. a finite composition of reduction or extension steps is called a * homotopy deformation * of the graph @xmath3 . * remarks . * + * 1 ) * since simple homotopy steps removing or adding vertices with contractible @xmath10 do not change the euler characteristic , it is a function on the homotopy classes @xcite . if we add a vertex for which @xmath10 is not contractible , we add a vertex with index @xmath75 which is a poincar - hopf index @xcite . given a function on the vertex set giving an ordering on the build up of the graph one gets the poincar - hopf theorem . + * 2 ) * examples like the bing house or the dunce hat show that homotopic to a one - point graph @xmath15 is not equivalent to contractible : some graphs might have to be expanded first before being contractible . this is relevant in lusternik - schnirelmann category @xcite . + * 3 ) * the discrete notion of homotopy builds an equivalence relation on graphs in the same way homotopy does in the continuum . the problem of classifying homotopy types can not be refined as one can ask how many types there are on graphs with @xmath40 vertices . + * 4 ) * the discrete deformation steps were put forward by whitehead @xcite in the context of cellular complexes . the graph version is due to @xcite and was simplified in @xcite . + * 5 ) * the definition of @xmath2-spheres and @xmath2-balls is inductive . introduced in @xcite we were puzzled then why this natural setup has not appeared before . but it actually has , in the context of digital topology @xcite going back to @xcite and we should call such spheres evako spheres . alexander evako is a name shortcut for alexander ivashchenko who also introduced homotopy to graph theory and also as i only learned now while reviewing his work found in @xcite a similar higher dimensional gauss - bonnet - chern theorem @xcite in graph theory . the induction starts with the assumption that the empty graph is the only @xmath14-sphere and that the graph @xmath15 is the only @xmath16-ball . a graph is called a * d - sphere * , if all its unit spheres @xmath10 are @xmath0 spheres and if there exists a vertex @xmath76 such that the graph generated by @xmath11 is contractible . a contractible graph @xmath3 is called a * @xmath2-ball * , if one can partition its vertex set @xmath77 into two sets @xmath78 is a @xmath0 sphere @xmath79 and @xmath80 is a @xmath0-ball @xmath79 such that @xmath81 generates a @xmath0-sphere called @xmath82 , the * boundary of @xmath3*. by induction , @xmath83 if @xmath3 is a @xmath2-sphere and @xmath84 if @xmath3 is a @xmath2-ball . + * examples . * + * 1 ) * the boundary sphere of the @xmath16-ball @xmath15 is the @xmath14 sphere @xmath85 , the empty graph . + * 2 ) * the boundary sphere of the line graph @xmath39 with @xmath86 vertices , is the @xmath16-sphere @xmath87 . line graphs are @xmath23-balls . + * 3 ) * the boundary sphere of the wheel graph @xmath88 with @xmath89 is the circular graph @xmath90 . the wheel graph is an example of a @xmath25-ball and @xmath90 is an example a @xmath23-sphere . + * 4 ) * the boundary sphere of the @xmath26-ball obtained by making a suspension of a point with the octahedron is the octahedron itself . + * 5 ) * we defined in @xcite platonic spheres as @xmath2-spheres for which all unit spheres are platonic @xmath0-spheres . this definition has been given already by evako . the discrete gauss - bonnet - chern theorem @xcite easily allows a classification : all @xmath23-dimensional spheres @xmath91 are platonic for @xmath92 , the octahedron and icosahedron are the two platonic @xmath25-spheres , the sixteen and six - hundred cells are the platonic @xmath26-spheres . as we only now realize while looking over the work of evako , we noticed that the gauss - bonnet theorem @xcite appears in @xcite . the @xmath2-cross polytop @xmath93 obtained by repeating suspension operations from the @xmath16-sphere @xmath87 is the unique platonic @xmath2-sphere for @xmath94 . the dimension of a graph is inductively defined as @xmath95 , where @xmath96 is the cardinality of the vertex set . the induction foundation is that the empty graph @xmath85 has dimension @xmath16 . the dimension of a finite simple graph is a rational number . * remarks . * + * 1 ) * this inductive dimension for graphs has appeared first in @xcite . it is motivated by the menger - uryson dimension in the continuum but it is different because with respect to the metric on a graph , the menger - uryson dimension is @xmath16 . + * 2 ) * much of graph theory literature ignores the whitney simplex structure and treat graphs as one dimensional simplicial complexes . the inductive dimension behaves very much like the hausdorff dimension in the continuum , the product @xcite is super additive @xmath97 like hausdorff dimension of sets in euclidean space . + * 3 ) * there are related notions of dimension like @xcite , who look at the largest dimension of a complete graph and then extend the dimension using the usual cartesian product . this is not equivalent to the dimension given above . + * examples . * + * 1 ) * the complete graph @xmath98 has dimension @xmath99 . * 2 ) * the dimension of the house graph obtained by gluing @xmath100 to @xmath101 along an edge is @xmath102 : there are two unit spheres of dimension @xmath16 which are the base points , two unit spheres of dimension @xmath103 corresponding to the two lower roof points and one unit sphere of dimension @xmath23 which is the tip of the roof . + * 3 ) * the expectation @xmath104 of dimension on erdoes - renyi probability spaces @xmath105 of all subgraphs in @xmath98 for which edges are turned on with probability @xmath32 can be computed explicitly . it is an explicit polynomial in @xmath32 given by @xmath106 @xcite . a finite simple graph @xmath8 is called a * geometric graph * of dimension @xmath2 if every unit sphere @xmath10 is a @xmath0-sphere . a finite simple graph @xmath3 is a * geometric graph with boundary * if every unit sphere @xmath10 is either a @xmath0 sphere or a @xmath0-ball . the subset of the vertex set @xmath77 , in which @xmath10 is a ball generates the * boundary graph * of @xmath3 . we denote it by @xmath82 and assume it to be geometric of dimension @xmath0 . * examples . * + * 1 ) * by definition , @xmath2-balls are geometric graphs of dimension @xmath2 and @xmath2-spheres are geometric graphs of dimension @xmath2 . + * 2 ) * for every smooth @xmath2-manifold one can look at triangulations which are geometric @xmath2-graphs . the class of triangulations is much larger . + * remarks . * + * 1 ) * geometric graphs play the role of manifolds . by embedding each discrete unit ball @xmath59 in an euclidean space and patching these charts together one can from every geometric graph @xmath3 generate a smooth compact manifold @xmath107 . similarly , if @xmath3 is a geometric graph with boundary , one can fill it up " to generate from it a compact manifold with boundary . + * 2 ) * the just mentioned obvious functor from geometric graphs to manifolds is analogue to the construction of manifolds from simplicial complexes . we do nt want to use this functor for proofs and remain in the category of graphs . one reason is that many computer algebra systems have the category of graphs built in as a fundamental data structure . an other reason is that we want to explore notions in graph theory and stay combinatorial . + * 3 ) * graph theory avoids also the rather difficult notion of triangularization . many triangularizations are not geometric . in topology , one would for example consider the tetrahedron graph @xmath108 as a triangulation of the @xmath25-sphere . but @xmath108 is not a sphere because unit spheres are @xmath101 which are not spheres etc . and @xmath108 is also not a ball . while it is contractible , it coincides with its boundary as it does not have interior points . free after euclid one could say that @xmath109 is a * @xmath2-dimensional point * , as it has no @xmath2-dimensional parts . + * 4 ) * every graph defines a simplicial complex , which is sometimes called the whitney complex , but graphs are a different category than simplicial complexes . algebraically , @xmath110 is a simplicial complex which is not a graph as it does not contain the triangle simplex . the graph completion @xmath101 described by @xmath111 however , is a graph . a geometric graph of dimension @xmath2 is called * orientable * if one can assign a permutation to each of its @xmath2-dimensional simplices in such a way that one has compatibility of the induced permutations on the intersections of neighboring simplices . for an orientable graph , there is a constant non - zero @xmath2-form @xmath112 , called * volume form*. it satisfies @xmath113 for the exterior derivative @xmath2 but which can not be written as @xmath114 . * remarks . * + * 1 ) * a connected orientable @xmath2-dimensional geometric graph has a @xmath23-dimensional cohomology group @xmath115 . this is a special case of poincar duality , assuring an isomorphism of @xmath116 with @xmath117 , which holds for all geometric graphs . + * 2 ) * for geometric graphs , an orientation induces an orientation on the boundary . stokes theorem for geometric graphs with boundary is @xmath118 @xcite , as it is the definition on each simplex . + * examples . * + * 1 ) * all @xmath2-spheres with @xmath119 are examples of orientable graphs . + * 2 ) * if a @xmath2-sphere @xmath3 has the property that antipodal points have distance at least 4 , then the antipodal identification map @xmath120 factors out a geometric graph @xmath121 , we get a discrete projective space @xmath122 . for even dimensions @xmath2 , this geometric graph is not orientable . + * 3 ) * the cylinder @xmath123 , with @xmath124 is orientable . one can get a sphere , a projective plane , a klein bottle or a torus from identifications of the boundary of @xmath125 in the same way as in the continuum . for example , the graph @xmath126 is obtained by taking the 25 polynomial monoid entries of @xmath127 as vertices and connecting two if one divides the other . + the following definition of the graph product has been given in @xcite : a graph @xmath8 with vertex set @xmath128 defines a polynomial @xmath129 , where @xmath130 with @xmath131 represents a complete subgraph of @xmath3 . the polynomial defines the graph @xmath37 with vertex set @xmath132 and edge set @xmath133 , where @xmath134 means @xmath7 divides @xmath135 , geometrically meaning that @xmath7 is a sub - simplex of @xmath135 . given two graphs @xmath136 , define its graph product @xmath137 . the graph @xmath138 is called the * enhanced graph * obtained from @xmath3 . * examples . * + * 1 ) * for @xmath139 we have @xmath140 and @xmath141 . + * 2 ) * for @xmath142 , we have @xmath143 and @xmath144 . + * 3 ) * for a graph without triangles , @xmath37 is homeomorphic to @xmath3 in the classical sense . + * remarks . * + * 1 ) * the graph @xmath37 has as vertices the complete subgraphs of @xmath3 . two simplices are connected if and only one is contained in the other . if @xmath3 is geometric , then @xmath37 is geometric . for example , if @xmath3 is the octahedron graph with @xmath145 vertices , @xmath146 edges and @xmath147 triangles , then @xmath37 is the graph belonging to the catalan solid with @xmath148 vertices and which has triangular faces . also @xmath37 is a @xmath25-sphere . + * 2 ) * in full generality , the graph @xmath37 is homotopic to @xmath3 and has therefore the same cohomology . the unit balls of @xmath37 form a weak ech cover in the sense that the nerve graph of the cover is the old graph @xmath3 and two elements in the cover are linked , if their intersection is a @xmath149 dimensional graph . to get from @xmath37 to @xmath3 , successively shrink each unit ball of original vertices analogue to a vietoris - begle theorem . if @xmath3 is geometric , it is possible to modify the cover to have it homeomorphic in the sense of @xcite so that if @xmath3 is geometric then @xmath3 and @xmath37 are homeomorphic . as the dimension of @xmath37 can be slightly larger in general , the property that @xmath37 and @xmath3 are homeomorphic for all general finite simple graphs does not hold . + * examples . * + * 1 ) * a hamiltonian path @xmath1 in the icosahedron @xmath3 , a @xmath25-sphere , does not leave any room for complementary domains . however , the graph @xmath36 in @xmath37 divides @xmath37 into two regions . the graph @xmath37 by the way is the disdyakis triacontahedron , a catalan solid with @xmath150 vertices . + * 2 ) * let @xmath3 be the octahedron , a @xmath25-sphere with @xmath71 vertices . assume @xmath151 are the vertices of the equator sphere @xmath100 and that @xmath152 are the north and south pole . define the finite simple curve @xmath153 of length @xmath154 . it is a circle @xmath1 in @xmath3 , but it is not embedded . in this case , the complement of @xmath1 has only one region @xmath155 . the jordan - brouwer theorem is false . but its only in this picture . if we look at the embedding of @xmath36 in @xmath37 , then this is an embedding which divides the catalan solid @xmath37 into two regions @xmath156 . a graph @xmath157 is a subgraph of @xmath8 if @xmath158 and @xmath159 . let @xmath160 be geometric graphs . a graph @xmath1 is * embedded * in an other graph @xmath3 if @xmath1 is a subgraph such that for any collection of unit spheres @xmath161 in @xmath3 , where @xmath162 is in the vertex set of some @xmath163 , the intersection @xmath164 is a sphere . * examples . * + * 1 ) * a @xmath16-sphere @xmath1 is embedded in a geometric graph if the two vertices are not adjacent . a @xmath23-sphere @xmath1 is embedded if two vertices of @xmath1 are connected in @xmath3 if and only if they are connected in @xmath1 . + * 2 ) * a graph @xmath165 can only be immersed naturally in @xmath90 if @xmath40 divides @xmath17 and @xmath89 . it is a curve winding @xmath166 times around @xmath90 . for example , @xmath167 can be immersed in @xmath168 and described by the homomorphism alebraically described by @xmath169 if @xmath90 is identified with the additive group @xmath170 . in this algebraic setting dealing with the fundamental group it is better to look at the graph homomorphism rather than the physical image of the homomorphism . + * 3 ) * if @xmath1 is embedded in @xmath3 , then also @xmath36 is an embedding of @xmath37 . but @xmath36 is an embedding in @xmath37 even if @xmath1 is only a subgraph of @xmath3 . see figure ( [ octahedron ] ) . + the following definition places the sphere embedding problem into the larger context of knot theory : a * knot of co - dimension @xmath17 * is an embedding of a @xmath18-sphere @xmath1 in a @xmath2-sphere @xmath3 . * remarks . * + * 1 ) * a knot can be called * trivial * if it is homeomorphic to the @xmath18-cross polytop embedded in the @xmath2-cross polytop in the sense of @xcite . + * 2 ) * as we do nt yet know whether there are graphs homeomorphic to spheres which are not spheres or whether there are graphs homeomorphic to balls which are not balls . + * 3 ) * the jordan - brouwer - schoenflies theorem can not be stated in the form that a @xmath0-sphere in a @xmath2 sphere is trivial . a * closed curve * in a graph is a sequence of vertices @xmath162 with @xmath171 and @xmath172 . a * simple curve * in the graph is the image of an injective graph homomorphism @xmath173 , where @xmath39 is the line graph . a * simple closed curve * is the image of an injective homomorphism @xmath174 with @xmath175 . it is an embedding of the circle if the image generates a circle . in general , a simple closed curve is not an embedding of a circular graph . * example . * + * 1 ) * a hamiltonian path is a simple closed curve in a graph @xmath3 which visits all vertices exactly once . such a path is not an embedding if @xmath3 has dimension larger than @xmath23 as illustrated in figure ( [ octahedron ] ) . + * 2 ) * while we mainly deal with geometric graphs , graphs for which all unit spheres are spheres , the notion of a simple closed curve or an embedding can be generalized for any pair of finite simple graphs @xmath160 : if @xmath1 is a subgraph of @xmath3 , then there is an injective graph homomorphism from @xmath1 to @xmath3 . if the intersection of an intersection of finitely many neighboring unit spheres with @xmath1 is a sphere , we speak of an embedding . an embedding of a graph @xmath1 in @xmath3 * separates * @xmath3 into two graphs @xmath4 if @xmath176 and @xmath177 are disjoint nonempty graphs . the two graphs @xmath4 are called * complementary subgraphs * of the embedding @xmath1 in @xmath3 . * examples . * + * 1 ) * the empty graph separates any two connectivity components of a graph . + * 2 ) * by definition , if a graph @xmath3 is @xmath17-connected but @xmath178-disconnected , there is a graph @xmath1 consisting of @xmath17 vertices such that @xmath1 separates @xmath3 . + * 3 ) * for @xmath179 , there is no subgraph @xmath1 which separates @xmath3 . + * 4 ) * the join of the @xmath23-sphere @xmath100 with the @xmath16-sphere @xmath87 is a 2-sphere , the disdyakis dodecahedron , a catalan solid which by the way is @xmath180 , where @xmath3 is the octahedron . see figure ( [ octahedron ] ) . + the @xmath100 subgraph embedded as the equator in the octahedron @xmath3 separates @xmath3 into two wheel graphs @xmath4 . + * remarks . * + * 1 ) * a knot @xmath1 of co - dimension @xmath25 in a @xmath26-sphere @xmath3 is a closed simple curve embedded in @xmath3 . classical knots in @xmath181 can be realized in graph theory as knots in @xmath2-spheres , so that the later embedding is the same as in the discrete version . the combinatorial problem is not quite equivalent however as it allows refined questions like how many different knot types there are in a given @xmath26-sphere @xmath3 and how many topological invariants are needed in a given @xmath26-sphere to characterize any homotopy type or a knot of a given co - dimension . . + * 2 ) * an embedded curve has some smoothness " . in a @xmath25-sphere for example , it intersects every triangle in maximally 2 edges . an extreme case is a hamiltonian graph @xmath1 inside @xmath3 which by definition generates the entire graph . a hamiltonian path @xmath1 which is a subgraph of a higher dimensional graph plays the role of a space - filling peano curve in the continuum , a continuous surjective map from @xmath182 $ ] to the @xmath25-manifold @xmath107 . for a simple curve @xmath1 which is not an embedding in @xmath3 , the complement of @xmath1 can therefore be empty . a * simple homotopy deformation * of a @xmath0-sphere @xmath1 in a @xmath2-sphere @xmath3 is obtained by taking a @xmath2-simplex @xmath7 in @xmath3 which contains a non - empty set @xmath183 of @xmath0-simplices of @xmath1 and replacing these simplices with @xmath184 , the set of @xmath0-simplices in @xmath7 which are in the complement @xmath183 . * examples . * + * 1 ) * if @xmath1 is a simple curve in a @xmath25-sphere @xmath3 and @xmath7 is a triangle containing a single edge @xmath185 of @xmath1 , replace @xmath185 with the two other edges of the triangle . this stretches the curve a bit . the reverse operation produces a shortcut " between two vertices @xmath186 visited by the curve initially as @xmath187 . + * 2 ) * if @xmath1 is a @xmath25-dimensional graph , then a homotopy step is done by replacing a triangle in the tetrahedron with the 3 other triangles of a tetrahedron . + * remarks . * + * 1 ) * it is allowed to replace an entire @xmath2-simplex with the empty graph to allow a smaller dimensional sphere to be deformed to the empty graph . this is not different in the continuum , where we deform curves to a point . + * 2 ) * a homotopy step does not honor embeddings in general . however , it preserves the class of simple curves with the empty curve included . + * 3 ) * we have in the past included a second homotopy deformation which removes or adds backtracking parts @xmath188 . since is only needed if one looks at homotopy deformations of general curves , we do nt use it . homotopy groups must be dealt with using graph morphisms , rather than graphs . the backtracking deformation steps would throw us from the class of simple curves . we say that a @xmath17-sphere @xmath1 is * trivial * in a @xmath2-sphere @xmath3 if there is a sequence of simple homotopy deformations of @xmath1 which deforms @xmath1 to the empty graph . if every @xmath23-sphere is trivial in @xmath3 , then @xmath3 is called * simply connected*. if every @xmath17-sphere is trivial in @xmath3 , we say the @xmath17th homotopy class is trivial . * remarks . * + * 1 ) * the set of simple closed curves is not a group , as adding a curve to itself would cross the same point twice . similarly , the set of simple @xmath17-spheres is not a group . in order to define the fundamental group , one has to look at graph homomorphism and not at the images . this is completely analogue to the continuum , where one looks at continuous maps from @xmath189 to @xmath3 . + * 2 ) * unlike in the continuum , where the zeroth homotopy set @xmath190 is usually not provided with a group structure , but @xmath190 can has a group structure . it is defined as the commutative group of subsets of @xmath77 with the symmetric difference @xmath191 as addition , modulo the subgroup generated by sets @xmath192 . it is of course @xmath193 where @xmath194 is the number of connectivity components . + * 3 ) * the hurewicz homomorphism @xmath195 maps a subset @xmath21 of @xmath77 to a locally constant function obtained by applying the heat flow @xmath196 on the characteristic function @xmath197 which is @xmath23 on @xmath21 and @xmath16 else , playing the role of a @xmath16-current = generalized function in the continuum . + * 4 ) * also the hurewicz homomorphism @xmath198 is explicit by applying the heat flow @xmath199 on the function on edges telling how many times the curve has passed in a positive way through the edge . + * examples . * + * 1 ) * every simple curve in a @xmath25-sphere is trivial if it can be deformed to the empty graph . this general fact for @xmath2-spheres is easy to prove in the discrete setup because by definition , a @xmath2-sphere becomes contractible after removing one vertex . the contraction of this punctured sphere to a point allows a rather explicit deformation of the curve to the empty graph . + * 2 ) * the deformation works also for @xmath16-spheres . in a connected graph , any embedded @xmath16 sphere can be homotopically deformed to the empty graph . so , a graph is connected if and only if every @xmath16-sphere in @xmath3 is trivial . fix a geometric @xmath2-dimensional graph @xmath3 . let @xmath200 denote the union of all graph homomorphisms from a graph in the set @xmath201 to @xmath3 . any such homomorphism @xmath202 defines the * homomorphism graph * , for which the vertices are the union of the vertices of @xmath1 and @xmath3 and for which the edge set is the union of the edges in @xmath51 and @xmath3 together with all pairs @xmath203 . two such homomorphisms are called homotopic , if the corresponding homomorphism graphs are homotopic . the homotopy classes @xmath204 define the * fundamental group * of @xmath3 . the @xmath16-element in the group is the homotopy class of a map from the empty graph to @xmath3 . the addition of two maps @xmath205 is a map from @xmath206 obtained in the usual way by first deforming each map so that @xmath207 is a fixed vertex @xmath208 , then define @xmath209 for @xmath210 and then @xmath211 for @xmath212 . * remarks . * + * 1 ) * if one would realize the graph in an euclidean space and see it as a triangularization of a manifold , then the fundamental groups of @xmath3 and @xmath107 were the same . the groups work also in higher dimensions . as we have to cut up a sphere at the equator to build the addition in the higher homotopy groups , it would actually be better to define the addition in the enhanced picture and look at maps @xmath213 where @xmath36 is the enhanced graph of the @xmath17-sphere @xmath1 and @xmath37 the enhanced version of the graph @xmath3 . a deformation of a graph @xmath1 to a graph @xmath214 is then geometrically traced as a surface . + * 2 ) * as a single basic homotopy extension step @xmath215 to @xmath216 keeps the map in the same group element of @xmath204 , the verification that the group operation is well defined is immediate . + there are various generalized notions of geometric graphs " , mirroring the definitions from the continuum . we mention them in the next definition , as we still explore discrete versions of questions related to schoenflies problem in the continuum . the main question is whether there are discrete versions of exotic spheres , spheres which are homeomorphic to a @xmath2-sphere but for which unit spheres are not spheres . a * homology @xmath2-sphere * is a geometric graph of dimension @xmath2 which has the same homology than a @xmath2-sphere . it is a geometric graph of dimension @xmath2 with poincar polynomial @xmath217 . a * homology graph * of dimension @xmath2 is a graph for which every unit sphere is a homology sphere . a * pseudo geometric graph * of dimension @xmath2 is a graph for which every unit sphere is a finite union of @xmath0 spheres . a * discrete @xmath2-variety * is defined inductively as a graph for which every unit sphere is a @xmath0-variety with the induction assumption that a @xmath14-variety is the empty graph . * examples . * + * 1 ) * an example of a homology sphere can be obtained by triangulating the dodecahedron and doing identifications as in the continuum . a suspension of a homology sphere is an example of a homology graph . + * 2 ) * a figure eight graph is an example of a pseudo geometric graph of dimension @xmath23 . + * 3 ) * the cube graph or dodecahedron graph are examples of discrete @xmath23-varieties ; their unit spheres are the @xmath16-dimensional graphs @xmath72 which are not @xmath16-spheres but @xmath16-varieties . two @xmath17-spheres @xmath136 in a @xmath2-sphere @xmath3 are called * geometric homotopic within @xmath3 * if there is a geometric @xmath178-dimensional graph with boundary @xmath107 in the @xmath218-dimensional graph @xmath38 such that @xmath219 and @xmath220 and such that the boundary of @xmath107 is included in the boundary of @xmath38 . * remarks . * + * 1 ) * given a @xmath0-sphere @xmath1 embedded in a @xmath2-sphere @xmath3 . the deformation @xmath221 is equivalent to a homotopy deformation of the complement . + * 2 ) * the above definition can can also be done for more general @xmath17-spheres ( where @xmath17 is not necessarily @xmath149 ) by taking intersections of unit spheres with @xmath178-sphere and performing the deformation within such a sphere . + * 3 ) * any homotopy deformation of @xmath1 within @xmath3 defines a deformation of the embedding of @xmath36 in @xmath37 and can be seen as a geometric homotopy deformation , a surface in @xmath222 . we will explore this elsewhere . let @xmath112 be the polynomial in the variables @xmath223 representing vertices in @xmath3 . the function @xmath224 describes the @xmath218-dimensional space @xmath222 in which we want to build a surface . let @xmath225 be the function describing the surface @xmath1 in @xmath3 , where @xmath226 are the vertices in @xmath1 . make the deformation at @xmath7 : define @xmath227 + b [ x + g_e(s(x ) ) + g(x_1,\dots , x_n)]$ ] . for example if @xmath228 , @xmath229 . + * 4 ) * in @xcite we wondered what the role of @xmath23 in the ring describing graphs could be . it could enter in reduced cohomology which is used in the alexander duality theorem @xmath230 ( going back to @xcite ) . here @xmath231 if @xmath232 and @xmath233 if @xmath234 . in the jordan case for example , where @xmath235 and @xmath236 one gets @xmath237 and @xmath238 . when embedding a @xmath25 sphere @xmath1 in a @xmath26-sphere @xmath3 , then @xmath239 and @xmath240 as well as @xmath236 and @xmath241 . this works for any @xmath2 as @xmath242 and @xmath243 by schoenflies . in this section , we put together three results which will be essential in the proof . the first is the triviality of the fundamental group in a sphere : every embedding of a @xmath23-sphere @xmath1 in a @xmath2-sphere @xmath3 for @xmath244 is homotopic to a point . look at the sphere @xmath36 in @xmath37 . remove a vertex @xmath7 disjoint from @xmath36 . by definition of a @xmath2-sphere , this produces a @xmath2-ball . the curve @xmath36 is contained in this ball @xmath12 . we can now produce a homotopy deformations at the boundary of @xmath12 until @xmath1 is at the boundary . note that @xmath12 does not remain a ball in general during this deformation as the boundary might not generate itself but a larger set . but @xmath245 remains a ball . once the @xmath1 is at the boundary switch an make homotopy deformations of @xmath1 until @xmath1 again in the interior of @xmath12 . continuing like that , perform alternating homotopy deformations of @xmath12 and @xmath1 . because the ball @xmath12 can be deformed to a point , we can deform @xmath1 to the empty graph . we now show that if sphere @xmath1 is a subgraph of @xmath3 , then the enhanced sphere @xmath36 is embedded . it is an important point but readily follows from the definitions : if @xmath1 is a @xmath0-sphere which is a subgraph of a @xmath2-sphere @xmath3 , then @xmath36 is embedded in @xmath37 . we have to show that @xmath246 is a @xmath247-sphere for every @xmath17 and every simplex @xmath7 with vertices @xmath248 in @xmath36 . any intersections are the same whether we see @xmath1 as part of @xmath3 or whether @xmath1 is taken alone . the reason is that any of the unit spheres @xmath249 consists of simplices which either contain @xmath250 or are contained in @xmath250 . none of these simplices invoke anything from @xmath3 . so , the statement reduces to the fact that the intersection of spheres @xmath249 with @xmath250 belonging to a simplex form a sphere . but this is true by induction . for one sphere it is the definition of a sphere . if we add an additional sphere , we drill down to a unit sphere in a lower dimensional sphere . finally , we have to look at an intersection number . at appears at first that we need a transversality condition when describing spheres @xmath251 of complementary dimension @xmath252 in a @xmath2-sphere @xmath3 . while we will not require transversality , the notion helps to visualize the situation . given an embedding of a @xmath0-sphere @xmath1 in a @xmath2-sphere @xmath3 and a simple curve @xmath51 in @xmath3 . we say it * @xmath51 crosses @xmath1 transversely * if for every @xmath253 such that @xmath254 , both @xmath255 and @xmath256 are not in @xmath1 . more generally : given a @xmath2-sphere @xmath3 , let @xmath214 be an embedded @xmath17-sphere and let @xmath1 be an embedded @xmath18 sphere . we say @xmath251 * intersect transversely * if @xmath257 intersect in a @xmath16-dimensional geometric graph . * remarks * + * 1 ) * given two complementary spheres @xmath136 in a sphere @xmath3 . look at the spheres @xmath257 in @xmath37 . there is always a modification of the spheres so that they are transversal . consider for example the extreme example of two identical @xmath23-spheres @xmath136 in the equator of the octahedron @xmath3 . the graphs @xmath257 are closed curves of length @xmath258 inside the catalan solid @xmath37 . now modify the closed curves by forcing @xmath36 to visit the vertices in @xmath37 corresponding to the original @xmath31 upper triangles and @xmath15 visit the vertices in @xmath37 corresponding to the original @xmath31 lower triangles . the modified curves now intersect in @xmath31 vertices . + * 2 ) * given a @xmath0-sphere @xmath1 embedded in a @xmath2-sphere @xmath3 and given a closed curve @xmath51 which is transverse to @xmath1 let @xmath162 be the finite intersection points . we need to count these intersection points . for example , if a curve is just tangent to a sphere , we have only one intersection point even so we should count it with multiplicity @xmath25 . when doing homotopy deformations , we will have such situations most of the time . as we can not avoid losing transversality when doing deformations , it is better to assign intersection numbers in full generality , also if we have no transversality . + for the following definition , we fix an orientation of the @xmath23-sphere @xmath51 which we only require to be a simple closed curve and not an embedding and we fix also an orientation on the @xmath0-sphere @xmath1 . since both are spheres and so orientable , this is possible . if the vertex set of @xmath51 is contained in the vertex set of @xmath1 , we define the * intersection number * to be zero . otherwise , let @xmath259 be a connected time interval in @xmath36 . let @xmath260 be the vertex in @xmath261 just before hitting @xmath36 and @xmath262 the vertex just after leaving @xmath36 . the vertex @xmath263 is contained in a @xmath2-dimensional simplex generated by the edge @xmath264 in @xmath51 and a simplex @xmath265 in @xmath36 containing @xmath263 . this @xmath2-simplex has an orientation from the orientation of the edge in the curve and the simplex @xmath265 in @xmath1 . if the orientation of the simplex generated by @xmath51 and @xmath1 at the point agrees with the orientation of the @xmath2-simplex coming from @xmath37 , define the * incoming intersection number * to be @xmath23 . otherwise define it to be @xmath16 . similarly , we have an outgoing intersection number which is again either @xmath16 or @xmath23 . the sum of all the incoming and outgoing intersection numbers is called the * intersection number * of the curve . in other words , if the curve @xmath51 touches @xmath1 and bounces back after possibly staying in @xmath1 for some time then the intersection number is @xmath25 or @xmath266 , depending on the match of orientation of this touch down " . if the curve @xmath51 passes through @xmath1 , entering on and leaving on different sides , then the intersection number belonging to this crossing " is assumed to be @xmath23 . if a curve is contained in @xmath1 , then the intersection number is @xmath16 . the intersection number depends on the choice of orientations chosen on @xmath51 and @xmath1 , but these orientations only affect the sign . if the orientation of the curve or the @xmath0-sphere is changed , the sign changes , but only by an even number . two homotopic curves in a sphere @xmath3 with embedded @xmath0 sphere have the same intersection number modulo @xmath25 . just check it for a single homotopy deformation step of the curve . as these steps are local , a change of the intersection number can only happen if a curve makes a touch down at @xmath1 before the deformation and afterwards does no more intersect . this changes the intersection number by @xmath25 . the following lemma will be used in jordan - brouwer and is essentially equivalent to jordan - brouwer : [ keylemma ] the intersection number of a closed curve @xmath214 with an embedded @xmath0-sphere @xmath1 in a @xmath2-sphere @xmath3 is always even . the sphere @xmath3 is simply connected , so that every curve @xmath51 can be deformed to a vertex not in @xmath1 . for the later , the intersection number is zero . in the case of transversal intersections , the result can be illustrated geometrically . look at the intersection of the @xmath25-dimensional deformation surface of the curve with the @xmath2-dimensional deformation cylinder @xmath267 . as long as the curve never intersects @xmath1 in more than two adjacent vertices , this is a @xmath23-dimensional geometric graph , which must consist of a finite union of circular graphs . a @xmath0-sphere @xmath1 embedded in a @xmath2-sphere @xmath3 separates @xmath3 into two complementary components @xmath4 . we use induction with respect to the dimension @xmath2 of the graph @xmath3 . for @xmath268 , the graph @xmath3 is the @xmath16-sphere and @xmath1 is the empty graph @xmath269 and the complementary components @xmath4 are both @xmath15 graphs which by definition are @xmath16-balls . + to prove the theorem , we prove a stronger statement : if @xmath1 is a @xmath0-sphere which is a subgraph of @xmath1 , then @xmath36 separates @xmath37 into two components @xmath4 . + take a vertex @xmath270 . when removing it , by definition of spheres , it produces the @xmath0-ball @xmath271 and the @xmath2-ball @xmath272 . by definition of embedding , the boundary of @xmath273 is @xmath274-sphere which is a subgraph of the boundary @xmath10 of @xmath275 . let @xmath280 be the path connected component in @xmath281 containing @xmath278 and let @xmath282 be the path connected component in @xmath281 containing @xmath279 . the union of @xmath280 and @xmath282 is @xmath3 if there was a third component it would have @xmath1 in its boundary and so intersect @xmath10 , where by induction only two components exist . the graphs @xmath283 and @xmath284 are @xmath2-dimensional graphs which cover @xmath37 and have the @xmath0 sphere @xmath36 as a common boundary . + it remains to show that the two components @xmath285 are disconnected . assume they are not . then there is a path @xmath51 in @xmath37 connecting a vertex @xmath286 with a vertex @xmath287 such that @xmath51 does not pass through @xmath1 . define an other path @xmath288 from @xmath7 to @xmath135 but within @xmath10 . the path @xmath288 passes through @xmath1 and the sum of the two paths @xmath289 form a closed path whose intersection number with @xmath1 is @xmath23 . this contradicts the intersection lemma ( [ keylemma ] ) . * remarks . * + * 1 ) * as we know @xmath290 and @xmath291 we have @xmath292 so that @xmath293 . we will of course know that @xmath294 but @xmath295 is for free . + * 2 ) * the proof used only that @xmath3 is simply connected . we do nt yet know that @xmath4 are both simply connected . at this stage , there would still be some possibility that one of them is not similarly as the alexander horned sphere . + * examples . * + * 1 ) * the empty graph separates the two point graph @xmath87 without edges into two one point graphs . * 2 ) * the two point graph @xmath87 without edges separates @xmath100 into two @xmath23-balls @xmath296 . + * 3 ) * the equator @xmath100 in an octahedron @xmath297 separates @xmath297 into two wheel graphs @xmath298 . + * 4 ) * the unit circle @xmath299 of a vertex @xmath7 in an icosahedron separates it into a wheel graph @xmath300 and a @xmath25-ball with 11 vertices . + * 5 ) * the @xmath0-dimensional cross polytop embedded in a @xmath2-dimensional cross polytop separates it into two @xmath2-balls . + * 6 ) * more generally , a @xmath2-sphere @xmath1 divides the suspension @xmath301 into two balls , which are both suspensions of @xmath1 with a single point . + jordan - brouwer gives two complementary domains @xmath4 in @xmath3 . they define complementary domains @xmath156 in @xmath37 , where @xmath156 have a nonzero number of vertices . the proof goes in two steps . we first show that @xmath156 are balls : to show that @xmath302 is a @xmath2-ball , ( the case of @xmath245 is analog ) , we only have to show that @xmath302 is contractible , as the boundary is by definition the sphere @xmath1 . it is enough to verify that we can make a homotopy deformation step of @xmath1 so that the number of vertices in @xmath302 gets smaller . this can be done even if the complement of @xmath1 has no vertices left , like for example if @xmath1 would be a hamiltonian path where @xmath21 is empty . a simple homotopy deformation of @xmath1 will reduce the number of vertices in @xmath302 and since @xmath302 only has a finite number of vertices , this lead to a situation , where @xmath36 has been reduced to a unit sphere meaning that @xmath1 has reduced to a simplex verifying that @xmath302 is a @xmath2-ball . the deformation step is done by taking any vertex @xmath7 in @xmath302 which belongs to a @xmath2-simplex @xmath7 in @xmath21 . the step @xmath303 produces the step @xmath304 in @xmath37 . this finishes the verification that @xmath156 are @xmath2-balls . but now we go back to the original situation and note that @xmath21 and @xmath302 originally are homeomorphic . there is a concrete cover of @xmath302 with @xmath2-balls such that the nerve graph is @xmath21 . so , also @xmath21 is contractible . * remarks . * + * 1 ) * without the enhanced picture of the embedding @xmath36 in @xmath37 , there would be a difficulty as we would have to require the embedded graph @xmath1 to remain an embedding . the enhanced picture allows to watch the progress of the deformation . + * 2 ) * as shown in the proof , one could reformulate the result by saying that if a @xmath0-sphere @xmath1 is a subgraph of a @xmath2-sphere @xmath3 ( not necessarily an embedding ) then @xmath36 separates @xmath37 into two complementary components @xmath156 which are both @xmath2-balls .

we present hst ghrs and stis observations of five qsos that probe the prominent high - velocity cloud ( hvc ) complex c , covering% of the northern sky . based upon a single sightline measurement ( mrk 290 ) , a metallicity [ s / h]=.05.12 has been associated with complex c by wakker ( 1999a , b ) . when coupled with its inferred distance (kpc ) and line - of - sight velocity ( to kms ) , complex c appeared to represent the first direct evidence for infalling low - metallicity gas onto the milky way , which could provide the bulk of the fuel for star formation in the galaxy . we have extended the abundance analysis of complex c to encompass five sightlines . we detect absorption in three targets ( mrk 290 , mrk 817 , and mrk 279 ) ; the resulting [ / ] values range from.36 ( mrk 279 ) to.48 ( mrk 817 ) to.10 ( mrk 290 ) . our preliminary fuse analysis of the mrk 817 sightline also supports the conclusion that metallicities as high as 0.3 solar are encountered within complex c. these results complicate an interpretation of complex c as infalling low - metallicity galactic fuel . ionization corrections for and can not easily reconcile the higher apparent metallicities along the mrk 817 and mrk 279 sightlines with that seen toward mrk 290 , since h emission measures preclude the existence of sufficient . if gas along the other lines of sight has a similar pressure and temperature to that sampled toward mrk 290 , the predicted h emission measures would bemr . it may be necessary to reclassify complex c as mildly enriched galactic _ waste _ from the milky way or processed gas torn from a disrupted neighboring dwarf , as opposed to low - metallicity galactic _ fuel_. # 1to 0pt#1

astro-ph0109220

the infall of low - metallicity gas onto the milky way is a crucial component of virtually all galactic formation and evolution scenarios . the simple closed - box model of chemical evolution assumes no infall , inflow , or outflow within the given region being modeled , and is successful at explaining many of the observational characteristics of the galactic disk . however , it fails to reproduce the metallicity distribution of nearby g- and k - dwarfs , the so - called `` g - dwarf problem '' ( pagel 1994 ; flynn & morell 1997 ) , related to the well - known paucity of metal - poor g- and k - dwarfs in the solar neighborhood . closed - box models , in which star formation proceeds as some functional form dependent upon local gas density , tend to overproduce low - metallicity stars before self - enrichment affects subsequent generations of stars . conversely , infall models are attractive because galactic disk formation proceeds by gradual accretion of pristine or partially processed material . the initial reservoir of @xmath8 gas in a given volume of the disk is presumed to be negligible , and the number of first generation stars is likewise negligible . the processed ejecta of this first generation mixes with infalling low - metallicity gas , and a second , more substantial , generation of star formation begins , now with @xmath9 . this process continues , presumably unabated , to the present day , regulated in a manner that recovers the local g - dwarf metallicity distribution and avoids the overproduction of metal - poor disk stars ( , pagel 1994 ; figure 23 ) . it should be stressed that infalling fuel does not necessarily need to be pristine in order to satisfy the g - dwarf problem . under the scenario explored by tosi ( 1988 ) , for example , infalling gas metallicities can be replaced by [ s / h ] for the arguments that follow . ] [ o / h]@[email protected] yield results essentially indistinguishable from primordial metallicity gas . infalling gas metallicities [ o / h]@[email protected] can be ruled out , owing to their inability to recover the present - day radial distribution of region oxygen abundances in the galactic disk ; intermediate - metallicities @xmath11.0@xmath10[o / h]@[email protected] are marginally consistent with this same region constraint . more sophisticated models , based upon the dual infall scenario of chiappini ( 1999 ) , are currently being examined and are discussed elsewhere ( gibson et al . 2001,2002 ) . while tenable theoretically , observational proof of the existence of this infalling low - metallicity galactic fuel had been difficult to obtain prior to the work of wakker ( 1999a , b ) . larson ( 1972 ) suggested that the population of galactic high - velocity clouds ( hvcs ) was the most likely culprit , a scenario that has been explored in detail by tosi ( 1988 ) . hvcs are defined as neutral hydrogen clouds travelling at velocities incompatible with those allowed by differential galactic rotation ( , wakker , van woerden & gibson 1999c , and references therein ) . despite the ubiquity of hvcs , with a sky covering factor as great as 37% ( murphy , lockman & savage 1995 ) , our ignorance concerning their distance and metallicity makes an unambiguous identification of the galactic hvcs as infalling galactic fuel very difficult . wakker ( 1999a , b ) derive a metallicity of [ s / h][email protected]@xmath20.12 for one portion of hvc complex c , using the background seyfert mrk 290 as a probe of the intervening gas . this metallicity is @xmath12 times lower , for example , than that encountered in the magellanic stream ( [ /][email protected] ; gibson 2000 ) . this metallicity is also consistent with tosi s infalling galactic fuel model . wakker conclude that complex c is a proto - typical example of infalling low - metallicity gas responsible for fueling galactic star formation . as they discuss , the _ source of this infalling gas could be primordial gas left over from the formation of the galaxy , or from mildly - processed gas stripped from local group dwarfs , and remains undetermined . more metal - rich hvcs must have another association , such as tidal disruption of the magellanic clouds ( lu 1998 ; gibson 2000 ) or galactic fountains ( richter 1999 ) ( galactic waste ) and are unlikely candidates for galactic fuel . because of the implications for galactic formation and evolution , we have undertaken further analysis of complex c to confirm the inferred low metallicity through studies of four additional sightlines . _ in [ data ] , we present the data and analysis for five hst ghrs and stis targets , mrk 290 , 817 , 279 , 501 , and 876 . each target probes the intervening gas associated with hvc complex c. because sulfur is little affected by depletion onto dust ( savage & sembach 1996 ; figure 6 ) , the column densities provided by the @xmath131250,1253,1259 lines yield a robust measure of the metallicity of complex c and form the basis for much of our analysis . the hvc gas probed by mrk 290 appears to be of low metallicity ( @xmath00.1@xmath6 solar ) , confirming the work of wakker ( 1999a , b ) . the mrk 817 and mrk 279 sightlines , however , probe complex c gas that appears to be more metal - rich ( @[email protected]@xmath6 solar ) . conversely , the neutral nitrogen abundance along the mrk 876 sightline is @xmath140.1@xmath6 solar . we attempt to reconcile these new observations with the mrk 290 data by invoking variable ionization conditions within complex c. to first order , such a reconciliation is difficult . we summarize our results in [ summary ] and briefly discuss the implications of known published and preliminary metallicity measurements of complex c made with fuse . we then suggest future observational tests with hst , fuse , and 21 cm telescopes that may unequivocally discriminate between the low - metallicity galactic fuel and associations appropriate for higher metallicity gas . the analysis of wakker ( 1999a , b ) was restricted to a single probe of complex c , provided by the background seyfert mrk 290 . a byproduct of the stocke - shull hst ghrs and stis program on the origin and physical conditions in the low - redshift ly@xmath7 forest ( hst pid#6593@xmath157345 ) has been the serendipitous detection of several major high- and intermediate - velocity cloud complexes ( penton , stocke & shull 2000 ) . our analysis of magellanic stream absorption features is described elsewhere ( gibson 2000 ) . five of the 31 targets in our full ghrs@xmath15stis sample , including mrk 290 , intersect the foreground hvc complex c. figure [ fig : map ] shows a 2100deg@xmath16 region of complex c , constructed from the leiden - dwingeloo survey ( hartmann & burton 1997 ; hereafter , lds ) . to avoid confusion with galactic low- and intermediate - velocity gas , we show only the velocity range @xmath17kms@xmath5 in figure [ fig : map ] . the five targets discussed in the present complex c abundance analysis ( mrk 290 , 817 , 279 , 501 , and 876 ) are labeled appropriately . the complex c column densities n ( ) along each sightline span an order of magnitude , ranging from @xmath18@xmath19 ( mrk 876 sightline ) to @xmath20@xmath19 ( mrk 290 ) . we use the effelsberg data of wakker ( 2001 ) in the analysis that follows ; the relevant column densities are listed in column 6 of table [ tbl : summary ] . it is important to bear in mind that these column densities are derived using the 10@xmath21 effelsberg beam , while the absorption measurements discussed below sample gas at far higher sub - arcsecond scale resolution . it remains an open question as to how accurately emission radio measurements represent the gas that is sampled in the pencil beam represented by absorption measurements . variations as large as a factor of five in n ( ) may be encountered at the arcminute level ( wakker & schwarz 1991 ) . however , for seven targets for which both @xmath141@xmath21 and 10@xmath21@xmath112@xmath21 data exists ( table 1 of wakker 2001 ) , the ratio of n()@xmath22/n()@xmath23 is only [email protected] ( with extrema of 0.75 and 1.24 ) . this might lead one to conclude that such spatial resolution limitations are not as critical as first imagined , although this conclusion is based upon only seven data points . comparisons have also been made of n ( ) derived from 21 cm mapping with measurements of n ( ) derived from ly@xmath7 absorption along the lines of sight to early type stars in the galactic halo ( lockman , hobbs & shull 1986 ) and toward extragalactic sources ( savage 2000 ) . the ratio n()@xmath24 / n()@xmath25 is somewhat less than unity , with a variation of less than @xmath250% , in agreement with the n()@xmath22/n()@xmath23 comparison noted above . in summary , we adopt a conservative first - order estimate of the uncertainty of the adopted n ( ) for each sightline due to this mismatch in resolution of @xmath20.2dex ( @xmath250% ) . it remains possible that we may have been unfortunate enough to encounter a line of sight with a 10@xmath21-inferred n ( ) that differs from the `` pencil beam '' n ( ) by a factor of five or more . higher resolution radio mapping will better address this possibility . table [ tbl : obs ] lists the 12 hst spectra , distributed over the five targets employed in the present study ; all but one ( mrk 876 ) are ghrs g160 m data . of the ghrs data , only one spectrum ( mrk 501 ) was a pre - costar observation . the data for mrk 290 ( pi : wakker ; pid # 6590 ) and mrk 876 ( pi cot : pid # 7295 ) were extracted from the hst archive , to supplement the three targets from our go programs . details concerning spectrum preparation are described by penton ( 2000 ) . in deriving the metallicity of complex c , we have employed two different techniques , one based upon the assumption of optically - thin absorption features , and the other based upon the apparent optical depth method of sembach & savage ( 1992 ) . in the weak - line ( optically - thin ) limit , the column density n@xmath26 ( in @xmath19 ) and equivalent width of a line w@xmath27 ( in m ) are related through @xmath28 where @xmath29 ( in ) is the rest wavelength of the line and @xmath30 is its oscillator strength ( , savage & sembach 1996 ; equation 3 ) . for the lines considered in this paper , the relevant values of @xmath29 and @xmath30 are provided in columns 2 and 3 of table [ tbl : atomic ] . under the apparent optical depth method of sembach & savage ( 1992 ) , the apparent column density n@xmath31 , in the limit of a finite number of data points , is given by @xmath32\,{\rm d}v_i , \label{eq : app_col}\ ] ] where @xmath33 and @xmath34 are the observed and estimated continuum intensities at velocity @xmath35 ( equations a21 and a29 of sembach & savage ) . the statistical noise uncertainty associated with the apparent column density of equation [ eq : app_col ] is @xmath36\,{\rm d}v_i \biggr ) , \label{eq : sig_app_col}\ ] ] based upon equations a27 and a30 of sembach & savage . uncertainties quoted in this analysis correspond to those of equation [ eq : sig_app_col ] ; those associated with continuum placement have not been considered here . for the high s / n ghrs data , the uncertainty associated with the continuum placement is only an additional @xmath05% beyond that of equation [ eq : sig_app_col ] ( penton 2000 ) and is neglected here . generally , our complex c metallicity determinations are made via the singly - ionized @xmath131250,1253,1259 lines . is a good species for these measurements because : ( i ) sulfur is only mildly depleted onto dust , and ( ii ) is the dominant ionization stage in the cold and warm photoionized portions of the interstellar medium ( savage & sembach 1996 ) . it remains true that we do not directly measure [ s / h ] , the sulfur abundance with respect to that of the sun , but rather s ii / h i. our assumed solar abundances for relevant elements may be found in table [ tbl : atomic ] . in [ mrk290]-[mrk876 ] , we present our ghrs@xmath15stis spectra for each of the five background probes of hvc complex c. we provide detections and upper limits for n ( ) , for mrk 290 , 817 , and 279 , which , in combination with the column densities of table [ tbl : summary ] , these yield the ratio /. uncertainties in the conversion of this ratio to the _ true metallicity [ s / h ] will be discussed for the individual sightlines . _ wakker ( 1999a , b ) were the first to derive a metallicity for hvc complex c. using the @xmath371253 and @xmath371259 lines , they found [ s / h]@xmath38[(@xmath15)/(@xmath15)]= @[email protected] , under the assumption that , within the portion of the cloud where hydrogen is fully ionized , sulfur is 2/3 and 1/3 , and n()=0.2n ( ) . the latter was derived empirically from their wham h@xmath7 emission measure analysis of the mrk 290 sightline , for which i(h@xmath7)=187@xmath210mr . the former assumption is more uncertain , as only upper limits on @xmath376716 emission were available for this field . however , with such a small inferred fraction , their conclusions were insensitive to the assumed / ratio in the region . figure [ fig : mrk290_ghrs ] shows a 10 region of the co - added ghrs g160 m spectrum used in our analysis . this spectrum has been smoothed by the post - costar line spread function ( gilliland 1994 ) to improve the signal - to - noise ( s / n ) , although we measure absorption feature equivalent widths from the raw spectra . the galactic ( g ) and complex c ( c ) @xmath371253 and @xmath371259 lines are clearly seen , as is the saturated blend of ( g@xmath15c ) @xmath371260 . the latter , because of this saturation , provides only a loose lower limit on the silicon abundance along this sightline . the @xmath371253 line resides on the redward wing of the broad , intrinsic mrk 290 ly@xmath7 emission line , complicating continuum placement . normalizing by the local continua , we obtain the @xmath371259 , @xmath371253 , and @xmath371260 profiles shown in figure [ fig : mrk290_stack ] . the abscissa has been transformed from the ghrs heliocentric wavelength scale to velocity with respect to the local standard of rest ( lsr ) . an _ a posteriori shift of @xmath1520kms@xmath5 was applied , in order to reconcile a residual systematic offset between the galactic and hvc absorption features ( middle two panels ) and emission features ( upper panel ) . the ghrs line profiles shown in figure [ fig : mrk290_stack ] are the raw data , unsmoothed by the post - costar ghrs line spread function . the effelsberg spectrum for this sightline is shown in the upper panel of figure [ fig : mrk290_stack ] , and was kindly provided by peter kalberla and bart wakker prior to publication ( wakker 2001 ) . _ two individual complex c components are clearly seenkms@xmath5 component of @xmath371259 is missing in the ghrs spectrum . we believe we have detected it , based upon the middle panel of figure [ fig : mrk290_stack ] . ] in both and @xmath371259 but only marginally so in @xmath371253 . the and centroids for both components agree to within 4kms@xmath5 . we have not deconvolved the line profiles in order to treat components 1 and 2 separately ; instead , we have simply integrated over the full line profile in both the effelsberg and hst datasets , just as we did in our earlier analysis of the magellanic stream ( gibson 2000 ) . in table [ tbl : features ] , the first three entries list the relevant information for the and detections of complex c in the mrk 290 sightline . the line centroid ( column 3 ) and velocity range over which the line profile was integrated ( column 4 ) yield the quoted equivalent width ( column 5 ) . these are supplemented by the previously mentioned two determinations of the inferred column density , n@xmath26 ( optically - thin assumption - column 6 ) and n@xmath31 ( apparent optical depth method of sembach & savage 1992 - column 7 ) . our @xmath131253,1259 equivalent widths are indistinguishable from those of wakker ( 1999a , b ) , and consistent with the expected theoretical ratio ( 1.5 ) , suggesting that optical depth effects are minimal . using equations [ eq : app_col ] and [ eq : sig_app_col ] , the apparent column densities of the complex c and features seen in the mrk 290 ghrs spectrum are n@xmath31(@xmath371259)= ( [email protected])@xmath610@xmath39@xmath19 , n@xmath31(@xmath371253)= ( [email protected])@xmath610@xmath39@xmath19 , and n@xmath31()@xmath401.0@xmath610@xmath39@xmath19 . the -weighted average is n@xmath31()@xmath38 n()=([email protected])@xmath610@xmath39@xmath19 . with a solar abundance a@xmath41 ( column 4 of table [ tbl : atomic ] ) and n()=([email protected])@xmath610@xmath42@xmath19 ( table [ tbl : summary ] ) , we can now write the and abundances for complex c ( sampled by mrk 290 ) , as @xmath43 } & = & -1.10\pm 0.06 \nonumber \\ { [ \rm si~ii / h~i ] } & > & -1.61\pm 0.03 \nonumber\end{aligned}\ ] ] the above quoted uncertainty only represents the random statistical component ; a conservative systematic uncertainty of @xmath20.10dex is assumed for both n ( ) ( wakker 2001 ) and the ghrs continuum placement ( penton 2000 ) , for a total systematic uncertainty of @xmath20.14dex . of potentially greater concern are the ionization corrections required to derive [ s / h ] from our [ / ] determination . based on their wham data , wakker ( 1999a , b ) find n()=([email protected])n ( ) , and under the assumption that within the portion of the cloud where hydrogen is fully ionized , 0@xmath1100% of the sulfur is in the form of , n()=([email protected]@xmath20.07)n ( ) . this implies a correction that may reach [email protected] , if n ( ) = 0 , in the sense of [ s / h]=[/]@xmath10.08 . however , one must bear in mind that the wham field of view is 1@xmath44 , and the same resolution uncertainties that plague column density determinations also affect the measurements . in addition , for systems with larger fractions of , uncertainties in the fraction of present in these ionized regions become increasingly important . because we have , as yet , no emission measure information for any of our remaining sightlines , in the wham data for the mrk 876 sightline ; we return to this in [ mrk876 ] . ] we generally restrict our discussion to [ / ] , but draw attention to important implications for future emission measure work , where appropriate . as an additional way to explore the uncertainties associated with converting [ / ] to [ s / h ] , we have constructed a grid of photoionization models using the code cloudy ( ferland 1996 ) . we make the very simplified assumption that the absorbers are plane - parallel slabs of gas illuminated on one side by incident ionizing radiation dominated by galactic ob associations . we approximate this integrated spectrum with a power - law @xmath45 spectrum between 1 and 4 rydbergs , with a factor of 100 drop beyond 4 rydbergs , which includes the harder extragalactic component . we assume a level of ionizing radiation compatible with complex c being at a distance of order 10@xmath120kpc , and exposed to approximately 5% of the ionizing radiation produced in the galaxy , that is , a one - sided , normally incident photon flux ( photons @xmath19 s@xmath5 ) of @xmath46 . while this power law is a reasonable approximation of the integrated spectrum of ob associations ( sutherland & shull 2001 ) , we also consider a softer spectrum , that based upon a t=35,000k kurucz model atmosphere . such a soft ionizing spectrum has been suggested to account for emission - line ratios in the warm interstellar medium ( sembach 2000 ) . figure [ fig : sii ] summarizes the results of our calculations . the top panel , which assumes the absorber has n()=10@xmath47@xmath19 , indicates that little ionization corrections are required for high column density lines of sight such as those toward fairall 9 ( gibson 2000 ) . even for the case when the density @xmath48@xmath49 and metallicities are 0.1 of solar , [ / ] provides only a 50% overestimate of [ s / h ] . for higher densities , the correction is negligible . for much lower column densities n()=10@xmath42@xmath19 , the bottom panel shows that the ionization correction remains small for @xmath50@xmath49 . figure [ fig : mrk817_ghrs ] shows a 16 region of our ghrs g160 m co - added spectrum of mrk 817 , centered on the intervening galactic @xmath371250 feature . the spectrum was smoothed with the post - costar line spread function ( gilliland 1994 ) . as was the case for mrk 290 , galactic ( g ) and complex c ( c ) features , here at 1250 and 1253 , are clearly seen in absorption . because the @xmath371253 features coincide with the peak of the intrinsic ly@xmath7 emission line associated with mrk 817 , identification of the local continuum is more challenging than it was for mrk 290 , although we retain confidence in the third - order polynomial employed . the normalized velocity stack of relevant absorption features is shown in figure [ fig : mrk817_stack ] . we employed a shift of @xmath1520kms@xmath5 to bring the centroids of the galactic features into alignment with those of the galactic ( upper panel of figure [ fig : mrk817_stack ] ) . the complex c column density along this sightline is n()=([email protected])@xmath610@xmath42@xmath19 ( wakker 2001 ) . the equivalent widths for the two complex c absorption features , w@xmath27(1253)=27@xmath23m and w@xmath27(1250)=13@xmath23m ( see relevant entries in table [ tbl : features ] ) , are consistent with the expected theoretical ratio ( 2.0 ) , again suggesting that saturation effects are minimal , as they were for the mrk 290 sightline . using equations [ eq : app_col ] and [ eq : sig_app_col ] , we find the column density to be n()=@xmath51@xmath19 . coupling the and column densities , we infer the complex c sulfur abundance along this mrk 817 sightline to be @xmath43 } & = & -0.48\pm 0.06 , \nonumber\end{aligned}\ ] ] a factor of 4 times greater than the inferred [ / ] along the line - of - sight to mrk 290 ( [ mrk290 ] ) . again , the quoted error budget reflects internal statistical uncertainties alone . it is tempting to now explore whether our mrk 817 result , [ /][email protected]@xmath20.06 , could be reconciled with the wakker ( 1999a , b ) result for mrk 290 , [ s / h][email protected]@xmath20.12 , via ionization corrections . we assume a simple model where the gas is either completely neutral in h or completely ionized in h. in the region , s is entirely , being photoionized by diffuse ultraviolet radiation . in the region , s is divided between and , with a ratio @xmath52/. in this case , it may be shown that the additional ionized hydrogen column density n ( ) which must be invoked to reconcile [ / ] with [ s / h ] is @xmath53 - [ s / h ] } - 1 \right ] . \label{eq : hiicorrection}\ ] ] under the assumption that the mrk 290 metallicity determination ( [ s / h][email protected] ) is correct , and @xmath54 , equation [ eq : hiicorrection ] implies that a column density n()=4.1n ( ) must be present to be consistent with [ /].=0 ( 1 ) , n()=2.7 ( 5.4 ) n ( ) . ] this estimate represents an unreasonably large amount of additional ionized hydrogen , for reasons that we now explain . combined with the effelsberg data , n()@xmath55=@xmath56@xmath19 , this implies n()@xmath57@xmath19 . if we assume that the temperature and pressure of the complex c gas probed by mrk 817 and 279 are similar to that inferred for the mrk 290 sightline , and adopt the wakker ( 1999a ) standard model of constant density , neutral core , and ionized shell , we can scale the observational results provided by wham ( see [ mrk290 ] ) , to yield a first - order prediction for the h@xmath7 emission measure i(h@xmath58 along the mrk 817 sightline , @xmath59 where n()@xmath60=@xmath61@xmath19 and i(h@xmath62=187@xmath210mr . for n()@xmath63=@xmath64@xmath19 , equation [ eq : wham ] predicts an h@xmath7 emission measure of i(h@xmath58@xmath651200mr . arbitrarily setting n()=0 ( , not allowing to track , contrary to expectations ) for mrk 817 , the inferred n ( ) would be reduced to 2.7n ( ) . the predicted h@xmath7 emission measure , under this scenario , is i(h@xmath58@xmath65800mr , still a surprisingly high result . we emphasize , however , that this is a highly tentative suggestion for the magnitude of i(h@xmath58 . if we assume a range in @xmath66 and propagate our formal uncertainties as well as the uncertainties in the wakker ( 1999a , b ) results through equations [ eq : hiicorrection ] and [ eq : wham ] , this would introduce a range in i(h@xmath58 @xmath65600 - 4000mr . inclusion of the systematic uncertainties in our measurements would broaden this range still further . we estimate that the true column density along the mrk 817 sightline would have to be n()@[email protected]()@xmath55 , in order to reconcile the apparent metallicity ( [ /][email protected] ) with that seen toward mrk 290 ( [ s / h][email protected] ) . as we have discussed , one can not exclude the suggestion that n()@xmath55 underestimates the true column density by a factor of 2.5 . however , despite the fact that n ( ) toward the mrk 290 sightline was measured with additional 2@xmath21 westerbork data , one similarly can not dismiss the suggestion that its column may be overestimated by the same factor of 2.5 , since the 2@xmath21 westerbork beam is still two orders of magnitude greater than the solid angle probed by ghrs . indeed , the intermediate velocity gas in the foreground of globular clusters such as m13 , m15 and m92 show factors of ten and more variations in n ( ) . one also sees variations approaching a factor of 10 in n ( ) , and possibly n ( ) , on 10@xmath68 angular scales . at at distance @xmath69pc , this corresponds to order - of - magnitude variations on 0.005pc spatial scales [ shaw 1996 ; meyer & lauroesch 1999 ; andrews , meyer , & lauroesch 2001 ( see references therein for similar studies using binary or single stars ) ] . the latter scale is three orders of magnitude smaller than the complex c spatial scale probed by the 2@xmath21 westerbork beam for @xmath70kpc . while these metal ions are generally not the dominant ion and so may not trace faithfully , they do suggest that significant column variations at sub - arcminute scales may complicate both mrk 290 and mrk 817 analyses . further evidence of significant sub - arcsecond column density variations in the interstellar medium are provided by davis , diamond , & goss ( 1996 ) . our co - added ghrs g160 m spectrum of mrk 279 is the final sightline for which we claim a detection of . figure [ fig : mrk279_ghrs ] shows a 14 region of the spectrum employed , again centered upon the galactic @xmath371250 absorption feature . complex c is seen in @xmath371250 absorption , as identified . the detection is admittedly significant only at the 4@xmath71 level ( penton 2000 ) , but was identified in a survey unbiased by any knowledge of corresponding . it is apparent in each of the four individual sub - exposures , and we currently see no compelling reason to dismiss the line assignment , despite its lower significance . the profile along this sightline is significantly more complex than that encountered in any of the others included in the present study . wakker ( 2001 ) decompose the full profile ( upper panel of figure [ fig : mrk279_stack ] ) into eight components at low- , intermediate- , and high - velocity . two complex c components can be identified ( centered at @xmath1102kms@xmath5 and @xmath1137kms@xmath5 ) , with combined column density n()=([email protected])@xmath610@xmath42@xmath19 . we have made no attempt to deconvolve the @xmath371250 absorption feature and simply integrate both the and profiles over the full velocity range spanned by the two components . the result of the integration is a measured equivalent width of 18@xmath2 3 m , where the uncertainty is a reflection of the formal statistical error . the middle and lower panels of figure [ fig : mrk279_stack ] show the normalized ghrs spectrum in the vicinity of @xmath371253 and @xmath371250 , respectively . a velocity shift of @xmath1520kms@xmath5 was applied , in order to align the centroids of the galactic @xmath371250,1253 lines with the galactic . it should be stressed that the appropriate correction to the velocity scale was difficult to ascertain . strong ly@xmath7 absorption intrinsic to mrk 279 is seen `` redward '' of the galactic @xmath371250 absorption , spanning @xmath72kms@xmath5 ) and complicating the velocity scale determination . while @xmath1520kms@xmath5 was the applied shift , we estimate that this is uncertain at the @xmath220 kms@xmath5 level . this additional uncertainty compounds the systematic uncertainty arising from the assumption that the lower angular resolution h i map is a faithful guide to the velocity profile . we have explored the extent of this uncertainty by stepping our integration from -200 to -110 km s@xmath5 to -160 to -70 km s@xmath5 . this can introduce high excursions in the measured equivalent width as the integration begins to include galactic absorption . the lowest excursion in measured equivalent width is 16.1@xmath23 m . while this is an extermal value , to be conservative we have added this 1.9 m excursion to the formal uncertainty . complex c absorption along this sightline is only apparent in the @xmath371250 line , at @xmath73kms@xmath5 ( lower panel of figure [ fig : mrk279_stack ] ) . the characteristics of the feature ( , centroid , equivalent width , etc . ) are listed in table [ tbl : features ] . the centroid of the complex c @xmath371250 absorption is offset by @xmath020kms@xmath5 from the , but the aforementioned uncertainty in the velocity scale makes the significance of this offset unclear . unfortunately , the predicted location of the expected complex c @xmath371253 absorption coincides with strong ly@xmath7 absorption intrinsic to mrk 279 at @xmath74kms@xmath5 ( middle panel of figure [ fig : mrk279_stack ] ) . our conclusions here are necessarily based upon the single @xmath371250 line . with n()=([email protected])@xmath610@xmath42@xmath19 and an apparent column density n(@xmath371250)=([email protected])@xmath610@xmath39@xmath19 , equations [ eq : app_col ] and [ eq : sig_app_col ] yield @xmath43 } & = & -0.36\pm 0.18 . \nonumber\end{aligned}\ ] ] recall that for the mrk 290 and 817 sightlines we found [ /][email protected]@xmath20.06 and [ /][email protected]@xmath20.06 , respectively . the inferred metallicity along this mrk 279 sightline seems surprisingly high . as was found to be the case for mrk 817 , it seems untenable to appeal to ionization corrections to reconcile our derived [ /][email protected]@xmath20.18 with the [ s / h][email protected]@xmath20.12 found for mrk 290 by wakker ( 1999a , b ) , using equation [ eq : hiicorrection ] for mrk 279 , we find n()=@xmath75@xmath19 , assuming /@xmath76 0.5 . via equation [ eq : wham ] , the predicted h@xmath7 emission measure in this case would be i(h@xmath58@xmath651700mr , @xmath040% greater than that predicted for the mrk 817 sightline ( [ mrk817 ] ) , and a factor of @xmath09 times greater than that seen toward mrk 290.85 times greater than that seen toward mrk 876 ( murphy 2000 ) . ] the marginal detection of @xmath371012 in our high s / n fuse spectrum implies n()@xmath65([email protected])@xmath610@xmath39@xmath19 , which is only 14% that of the column density . this would imply n()@xmath651.4@xmath610@xmath47@xmath19 and i(h@xmath7)@xmath651300mr . even in the limit of n()=0 in the ionized regions of the cloud , an column density of n()@xmath1110@xmath47@xmath19 is needed to bring the mrk 279 metallicity into formal 1@xmath71 agreement with that inferred from mrk 290 . in this case , n ( ) is still a factor of 5 greater than that seen along the mrk 290 sightline . since the @xmath371250 absorption is real and visible in each of the four sub - exposures , we are left with two options . either our line assignment is valid and the inferred metallicity is a factor of @xmath05 times higher than that along the mrk 290 line of sight , or the 1250 feature is an intrinsic ly@xmath7 feature , and ( perhaps ) there is no metallicity discrepancy . the detection of @xmath131122,1144 and @xmath371020 with fuse lends indirect support to our line assignment . regardless , without @xmath131253,1259 , we are unable to unequivocally rule out a line misidentification for our claimed @xmath371250 hvc feature . for completeness , we include here our ghrs pre - costar g160 m spectrum of mrk 501 . this background seyfert probes a small clump of complex c , approximately 30@xmath44 from mrk 290 ( see figure [ fig : map ] ) . figure [ fig : mrk501_ghrs ] shows a 16 region of the spectrum , smoothed with the pre - costar line spread function . the identification of line - free continuum regions in figure [ fig : mrk501_ghrs ] is clearly difficult , complicated by the significant undulations seen throughout the spectrum . the latter can masquerade as broad , shallow , absorption features . the normalized spectra in the vicinity of @xmath371253 and @xmath371250 are shown in the middle and lower panels , respectively , of figure [ fig : mrk501_stack ] ; the upper panel shows the effelsberg profile along this line of sight . the two - component structure of complex c is apparent , as it was for mrk 279 ( figure [ fig : mrk279_stack ] ) . the total column density , integrating over both components , is n()=([email protected])@xmath610@xmath42@xmath19 . the continuum near @xmath371250 is particularly poorly constrained , as witnessed by the middle panel of figure [ fig : mrk501_stack ] , so we will restrict the following discussion to the _ marginally better - determined continuum near @xmath371253 . with a local continuum defined over the @xmath11400@xmath77800kms@xmath5 region surrounding @xmath371253 , the non - detection of complex c @xmath371253 absorption sets an upper limit to n(@xmath371253 ) of 2.0@xmath610@xmath39@xmath19 . this conservative upper limit sets a fairly unrestrictive upper limit to the sulfur abundance of @xmath43 } & < & -0.16\pm 0.06 . \nonumber\end{aligned}\ ] ] because of the uncertain local continuum identification , we do not wish to belabor further the mrk 501 constraint . complicating the analysis still further is the fact that there appears to be an ivc along this sightline whose properties may have contaminated our analysis ( see wakker 2001 for details ) . _ mrk 876 was observed by cot ( pid # 7295 ) , as part of their stis g140 m program designed to probe the outer disk / halo of nearby spirals . as figure [ fig : mrk876_stis ] shows , cot have been successful in detecting ly@xmath7 absorption associated with the halo of ngc 6140 . a second , serendipitous , detection of ly@xmath7 associated with a small group of galaxies at @xmath78kms@xmath5 is also seen . because the cot program was centered on relatively low - redshift ly@xmath7 , their stis g140 m wavelength range did not encompass the lines . instead , we will restrict ourselves to the constraints imposed by the detection of complex c in @xmath371199.5 and @xmath371206 . the s / n in this single stis spectrum is only @xmath08 , with a 3@xmath71 detection limit of @xmath058m near both of these features . despite the low s / n , our analysis does provide a useful complement to the fuse early release observations described by murphy ( 2000 ) . murphy ( 2000 ) found [ /][email protected]@xmath20.19 and [ @xmath15/]@[email protected]@xmath20.10 , for complex c along the mrk 876 line of sight . this apparent high iron abundance is at odds with the sulfur abundance derived toward mrk 290 , but in agreement with the higher sulfur values derived from our mrk 817 and 279 data . wakker ( 2001 ) favors a photo - ionization interpretation to the fuse mrk 876 iron results , such that n()@xmath653n ( ) and n(fe)@xmath65n ( ) , and therefore [ fe / h]@[email protected] . such ionization ratios are unusual in the ism . the non - detection of h@xmath7 ( murphy 2000 ) is also somewhat surprising ( see the related discussions of predicted emission measures in [ mrk817 ] and [ mrk279 ] ) . while the overall shape of the continuum near @xmath371199.5 and @xmath371206 is well - defined , the low s / n makes setting its `` zero point '' somewhat uncertain . regardless , normalizing by the best - fit third - order polynomial leads to the @xmath371199.5 and @xmath371206 profiles shown in the middle and lower panels , respectively , of figure [ fig : mrk876_stack ] . the upper panel shows the same two - component structure seen in previous sightlines ; the total column density is n()=([email protected])@xmath6 10@xmath42@xmath19 . the first thing to note from figure [ fig : mrk876_stack ] is that both galactic and complex c @xmath371206 absorption features are clearly saturated and blended . we can do little more than set an unrestrictive lower limit on [ / ] . integrating the line profile over the range @xmath1250@xmath79@xmath195kms@xmath5 yields the equivalent width ( column 5 ) and apparent column densities ( column 7 ) listed in the final entry of table [ tbl : features ] . these data imply @xmath80 } & > & -1.30\pm 0.04 . \nonumber\end{aligned}\ ] ] unlike sulfur , silicon is easily depleted onto dust , so an upward correction of [email protected] would not be unexpected ( savage & sembach 1996 ) . both ionization and line saturation effects act in the direction of increasing [ / ] , although it should be noted that is not expected to coincide spatially with the . the middle panel of figure [ fig : mrk876_stack ] is perhaps more intriguing . while the data are noisy , there is clear evidence for an extended `` blue '' wing to the saturated galactic @xmath371199.5 absorption feature , coinciding with the position of the complex c emission ( upper panel ) . the inferred apparent column density ( table [ tbl : features ] ) , coupled with the effelsberg n ( ) , yields a measure of the nitrogen abundance for complex c of @xmath81 } & = & -1.41\pm 0.08 . \nonumber\end{aligned}\ ] ] in the absence of any ionization correction , this inferred metallicity is surprisingly low in comparison with the previous sightlines discussed here ( [ mrk290]-[mrk501 ] ) . a grid of cloudy models ( see [ mrk290 ] ) was constructed to explore the magnitude of any potential ionization correction . for n()@xmath4010@xmath42@xmath19 and density n@[email protected]@xmath49 , we find [ /]= ( [email protected])[n / h ] . for n@[email protected]@xmath49 , [ / ] is equivalent to [ n / h ] , to within 10% . thus , for a wide range of ionization and density conditions , the ratio of neutral nitrogen to hydrogen is a very good representation of the true nitrogen metallicity . this result is also insensitive to variations in the shape of the spectrum of the incident radiation field . since nitrogen is not depleted onto dust ( savage & sembach 1996 ) and the @xmath371199.5 is not severely saturated , the current constraint upon [ / ] appears robust.1134.1 fuse analysis of murphy ( 2000 ) , who found n(@xmath371134.1)=([email protected])@xmath610@xmath39@xmath19 , in comparison with our value n(@xmath371199.5)=([email protected])@xmath610@xmath39@xmath19 . ] since we expect that [ / ] should trace [ n / h ] , we are left with three possibilities for the low [ / ] metallicity inferred along this sightline : ( i ) either there are significant and unexpected ionization corrections for higher - ion species of nitrogen , without a consequent correction for ionized hydrogen ; ( ii ) the nucleosynthetic history of complex c nitrogen is significantly different from that of its sulfur enrichment ; or ( iii ) our inferred column density is underestimated by approximately a factor of two . in the absence of proof to the contrary concerning ( iii ) , we believe that option ( ii ) is most likely , since nitrogen and sulfur have different nucleosynthetic origins . sulfur is produced in equal measure by type ia and type ii supernovae ( gibson , loewenstein & mushotzky 1997 ) , while nitrogen is primarily produced by hot - bottom burning in thermally - pulsating asymptotic giant branch ( agb ) stars ( gibson & mould 1997 ) . primary nitrogen might be a byproduct of primordial metallicity , high - mass stars , but it is not the dominant nitrogen production mechanism . if the mrk 876 sightline probes a region of complex c that , for whatever reason , is free of agb stellar `` pollution '' , a significantly lower [ n / s ] or [ n / fe ] might be expected . the murphy ( 2000 ) analysis of the fuse early release observations found [ /][email protected] , in agreement with our ( current ) favored interpretation of this low s / n stis data . we have undertaken a self - consistent ghrs and stis analysis of five different probes of hvc complex c. it has been suggested that complex c represents the best example of the predicted infalling low - metallicity gas responsible for fueling the bulk of star formation in the galaxy . such infalling gas is a necessary component of galactic chemical evolution models , in order to reconcile the paucity of observed low - metallicity stars in the solar neighborhood with those predicted upon theoretical grounds . based solely upon the mrk 290 sightline , we confirm the earlier analysis of wakker ( 1999a , b ) , who found [ s / h][email protected]@xmath20.12 and support their low - metallicity claim . however , both the mrk 817 and mrk 279 sightlines appear to probe significantly more metal - rich gas , at odds with the mrk 290 result . for mrk 817 , we found [ /][email protected]@xmath20.06 , a factor of four higher metallicity than that found toward mrk 290 , in keeping with metallicities observed in several other hvcs ( lu 1998 ; richter 1999 ; gibson 2000 ) . for mrk 279 , [ /][email protected]@xmath20.18 , a factor of five higher than that toward mrk 290 . despite the detection of @xmath131122,1144 and @xmath371020 along this line of sight , we can not rule out the possibility that our @xmath371250 feature is actually ly@xmath7 intrinsic to mrk 279 . the poorly constrained continuum near the expected mrk 501 s ii features allowed us to place only approximate upper limits to [ / ] . finally , for mrk 876 , we were able to place limits on the nitrogen abundance along this complex c sightline . our result , [ /][email protected]@xmath20.08 , suggests that scaled - solar abundance ratios are not encountered in complex c , a suggestion supported by murphy et al.s ( 2000 ) conclusion that [ /]@xmath140 . based on [ / ] from mrk 876 and [ / ] from mrk 290 and 817 , it would appear that [ /][email protected]@xmath20.4 , suggestive of mild @xmath7-element enrichment from supernovae , with only marginal nitrogen pollution from hot - bottom burning in intermediate mass stars . table [ tbl : summary ] summarizes the extant ghrs and stis data pertaining to hvc complex c. we reference all of our results to the column densities provided by the effelsberg spectra of wakker ( 2001 ) . while it is tempting to attach a @xmath20.2dex systematic uncertainty to the abundances due to the resolution mismatch , counter - examples do exist in the literature which show that significantly larger variations in column might be encountered at the arcsecond level . we emphasize that our results are limited by the 9@xmath21 beam size data used here . we will revisit our conclusions once higher spatial resolution 21 cm data become available . we attempted to reconcile the higher inferred sulfur abundances along the mrk 817 and 279 sightlines with that seen toward mrk 290 , by invoking an ionization model similar to that of wakker ( 1999a ) . the predicted h@xmath7 emission measures are unrealistically high , 1200 and 1700mr , respectively . we could also construct unrealistic ionization models in which sulfur remains in the state , even in regions of the cloud in which all of the hydrogen is ionized . this would bring both the mrk 817 and mrk 279 sightlines into @xmath01@xmath71 agreement with the lower metallicity mrk 290 data . the predicted h@xmath7 emission measures , i(h@xmath7)@xmath65900mr , are substantially lower , although still fairly high . it is also quite possible that the other sightlines have different density distributions , which introduces still more uncertainties into an estimate of the expected i(h@xmath7 ) . the remaining ( large ) uncertainties in the column density determination at arcsecond levels suggest that detailed discussion of ionization conditions is not fruitful at this stage . the mrk 290 sightline through hvc complex c does support the suggestion that this cloud may be low - metallicity galactic fuel . however , our analysis of the nearby mrk 817 and and mrk 279 sightlines does not , nor does the iron abundance found in the fuse analysis of mrk 876 ( murphy 2000 ) . on the other hand , the analysis of richter ( 2001 ) of the sightline toward pg 1259 + 593 indicates an [ o i / h i ] = @[email protected] , an abundance determination that is robust to ionization or depletion corrections . our preliminary analysis of three o i absorption profiles toward mrk 817 in the fuse band yields [ o i / h i ] @xmath84 to @xmath85 , assuming the same grevesse & noels ( 1993 ) solar oxygen abundence o / h @xmath86 as richter et al . ( 2001 ) ; adopting the revised holweger ( 2001 ) solar oxygen abundance results in a complex c oxygen abundance of [ o i / h i ] @xmath87 to @xmath88 . in other words , _ the oxygen abundance along the mrk 817 sightline is @[email protected]@xmath6 solar . if complex c has a single characteristic metallicity , a compelling case does not exist to reconcile all observations with either the low - metallicity galactic fuel scenario ( [ s / h]@[email protected] ) , or the higher - metallicity galactic waste / disrupted processed gas scenario ( with [ s / h]@[email protected] ) . _ in the absence of a compelling argument for retaining the above assumption of mono - metallicity , it is worth considering the hypothesis that the observed differences may simply reflect internal abundance variations within the hvc . a useful example of such variations is provided by the smc . first , the mean metallicity of the smc is @xmath00.2 times solar ( gibson 2000 ; table 6 ) , which is comparable to that of complex c. second , its linear extent is also comparable to that of complex c , assuming the latter is at a distance of @xmath010kpc ( wakker 1999a , b ) . our analysis of the mrk 290 , 817 , and 279 sightlines , in conjunction with that of mrk 876 by murphy ( 2000 ) , implies the presence of complex c abundance differences on the order of factors of 3@xmath14 over scales of 1@xmath12kpc , the typical impact parameter between these four probes . for comparison , both f - type supergiants ( russell & bessell 1989 ) and b - stars ( rolleston 1999 ) in the smc show abundance variations on the order of a factor of @xmath03 . in other words , a nearby system of both comparable size and metallicity to complex c shows abundance variations at the level of @xmath00.5dex , similar to the magnitude of abundance variations implied by table [ tbl : summary ] and murphy ( 2000 ) . this argument does not prove that abundance variations in complex c exist , but is simply provided to show that they can not be excluded at this point , and indeed may not be unexpected . between latitudes @xmath1515@xmath10@xmath89@xmath10@xmath1530 , it has long been recognised ( e.g. davies 1972 ; verschuur 1975 , and references therein ) that complex c connects smoothly ( in terms of kinematics ) to what is now dubbed the outer arm complex ( adopting the wakker & van woerden 1991 nomenclature ) of the milky way ( in the galaxy s first quadrant ) . the models of davies and verschuur place complex c ( and indeed , many of the northern hvcs ) in the outer galaxy , either as a high - scale height extension of the outer arm ( where the maximum in the disk s warp occurs ) or a separate spiral arm . in contrast , haud ( 1988 ) suggests the milky way is a polar ring galaxy , with the magellanic stream and complex c part of a lengthy , near - continuous feature , surrounding the galaxy . metallicity differences between the two hvcs ( compare gibson et al . 2000 with the results presented here ) likely make the haud scenario untenable ( as does the fact that the stream and complex c have significantly different position angles on the sky ) ; on the other hand , the metallicity of complex c does match that of the outer disk ( e.g. gibson et al . 2002 ; figure 2 ) , as one might expect if the connection to the outer arm was true . it should be stressed though that this metallicity match may simply be coincidental . regardless , a revisit to the classic davies and verschuur models appears timely . we are currently pursuing 3-dimensional hydrodynamical simulations of the outer arm complex and complex c , in an attempt to unravel the kinematics in this region of the galaxy . at this point , we urge that caution be exercised in identifying the hvc complex c as the long - sought low - metallicity infalling gas . future observations with fuse and hst / stis ( e.g. mrk 205 by bowen et al . ) of complex c should clarify many of the unanswered questions raised here . we wish to express our gratitude to bart wakker , who provided feedback above and beyond the call of duty . special thanks are also due peter kalberla who provided spectra in advance of their publication . jms , mlg , and bkg acknowledge the financial support at the university of colorado through the nasa long - term space astrophysics program ( nag5 - 7262 ) and the fuse science team ( nas5 - 32985 ) . we also thank the hst program go-0653.01 - 95a for use of the ghrs spectra . gibson , b. k. , giroux , m. l. , stocke , j. t. , & shull , j. m. 2001 , in gas & galaxy evolution , ed . j. e. hibbard , m. p. rupen , & j. h. van gorkom , ( san franciso : asp conference series ) , in press ( astro - ph/0009050 ) gibson , b. k. , fenner , y. , maddison , s. t. , & kawata , d. 2002 , in extragalactic gas at low redshift , ed . j. s. mulchaey & j. t. stocke , ( san franciso : asp conference series ) , in press ( astro - ph/0107496 ) llclrccrr 6590/wakker & z3kh0104 t & 01/10/97 & mrk 290 & 1622.0 & ghrs & g160 m & 1231.8 & 1269.0 + 6590/wakker & z3kh0105 t & 01/10/97 & mrk 290 & 2433.0 & ghrs & g160 m & 1231.8 & 1269.0 + 6590/wakker & z3kh0107 t & 01/10/97 & mrk 290 & 2230.3 & ghrs & g160 m & 1231.8 & 1269.0 + 6593/stocke & z3e70104 t & 01/12/97 & mrk 817 & 8870.4 & ghrs & g160 m & 1222.6 & 1258.7 + 6593/stocke & z3e70106 t & 01/12/97 & mrk 817 & 8236.8 & ghrs & g160 m & 1222.6 & 1258.7 + 6593/stocke & z3e70108 t & 01/12/97 & mrk 817 & 7907.3 & ghrs & g160 m & 1222.6 & 1258.8 + 6593/stocke & z3e70304 t & 01/17/97 & mrk 279 & 5702.4 & ghrs & g160 m & 1222.5 & 1258.7 + 6593/stocke & z3e70306 t & 01/17/97 & mrk 279 & 5702.4 & ghrs & g160 m & 1222.6 & 1258.7 + 6593/stocke & z3e70308 t & 01/17/97 & mrk 279 & 5702.4 & ghrs & g160 m & 1222.6 & 1258.7 + 6593/stocke & z3e7030at & 01/17/97 & mrk 279 & 1419.3 & ghrs & g160 m & 1222.6 & 1258.8 + 3584/stocke & z1a65204 m & 02/28/93 & mrk 501 & 29196.3 & ghrs & g160 m & 1221.4 & 1257.6 + 7295/cot & o4n30801o & 09/19/98 & mrk 876 & 2298.0 & stis & g140 m & 1194.8 & 1249.2 + llll @xmath371199 & 1199.5496 & 0.1328 & @xmath90 + @xmath371250 & 1250.578 & 0.00545 & @xmath91 + @xmath371253 & 1253.805 & 0.0109 & @xmath92 + @xmath371259 & 1259.518 & 0.0162 & @xmath92 + @xmath371206 & 1206.500 & 1.669 & @xmath93 + @xmath371260 & 1260.4221 & 1.007 & @xmath92 + llccrcr mrk 290 & s ii@xmath371253 & 1253.2 & @xmath1170@xmath94@xmath185 & 23@xmath23@xmath95 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 290 & s ii@xmath371259 & 1258.9 & @xmath1170@xmath94@xmath185 & 37@xmath25@xmath95 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 290 & si ii@xmath371260 & 1259.9 & @xmath1180@xmath94@xmath160 & 399@xmath25@xmath95 & ( [email protected])@xmath610@xmath96 & @xmath401.0@xmath610@xmath39 + mrk 817 & s ii@xmath371250 & 1249.9 & @xmath1140@xmath94@xmath180 & 13@xmath23@xmath95 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 817 & s ii@xmath371253 & 1253.3 & @xmath1140@xmath94@xmath180 & 27@xmath23@xmath95 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 279 & s ii@xmath371250 & 1249.8 & @xmath1180@xmath94@xmath190 & 18@xmath23@xmath95 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 876 & n [email protected] & 1198.9 & @xmath1210@xmath94@xmath195 & 134@xmath215 & ( [email protected])@xmath610@xmath39 & ( [email protected])@xmath610@xmath39 + mrk 876 & si iii@xmath371206 & 1205.8 & @xmath1250@xmath94@xmath195 & 363@xmath212 & ( [email protected])@xmath610@xmath96 & @xmath404.1@xmath610@xmath96 + lcccccccccc mrk 290 & [email protected] & & @xmath401.0@xmath97 & & [email protected] & @[email protected] & & @[email protected]@xmath20.03 & & 187@xmath210 + mrk 817 & [email protected] & & & & [email protected] & @[email protected] & & & & + mrk 279 & [email protected] & & & & [email protected] & @[email protected] & & & & + mrk 501 & @xmath142.0@xmath98 & & & & [email protected] & @[email protected]@xmath20.06@xmath99 & & & & + mrk 876 & & [email protected] & & @xmath400.4@xmath97 & [email protected] & & @[email protected] & & @[email protected]@xmath20.04 & @xmath1420 +

the first astrophysical objects shaped the cosmic environment by reionizing and heating the intergalactic medium ( igm ) . in particular , x - rays are very efficient at heating the igm before it became completely ionized , an effect that can be measured through the redshifted 21 cm line of neutral hydrogen . high - mass x - ray binaries ( hmxbs ) , known to be prolific x - ray sources in star - forming galaxies at lower redshifts , are prime candidates for driving the thermal evolution of the igm at redshifts . despite their importance , the formation efficiency of hmxbs from the first stellar populations is not well understood as such , their collective x - ray emission and the subsequent imprint on the 21 cm signature are usually evaluated using free parameters . using-body simulations , we estimate the rate of hmxb formation via mutual gravitational interactions of nascent , small groups of the first stars ( population iii stars ) . we run two sets of calculations : one in which stars form in small groups of five in nearly keplerian initial orbits , and another in which two such groups collide ( an expected outcome of mergers of host protogalaxies ) . we find that hmxbs form at a rate of one per in newly born stars , and that they emit with a power of in the kev band per solar mass per year of star formation . this value is a factor larger than what is observed in star forming galaxies at lower redshifts ; the x - ray production from early hmxbs would have been even more copious , if they also formed _ in situ _ or via migration in protostellar disks . combining our results with earlier studies suggests that early hmxbs were highly effective at heating the igm and leaving a strong 21 cm signature . we discuss broader implications of our results , such as the rate of long gamma - ray bursts from population iii stars and the direct collapse channel for massive black hole formation . [ firstpage ] cosmology : theory early universe cosmology : dark ages , reionization , first stars stars : population iii intergalactic medium x - rays : binaries stars : kinematics and dynamics

1509.05427

a major outstanding goal in cosmology is to piece together the history of the universe between cosmic dawn , the emergence of the first stars and galaxies , and the end of reionization , when the radiation from these objects had ionized the intergalactic medium ( igm ) . advances in numerical techniques , combined with exquisite measurements of the `` initial '' conditions ( at a redshift @xmath6 ; @xcite , @xcite ) , have led to remarkable simulations ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) of the conditions leading up to the former milestone , occurring at @xmath7 , when the universe was @xmath8 old . however , reconstructing the subsequent several hundred myr of cosmic history has proved far more challenging , due to the difficulties in reliably modeling the numerous forms of feedback from the first astrophysical objects ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in particular , x - rays from the first galaxies can act as a powerful source of feedback ( e.g. * ? ? ? * ; * ? ? ? * ) that exerts influence over a wide range of distance scales . because hard x - rays ( energies @xmath9 ) have mean free paths comparable to the hubble horizon , they can isotropically heat and partially reionize the early igm ( e.g. @xcite ) . in fact , they are expected to be the dominant agent in heating the igm . such heating may suppress star formation @xcite and massive black hole ( bh ) growth @xcite inside low - mass dark matter haloes by raising the jeans and filtering masses of the igm @xcite . on galactic and circum - galactic scales , soft x - rays ( @xmath10 ) can affect the formation of stars and possibly massive black holes ( bhs ) by promoting the formation of molecular hydrogen via electron - catalyzed reactions ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? in addition to their suspected roles in early galaxy evolution , x - rays are important also because they can leave an observable signature that can be exploited to probe the cosmological epoch in question @xcite . their thermal impact on the early universe should be measurable through the redshifted @xmath11 transition line of neutral hydrogen , which is observed in emission or absorption depending on the relative temperature of the igm with respect to that of the cosmic microwave background ( cmb ) . several studies have investigated how forthcoming observations of the sky - average amplitude and power spectrum of the relic @xmath11 line ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) could be used to constrain the astrophysical agent ( or agents ) responsible for heating the early igm . there are sound reasons to expect that the first galaxies produced x - rays in abundance , and rapidly heated the igm . there are two dominant x - ray sources in present - day galaxies both powered by gas accretion onto bhs , and both plausibly prominent shortly after cosmic dawn : gas feeding massive bhs shining as active galactic nuclei ( agn ) , and x - ray binaries , powered by a stellar - mass bh gradually cannibalizing a companion star . estimates of the mass accumulated by nuclear bhs prior to @xmath12 @xcite , the existence of very massive bhs at @xmath13 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , as well as the observed ( e.g. * ? ? ? * ) and theoretically expected @xcite increase in their duty cycles toward higher redshifts , all hint that x - ray agn may have been much more common during this epoch . likewise , high - mass x - ray binaries ( hmxbs ) dominate the x - ray emission of star - forming galaxies ; the low metallicity and rapid baryonic mass accretion of the earliest galaxies both lend credence to the notion that they were rife with hmxbs . theoretical models suggest that either type of x - ray source could heat the igm to above the cmb temperature as early as @xmath14 , and that this transition should be measurable by the planned @xmath11 experiments . at present , there are too many theoretical uncertainties to determine from the future data which type of x - ray source agn or hmxbs was responsible for driving the thermal evolution of the @xmath15 igm . modeling the early agn x - ray emission is particularly difficult , because the conditions for triggering agn activity are not fully understood even at low redshifts ( e.g. * ? ? ? * ; * ? ? ? because of the uncertainty in the fraction of x - ray photons that is released into the igm as opposed to being trapped inside the accretion flow or reprocessed into the infrared ( e.g. * ? ? ? * ; * ? ? ? * ) ; and because the epoch , initial masses , and birthplaces of the massive bh `` seeds '' are not yet constrained by observations ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? similarly , studies usually estimate the x - ray contribution from early hmxbs by simply inferring empirical relations between x - ray luminosity and star formation rate ( sfr ) in local galaxies ( and modeling the sfr using a semi - analytic cosmological model ) , or combining such relations with one or more free parameters ( e.g. * ? ? ? * ; * ? ? ? the goal of this study is to alleviate the uncertainties in the formation rate and x - ray output of hmxbs in the early universe , by using @xmath1-body simulations of nascent groups of the first ( population iii , henceforth pop iii ) stars . we choose the properties of the star groups in our simulations to reflect those found in hydrodynamical simulations of pop iii star formation at @xmath16 ( @xcite , @xcite ) . we follow the formation and dynamical evolution of compact binaries over thousands to millions of years , including the effects of the background gravitational potential and dynamical friction . this allows us to compute the fraction of pop iii stars that form stable , compact binaries , and eventually undergo an x - ray bright phase . the end result is an estimate of the formation rate of hmxbs in the first protogalaxies , as well as the amount of x - rays they generate per unit star formation . to our knowledge , this is the first published estimate of this type . our simulations predict a binary formation rate which is similar to what is observationally inferred in present - day galaxies . however , we derive a hmxb energy output ( normalized to the star formation rate ) that is a factor @xmath17 higher than in present - day star - forming galaxies , if the hmxb duty cycle is similar to the one in the local universe . we find that the x - ray output does not change significantly within the wide variety of simulation setups considered such as different orientations for collisions between star groups , and ambient gas density and submit that this is a robust estimate . the findings of this study can be used as model inputs in estimating the @xmath11 global signature and power spectrum , but have wider applications . as stated above , the x - ray output of the first galaxies are also of interest for studying feedback on smaller scales , such as subsequent star formation and massive bh formation . our work is also relevant for predicting the rates of long - duration gamma - ray bursts ( lgrbs ) from pop iii stars . lgrbs are important probes that can shed light on the universe out to @xmath18 ( e.g. * ? ? ? according to the collapsar model @xcite , progenitors of lgrbs require rapid rotation of the he core and removal of the h envelope . both criteria are satisfied by pop iii hmxbs , and it is plausible that massive pop iii stars in binary systems are dominant lgrb progenitors in the early universe @xcite . our results on hmxb formation rates can therefore be used to predict and interpret observations of high - redshift lgrbs . the paper is organized as follows . we start in [ sec : dynamics ] by discussing the problem to be solved beginning with the equations of motion , followed by the description of our @xmath1-body code , our choices for the initial conditions , and how the data is interpreted for hmxb formation . we present our results in [ sec : results ] . in [ sec : discussion ] , we discuss the implication of our work for the x - ray output of the first galaxies , as well as for other topics such as lgrbs and smbh formation . we conclude with a summary of our findings in [ sec : summary ] . here , we provide an overview of our simulations namely : the equations of motion that are solved to simulate the dynamical evolution of the star groups ; the numerical scheme we use to solve the equations ; the different types of initial conditions we adopted , as well as the reasoning behind our choices ; and finally , how the results are interpreted for hmxb formation . our @xmath1-body code computes the motion of @xmath1 objects with ( generally different ) masses @xmath19 , moving under their mutual gravitational influence , a dissipative dynamical friction force , and a background gravitational potential . we numerically integrate the equations of motion @xmath20 the first term on the right - hand side of equation ( [ eq : eom ] ) is the specific force due to newtonian gravity , @xmath21 where @xmath22 is the gravitational constant , @xmath23 is the displacement of the @xmath24 star from the center of the host dark matter halo , and @xmath25 . we adopt the plummer softening kernel @xmath26 ( e.g. * ? ? ? * ) , @xmath27 where we take @xmath28 . the second term on the right - hand side of equation ( [ eq : eom ] ) , @xmath29 , is the specific drag force due to dynamical friction . for collisionless systems , the standard chandrasekhar formula for dynamical friction is @xcite , @xmath30 where @xmath31 @xmath32 is the speed of the @xmath33 star with respect to the background , @xmath34 , @xmath35 is the velocity dispersion , @xmath36 is the coulomb logarithm and @xmath37 is the local gas density . we adopt the modified formula for gaseous medium used in @xcite . this prescription incorporates behaviors found in numerical simulations for subsonic and supersonic regimes @xcite . the specific drag force vector always points opposite to the direction of motion , and is given by : @xmath38 with @xmath39 \\ \hspace{100pt } 0\leqslant { \mathcal m}_i \leqslant 0.8;\\ 1.5~\ln\lambda \big[{\rm erf}\left ( \frac{{\mathcal m}_i}{\sqrt{2}}\right ) -\sqrt{\frac{2}{\pi}}{\mathcal m}_i~ \exp\left(-\frac{{\mathcal m}_i^{2}}{2}\right)\big ] \\ \hspace{93pt } 0.8\leqslant { \mathcal m } \leqslant \mathcal{m_{\rm eq } } ; \\ \frac{1}{2}\ln\big(1-\frac{1}{{\mathcal m}_i}\big)+\ln\lambda \\ \hspace{115pt } { \mathcal m}_i > \mathcal{m_{\rm eq}}.\\ \end{cases}\ ] ] above , @xmath40 is the mach number , and @xmath41 is the sound speed . we use @xmath42 and the corresponding value of @xmath43 . in our simulations , the motion of the stars with respect to the background gas is supersonic . in this regime , the characteristic dynamical friction timescale , for a circular keplerian orbit of two bodies with for @xmath44 and @xmath45 , is @xmath46 where @xmath47 is the frictional dissipation power ( @xmath48 ) , and @xmath49 the orbital energy ( @xmath50 ) . the third and last term , @xmath51 , is the specific force due to the background potential , which is dominated by gas . the background potential provides an additional inward force whose functional form depends on the density profile . for simplicity , here we use a constant density and explore different values in our simulations . the force due to the background potential is then @xmath52 where here @xmath23 is the vector pointing from the center of the halo to the @xmath53-th star . the equation of motion is solved iteratively , with the positions , velocities and accelerations of each star updated at every time step . we describe our computational method below . we perform 3-dimensional , @xmath1-body simulations with 4th - order & 5-stage runge - kutta - fehlberg methods ( rkf45 method , @xcite ) using adaptive time steps . the rkf45 is a very precise and stable integration method among the large class of runge - kutta schemes , particularly by adapting the butcher tableau for fehlberg s 4(5 ) method . we solve equation ( [ eq : eom ] ) , as described in the preceding text , updating the position and velocity components of the stars while treating the background density of the gas as a uniform and static distribution . to ensure numerical precision , our computational scheme varies the value of each subsequent time step analytically , so that numerical errors for each variable in the simulation do not exceed @xmath54 times the size of the variable . in some cases , however , this method leads to excessive computational effort for calculating relatively trivial interactions . for example , near the pericenter of hyperbolic or highly elliptical encounters , the time steps become increasingly small in a runaway fashion to compensate for the steep rise in acceleration and associated errors . in order to avoid such situations , we implement the following two numerical shortcuts to keep computation times tractable . the first shortcut is to use analytic approximations for very close 2-body encounters . this is justified in cases where pairs of stars are sufficiently close to each other and isolated from the other stars in the simulation , so that ( i ) the gravitational pull from the other stars and the background potential are negligible compared to the mutual gravitational pull of the pair , and ( ii ) the orbital motion is supersonic and the dynamical friction force can be treated as a linear perturbation to the 2-body keplerian problem . the code employs analytic approximations for any close stellar pairs that satisfy these conditions , and reverts to the rkf algorithm when the conditions stop being satisfied . we derived the following approximations for the semi - major axis and the eccentricity : @xmath55^{-2/3},\ ] ] @xmath56 above , @xmath57 is the reduced mass , @xmath58 and @xmath59 are respectively the semi - major axis and the eccentricity at @xmath60 , @xmath61 is a dimensionless constant , and @xmath62 is a function of the mass ratio that is symmetric about @xmath63 . note that the variable @xmath64 with subscript indicates the specific force , while without subscript it represents the semi - major axis . to derive the analytical expression for the time - evolution of the orbital distance in equation ( [ eq15 ] ) , we integrated the equation of the motion of a star under the influence of a dynamical friction torque . equation ( [ eq16 ] ) then follows from the definition of orbital eccentricity in terms of orbital energy and angular momentum . note that as the radial distance decreases , the eccentricity increases . here , @xmath61 and @xmath62 are free parameters and we use @xmath65 and @xmath66 . the stellar coordinates and velocities are recovered as functions of @xmath64 and @xmath67 using standard expressions for keplerian orbits @xcite . because we are dealing with systems with more than three bodies of different masses , stars frequently form hierarchical triple systems whose motions are affected by the kozai mechanism @xcite . since our analytic solutions above do not account for such changes in mutual inclination , we limit the use of these solutions to situations where @xmath68 , where @xmath69 is the dynamical timescale for the outer pair in the hierarchical triple , and @xmath70 is the time scale for the kozai mechanism , @xmath71 the subscript 1 above indicates the primary star of the inner compact binary along with the satellite star denoted by the subscript 3 , while the hierarchical tertiary star is indexed by the subscript 2 , and @xmath72 is the orbital period . our second shortcut for keeping the simulation runtimes manageable is to set a minimum value for the time step . we choose a physically motivated value , @xmath73 , where @xmath74 is the smallest value of the dynamical time between any two stars in the simulation ( that are not being treated by the analytic shortcut above ) at a given time step . this procedure is necessary when there are three or more stars interacting at small separations , in which case the analytic approximations above can not be used . we note that such situations are rare compared to the places where the analytic shortcut is applicable . it is assumed that a hmxb has formed if both of the following criteria are satisfied : 1 . _ one of two stars forming a binary turns into a black hole . _ in order to determine which stars turn into black holes , we need to compare the typical lifetime ( @xmath75 ) of a massive star with the time ( @xmath76 ) in the simulation ( taken to coincide with the time at which stars are born ) . if @xmath77 , the star is marked as a black hole in the simulation . for main - sequence stars , it is possible to estimate this lifetime from the mass - luminosity relation , that is @xmath78 and @xmath79 , leading to @xmath80 with @xmath81 . however , due to the uncertain value of @xmath82 for pop iii stars , we rather prefer to use here the nuclear time scale of pop iii stars estimated by @xcite and @xcite . they calculate the h - burning nuclear time scale for these stars with a stellar evolution code . we assume that a bh forms if the stellar mass is greater than @xmath83 . the two stars are close enough so that the accretion occurs through roche - lobe overflow ( rlof)_. this criterion is simply written as @xmath84 , where @xmath85 is the stellar radius , and @xmath86 is the roche - lobe radius of the most massive star calculated from the center of the star to the inner lagrange point . an approximate analytic formula to the roche - lobe radius of star 1 for a wide range of the mass ratio is @xcite , @xmath87 where @xmath88 is the orbital distance and @xmath89 is the mass ratio . although the above expression was derived in a study of binaries in circular orbits , @xcite found that the roche lobe radius does not differ very much for eccentric binaries . we therefore apply equation ( [ eq7 ] ) to all binaries found in our simulation . if a binary satisfies these two criteria , the binary is marked as a hmxb . since the two criteria are independently checked at every time step , which criterion is satisfied first is not important . we design our simulations with the @xmath90 universe in mind . this is the redshift range where the dark matter haloes reach virial temperatures @xmath91 , the expected condition for pop iii formation , at the highest rates . the lifetimes of the stars are several myr , which is comparable to the timescales on which the host haloes undergo mergers with other pop iii forming haloes @xcite . the initial positions of the stars are generated quasi - randomly via a monte carlo realization as follows , motivated by the assumption that they formed inside a shared keplerian gas disk at the center of the host dark matter halo . their initial radial positions in the disk plane are chosen randomly from a uniform distribution whose amplitude depends on the characteristic size scale of the star - forming cloud ( see below ) . their azimuthal distribution is chosen to be nearly uniformly distributed , so that the azimuthal angular position of the @xmath92-th star is given by @xmath93 , where @xmath1 is the total number of stars in the group . this choice is made to minimize the radial gravitational pull between the stars , so that they do not immediately fly apart . naturally , the azimuthal positions do not remain evenly spaced , but become mixed quickly . finally , we allow the stars to be displaced out of the plane of the disk . their positions perpendicular to the plane are chosen randomly , so that their vertical displacement out of the disk plane is no more than @xmath94 above or below their initial radial displacement from their shared center of mass . the initial velocities of the stars are assigned to be circular , parallel to the disk plane , and keplerian at the instant the simulation begins . for this study , we investigated two different size scales of star forming regions . 1 . _ large scale _ : following the results from @xcite , where pop iii protostars are formed inside star - forming regions with a size of a few thousand au , we have performed simulations where the stars are placed inside a region of size @xmath95 . + we explore two different values of uniform , constant gas density , @xmath96 ( denoted @xmath97 hereafter ) and @xmath98 ( @xmath99 ) . + for gas of primordial composition , the molecular weight @xmath100 can vary between @xmath101 and @xmath102 , depending on the ionization fraction . we adopt @xmath103 for simplicity ; this choice does not qualitatively affect our results , since our values for @xmath92 are selected arbitrarily with the goal of exploring the qualitative dependence on @xmath92 . + each simulation is run for @xmath104 . the masses of the stars follow the initial mass function ( imf ) they provide with @xmath105 ( @xmath106 ) , @xmath107 and @xmath108 . for these simulations , we consider @xmath109 stars . small scale _ : in the simulations by @xcite , multiple protostars formed several au apart from each other . we therefore run a second set of simulations , where stars form within a 10 au radius . since @xcite do not provide a slope for the imf , we use @xmath105 as above . that study found that more than half of the mass accreted during the protostar phase goes to the most massive protostar in the group . + to mimic this behavior , we first generate a star with @xmath110 , then generate each subsequent star with @xmath111 [ the sum of the masses of the previously generated stars ] . + just as with the large - scale case , we run simulations with number densities @xmath97 and @xmath99 . because this case is the more relevant one for forming hmxbs , we explore several different configurations : cases with a star group of @xmath109 stars , a star group of @xmath112 stars , and collisions between two groups of @xmath109 stars . the last case is motivated by the fact that pop iii - forming minihaloes undergo frequent mergers , which suggests that the nascent star groups themselves undergo close encounters . here , we summarize the findings of our @xmath1-body simulations , focusing in particular on the properties of the most compact binaries found for each set of runs . [ sec : results ] we ran 20 simulations for @xmath97 and 23 for @xmath99 . for @xmath97 , in 12 out of 20 runs , the two most massive stars ( @xmath113 and @xmath114 , where @xmath115 is the @xmath53-th most massive star ) form the most compact binary we denote such a binary with the notation @xmath116 . similarly , binaries of type @xmath117 ( i.e. made up of the most massive star @xmath113 and the third most massive star @xmath118 ) form the most compact binary in 6 of the runs . the left panel of figure [ fig:1 ] shows a sample set of stellar trajectories for one of the large scale @xmath119 runs . throughout the simulation , the compact binary tends to remain near the center of the halo , as less massive stars repeatedly undergo 3-body interactions with the binary . the fact that they do not stray far from the center of the halo is a combined effect of the background potential and the gravitational potential of the massive binary . it is these 3-body encounters that cause the most compact binary in the simulations to end up as type @xmath116 or @xmath117 . in two of the runs , the most compact binaries after @xmath104 consist of less massive stars . the average value of the semimajor axis after @xmath120 in the large scale , @xmath119 runs is @xmath121 , and the minimum value of @xmath64 across the 20 runs is 60 au . these values for @xmath64 are much larger than that necessary for rlof to take place , @xmath122 au . their characteristic dynamical friction time scales are roughly @xmath123 yrs , which is much longer than their lifetimes ( see equation [ eq7 ] when @xmath124 and equation [ eq11 ] ) . we therefore conclude that in systems of @xmath109 stars and @xmath125 , with the stars initially separated at hundreds of au , hmxbs are unlikely to form . 1c c c large scale & @xmath97 & @xmath99 + number of runs & 20 & 23 + @xmath126 $ ] & 0.60(0.90 ) & 0.56(0.74 ) + @xmath127[au ] & 270 & 340 + for the @xmath99 case , the fraction of runs where the most compact binary after @xmath104 is type @xmath116 or @xmath117 are similar to the @xmath119 simulations : @xmath128 for @xmath116 , and @xmath129 for @xmath116 _ or _ @xmath117 . however , the dynamical evolution is quite different , as can be seen by the right panel of figure [ fig:1 ] . the lower gas densities lead to weaker forces due to dynamical friction and background potential , and stars ( especially less massive ones ) tend to be scattered farther from their initial position , as far as @xmath130 . this also results in less frequent three - body interactions in general , and may explain the larger average value of the semi - major axis after @xmath104 , @xmath131 . the formation of hmxbs appears even more intractable for the @xmath132 case . sample stellar trajectories for the @xmath119 and @xmath132 large scale cases are shown in figure [ fig:1 ] , and a summary of the results is given in table [ tab : tab1 ] . 0.7c c c small scale & @xmath97 & @xmath99 + runs & 86 & 86 + @xmath133 [ au ] & 1.37 & 1.42 + @xmath134[yr ] & @xmath135 & @xmath136 + companion stars & + @xmath137 & 0.070 & 0.070 + @xmath138 [ @xmath139 & 4.6 & 4.6 + for the small scale case , we performed 86 runs of 5-body simulations for each of the number density values @xmath119 and @xmath132 . due to the smaller initial separations of the stars , we run the simulations for a shorter amount of time . we run each simulation for a minimum of @xmath140 , but stop the run if a stable binary forms , and if no further significant dynamical changes are observed . if such a binary does not form , we run the simulations to a maximum duration of @xmath141 . in order to properly compare the results from the @xmath97 calculations with those from the @xmath99 calculations , we use the same initial conditions for each set of runs . in the @xmath97 calculations , the most common scenario is that @xmath113 always forms the most compact binary almost immediately , while stellar scatterings are most common during the first few years to about 40 years . thereafter , a multiple system usually survives and stabilizes , while less massive stars are ejected . a difference between the large - scale and the small - scale scenarios is that , whereas most cases in the large - scale calculations end up with one binary and unbound single stars ( in @xmath142 out of @xmath143 cases ) , in the small - scale runs a multiple system such as triple or quartet ( rather than a simple binary ) forms ( in 59 of 86 cases ) less massive stars take some energy from the multiple system and convert it into their kinetic energies while causing the binary to harden . as an example , figure [ fig:7 ] shows the evolution in the semi - major axis of a compact binary in one of the runs . one can easily notice the quick decrease in the semi - major axis during violent stellar interactions ( before 10 years ) , which clearly indicates the hardening . after a couple of stars are cast away and the multiple system is stabilized and isolated ( after about 10 years ) , dynamical friction plays the main role in the decrease of the semi - major axis ; after this point , we do not expect further dramatic hardening of the binary . in our 86 runs , the last surviving binary is typically of type @xmath144 ( pairing of the most massive and third most massive stars ) . with total mass of @xmath145 and @xmath146 . @xmath147 and the minimum semi - major axis is @xmath148 . the corresponding dynamical friction time scale ( equation [ eq11 ] ) is @xmath149 yr . the ejected stars have speeds of @xmath150 , and @xmath151 . six of the hmxb candidates ( hmxbc hereafter ) have been formed in all runs ( @xmath137 = 0.070 ) . for later use , let us define @xmath138 as the number of hmxbc formed across all simulations , normalized by the total mass of the stars . then @xmath152 ( this term will be used to estimate the x - ray luminosity and is one of the primary results of our study ) . the outcomes of the @xmath132 simulations are quite similar : a triple or higher multiple forms in 65 of 86 runs , and @xmath153 . furthermore , due to the same number of hmxbc , @xmath137 and @xmath154 are the same . one notable difference is that for @xmath132 , the dynamical friction time scale is longer by 2 orders of magnitude compared to @xmath97 , because this quantity is inversely proportional to the number density . interestingly , there are four runs ( for each density value ) in which the most compact binary is eccentric , and inside the requisite separation for rlof at pericenter but outside it at apocenter . we consider only two of them as hmxbc and rule out the other two binaries since the mass of the more massive star is smaller than @xmath155 @xcite . we present the distribution of eccentricity for pericenter distance in figure [ fig:8 ] . in particular , the _ left _ panel shows the distribution for @xmath97 calculations and the _ right _ panel for @xmath99 . the circled point indicates hmxbc with distance at pericenter shorter than the corresponding roche - lobe radius . how accretion proceeds in a highly eccentric binary system under these conditions remains an unsettled issue to date , as studies have claimed that the orbital semimajor axis and eccentricity can either increase or decrease depending on the binary properties at pericenter ( @xcite and @xcite ) . if the rlof does induce circularization , then accretion proceeds normally ( i.e. steadily ) . however , if the rlof instead increases the eccentricity of the system , then , whether accretion can proceed steadily rather than intermittently will depend on the relative timescale between the disk lifetime @xmath156 ( on the order of the viscous timescale ) , and the orbital period of the binary @xmath157 . we computed these timescales for all the eccentric binaries in our simulations ( for which the roche - lobe radius straddles the pericenter and apocenter ) , and found that @xmath158 in all cases but one . this implies that the fraction of binaries whose eccentricities cause intermittent rlof is small , and that as a global average , rlof is steady to a good approximation . the magnitude of the gas density considerably influences the characteristics of the dynamical interactions . while on the one hand @xmath159 of the simulations end up forming the same triples for both density values , on the other hand their trajectories and the center of mass movements differ significantly . figure [ fig:4 ] shows two sample trajectories of 5 stars with low number density ( left panel ) and high number density ( right panel ) after 1000 yr . they were given identical initial conditions for the run , but their trajectories have developed differently . in figure [ fig:3 ] , at 800 yr , even though the same stars form a triple and the same star is ejected for both number densities , their trajectories are clearly different . as can be seen in figure [ fig:2 ] , which shows the change of forces per unit mass exerted on a typical kicked - off star , @xmath160 and @xmath161 are negligible compared to @xmath162 before about 5 years . note that @xmath160 and @xmath161 are synchronized at early times because the motions of the stars are close to the keplerian motion , @xmath163 . the star in figure [ fig:2 ] is ejected at @xmath164 for the high number density ( and at @xmath165 for the low number density ) . this can be understood from the fact that @xmath162 monotonically decreases and @xmath161 increases ( @xmath166 where @xmath88 is the distance from the center of mass ) . at the same time , @xmath167 barely changes since the star moves in the supersonic regime and the speed decreases slowly . there is no noticeable difference in the overall results between the @xmath132 and @xmath119 runs quantities such as @xmath168 , the dynamical friction time scale @xmath169 , the total mass of the most compact binary ( or which star forms the compact binary with @xmath113 ) as summarized in table [ tab : tab2 ] . for both number densities , the companion star of the binary is typically the third most massive star @xmath118 . to sum up , we find that @xmath170 of our simulations form hmxbc , regardless of the gas density value . normalized to the total stellar mass in the simulations , the number of hmxbc formed per stellar mass is @xmath171 . we now explore several different configurations for the star group , and run several sets of simulations with 10 stars ( instead of 5 ) . these are : ( 1 ) 10-body version of the small scale calculation presented above ; ( 2 ) head - on crash of two star groups containing 5 stars each ; ( 3 ) a close encounter and subsequent inspiral and merger of two star groups containing 5 stars each . the latter two scenarios are motivated by the fact that the merger timescales and mass accretion timescales of pop iii host haloes , as well as the lifetimes of the massive pop iii stars themselves , are of the same order , @xmath172 . this suggests that merging haloes will be continuously forming new stars ( perhaps pop ii instead of pop iii ) as they merge with other haloes , and that close interactions and mergers of nascent star groups may be relatively common . we generate stellar masses in the same way as for the 5-body case , but with a larger value for the parameter @xmath173 . we have run each simulation for 500 yr . the results of these simulations are summarized in table [ tab : tab3 ] . we briefly discuss each one , as follows . 0.6c c c c c c scenario & 1 & 2 & 3 & 4a & 4b + runs & 54 & 30 & 30 & 30 & 30 + @xmath174 @xmath175 $ ] & 1.0 & 1.1 & 0.90 & 1.8 & 1.6 + @xmath176 @xmath175 $ ] & 1.6 & 1.5 & 1.8 & 2.4 & 2.0 + @xmath177 @xmath175 $ ] & 0.72 & 0.15 & 0.42 & 1.2 & 0.23 + @xmath178 & 0.33 & 0.33 & 0.27 & 0.13 & 0.27 + @xmath179 [ @xmath139 & 15 & 11 & 9.0 & 4.2 & 8.4 + we set up the simulations as in the 5-body calculations , but with 10 stars . we find that @xmath180 . interestingly , there are 18 out of 54 cases in which @xmath181 . there is a large difference in scale between @xmath182 and @xmath183 , where @xmath182 is the semi - major axis of @xmath116 ( binary made up of the two most massive stars ) and @xmath183 is the semi - major axis of the binary stars other than @xmath113 and @xmath114 . this is a common feature of the 10-body simulations : they often end up with triples whose inner binary is @xmath184 while the outer binary is @xmath185 . our simulations yield @xmath186 and @xmath187 . also note that , in 14 out of 18 hmxbc , the binary is @xmath188 ( i.e. it is not made up of the two most massive stars ) . since the compact binaries in these simulations form quickly and we only follow them for 500 years , it is technically possible that they will be disrupted before one of the stars turns into a bh . however , our simulations for the 5-body scenario showed that compact , quasi - steady binaries are unlikely to be disrupted , and for practical purposes we extrapolate this qualitative result to the 10-body case . we find that a hmxbc forms in a larger fraction of these simulations than in the 5-body case , @xmath189 , for the obvious reason that there are more stars . per unit stellar mass in the simulations , the number of hmxb candidates is @xmath190 , which is a factor @xmath191 higher than we found for the 5-body case . two groups of 5 stars are set up with random initial conditions , in the same manner as for the previous simulations of 5-body groups . the two groups are then arranged to collide head - on , as follows : they are placed at a separation of two to three times their sizes and their disks are aligned so that the mutual inclination is zero . the initial relative speed of the groups is roughly the speed of sound and the center of mass of one group is set to move directly toward the center of mass of the other group . we find that prior to colliding , each group forms a compact binary of type @xmath116 ( the most and second - most massive star ) . when the two groups collide , those two binaries that existed before the collision were broken and the two most massive stars of each group form a new compact binary with high chances . the average @xmath192 is @xmath193 , but @xmath194 and @xmath195 , meaning that the most compact binaries are not formed from the most massive stars . the shorter average separations may be a result of a larger number of early 3-body scatterings , which act to harden the group as a whole . hmxbc form in 10 out of 30 runs , and they are not of type @xmath116 . however , we find a rate of hmxb formation per stellar mass @xmath196 ; this is higher than in the 5-body case and comparable to scenario 1 above . sample trajectories for one of the simulations of scenario 2 are depicted in figure [ fig:10 ] . we have used the same input parameters for the two groups as in scenario 2 , except that we now set the impact parameter to be of order the size of the group , whereas it was set to zero in scenario 2 . the groups are given opposite velocities of @xmath197 along the @xmath198-direction , and are offset by a displacement along the @xmath199-direction that is @xmath200 times the typical size of the star group ( @xmath201 ) so that they merge with a spiral motion . we find that @xmath202 , @xmath203 and @xmath204 . the average separation lies between what we find in scenarios 1 and 2 . this can be interpreted as being due to the fact that these simulations ( in which the two groups merge gradually via inspiral ) have more close 3-body interactions than in scenario 1 ( in which 10 stars in quasi - keplerian orbits evolve in isolation ) but fewer such interactions than in scenario 2 ( in which the two groups merge head - on ) . hmxbs form in 8 out of 30 runs and , as with scenarios 1 and 2 , none of the hmxbs are made up of the two most massive stars . we find a similar hmxb formation rate per stellar mass , @xmath205 . sample trajectories from one of the simulations for scenario 3 are shown in figure [ fig:11 ] . in these two scenarios , we again set two groups of five stars each on a collision course . the difference is that the orbital plane of one star group is tilted , so that the two stellar disks have a mutual inclination @xmath53 . we set the inclination at @xmath206 ( nearly co - rotating ) for scenario 4a , and @xmath207 ( nearly counter - rotating ) for scenario 4b . the groups are initially placed at a separation of two to three times their sizes , and set in motion at the same speeds as for the inspiral case ( scenario 3 ) . we find that @xmath208 , @xmath209 , and @xmath210 for scenario 4a . in scenario 4b , we find @xmath211 , @xmath212 , and @xmath213 . in @xmath214 of the runs for both scenarios , the two most massive stars form the most compact binary . stellar binaries end up with somewhat closer separations in the nearly counter - rotating case , due to the fact that the net angular momentum of the merged star group is smaller . indeed , we find a total of four hmxbc across all the @xmath206 simulations , and eight in a same number of @xmath207 simulations . for the same reason , we find a larger fraction of hmxb candidates per stellar mass simulated in the nearly counter - rotating case ( @xmath215 ) compared to the nearly co - rotating case ( @xmath216 ) . the overall formation rate of hmxb candidates is lower for both cases than the cases in which all the stellar orbits were nearly coplanar ( scenarios 1 , 2 and 3 ) , plausibly due to the additional degree of freedom in the stellar orbits . still , the value of @xmath138 is within a factor of a few for all of our simulations . sample trajectories from runs for scenarios 4a and 4b are depicted in figure [ fig:12 ] and figure [ fig:13 ] , respectively . our large - scale simulations show that the if pop iii stars form hundreds of au apart , as in the simulations of @xcite , then the timescales required to make hmxbs via stellar scatterings are simply too long . on the other hand , if protostellar clouds fragment and form stars in close groups on scales of @xmath217 , as in @xcite , then a small fraction of groups can form hmxbs . we briefly discuss the dynamics of hmxb formation in our simulations , then move on to discuss the astrophysical implications of our findings . the simulations indicate that , as expected , scatterings play a major role in making a compact binary . the background potential and the dynamical friction play a secondary role , by allowing the most compact binary ( or triple ) to remain near the center of mass of the halo and for other stars to return and scatter again and again . we find that on average , the number of hmxb candidates formed per stellar mass , @xmath138 , is a function of the number and orientation of close 3-body encounters . our results indicate that @xmath138 may be somewhat higher in configurations that result in fewer ejections of stars , and if the interactions are coplanar . while we are able to interpret this , as well as trends in the average separation between stellar pairs , in terms of the initial kinematic setup of the various scenarios simulated ( see [ subsec:10-body ] above ) , the value of @xmath138 does not vary by more than a factor @xmath191 . we interpret this lack of a significant variation in @xmath138 , for such a diverse set of initial conditions and ambient gas densities , to mean that our values are not far from the one that results from similar stellar encounters in nature . another way in which a nascent stellar group could harden is migration through a gaseous disk . the migration could occur as the protostars form@xcite found significant accretion from the protostellar disk onto the most massive protostar , and did not follow the evolution of the system beyond this stage . ( it could also occur to a lesser degree in a vestigial gas disk , after the stars are in place . ) we evaluate the possible role of disk migration on the separation of pop iii stars by considering a steady , geometrically thin disk with an @xmath218 viscosity ( @xcite ; see also @xcite ) . we adopt a disk with @xmath219 and an accretion rate @xmath220 , following @xcite and @xcite , who considered the structure of accretion disks around pop iii stars at high redshifts . we estimate the migration timescale @xmath221 following @xcite . we take a binary system with primary mass @xmath222 and secondary mass @xmath223 , based on the mean values found across our simulations . for these masses and disk parameters , the secondary is able to clear a gap around its orbital path ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , by satisfying both of the following two conditions : @xmath224 @xmath225 where @xmath226 is the distance of the secondary from the primary , @xmath227 is the scale height of the disk at that location , and @xmath228 is the binary s mass ratio . if condition ( 1 ) is violated , the gap will be closed by the radial pressure gradient . condition ( 2 ) relates the gap width and the roche radius of the secondary ; this ensures that the secondary acts to transfer orbital angular momentum through the disk , rather than accreting mass via rlof . we find that inside @xmath229 , for the disk properties stated above , @xmath230 is smaller than the right - hand side of equation ( [ eq26 ] ) by a factor @xmath231 , and smaller than the right - hand side of equation ( [ eq21 ] ) by a factor @xmath232 , indicating that the secondary is easily able to open a gap . the migration timescale of the secondary depends on the dimensionless parameter @xcite @xmath233 where @xmath234 is the local surface density of a steady - state disk around the primary , in the absence of perturbations by the secondary . for @xmath235 , the gas in the disk is able to dynamically dominate over the gravitational influence of the secondary , and the secondary is pushed inward on the viscous diffusion timescale of the disk , @xmath236 for @xmath237 , @xcite found that the migration timescale is longer , @xmath238 in figure [ fig:6 ] , we plot the local disk mass near the secondary , @xmath239 , alongside the typical secondary mass , @xmath240 ( the numerator and denominator , respectively , for the ratio @xmath241 ) . for @xmath242 , @xmath237 and the migration timescale is expected to slow . we plot the two migration timescales @xmath243 and @xmath244 in figure [ fig:9 ] . both of these timescales are shorter than both the typical lifetimes of protostellar disks , as well as of the stars themselves . this suggests that disk migration could , in principle , lead to initial stellar separations smaller than what we have assumed , making the formation of hmxbs via stellar scatterings more favorable . on the other hand , radiative feedback from the stars could blow away the disk before significant migration can occur . we submit that the hmxb formation rates inferred from our simulations be taken as a conservative estimate , with possible additional contributions from channels other than stellar scattering . as discussed in [ sec : intro ] , hmxbs are believed to be a major source of x - rays in the early universe . observations of nearby star - forming galaxies suggest that their x - ray luminosities ( which are dominated by hmxbs ) scale linearly with their star formation rate @xcite . @xcite find many studies support a linear proportionality between the x - ray luminosity of hmxbs and the star formation rate ( sfr ) ( @xcite , @xcite , @xcite and @xcite ) . in the linear regime , the x - ray luminosity of the local universe is given by @xcite , @xmath245 where sfr is the star formation rate . how the ratio of x - ray luminosity to sfr evolves with redshift is a key question in evaluating the properties of galaxies and young stellar populations , and in linking the earliest galaxies ( most of which should be actively forming stars , based on their mass accretion rates ) with their x - ray luminosities . the question is as yet unresolved by current observations @xcite , and is often treated as a free parameter in studies estimating the x - ray production of the first galaxies ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in the following , based on the results of our simulations , we are going to quantitatively evaluate the relation between x - ray luminosity and sfr , and compare it with equation ( [ eq29 ] ) . we can write the x - ray luminosity as @xmath246 below , we discuss each quantity in equation ( [ eq28 ] ) . 1 . @xmath247 , the eddington luminosity , which scales with the typical mass of the bh engine @xmath248 as @xmath249 . @xmath250 , the typical ratio of the total radiative power emitted by hmxbs ( the bolometric luminosity ) to @xmath247 . if the typical luminosity of a hmxb during an active phase is eddington , then @xmath250 is effectively the mean duty cycle . other studies have adopted values ranging from @xmath251 to @xmath252 @xcite ; we take as our fiducial value @xmath253 . @xmath254 , the fraction of the bolometric luminosity that is emitted between @xmath255 and @xmath256 . observational estimates vary between @xmath251 and @xmath257 ( e.g. * ? ? ? * ; * ? ? ? because the bh masses in pop iii hmxbs are expected to be higher than in present - day populations , and the peak energies of accretion disks scale with the mass of the central engine as @xmath258 , their characteristic spectra could be somewhat softer . we do not expect this to significantly affect our estimates here . @xmath259 , the time that a massive binary spends as a hmxb , with the less massive star donating mass to the more massive bh companion . if the two stars form simultaneously and form a compact binary before the more massive member dies to become a bh , then this is simply @xmath260 , the difference in the lifetimes of the stars . we use this ( somewhat arbitrary ) estimate . however , because the lifetimes of the less massive star ( @xmath261 for stars with masses @xmath262 ) are comparable to the expected specific star formation rate in galaxies at this redshift , we argue that any prescription that satisfies @xmath263 is a reasonable order - of - magnitude estimate . @xmath264 , the fraction of hmxb candidates identified in our simulations that actually survive to become hmxbs . this quantity accounts for possible disruptions of binaries , due to ( a ) the merger of the stars during main sequence and post - main sequence evolution @xcite ; ( b ) the more massive star getting kicked following a supernova explosion ( e.g. * ? ? ? * ; * ? ? ? * ) ; and ( c ) subsequent disruptions by stellar scatterings that were not captured by our simulations . ( our simulations follow the evolution of star groups until the formation of a stable compact binary , but not until stellar death millions of years later . ) theoretical estimates typically yield @xmath265 ( e.g. * ? ? ? * ; * ? ? ? @xmath266 , the fraction of x - rays that escape the galaxy . unless the environment of the hmxbs are compton - thick , which is unlikely for the low - mass galaxies of interest , we expect @xmath267 . @xmath138 , the number of hmxbs formed per stellar mass . this is the main output of our simulations . whereas previous theoretical works had arrived at this value by extrapolating the locally observed value with an assumed redshift evolution , or with free parameters , here we provide an estimate based on suites of @xmath1-body simulations whose initial conditions are motivated by cosmological simulations of pop iii star formation . ( note that this quantity has units @xmath268 ; we use the capital letter to distinguish it from the dimensionless fractions represented by @xmath269 ) across all of our small - scale simulations varying the number of stars in the group , whether groups evolved in isolation or through several different orientations of mutual collisions , and exploring two values for the ambient gas density separated by two orders of magnitude we find @xmath270 , varying by less than a factor of @xmath271 between the lowest and the highest values ( see table [ tab : tab4 ] ) . finally , we can write @xmath272 as follows : @xmath273 based on our choices of the factors , @xmath253 , @xmath274 , @xmath275 and @xmath276 , we can estimate the normalized x - ray luminosities per sfr , @xmath277 . we report this quantity for each of our models in table [ tab : tab4 ] . it varies from a minimum of 37 to a maximum of 450 among the studied scenarios . these @xmath278-to - sfr ratios are @xmath279 higher than what is observed in the local universe . this result is qualitatively consistent with the findings of @xcite and @xcite , who find an increase in the @xmath280-to - sfr ratios toward @xmath281 . our high @xmath278-to - sfr values stem from the large mass of the hmxbc primary , and the relatively low mass of the secondary . the former leads to a higher eddington luminosity compared to typical stellar - mass bhs ( @xmath282 ) in the local universe , and the latter results in long stellar lifetimes , which in turn leads to longer @xmath283 . note that we used the stellar mass as a proxy for the bh mass for simplicity , due to the theoretical uncertainties in evaluating the mass loss due to winds and during the transition to a bh . any significant mass loss ( e.g. simulations by @xcite suggest that the bh remnants of massive metal - free stars end up with @xmath284 of the progenitor mass ) would be directly translatable to the estimated x - ray luminosities reported in table [ tab : tab4 ] . a higher @xmath278-to - sfr ratio implies that igm heating will occur earlier than commonly thought . the thermal history of the igm can be probed in the @xmath11 line , which is observable in absorption ( or in emission ) , depending on whether the spin temperature of the igm is below ( or above ) the cmb temperature . if the igm heats early , as suggested by our estimates of the x - ray emission of early galaxies , the 21 cm absorption line appears earlier , and the `` dip '' as a function of redshift caused by adiabatic cooling is not as deep and not as sharp as in the case of late heating as it would otherwise ( see figure 2 in @xcite ) . another consequence of early , intense heating is that the temperature of the igm could become high enough , to suppress the formation of low - mass galaxies @xcite and the growth of their nuclear bhs @xcite . 1c | c c | c c c c c c & 5-body(@xmath285 & 5-body(@xmath99 ) & _ sce . 1 _ & _ sce . 2 _ & _ sce . 3 _ & _ sce . 4a _ & _ sce . + @xmath286[@xmath287(@xmath288 $ ] ) & 110(2.5 ) & 88(2.9 ) & 110(2.5 ) & 120(2.4 ) & 110(2.4 ) & 160(2.0 ) & 110(2.4 ) + @xmath289[@xmath287 ( @xmath290 $ ] ) & 12(17 ) & 11(18 ) & 18(11 ) & 45(4.8 ) & 28(7.1 ) & 63(3.7 ) & 63 ( 3.7 ) + @xmath291[myr ] & 14 & 15 & 8.6 & 2.5 & 4.7 & 1.7 & 1.3 + @xmath138 [ @xmath139 & 4.6 & 4.6 & 15 & 11 & 9.0 & 4.2 & 8.4 + @xmath292 @xmath293 & 220 & 200 & 450 & 107 & 160 & 37 & 40 + @xmath294 & 75 & 67 & 150 & 36 & 52 & 12 & 13 + @xmath295 & 2.2 & 2.2 & 7.0 & 5.4 & 4.3 & 2.0 & 4.0 + as discussed in [ sec : intro ] , the fraction of hmxbs at high redshifts has potential implications for the expected rates of lgrbs from pop iii stars . according to the collapsar model @xcite , for an exploding massive star to yield a grb , several conditions need to be satisfied , namely : 1 . the core of the star must collapse to a bh . this is realized by most pop iii stars , given their large masses . the hydrogen envelope of the progenitor star must be stripped , so that a relativistic jet can penetrate and exit the remaining envelope . 3 . the bh should be surrounded by an accretion disk of high angular momentum material . this is realized if the core of the progenitor star has retained sufficient angular momentum during the evolution . binary systems more easily satisfy the last two conditions with respect to single stars ( see e.g. * ? ? ? in fact , for single stars to end their lives as lgrbs , they need to be born with large initial rotation ( since they are less likely to be spun up by other stars ) , and also avoid being slowed down by magnetic torques ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in contrast , binary stars can spin up the helium core of the progenitor star via tidal coupling and spin - orbit locking . further , rlof can strip the hydrogen envelope during a common - envelope phase without reducing the rotation of the helium core @xcite . this is especially important for pop iii stars , whose heavier hydrogen envelopes would be more difficult to shed in isolation . therefore , compact binary systems , or hmxbs , constitute a promising channel to produce lgrbs from pop iii stars . we can use our results for the formation rates of pop iii hmxbs to estimate the fraction of lgrbs from pop iii stars . @xcite quantified the grb formation efficiency as @xmath296 where @xmath297 is the number of bh - forming stars resulting from a given total stellar mass , @xmath298 is the binary fraction and @xmath299 is the fraction of sufficiently close binaries to undergo rlof . for pop i / ii stars they calculated @xmath300 . combining this value with adopted values for the other parameters@xmath301 , @xmath302 , and @xmath303yields @xmath304 . @xcite noted that it was only to make educated guesses for the pop iii case , due to an absence of detailed calculations for the fraction of close binaries . our work is a first attempt to fill the gap in our theoretical knowledge . we can write @xmath305 . adopting for comparison the same value of @xmath303 , we then infer @xmath306 for the interacting 5-star case , and @xmath307 for the ( most favorable ) 10-body scenario . therefore , our results suggest that lgrb rates from pop iii stars could be comparable to or somewhat higher than the rates from pop i / ii stars . our suite of @xmath1-body simulations , spanning diverse sets of initial conditions for pop iii stars and their environments , point that hmxbs form in higher fractions in the earliest galaxies than at low redshifts , and make significant contributions to the thermal history of the igm , the @xmath11 signature at @xmath16 , and to the rates of lgrbs . the fact that the formation rates of hmxbs varied little between the simulations suggest that our estimates for @xmath138 is reasonably robust . however , we here point out several uncertainties of our work that could affect our conclusions . one important factor is the imf of the stars ( e.g. * ? ? ? we adopted the imf of @xcite , which were based on the masses of protostars roughly @xmath141 after the formation of the protostellar seeds . however , our key results are based on fragmentation of the protostellar clouds on smaller scales , as found in the simulations by @xcite . in those simulations , the most massive protostars had the highest accretion rates , suggesting that the imf slope may be steeper than what we assumed . we also did not account for changes in the masses of stars as they evolved . these are important considerations , as the masses of the stars play a critical role in determining the time that a binary spends transferring mass as a hmxb . we made an effort to account for this limitation by using free parameters such as @xmath264 , which accounts for the mass loss during the sn explosion . another factor that could be more carefully treated in future studies is the spin of the star . in a binary system , this impacts the circularization of the orbit , followed by low eccentricity and the synchronization of spin with orbital phase . due to tidal dissipation and circularization , this could additionally decrease the orbital distance of a binary , boosting the formation of hmxbs and change the average value of the eccentricity . in addition to reducing the orbital separation , the spin makes a difference in mass transfer rates . our simulations generally produce binaries with large eccentricities . for highly eccentric orbits , mass transfer occurs only near pericenter ( @xcite and @xcite ) , and the rates depend on whether the orbital angular speed of the star is super - sychronous or sub - synchronous with rotational angular speed @xcite . we found that the viscous timescales in our eccentric binaries were longer than the orbital timescales in the majority of cases , and used this fact to conclude that their duty cycle should be the same as for nearly circular binaries . however , we did find one exception , in which a rlof accretion event would be sufficiently short - lived to be episodic . a more careful study focused on stellar rotation effects may be necessary to more conclusively estimate the duty cycle of eccentric pop iii hmxbs . finally , our simulations are an attempt to model the stellar dynamics as a gravitational @xmath1-body problem with perturbative forces due to a fixed , smooth gaseous background . more detailed simulations that include detailed stellar feedback , as well as the dynamics and thermodynamics of the gaseous environment , could lead to additional revelations about the early evolution of pop iii star groups . in this study , we used @xmath1-body simulations of the first stars to explore the formation , evolution , disruption and energy output of pop iii hmxbs . the code includes gravitational scattering of stars , dynamical friction , and the gravitational potential of ambient gas . the initial conditions for the simulations ( i.e. imf , typical star separation in the host haloes , ambient densities ) are taken from two different sets of cosmological simulations of pop iii formation , namely by @xcite ( large - scale , i.e. a few thousands of au ) , and @xcite , ( small - scale , i.e. a few tens of au ) . these provide two complementary sets in that they explore different physical scales for star formation ( for details , see [ sec : initialsetup ] ) . for each of the two scenarios , we investigated star evolution in two backgound gas densities , a high - density case ( @xmath308 ) , and a lower - density one ( @xmath309 ) . 1 . _ 5-body simulations _ : if stars form in quasi - keplerian disk configurations with initial separations of hundreds of au , hmxbs are highly unlikely to form . in contrast , if stars form in compact groups separated by @xmath310 , as is expected from turbulent fragmentation , stellar scatterings lead to a significant hmxb formation rate . in particular , we found that hmxbs form at a rate of a few per @xmath311 of stars formed , independent of the ambient gas density . 10-body simulations _ : we simulated 10 stars on separations of @xmath310 , and evolved them as isolated quasi - keplerian disks , or as two colliding groups with 5 stars each . for the latter , we ran several different sets of collision geometries . + we found that the hmxb formation rate was a factor @xmath312 times higher than for the @xmath313-body simulations , mainly due to the fact that the larger number of stars allowed for more hardening via stellar scattering . all of the small - scale simulations suggest an x - ray luminosity per unit star formation that is a factor @xmath5 higher than what is observed in the local universe ( under the assumption that other variables such as the x - ray escape fraction from galaxies and the duty cycle of hmxbs do not differ significantly ) . these results are mostly due to _ the large mass of the most massive star of the hmxbc compared to that of the companion star , implying both a large @xmath283 as well as a higher luminosity of the remnant bh_. the fact that we found little variation in this quantity across all of our simulations suggests that this is a robust estimate . additional factors , such as in - disk migration of nascent stars , could further increase the hmxb formation efficiency . a direct consequence is that x - rays can heat the igm rapidly at cosmic dawn . signals of early heating can be probed via the 21 cm line radiation : the absorption line signal is expected to show a broader , shallower minimum due to the shorter gas cooling time , while the emission line would be observed earlier because of the higher gas temperature , at earlier times . several studies have modeled the 21 cm signature of the first hmxbs , but relied on assumptions for their @xmath314 relation relative to the empirical value found at lower redshifts . our work provides a theoretically driven estimate for this quantity . in addition to the implications for the thermal history of the igm , these high formation rates of hmxbs per stellar mass imply a higher grb formation efficiency from pop iii stars in binaries . this predictions can be tested with a long baseline of observational data from _ swift_.

we use _ spitzer space telescope _ observations from the _ spitzer _ survey of the small magellanic cloud ( smc ) to study the young stellar content of n66 , the largest and brightest region in the smc . in addition to large numbers of normal stars , we detect a significant population of bright , red infrared sources that we identify as likely to be young stellar objects ( ysos ) . we use spectral energy distribution ( sed ) fits to classify objects as ordinary ( main sequence or red giant ) stars , asymptotic giant branch stars , background galaxies , and ysos . this represents the first large - scale attempt at blind source classification based on _ spitzer _ seds in another galaxy . we firmly identify at least 61 ysos , with another 50 probable ysos ; only one embedded protostar in the smc was reported in the literature prior to the smc . we present color selection criteria that can be used to identify a relatively clean sample of ysos with irac photometry . our fitted seds indicate that the infrared - bright ysos in n66 have stellar masses ranging from 2 m to 17 m , and that approximately half of the objects are stage ii protostars , with the remaining ysos roughly evenly divided between stage i and stage iii sources . we find evidence for primordial mass segregation in the region , with the most massive ysos being preferentially closer to the center than lower - mass objects . despite the low metallicity and dust content of the smc , the observable properties of the ysos appear consistent with those in the milky way . although the ysos are heavily concentrated within the optically bright central region of n66 , there is ongoing star formation throughout the complex and we place a lower limit on the star formation rate of m yr over the last myr .

0707.3998

star formation is one of the most important astrophysical processes , but because it takes place on small physical scales and behind heavy optical obscuration , observations of the early phases of star formation have generally been limited to the milky way . only in recent years have these studies begun to be extended to extragalactic objects . now , the high sensitivity , high angular resolution , and large field of view provided by the _ spitzer space telescope _ at mid - infrared wavelengths open a new window on recently formed stars in nearby galaxies like the magellanic clouds . studying these young stellar objects ( ysos ) provides an excellent complement to observations of star formation in the milky way because the physical conditions they are subject to are quite different from those in the galaxy . in particular , the metallicity of the magellanic clouds is well below solar , and their dust content is correspondingly lower as well @xcite . most star formation today is taking place in galaxies smaller than the milky way ( for which the magellanic clouds are good prototypes ) , and star formation at high redshift occurred in low - metallicity environments , so investigating the effects of these differences on how star formation works will be an important step towards understanding how many of the stars in the universe were formed . the small magellanic cloud ( smc ) , and its more massive companion the large magellanic cloud ( lmc ) , are the two nearest star - forming dwarf galaxies . located at a distance of 61.1 kpc @xcite , and with a luminosity of @xmath5 l@xmath1 @xcite and a metallicity of @xmath6 solar @xcite , the smc is perhaps the best local analog of primitive galaxies . the smc is actively forming stars at a rate of @xmath7 m@xmath1 yr@xmath3 @xcite , and is populated by well studied regions and young star clusters . the most active star - forming region is ngc 346 ( alternately known as n66 ; * ? ? ? * ) , located toward the northern end of the smc bar . the ob association powering n66 contains 33 spectroscopically confirmed o stars , and a similar number of bright blue stars without spectra that are likely to be o stars as well @xcite in this paper , we employ _ spitzer _ observations from the _ spitzer _ survey of the small magellanic cloud ( s@xmath8mc ; @xcite , hereafter b07 ) to locate objects in n66 with mid - infrared excesses that we identify as candidate ysos . in the following section , we very briefly describe the observations and our photometry . in [ results ] , we classify the detected sources by fitting their spectral energy distributions and study their locations in color - color space . in [ discussion ] we discuss some of the implications of our results and compare to recent optical studies of n66 . we summarize our findings in [ conclusions ] . the s@xmath8mc is a project to map the star - forming body of the smc with _ spitzer _ in all seven infrared array camera ( irac ; * ? ? ? * ) and multiband imaging photometer for _ spitzer _ ( mips ; * ? ? ? * ) bands . the images cover an area of @xmath9 deg@xmath10 including the entire bar and wing of the smc , with an average exposure time of 144 s at each position . the mips data were obtained in 2004 november and the irac data in 2005 may . the data were processed with version s11.4 of the automated _ spitzer _ pipeline . we constructed mosaic images from the individual basic calibrated data frames using the mosaicking and point source extraction ( mopex ) software provided by the _ spitzer _ science center ( ssc ) . b07 describe further details of the observations and the data processing . we performed photometry on the mosaic images with the astronomical point source extraction ( apex ) tasks in the mopex package @xcite . we selected a set of @xmath11 bright stars in each band that were as isolated as possible and constructed point response functions ( prfs ) directly from the data . we then fit these prfs to every detected source in the images to determine fluxes . the images contain extensive diffuse emission that must be separated from point sources , so we used a small median filter ( @xmath12 ) to remove the background , and then detected sources on the background - subtracted image . because the median filtering removes some flux even from point sources , the psf fitting then took place on the original ( not background subtracted ) image . we found that this technique offered the best compromise between detecting bright sources on top of diffuse emission and detecting faint sources in background - free regions . for the 24 @xmath13 m mips data , in which the extended emission dominates over the point sources , we used a somewhat larger median filter ( @xmath14 ) and switched to the ` combo ' algorithm for separating clusters of bright emission into individual sources in place of the ` peak ' algorithm in the task detect . we used observations of bright , isolated point sources in the smc to measure the aperture corrections for the photometry . to provide near - infrared fluxes , we used vlt / isaac j and k@xmath15 imaging of the center of n66 ( * ? ? ? * rubio & barb , in preparation ) and two micron all sky survey ( 2mass ; * ? ? ? * ) measurements in the outer regions . we also added _ hst _ @xmath16- and @xmath17-band photometry of the central region from @xcite , but because the much higher angular resolution of the acs images can cause confusion within the _ spitzer _ resolution element , we only used these data for bright stars ( @xmath18 ) . the assumed photometric uncertainty associated with each flux measurement is very important for determining the relative weights given to the various data points in the sed fitting that we carry out in [ sedfitting ] . although the statistical uncertainties on all of the photometric measurements were quite small ( generally a few percent or less ) , we imposed larger minimum uncertainties on all of the data points to account for systematics . for example , the _ hst _ , 2mass , vlt , irac , and mips observations were made at various times over a period of 7 years , providing plenty of time for the fluxes of the ysos , which may be variable , to change . the absolute photometric accuracy of irac is 10% @xcite , so we added 10% of the flux in quadrature to the measured uncertainty for each source . mips also has a stated accuracy of 10% @xcite , so when combined with the irac calibration accuracy and the 6 month time baseline between the observations , we used a 20% minimum error for mips fluxes . since the near - ir observations took place @xmath19 years earlier , we imposed larger minimum errors of 35% ( for the deep vlt data ) and 50% ( for the shallow 2mass data ) on those data points . the optical measurements are more recent ( 2004 july ) and more accurate , so we assumed 10% uncertainties on the _ hst _ photometry . for the purposes of this study , we limited our point - source fitting to a @xmath20 region covering the full extent of the hot dust emission from n66 seen in our 24 @xmath13 m image ( from 005751 to 010037 and from @xmath21721648 to @xmath21720419 [ all coordinates j2000.0 ] ) . in this box , we detect 6544 sources at 3.6 @xmath13 m , 5836 at 4.5 @xmath13 m , 1784 at 5.8 @xmath13 m , 1718 at 8.0 @xmath13 m , 101 at 24 @xmath13 m , and 15 at 70 @xmath13 m , for a total of 8011 unique objects . a significant number of these objects have _ spitzer _ colors that are redder than normal stars should be at these wavelengths , suggesting that they may be ysos or background galaxies ( b07 ) . there are two possible approaches to determining the nature of individual _ spitzer _ sources . one could simply use the observed colors and magnitudes to classify the sources into various categories . stars of nearly all kinds ( except those with dusty atmospheres ) have colors near zero for any combination of _ spitzer _ bands because the irac and mips bands are on the rayleigh - jeans tail of their spectral energy distributions ( seds ) . ysos , because of the emission from warm ( @xmath22 k ) dust around the central protostar , have red colors throughout the mid - ir . irac colors for theoretical yso models are given by , e.g. , @xcite and @xcite . alternatively , one can compare the full seds to a variety of source models and find the best match for each object . the advantage of the color selection strategy is its simplicity , but it also has some drawbacks . it fails to use all of the available information about each object ( since we also have mips , optical , and near - ir fluxes for many of the sources ) , and it is very difficult to learn about the detailed properties of individual sources from only their colors . color selection also does not offer a way to test the assumption that sed models designed for milky way sources offer a good description of ysos that form in the metal- and dust - poor environment of the smc ; this is a disadvantage for studies seeking to determine how star formation proceeds at low metallicity , but could also be an advantage in that it enables the selection of ysos without regard to the properties of their environment . because the sed classification should be more accurate , we begin with that technique and then compare the results to color selection in [ colors ] . the sed fitting tool used for this study employs a linear regression method to find all the seds from a large grid of models that fit the data within a specified @xmath23 @xcite . the grid of models consists of 7853 stellar atmospheres @xcite encompassing all available metallicities and effective temperatures , a limited number of _ infrared space observatory _ ( iso ) spectra of galaxies @xcite and agb stars @xcite , and the 20,000 yso models from @xcite computed using the radiation transfer codes from @xcite . each yso model outputs seds for 10 viewing angles , so the yso grid effectively contains @xmath24 seds . the foreground extinction , @xmath25 , is fit simultaneously using an extinction law derived from glimpse observations ( indebetouw et al . 2006 ) . at ultraviolet and visible wavelengths , this extinction law is not appropriate for the smc , but in the near - ir and mid - ir the differences in extinction between the smc and galaxy are small @xcite . the fitter is run first using only the stellar atmosphere grid . it is then run three more times using the yso grid , iso galaxy spectra , and iso agb spectra on the sources that are not well - fit by stellar atmospheres . based on all the successful fit results , defined by their @xmath23 values ( see below ) , we can classify sources and calculate best estimates and uncertainties for each model parameter . we fit an sed for every object for which we had at least four flux measurements . to better constrain the fits at long wavelengths where most of the sources were not detected , we added 24 @xmath13 m upper limits of 1 mjy ( 5 times the limiting sensitivity of the 24 @xmath13 m data ) for each source that was not detected at 24 @xmath13 m . because there are many more model parameters than data points that go into each observed sed , the traditional reduced @xmath23 statistic can not be used to evaluate the success of the fits . as a proxy , we calculate the @xmath23 per data point ( where each sed has between 4 and 11 data points ) . given this definition , it is difficult to establish _ a priori _ the division between good and poor fits . we inspected the fits by eye and determined that fits with a @xmath23 per data point of less than 2.2 could reasonably be interpreted as good fits , while higher @xmath23 values indicated fits that were not a good match to the templates . we will therefore use this cutoff point to separate good and poor fits throughout the paper . note that this definition depends on the flux uncertainties imposed in [ observations ] . out of the 1645 sources with four or more fluxes , 1322 ( 80.4% ) are succesfully fit with stellar seds ( see above for what we mean by a `` successful fit '' ) . of the remaining 323 sources , 61 can only be fit successfully by yso models , 27 are background galaxies , 6 are agb stars , 81 can be fit by multiple classes of models , and 148 can not be fit by any of the available templates . we display one example yso sed in figure [ fits ] . for the sources that can be well - described by multiple types of templates , we used the @xmath23 value of the best fit for each template type to attempt a more accurate classification . we set a cutoff of 1.5 times the best @xmath23 value , and if , for example , the best yso fit had @xmath26 and the best galaxy fit had @xmath27 then we classified the object as a probable yso . this choice of a @xmath23 cutoff indeed results in the majority of the uncertain objects being classified as ysos and leaves very few objects with unknown types , which appears to be a reasonable outcome . while the exact value chosen for the cutoff is arbitrary , it is clear from the spatial distribution of the various object types that many of the sources classified as possible or definite galaxies must actually be ysos . the results of this paper do not depend strongly on the exact value of the cutoff . using this method , we were able to obtain probable classifications for 68 of the 81 objects of uncertain type , including 50 ysos , 13 galaxies , and 5 agb stars . thus , the final object counts are 111 ysos , 40 galaxies , and 11 agb stars . the remaining 13 objects had @xmath23 values for multiple object types that were too close to distinguish reliably . we list selected parameters from the fits for all 111 ysos in table [ ysotable ] , and we summarize the source classification results in table [ sourcetypes ] . we inspected each of the 148 poorly fit sources and the various attempted fits to their seds to determine why the fitting failed for them . we found that slightly more than 1/3 of the objects had seds at short wavelengths that appeared stellar in origin , but the long wavelength data ( often at 24 @xmath13 m ) were significantly in excess of the extrapolated photospheric emission . the fits to these sources appear to have failed for a number of reasons including slight mismatches between the optical and ir photometry ( either from variability or measurement error ) , confusion , misidentification of the long wavelength counterparts , and contamination of the photometry by underlying dust emission ( revealed by pah features ) . however , some of these sources may also have true infrared excesses indicating the presence of circumstellar material . another @xmath28 of the bad fits were generally faint sources that are spatially coincident with dust filaments in the _ spitzer _ images , again causing significant contamination of their seds by pah features in the irac bands . some of these objects are probably point sources ( either stars or ysos ) , while others may simply be unresolved dust knots . the final 1/3 of the bad fits was a mixed population whose seds could not be straightforwardly interpreted . photometric errors in one or more bands and blending were likely responsible for the failure to fit these objects , but they represented only 2.6% of the total sample . using the results of the sed fitting , we can now investigate the colors of the objects classified as ysos and as stars . we display four color - color plots in figure [ ysocolors ] to illustrate the possibilities for color selection . stars , as expected , lie in the cloud of points centered near ( 0 , 0 ) , while sources that are red in one or both colors may be ysos . note that the frequently used [ 3.6]@xmath29[4.5 ] vs. [ 5.8]@xmath30[8.0 ] color - color plot ( fig . [ ysocolors]a ) does not cleanly separate ysos from other types of sources , particularly stars with modest ir excesses and sources with pah contamination . better separation can be achieved using different combinations of colors , as shown in figure [ ysocolors]b . these plots take advantage of longer color baselines and the abrupt change in yso spectra between the 4.5 @xmath13 m and 5.8 @xmath13 m bands to distinguish ysos from stars , galaxies , and pahs . these results demonstrate that while color selection can be a useful technique for identifying ysos , it does not appear possible to obtain a yso sample that is both complete and clean with a simple set of color diagnostics . sed fitting is a more comprehensive way to determine the nature of sources if measurements in enough bands are available . nevertheless , because in many cases it is desirable to classify objects with easily applicable techniques , we use our data to provide guidance for color selection of ysos . assuming that the purple points in figure [ ysocolors ] are indeed ysos , one can select ysos with the following set of criteria : @xmath31 - [ 4.5 ] & > & 0.6\times([4.5 ] - [ 8.0 ] ) - 1.0 \\ \nonumber [ 4.5 ] - [ 8.0 ] & < & 2.8 \\ \nonumber [ 3.6 ] - [ 4.5 ] & < & 0.6\times([4.5 ] - [ 8.0 ] ) + 0.3 \\ { [ 3.6]}- [ 4.5 ] & > & -([4.5 ] - [ 8.0 ] ) + 0.85 . \label{colorselection1}\end{aligned}\ ] ] these criteria were defined to maximize completeness ; one could alternatively choose to minimize contamination at the cost of increased incompleteness , but given the distribution of sources in figure [ ysocolors]_b _ , the differences would be small . this selection is only 7% incomplete for ysos in our sample that have measured fluxes at 3.6 @xmath13 m , 4.5 @xmath13 m , and 8.0 @xmath13 m , and has a contamination of less than 27% ( the majority of the objects in this color box with uncertain classifications may still be ysos ) . approximately equivalent results can be obtained by substituting the [ 3.6]@xmath21[8.0 ] color for the [ 4.5]@xmath21[8.0 ] color , but it is slightly more difficult to avoid picking up stars on the blue end of the [ 3.6]@xmath21[8.0 ] color axis . b07 identified a set of very bright ysos across the entire smc based on their 8.0 @xmath13 m magnitudes and [ 5.8]@xmath21[8.0 ] colors . if we select the same region of the color - magnitude diagram in n66 , we find 15 sources , only 3 of which are conclusively classified as ysos . almost all of the remaining objects have seds that are not well - fit by any of the models we apply , primarily because of pah contamination . the pah contamination across the rest of the galaxy should be much lower than in n66 , so these numbers clearly represent a lower limit to the fraction of actual ysos in the b07 sample . ysos are traditionally divided into classes based on their observed spectral indices as originally defined by @xcite . because spectral indices can vary with inclination angle as well as with evolutionary state , @xcite described their yso models in terms of `` stages '' , which are analogous to the usual classes , but are based on the physical quantities that define the evolutionary stage of the models . when comparing these models to data ( rather than considering only observations ) , it therefore makes sense to use the stage system . the definitions of each stage rely on the ratio of the disk mass and envelope accretion rate to the central stellar mass . stage i sources have @xmath32 , stage ii sources have @xmath33 and @xmath34 , and stage iii sources have @xmath33 and @xmath35 , where @xmath36 is the envelope accretion rate . note that we assumed a dust - to - gas ratio of @xmath37 ( 1/5 of the milky way value ) for n66 @xcite . using this classification scheme , we have grouped the 111 definite and probable ysos in n66 into stages . the seds of most objects can be fit by more than 1 yso model , so we first selected the models that produce a @xmath23 that is within 1 of the best @xmath23 for each object . we then computed a stage for each of the acceptable models and calculated a weighted average stage , using the @xmath23 values as weights . the averages were rounded to the nearest integer to produce a classification . we find that of the 111 ysos , 33 are stage i , 50 are stage ii , and 28 are stage iii . we also calculated yso masses with the same weighted averaging scheme , and defined the uncertainty on the mass to be the weighted standard deviation of the masses of the acceptable models . we found ysos with a range of masses from @xmath38 m@xmath1 , including 19 objects that appear to be proto - ob stars ( @xmath39 m@xmath1 ) . we list the names , positions , luminosities , luminosity uncertainties ( @xmath40l , the weighted standard deviation of the luminosities of the acceptable models ) , masses , mass uncertainties ( @xmath40 m , as defined in the first sentence of this paragraph ) , stages , and stage uncertainties ( @xmath40stage , the weighted standard deviation of the stages of the acceptable models ) of all of the ysos in table [ ysotable ] . in figure [ spatialdist ] we plot the spatial distribution of the ysos in n66 . the protostars are obviously highly concentrated towards the peaks of the 8.0 @xmath13 m emission . however , there is also star formation taking place outside of the dense dust cloud that marks the optical region , particularly to the south and southeast . @xcite hypothesized that star formation in n66 has proceeded from the southwest to the center of the present - day region , but we do find a few probable ysos southwest of n66 , indicating that at least modest star formation has taken place there within the past few million years . the distribution of ysos throughout the region as a function of mass and stage is not uniform . we find that the most - embedded objects ( stage i ) are slightly more concentrated towards the center of the region than the more advanced ( and presumably older ) ysos . we also see evidence for mass segregation , with the most massive objects exhibiting a strong preference for locations close to the center ( see figure [ massdist ] ) . all but 2 of the ysos with @xmath41 m@xmath1 lie on top of bright dust filaments in the main region , and many of them are coincident with molecular peaks and optical star clusters @xcite . if these massive ysos are indeed single objects rather than multiple unresolved sources , then the mass segregation must be primordial in origin , as the ysos have not had time to move very far from their birthplaces . the census of previously known embedded protostars in the smc consists of a single object discovered by @xcite . @xcite and @xcite used variability data and imaging to uncover a small sample of herbig ae / be stars in the smc , and @xcite discovered several hundred low - mass pre - main - sequence stars in n66 via isochrone fits to _ hst _ photometry . these objects , however , are generally much more evolved than the ysos identified by this study , many of which are still embedded in their natal dust clouds and are therefore faint or invisible at optical wavelengths . the distribution of ysos is concentrated towards clumps of molecular gas traced by the co(2 - 1 ) emission line @xcite and the peaks of the dust emission at 7 @xmath13 m @xcite . @xcite showed that dense h@xmath42 knots are associated with these molecular clumps and suggested that massive star formation could be taking place there . the large number of embedded ysos found in this study confirms that prediction . moreover , ir spectroscopy of the 3 brightest embedded sources detected with ground - based near - ir imaging confirm that these sources are ysos ( rubio and barb ' a , in preparation ) . the sample of stage i , ii , and iii ysos that we have identified in n66 presents the first opportunity for studying in detail a large sample of embedded ysos in another galaxy ( see also * ? ? ? * ; * ? ? ? * ) , and more importantly , one whose ism properties differ substantially from those of the milky way . the fact that we are able to identify over 100 ysos in n66 , as well as obtain successful sed fits for almost all of the other sources , suggests ( perhaps surprisingly ) that protostars in the smc resemble the yso models constructed by @xcite for milky way objects , even though the metallicity and dust - to - gas ratio are a factor of @xmath43 lower in the smc . the only potential difference between milky way and smc ysos that is evident in our results is that some of the sources are best fit by relatively cool photospheres , but the yso models with such photospheres do not have high enough luminosities to match their observed brightnesses at the distance of the smc . if these objects are actually ysos , this problem could be an indication that accretion is continuing even after these stars have reached the main sequence , which leads to expanded photospheres and lower temperatures compared to normal pre - main sequence tracks and zero - age main - sequence photospheres @xcite . an alternative possibility is that these sources actually consist of multiple cool protostars , which would explain their unusually high luminosities . the luminosities of these objects are @xmath44 orders of magnitude higher than the expected luminosities of individual ysos of the same temperature , however , which makes the multiplicity explanation appear unlikely . finally , evolved stars in the post - agb phase also have expanded cool photospheres , so it is possible that some of the `` cool luminous '' sources could be post - agb stars . additional modeling and observations of some representative objects may be necessary to resolve this issue . the 111 ysos in our sample have a combined stellar mass of 692 m@xmath1 . the observed mass function ( see figure [ imf ] ) turns over at @xmath45 m@xmath1 , indicating that incompleteness becomes serious at this point . artificial star tests in our photometry show that we are 90% complete even in the confused central region of n66 down to flux levels of 300 @xmath13jy , 200@xmath13jy , 200@xmath13jy , and 300 @xmath13jy from 3.6 - 8.0 @xmath13 m . applying these limits to the full library of yso models confirms that our incompleteness is severe below 4 m@xmath1 . if we assume that star formation in n66 follows a @xcite initial mass function ( imf ) down to 0.1 m@xmath1 , we calculate that the total mass in protostars for the entire region is @xmath46 m@xmath1 . in reality , this is a lower limit to the mass because even at 4 m@xmath1 the data are somewhat incomplete . if these ysos have all formed within the last @xmath47 myr , then the average star formation rate over that time is @xmath2 m@xmath1 yr@xmath3 . thus , we find that n66 comprises at least @xmath48% of the total current star formation in the smc . @xcite , @xcite , and @xcite used _ hst _ imaging to study the young stellar population of n66 in the optical . @xcite identified 16 subclusters of pre - main sequence stars ( their figure 8 ; several of these clusters were also pointed out by @xcite ) in the region , and we find that all but two of these also have ysos associated with them ( see below ) . @xcite estimated ages of @xmath9 myr for the first 15 of the subclusters , but our detection of ysos demonstrates that star formation has continued until the present day in these areas . subcluster 1 ( sc 1 ) is nearly coincident with an extremely bright _ spitzer _ source ( the 4th - most luminous object in the field at 8.0 @xmath13 m ) , ssts3mc 14.7725 - 72.1766 , which we are able to fit with both yso and agb seds with similar @xmath23 values . given the location of this object at the center of the ngc 346 cluster , surrounded by numerous very young massive stars , it is most likely a yso ( nevertheless , since it does not formally meet our selection criteria it is not included in our analysis ) . if so , the fitted seds suggest that this is a stage i object with a luminosity of @xmath49 l@xmath1 and a mass of 14.7 m@xmath1 . however , it is important to remember that because of the very high source density here we may actually be seeing multiple unresolved ysos . in that case , we would be likely to overestimate the mass of the most massive yso and underestimate the total mass of ysos contained in this source . this type of source confusion should not have a strong effect on the sed fitting results , because the observed sed will be dominated by that of the most massive embedded yso . a mid - ir spectrum of this object is displayed in ( * ? ? ? * peak c ) . another bright _ spitzer _ source , ssts3mc 14.7748 - 72.1749 , is located within 3 of sc 1 and has a steeply - rising mid- and far - ir sed that we are unable to fit successfully . if we remove the 24 @xmath13 m upper limit and either the 70 @xmath13 m or @xmath50- and @xmath51-band detections ( perhaps justified because of confusion in this very densely populated region ) , then this source has the sed of a very massive ( @xmath52 m@xmath1 ) , early - stage yso . we do not detect any non - stellar sources in sc 3 ( immediately south of sc 1 ) , but confusion as a result of the very bright source just north of the cluster may play a role in this non - detection . sc 2 is located very close to 2 _ spitzer _ sources . ssts3mc 14.7574 has a stellar sed with a strong 24 @xmath13 m detection ( indicating either confusion or circumstellar dust ) and ssts3mc 14.7580 - 72.1763 is formally a poor fit as a result of a low @xmath51-band flux and slight pah contamination , but appears to have an sed consistent with being an early stage yso . scs 4 - 6 each have bright infrared counterparts . sc 4 is coincident with ssts3mc 14.7605 - 72.1687 , a massive embedded yso . sc 5 is ssts3mc 14.7514 - 72.1681 , another object that can be fitted by both yso and agb seds . again , given its position in a very young cluster , it is most likely a massive stage i yso rather than an evolved star . and sc 6 contains 2 blended _ spitzer _ sources , ssts3mc 14.7371 - 72.1651 and ssts3mc 14.7380 - 72.1651 , both very high - mass stage i ysos . the caveat mentioned above about multiple unresolved sources applies here as well . located on the southern edge of sc 7 , very close to the prominent dust lane that arcs nearly halfway around n66 , is ssts3mc 14.7733 - 72.1835 , yet another very massive ( 18.6 m@xmath1 ) stage i protostar that can also be fit by agb models . slightly further to the southwest are 2 additional , lower - mass ysos , ssts3mc 14.7673 - 72.1834 and ssts3mc 14.7698 - 72.1846 . sc 8 has a faint _ spitzer _ counterpart , ssts3mc 14.7816 - 72.1802 , which shows a stellar sed in the optical and near - ir and excesses in the irac bands , but is not fit well by any of our yso models . it is possible that the optical and ir emission is coming from different sources . scs 9 - 11 are located around a quintet of ysos . the 2 brightest of these , ssts3mc 14.8112 - 72.1843 and ssts3mc 14.8130 - 72.1840 , lie within sc 10 , but it is not clear whether these are truly blended ysos as opposed to a bright , extended clump of dust . the other 3 sources classified as ysos in this region are ssts3mc 14.8041 - 72.1867 , ssts3mc 14.8068 - 72.1856 , and ssts3mc 14.8174 - 72.1866 . sc 12 contains two bright _ spitzer _ sources ( ssts3mc 14.8318 - 72.1890 and ssts3mc 14.8353 - 72.1892 ) that are not fit well by any of the available models . the rising seds of these objects towards long wavelengths strongly suggest that there are ysos present here , but blending may be a problem . this cluster is coincident with the source labeled peak i by @xcite . scs 13 - 15 lie in the northward extension of the region ( n66a ) . clusters 13 and 14 each overlap with a yso ( ssts3mc 14.8009 - 72.1663 = @xcite peak f and ssts3mc 14.8080 - 72.1578 = @xcite peak g ) , but as with sc 10 the source in sc 14 does not appear pointlike in the irac bands , and so may not be a single object . sc 15 is associated with ssts3mc 14.8205 - 72.1544 , which shows very strong pah emission but is not well - fit by yso models . finally , sc 16 contains 2 _ spitzer _ sources , but both of these are well - fit by normal stellar models , consistent with the older age of this cluster derived by @xcite . out of the 44 ysos in our sample that lie within the acs images of @xcite , @xmath53 are spatially coincident with gas or dust pillars strongly resembling those made famous by _ hst _ imaging of the `` pillars of creation '' in the eagle nebula @xcite . higher spatial resolution near - ir and mid - ir imaging of these objects may reveal exactly how the ysos are related to these features . we have obtained mid - infrared imaging of the region n66 ( ngc 346 ) in the smc with the irac and mips instruments on the _ spitzer space telescope_. we detected 8011 unique sources , with photometric coverage extending from @xmath16-band to 24 @xmath13 m ( 70 @xmath13 m in a few cases ) . most of these sources have colors and seds consistent with being normal stars , but we also detect a significant population of objects that are very red in the mid - ir . sed fitting of the 1645 sources with photometric measurements in at least four bands yielded 111 objects with seds that are best fit by yso models rather than stars or background galaxies . these data represent the first significant sample of embedded ysos identified in an external galaxy . we show that these ysos can be mostly , but not completely , separated from stars on the basis of their irac colors . however , sed fitting is necessary to significantly constrain the properties of individual objects . we find that the yso models of @xcite , which were designed to represent milky way ysos with solar metallicities , fit most of the ysos in n66 well , despite its much lower metallicity and dust - to - gas ratio . these results suggest that if low metallicity causes significant changes in the star formation process , the threshold for those effects must lie below the metallicity of the smc ( 12 + log[o / h ] @xmath54 8.0 ; @xcite ) . the one possible difference between smc and milky way ysos is that the smc appears to contain a population of very luminous but cool objects that have not been seen in the milky way . these objects may have expanded photospheres and hence lower temperatures than normal because they are still accreting material from their protostellar disks when they reach the main sequence . alternatively , these sources could be multiple cool ysos that are unresolved by _ spitzer_. we calculate masses and stages ( analogous to the usual yso classes ) for each of the n66 ysos , finding a range of masses from 2.4 to 16.6 m@xmath1 ( including 19 objects with masses above 8 m@xmath1 ) . almost half ( 45% ) of the ysos are stage ii objects , 30% are stage i , and the remaining 25% are evolved stage iii sources . we examine the spatial distribution of the ysos and find that they are strongly concentrated in the center of the region where bright and dust emission is seen , but there are also small numbers of ysos in the surrounding region with much less diffuse ism emission . the most massive ysos are preferentially located closer to the center of n66 , indicating that mass segregation is taking place . we compare our yso map to the clusters of pre - main sequence stars identified in the optical by @xcite and find that all but 2 of the clusters have associated ysos . using a salpeter imf , we calculate that a total of at least 3160 m@xmath1 of ysos have been formed in the last @xmath55 yr , representing @xmath56% of the current star formation in the smc . we would like to thank the anonymous referee for a careful reading of the paper that produced valuable feedback . j.d.s . gratefully acknowledges the support of a millikan fellowship provided by the california institute of technology . this research was partially funded by nasa through an award issued by jpl / caltech ( nasa - jpl _ spitzer _ grant 1264151 awarded to cycle 1 project 3316 ) . b.w . was supported by nasa astrophysics theory program grant nng05gh35 g and the _ spitzer _ theoretical research program under subcontract 1290701 , m.r . was supported by the chilean _ center for astrophysics _ fondap no . 15010003 , and financial support from fondecyt no . 1050052 is acknowledged by r.h.b . we thank you - hua chu , robert gruendl , lynne hillenbrand , jacco van loon , adam leroy , and bob benjamin for helpful conversations , and we also thank dimitrios gouliermis for providing us with the _ hst _ photometry in advance of publication . 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 . this publication makes use of data products from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center / california institute of technology , funded by nasa and the national science foundation . this research has also made use of nasa s astrophysics data system bibliographic services and the simbad database , operated at cds , strasbourg , france . cccccccccccc 1 & [email protected] & 00 58 00.94 & @xmath2172 05 22.3 & 1290 & 556 & 6.5 & 0.8 & iii & 0.6 & 0.26 & + 2 & [email protected] & 00 58 13.89 & @xmath2172 09 19.5 & 131 & & 5.1 & & i & & 1.76 & + 3 & [email protected] & 00 58 30.26 & @xmath2172 08 41.5 & 328 & 107 & 4.6 & 0.4 & ii & 0.4 & 0.12 & + 4 & [email protected] & 00 58 40.31 & @xmath2172 09 01.0 & 1590 & 1600 & 6.0 & 2.5 & ii & 0.5 & 0.40 & + 5 & [email protected] & 00 58 48.43 & @xmath2172 11 00.2 & 35 & 2 & 2.8 & 0.5 & i & & 4.06 & + 6 & [email protected] & 00 58 50.25 & @xmath2172 10 20.9 & 3300 & & 8.3 & & iii & & 5.53 & + 7 & [email protected] & 00 58 51.55 & @xmath2172 14 33.9 & 671 & 353 & 5.3 & 0.9 & iii & 0.5 & 0.95 & + 8 & [email protected] & 00 58 53.35 & @xmath2172 10 04.3 & 3520 & 789 & 8.9 & 1.5 & i & & 1.39 & + 9 & [email protected] & 00 58 55.56 & @xmath2172 10 24.3 & 4430 & & 9.0 & & iii & & 1.37 & + 10 & [email protected] & 00 58 55.90 & @xmath2172 14 38.1 & 261 & & 6.4 & & i & & 6.15 & + 11 & [email protected] & 00 58 58.30 & @xmath2172 10 26.9 & 4340 & & 7.9 & & i & & 3.55 & + 12 & [email protected] & 00 58 59.63 & @xmath2172 11 03.3 & 348 & & 4.5 & & ii & & 1.68 & + 13 & [email protected] & 00 59 00.22 & @xmath2172 10 47.0 & 1940 & & 7.3 & & iii & & 2.37 & + 14 & [email protected] & 00 59 00.81 & @xmath2172 10 58.6 & 388 & 1 & 4.7 & & ii & & 5.48 & + 15 & [email protected] & 00 59 01.58 & @xmath2172 10 19.8 & 2230 & & 7.5 & & ii & & 0.68 & + 16 & [email protected] & 00 59 01.96 & @xmath2172 16 17.5 & 306 & 813 & 4.6 & 1.1 & ii & 0.7 & 0.41 & + 17 & [email protected] & 00 59 02.53 & @xmath2172 10 07.4 & 5120 & 2210 & 10.3 & 0.8 & i & 0.4 & 3.69 & + 18 & [email protected] & 00 59 03.34 & @xmath2172 12 24.3 & 197 & 46 & 3.9 & 0.2 & ii & & 0.56 & + 19 & [email protected] & 00 59 04.75 & @xmath2172 11 04.7 & 684 & 7 & 5.5 & & ii & & 0.50 & + 20 & [email protected] & 00 59 06.37 & @xmath2172 13 48.0 & 369 & 214 & 4.5 & 0.7 & ii & 0.5 & 0.00 & + 21 & [email protected] & 00 59 07.19 & @xmath2172 08 43.4 & 261 & & 6.4 & & i & & 9.19 & + 22 & [email protected] & 00 59 07.90 & @xmath2172 12 32.9 & 459 & 151 & 4.9 & 0.4 & ii & & 6.59 & + 23 & [email protected] & 00 59 09.03 & @xmath2172 10 37.0 & 4430 & & 9.0 & & iii & & 1.44 & + 24 & [email protected] & 00 59 09.13 & @xmath2172 11 11.0 & 9880 & & 11.6 & & iii & & 13.16 & + 25 & [email protected] & 00 59 09.25 & @xmath2172 10 57.8 & 30300 & & 16.6 & & i & & 6.79 & + 26 & [email protected] & 00 59 09.55 & @xmath2172 10 39.6 & 1620 & & 7.0 & & iii & & 3.99 & + 27 & [email protected] & 00 59 10.40 & @xmath2172 11 34.9 & 6540 & 12500 & 7.9 & 8.3 & i & & 12.36 & + 28 & [email protected] & 00 59 12.22 & @xmath2172 09 58.8 & 5060 & & 10.1 & & i & & 9.19 & + 29 & [email protected] & 00 59 12.98 & @xmath2172 11 12.3 & 176 & 15 & 3.8 & & ii & & 0.81 & + 30 & [email protected] & 00 59 13.33 & @xmath2172 16 19.1 & 456 & 245 & 4.8 & 0.7 & iii & 0.5 & 0.01 & + 31 & [email protected] & 00 59 13.46 & @xmath2172 11 30.5 & 5990 & 2100 & 13.1 & 3.1 & i & & 6.52 & + 32 & [email protected] & 00 59 13.62 & @xmath2172 11 08.0 & 540 & 131 & 5.2 & 0.4 & ii & & 0.18 & + 33 & [email protected] & 00 59 13.85 & @xmath2172 08 47.7 & 728 & 159 & 5.6 & 0.3 & iii & & 5.52 & + 34 & [email protected] & 00 59 13.92 & @xmath2172 09 27.9 & 8380 & 1470 & 11.0 & 0.6 & ii & 0.7 & 8.32 & + 35 & [email protected] & 00 59 14.49 & @xmath2172 09 43.7 & 387 & & 4.7 & & ii & & 7.82 & + 36 & [email protected] & 00 59 14.68 & @xmath2172 11 03.4 & 13000 & 7940 & 15.1 & 2.9 & i & & 1.17 & + 37 & [email protected] & 00 59 15.11 & @xmath2172 11 02.2 & 9560 & 2180 & 11.4 & 0.8 & ii & 0.5 & 5.23 & + 38 & [email protected] & 00 59 15.36 & @xmath2172 11 30.0 & 3300 & & 8.3 & & iii & & 0.42 & + 39 & [email protected] & 00 59 16.18 & @xmath2172 11 11.6 & 210 & 61 & 4.0 & 0.2 & ii & & 1.00 & + 40 & [email protected] & 00 59 17.07 & @xmath2172 10 55.8 & 87 & 10 & 4.5 & 0.3 & i & & 6.43 & + 41 & [email protected] & 00 59 18.82 & @xmath2172 09 24.3 & 1180 & & 6.4 & & ii & & 4.08 & + 42 & [email protected] & 00 59 19.00 & @xmath2172 09 30.5 & 656 & & 5.5 & & ii & & 4.88 & + 43 & [email protected] & 00 59 19.25 & @xmath2172 14 38.8 & 453 & 174 & 4.9 & 0.8 & iii & 0.5 & 0.00 & + 44 & [email protected] & 00 59 20.95 & @xmath2172 08 41.5 & 922 & 345 & 6.2 & 0.4 & ii & 0.3 & 0.15 & + 45 & [email protected] & 00 59 23.16 & @xmath2172 08 48.4 & 4430 & 1780 & 8.8 & 1.7 & iii & 0.3 & 2.67 & + 46 & [email protected] & 00 59 23.69 & @xmath2172 08 47.1 & 204 & 60 & 3.9 & 0.3 & ii & & 2.78 & + 47 & [email protected] & 00 59 25.22 & @xmath2172 08 04.1 & 1340 & & 6.6 & & iii & & 0.40 & + 48 & [email protected] & 00 59 28.83 & @xmath2172 12 31.9 & 131 & & 5.1 & & i & & 6.78 & + 49 & [email protected] & 00 59 31.91 & @xmath2172 13 35.3 & 206 & & 5.9 & & i & & 5.28 & + 50 & [email protected] & 00 59 33.61 & @xmath2172 16 43.9 & 678 & 215 & 5.5 & 0.5 & iii & & 2.11 & + 51 & [email protected] & 00 59 36.94 & @xmath2172 09 03.2 & 77 & 11 & 4.6 & 0.3 & i & 0.4 & 3.33 & + 52 & [email protected] & 00 59 36.98 & @xmath2172 14 23.2 & 1000 & 495 & 5.8 & 1.5 & ii & 0.2 & 0.22 & + 53 & [email protected] & 00 59 44.64 & @xmath2172 12 44.0 & 2150 & & 7.5 & & iii & & 7.94 & + 54 & [email protected] & 00 59 50.48 & @xmath2172 16 35.2 & 644 & 249 & 5.4 & 0.6 & iii & 0.5 & 0.51 & + 55 & [email protected] & 00 59 52.12 & @xmath2172 15 40.8 & 529 & 376 & 5.0 & 0.9 & ii & 0.5 & 0.21 & + 56 & [email protected] & 00 59 56.59 & @xmath2172 14 43.5 & 683 & 177 & 5.9 & 0.8 & ii & 0.4 & 11.79 & + 57 & [email protected] & 01 00 15.27 & @xmath2172 16 28.7 & 1000 & & 6.2 & & iii & & 5.99 & + 58 & [email protected] & 01 00 15.72 & @xmath2172 04 43.6 & 360 & 87 & 4.6 & 0.4 & ii & & 0.26 & + 59 & [email protected] & 01 00 20.62 & @xmath2172 13 23.6 & 488 & 119 & 5.0 & 0.3 & ii & 0.5 & 7.91 & + 60 & [email protected] & 01 00 22.32 & @xmath2172 09 58.2 & 2910 & 944 & 8.0 & 0.7 & i & 0.5 & 0.40 & + 61 & [email protected] & 01 00 33.97 & @xmath2172 15 26.5 & 1750 & 99 & 8.8 & 0.4 & i & & 5.55 & + 62 & [email protected] & 00 57 57.49 & @xmath2172 14 59.7 & 12700 & 10900 & 11.4 & 5.3 & iii & & 1.95 & 3.49 + 63 & [email protected] & 00 58 02.00 & @xmath2172 05 00.7 & 2000 & 2130 & 7.0 & 2.3 & ii & 0.4 & 0.04 & 8.60 + 64 & [email protected] & 00 58 02.50 & @xmath2172 12 07.0 & 573 & 536 & 5.1 & 1.5 & ii & 0.8 & 0.02 & 4.23 + 65 & [email protected] & 00 58 09.85 & @xmath2172 08 41.2 & 160 & 107 & 4.0 & 0.8 & ii & 0.5 & 0.28 & 1.45 + 66 & [email protected] & 00 58 21.87 & @xmath2172 11 53.8 & 114 & & 5.0 & & i & & 5.85 & 10.50 + 67 & [email protected] & 00 58 23.01 & @xmath2172 14 50.0 & 95 & 93 & 4.3 & 0.9 & i & 0.4 & 0.60 & 7.06 + 68 & [email protected] & 00 58 38.23 & @xmath2172 14 37.0 & 210 & 300 & 3.7 & 1.1 & ii & 0.5 & 0.01 & 2.29 + 69 & [email protected] & 00 58 47.45 & @xmath2172 09 03.2 & 2290 & 188 & 7.6 & 0.2 & iii & & 5.99 & 14.70 + 70 & [email protected] & 00 58 49.60 & @xmath2172 06 58.0 & 1470 & 2230 & 6.5 & 1.2 & ii & 0.3 & 0.05 & 3.93 + 71 & [email protected] & 00 58 52.04 & @xmath2172 10 43.8 & 517 & 900 & 4.8 & 0.9 & ii & 0.2 & 2.20 & 7.18 + 72 & [email protected] & 00 58 52.16 & @xmath2172 07 45.4 & 260 & 317 & 4.3 & 0.6 & ii & 0.2 & 0.07 & 1.78 + 73 & [email protected] & 00 58 52.19 & @xmath2172 12 29.1 & 80 & 16 & 4.1 & 1.0 & i & & 4.49 & 12.55 + 74 & [email protected] & 00 58 52.24 & @xmath2172 12 22.6 & 37 & 69 & 2.9 & 0.8 & ii & 0.6 & 0.00 & 0.32 + 75 & [email protected] & 00 58 53.00 & @xmath2172 09 12.9 & 1640 & 307 & 7.0 & 0.4 & iii & & 0.06 & 1.01 + 76 & [email protected] & 00 58 54.36 & @xmath2172 11 44.6 & 59 & 105 & 2.4 & 4.6 & i & & 0.01 & 2.38 + 77 & [email protected] & 00 58 54.45 & @xmath2172 16 25.1 & 1170 & 861 & 6.4 & 2.3 & iii & 0.5 & 0.00 & 2.29 + 78 & [email protected] & 00 58 55.85 & @xmath2172 08 31.3 & 236 & & 4.0 & & i & & 3.05 & 11.44 + 79 & [email protected] & 00 58 56.89 & @xmath2172 09 54.5 & 9390 & 3200 & 13.5 & 1.8 & i & & 3.54 & 10.49 + 80 & [email protected] & 00 58 57.11 & @xmath2172 09 54.3 & 14800 & 7190 & 15.5 & 2.4 & i & & 0.18 & 1.96 + 81 & [email protected] & 00 58 57.19 & @xmath2172 08 41.5 & 231 & 370 & 4.1 & 1.0 & ii & 0.5 & 0.02 & 1.76 + 82 & [email protected] & 00 58 57.76 & @xmath2172 16 34.0 & 1210 & 874 & 6.4 & 0.6 & ii & 0.6 & 0.48 & 1.57 + 83 & [email protected] & 00 58 58.23 & @xmath2172 10 12.4 & 3290 & & 8.3 & & iii & & 0.32 & 1.64 + 84 & [email protected] & 00 59 01.01 & @xmath2172 13 30.1 & 257 & 469 & 4.2 & 0.8 & ii & 0.4 & 0.00 & 3.06 + 85 & [email protected] & 00 59 01.65 & @xmath2172 16 21.1 & 251 & & 6.5 & & i & & 2.77 & 8.52 + 86 & [email protected] & 00 59 01.86 & @xmath2172 16 17.4 & 726 & 1020 & 5.2 & 1.4 & ii & 0.6 & 0.00 & 6.20 + 87 & [email protected] & 00 59 04.14 & @xmath2172 11 00.1 & 2290 & 1870 & 7.0 & 1.9 & ii & 0.7 & 2.42 & 8.66 + 88 & [email protected] & 00 59 04.84 & @xmath2172 12 32.5 & 310 & 47 & 6.7 & 0.5 & i & & 0.96 & 3.43 + 89 & [email protected] & 00 59 07.54 & @xmath2172 08 12.8 & 176 & 151 & 3.6 & 0.7 & iii & 0.5 & 0.50 & 2.90 + 90 & [email protected] & 00 59 08.93 & @xmath2172 10 41.2 & 2360 & & 7.6 & & ii & & 0.20 & 6.33 + 91 & [email protected] & 00 59 09.17 & @xmath2172 08 38.7 & 201 & 914 & 3.9 & 1.2 & ii & 0.5 & 0.12 & 1.09 + 92 & [email protected] & 00 59 09.43 & @xmath2172 08 24.2 & 80 & 40 & 3.6 & 0.5 & ii & 0.5 & 0.10 & 7.14 + 93 & [email protected] & 00 59 09.60 & @xmath2172 05 27.9 & 821 & 226 & 5.8 & 0.4 & iii & 0.7 & 0.48 & 3.25 + 94 & [email protected] & 00 59 09.80 & @xmath2172 12 31.9 & 57 & 23 & 2.7 & 0.2 & ii & & 0.38 & 8.03 + 95 & [email protected] & 00 59 09.81 & @xmath2172 11 49.1 & 2270 & 1610 & 7.2 & 1.7 & iii & 0.5 & 0.49 & 9.59 + 96 & [email protected] & 00 59 09.87 & @xmath2172 12 18.8 & 162 & 74 & 3.7 & 0.3 & ii & 0.3 & 0.05 & 6.49 + 97 & [email protected] & 00 59 10.93 & @xmath2172 11 11.5 & 114 & 373 & 3.2 & 1.5 & ii & & 0.01 & 3.12 + 98 & [email protected] & 00 59 11.72 & @xmath2172 08 13.7 & 114 & & 5.0 & & i & & 6.32 & 12.05 + 99 & [email protected] & 00 59 12.13 & @xmath2172 11 38.1 & 171 & 54 & 4.1 & 0.7 & i & 0.5 & 1.69 & 8.27 + 100 & [email protected] & 00 59 12.25 & @xmath2172 11 36.7 & 305 & 228 & 4.2 & 0.8 & ii & & 0.07 & 7.34 + 101 & [email protected] & 00 59 22.23 & @xmath2172 14 45.8 & 275 & 170 & 4.7 & 0.6 & ii & 0.2 & 0.00 & 0.12 + 102 & [email protected] & 00 59 23.19 & @xmath2172 09 04.5 & 356 & 109 & 4.5 & 0.6 & ii & & 2.22 & 6.76 + 103 & [email protected] & 00 59 24.77 & @xmath2172 10 00.8 & 1160 & 90 & 7.0 & 0.7 & ii & 1.0 & 0.70 & 7.04 + 104 & [email protected] & 00 59 29.15 & @xmath2172 10 38.6 & 490 & 58 & 7.3 & 0.3 & i & & 2.46 & 5.53 + 105 & [email protected] & 00 59 29.49 & @xmath2172 08 21.6 & 123 & 45 & 4.4 & 0.8 & i & & 2.10 & 4.92 + 106 & [email protected] & 00 59 30.74 & @xmath2172 10 09.9 & 441 & 385 & 4.4 & 1.5 & iii & 0.5 & 0.97 & 9.78 + 107 & [email protected] & 00 59 31.60 & @xmath2172 14 18.6 & 540 & 963 & 4.6 & 1.6 & ii & 0.7 & 0.16 & 6.91 + 108 & [email protected] & 00 59 32.35 & @xmath2172 04 54.2 & 98 & 47 & 4.2 & 1.1 & i & & 1.85 & 6.99 + 109 & [email protected] & 00 59 53.70 & @xmath2172 06 36.3 & 201 & 215 & 4.3 & 0.7 & ii & 0.6 & 0.05 & 4.07 + 110 & [email protected] & 01 00 21.49 & @xmath2172 10 27.5 & 80 & 30 & 3.9 & 0.9 & i & 0.5 & 0.14 & 3.10 + 111 & [email protected] & 01 00 28.78 & @xmath2172 08 15.6 & 847 & 166 & 5.9 & 0.3 & iii & & 0.06 & 6.27 + all sources & 1645 + normal stars & 1322 + ysos & 61 + galaxies & 27 + agb stars & 6 + probable ysos & 50 + probable galaxies & 13 + probable agb stars & 5 + multiple good fits ; unable to classify & 13 + poor fits & 148 [ sourcetypes ]